&EPA
United States
Environmental
Protection Agency
Office of
Pollution Prevention
and Toxics
EPA 747-R-94-004A
March 1995
Laboratory Evaluation of Dust
and Dust Lead Recoveries for
Samplers and Vacuum Cleaners
Volume I:
Objectives, Methods, and Results
-------
March 1995
EPA 747-R-94-004A
LABORATORY EVALUATION OF DUST
AND DUST LEAD RECOVERIES FOR
SAMPLERS AND VACUUM CLEANERS
VOLUME I: OBJECTIVES, METHODS, AND RESULTS
Technical Program Branch
Chemical Management Division
Office of Pollution Prevention and Toxics
Office of Prevention, Pesticides, and Toxic Substances
U.S. Environmental Protection Agency
Washington DC 20460
Racyclad/Recyclabl* •Printed with Vegetable Based Inks on Recycled Paper (20% Posteonaumer)
-------
The material in this document has been subject to Agency technical
and policy review and approved for publication as an EPA report.
Mention of trade names, products, or services does not convey, and
should not be interpreted as conveying, official EPA approval,
endorsement, or recommendation.
11
-------
ACKNOWLEDGEMENTS
The project planning, laboratory sampling, chemical analysis, and statistical analysis
described in this report represent the joint effort of several organizations and many
individuals. The names of the principal authors and the contributors of the various
organizations are listed below.
Westat, Inc.
Designed the study and wrote portions of and compiled the
Quality Assurance Project Plan for the project. Upon com-
pletion of the laboratory and chemical analysis, Westat
performed the statistical analysis and wrote portions of the
final report. John Rogers was the Work Assignment
Manager, responsible for the management of the work and
for the statistical analysis. Rick Rinehart provided expertise
on the design, interpretation, and review of the study results
and wrote sections of the report. Brian Dietz was responsible
for the preparation of the data files and review of the statisti-
cal procedures and statistical sections of the report.
Midwest Research
Institute (MRI)
Conducted the laboratory sampling and chemical analysis for
the pilot study and full study and wrote portions of the
Quality Assurance Project Plan and the final report. Paul
Gorman was the Project Leader with responsibility for
testing, and directing activities of laboratory personnel. Paul
Constant had overall management and fiscal responsibility
and reviewed all work products. Jack Balsinger and Karin
Bauer, under his direction, were responsible for reviewing
the data, conducting quality assurance audits, writing
sections on the QA results, and reviewing the reports.
EPA
EPA Staff had overall responsibility for this project and
played an active role in directing the pilot study and full
study. Principal EPA contributors included Ben Lim, Work
Assignment Manager, and John Schwemberger, Deputy
Work Assignment Manager.
111
-------
The authors of this report would like to acknowledge the many people that
contributed to this study. Staff members from several organizations were involved
in all steps of this project, formerly known as the Wipe and Vacuum Study, from
the preparation of the Quality Assurance Project Plan to the write-up of the Final
Report. First, we gratefully acknowledge the invaluable technical assistance and
administrative support of Janet Remmers, Jill Hacker, and Philip Robinson, EPA
Project Officers. For overall project management support, we thank Paul Constant
of MRI and Stephen Dietz of Westat.
The success of the project also depended directly on the staff which carried out the
various tasks. We wish to thank all EPA, MRI, and Westat staff who collected and
donated household dust. We also wish to acknowledge the Midwest Research
Institute staff responsible for laboratory sampling and chemical analyses, including
Dave Miles, Laboratory Supervisor, and Rick Moore and D. Chambers, Laboratory
Technicians.
Many other individuals contributed their time and effort to this challenging project.
The Quality Assurance Project Plan preparation was directed by Bill Devlin of
Westat and reviewed by Jay Glatz of EPA, Jack Balsinger of MRI, and David
Morganstein of Westat. Technical editing was provided by the Westat Editorial
Support Group.
IV
-------
CONTENTS
EXECUTIVE SUMMARY xv
1 INTRODUCTION 1
1.1 Purpose of the Project 2
1.2 Overview of the Report 3
2 BACKGROUND 5
2.1 Importance of House Dust Particle Size 5
2.2 House Dust Sampling and Collection Methods 6
2.2.1 Wipe and Vacuum Sampling Methods Used in this Study 6
2.2.2 History of Dust Sampling Methods Used by OPPT 7
2.2.3 Commercial Vacuum Cleaners 8
2.3 Previous Studies that Compared Samplers and Vacuum
Cleaners in the Laboratory and Field 8
2.3.1 Laboratory Comparison Studies: Real versus Artificial
House Dust 9
2.3.2 Laboratory Comparison Studies: Test Surfaces and
Collection Efficiency 11
2.3.3 Field Comparison Studies 11
3 STUDY OBJECTIVES 13
3.1 Questions to be Answered with the Data 13
3.1.1 Samplers 14
3.1.2 Vacuum Cleaners 14
3.2 Data Quality Objectives 15
4 STUDY DESIGN AND SAMPLE COLLECTION PROCEDURES 17
4.1 Study Design 17
4.2 Description of House Dust Used in Study 34
4.3 Fiber Preconditioning of Carpet and Upholstery 35
4.4 Dust Preconditioning of all Substrates.... 36
4.5 Statistical Design 37
4.6 Vacuum Cleaner Tests 39
4.7 Sampler Tests 42
4.8 Vacuum Cleaner Exhaust Emission Testing 47
-------
CONTENTS (Continued)
5 LABORATORY ANALYSIS PROCEDURES 49
5.1 Lead Analysis of Sieved Dust 49
5.2 Lead Analysis of Dust Samples from Vacuum Cleaner and
Sampler Tests 49
6 RESULTS 51
6.1 Summary of Results from the Pilot and Preconditioning Data 51
6.1.1 Pilot Test Results 51
6.1.2 Preconditioning Results 52
6.2 Test Dust Characteristics 52
6.2.1 Dust Recovery by Particle Size Class for Older and Newer
Homes 52
6.2.2 Lead Concentration by Particle Size Class for Older and
Newer Homes 55
6.3 Samplers 55
6.3.1 Sampler Dust Recovery 58
6.3.2 Sampler Lead Recovery 62
6.4 Commercial Vacuum Cleaners 67
6.4.1 Dust Recovery 69
6.4.2 Vacuum Cleaner Lead Recovery 73
6.4.3 Effect of Cleaning Effort 76
6.4.4 Exhaust Emissions 78
6.5 Sampling and Measurement Error 80
7 DISCUSSION OF RESULTS 81
7.1 Test Dust Characteristics 81
7.2 Samplers 82
7.3 Commercial Vacuum Cleaners 83
7.4 Effect of Sampling Method on Estimates from the National
Survey of Lead-Based Paint in Housing (HUD National Survey) 86
7.5 Additional Questions 88
7.6 Final Comments 89
8 DATA PROCESSING AND STATISTICAL ANALYSIS PROCEDURES 91
8.1 Data Entry and Data Processing Procedures 91
8.2 Statistical Analysis Procedures 94
8.2.1 Overview of the Statistical Analysis Procedures 94
VI
-------
CONTENTS (Continued)
8.2.2 Statistical Analysis of Sieved Dust Lead Concentration 101
8.2.3 Statistical Analysis of Gravimetric and Lead Analysis Data
for Samplers 105
8.2.4 Statistical Analysis of Gravimetric and Lead Analysis Data
for Vacuum Cleaners 108
8.2.5 Statistical Analysis of Vacuuming Effort Data 115
8.2.6 Statistical Analysis of Sampling and Measurement
Precision 117
9 QUALITY ASSURANCE 127
9.1 System Audit 127
9.1.1 Vacuuming Task 127
9.1.2 Analytical Task 127
9.2 Performance Audits 127
9.2.1 Performance Evaluation Sample Results 128
9.3 Data Audit 132
9.3.1 Vacuum Weight Data 135
9.3.2 Analytical Data 135
9.4 Data Assessment 136
9.4.1 Sample Preparation QC Data 136
9.4.2 Instrumental Analysis QC Data 140
9.4.3 Statistical Analysis QC Results 144
APPENDIX A: PILOT TESTS RESULTS FOR THE WIPE AND VACUUM
STUDY A-l
Al.O INTRODUCTION A-l
Al.l Objectives A-2
A2.0 STUDY DESIGN AND PROCEDURES A-4
A2.1 TASK 1—Determine Stability of Tare Weights for New
Clean Vacuum Cleaner Bags A-5
A2.2 TASK 2—Demonstrate Method of Securing Carpet and
Upholstery Substrates for Testing Vacuum Cleaners A-7
A2.3 TASK 3—Determine if Preconditioning Procedures are
Feasible for Using New Carpet in the Laboratory Tests, and
Determine if use of the Same Carpet for Each Series of
Tests is Feasible A-7
A2.4 TASK 4—Determine the Amount of Dust Needed for the
Tests A-ll
A2.5 TASK 5—Develop and Demonstrate Method for
Measuring Exhaust Emissions from Vacuum Cleaners A-13
A3.0 DATA COLLECTION A-15
vn
-------
CONTENTS (Continued)
A3.1 Pilot Test Data Collection A-15
A3.2 Quality Assurance Activities A-15
A4.0 STATISTICAL DATA ANALYSIS A-16
A4.1 Analysis of Task 1 A-16
A4.2 Analysis of Task 2 A-23
A4.3 Analysis of Task 3 A-23
A4.4 Analysis of Task 4 A-30
A4.5 Analysis of Task 5 A-42
A5.0 DISCUSSION OF RESULTS A-49
A6.0 LABORATORY DATA A-52
A6.1 TASK 1: Determine the stability of tare weights for new,
clean vacuum cleaner bags A-52
A6.2 TASK 2: Demonstrate method of securing carpet and
upholstery substrates for testing vacuum cleaners A-58
A6.3 TASK 3: Determine if preconditioning procedures are
feasible for using new carpet in the laboratory tests, and
determine if the use of the same substrate for each series
of tests is feasible A-59
A6.4 Determine the amount of dust needed for the tests A-64
A6.5 Develop and demonstrate a method for measuring
exhaust emissions A-69
APPENDIX B: PRECONDITIONING DATA B-l
Bl Fiber Preconditioning B-l
Bl.l Fiber Preconditioning on Carpets B-l
B1.2 Fiber Preconditioning on Upholstery B-4
B1.3 Fiber Preconditioning Data B-9
B2 Dust Preconditioning B-16
B2.1 Dust Preconditioning Data B-19
APPENDIX C: SIEVED DUST DATA C-l
APPENDIX D: SAMPLER DATA D-l
APPENDIX E: VACUUM CLEANER DATA E-l
Vlll
-------
TABLES
Table 4-1 Factors affecting dust and lead recovery 19
Table 4-2 Combinations of substrate and dust particle class tested in the
study 38
Table 4-3 Test sequence for vacuum cleaner tests by team 1 40
Table 4-4 Test sequence for vacuum cleaner tests by team 2 41
Table 4-5 Test sequence for sampler tests by team 1 45
Table 4-6 Test sequence for sampler tests by team 2 46
Table 6-1 Percent of dust in each particle size class, for older and newer
homes 53
Table 6-2 Geometric mean dust lead concentration (mg/g) by dust particle
size, with approximate 95% confidence intervals 56
Table 6-3 Average sampler dust recovery by sampler, with 95% confidence
intervals 59
Table 6-4 Sampler dust recovery by sampler and dust particle size class,
with 95% confidence intervals 61
Table 6-5 Average sampler lead recovery by sampler, dust particle size
class, and dust loading 66
Table 6-6 Sampler concentration ratio by dust particle size, with 95%
confidence interval 68
Table 6-7 Predicted average vacuum cleaner dust recovery for tested
substrates and vacuum cleaners, with 95% confidence intervals 72
Table 6-8 Average vacuum cleaner lead recovery by vacuum cleaner and
substrate, with 95% confidence intervals 75
Table 6-9 Average vacuum cleaner exhaust dust concentrations by
vacuum cleaner 79
Table 8-1 Critical values for the extreme studentized residual (5% level) 98
Table 8-2 Variance components for lead measurements 118
Table 8-3 ICP variance components which can be estimated from each type
of sample 119
IX
-------
TABLES (Continued)
Table 9-1 Percent recoveries of blind performance evaluation samples 129
Table 9-2 SKM 1646: percent recovery of 16 blind control samples 133
Table 9-3 Method digestion blank results 137
Table 9-4 Method spike replicate results 138
Table 9-5 Initial calibration verification sample (ICV) results 141
Table 9-6 Continuous calibration verification (CCV) sample results 142
Table A-l Regression estimates for predicting the weight of vacuum
cleaner bags as a function of time, relative humidity, and
temperature for data collected on Day 1 A-18
Table A-2 Regression estimates for predicting the weight of vacuum
cleaner bags as a function of time, relative humidity, and
temperature for data collected on Day 2 A-21
Table A-3 Dust recovery on the first vacuuming as a function of the dust
loading A-41
Table A-4 Average exhaust concentrations for each vacuum cleaner
exhaust test A-47
Table A-5 Tare weight of vacuum cleaner bags at one minute intervals for
Taskl A-54
Table A-6 Tare weight of sampling cassette at one minute intervals for
Taskl A-55
Table A-7 Tare weight of vacuum cleaner bag at 30 minute intervals for
Taskl A-56
Table A-8 Tare weight of vacuum cleaner bags after one minute of use, for
Taskl A-57
Table A-9 Fiber preconditioning, Task 3 A-61
Table A-10 Dust carry-over test for Task 3 A-62
Table A-ll Testing different amounts of dust and vacuuming the wand for
Task 4 A-66
Table A-12 Exhaust emissions data for Task 5 A-70
-------
TABLES (Continued)
Table B-l Predicted 40-second fiber uptake from carpets by substrate sample
and vacuum cleaner B-5
Table B-2 Predicted 40-second fiber uptake from upholstery by substrate
sample and vacuum cleaner B-8
Table B-3 Fiber Preconditioning Data B-10
Table B-4 Average dust recovery by vacuum and substrate, with 95%
confidence intervals B-18
Table B-5 Dust Preconditioning Data B-20
Table C-l Seived Dust Data C-2
Table D-l Sampler Data D-2
Table E-l Vacuum Cleaner Data E-2
FIGURES
Figure 4-1 Vacuum cleaner A 20
Figure 4-2 Vacuum cleaner B 21
Figure 4-3 Vacuum cleaner C 22
Figure 4-4 Vacuum cleaner D 23
Figure 4-5 Wipe sampling 24
Figure 4-6 Blue Nozzle sampler 25
Figure 4-7 BRM sampler 26
Figure 4-8 CAPS sampler 27
Figure 4-9 Tile substrate 28
Figure 4-10 Linoleum (sheet vinyl) substrate 29
Figure 4-11 Wood flooring substrate 30
Figure 4-12 Upholstery substrate 31
XI
-------
FIGURES (Continued)
Figure 4-13 Carpet substrate 32
Figure 4-14 Platform for substrates 33
Figure 6-1 Histogram of relative dust weight by dust particle size for
composite dust samples from newer and older homes 54
Figure 6-2 Geometric mean dust lead concentration by dust particle size,
with approximate 95% confidence interval 56
Figure 6-3 Histogram of relative lead weight by dust particle size for
composite dust samples from newer and older homes 57
Figure 6-4 Sampler dust recovery by sampler, with 95% confidence
intervals, averaged across all substrates 59
Figure 6-5 Sampler dust recovery by sampler and dust particle size 60
Figure 6-6 Average sampler lead recovery by dust particle size class, with
95% confidence intervals, and by dust particle size class and dust
loading 63
Figure 6-7 Average sampler lead recovery by dust particle size class, with
95% confidence intervals, and by sampler and dust particle size
class 64
Figure 6-8 Average sampler lead recovery by dust loading, with 95%
confidence intervals 65
Figure 6-9 Sampler concentration ratio by dust particle size, with 95%
confidence interval 68
Figure 6-10 Predicted average vacuum cleaner dust recovery for tested
substrates with 95% percent confidence intervals and average
dust recovery using alternate definitions of dust recovery 70
Figure 6-11 Predicted average dust recovery for tested vacuum cleaners with
95% percent confidence intervals 71
Figure 6-12 Average vacuum cleaner lead recovery by vacuum cleaner and
substrate, with 95% confidence intervals 74
Figure 6-13 Dust recovery versus vacuuming effort for six substrates 77
Figure 7-1 Estimated number of priority homes with children under seven
as a function of the recovery of other samplers relative to the
Blue Nozzle sampler 87
XII
-------
FIGURES (Continued)
Figure 8-1 Flow chart for data entry and verification of gravimetrics data 92
Figure 8-2 Flow chart for data entry and verification of lead analysis data 93
Figure 8-3 Relative standard deviation of lead concentration
measurements for sieved dust by dust particle size class and age
of home, as predicted by the theoretical model 104
Figure 8-4 Standard deviation of the variance components as a function of
lead concentration in the instrument sample after any dilution 121
Figure 8-5 Predicted coefficient of variation (cv) associated with lead
measurement as a function of the instrument response and
histogram of the observed instrument response for the dust
samples in the study 123
Figure 8-6 Measurement bias for each instrument batch as a function of the
known lead concentration 124
Figure 9-1 Historical X-bar (mean) QC chart for SRM 1646 recovery 130
Figure 9-2 X-Bar (mean) QC chart for SRM 1646 recovery showing revised
control and warning limits 131
Figure 9-3 Historical X-bar (mean) QC chart for SRM 2704 recovery 134
Figure 9-4 Method spike replicates: recovery (%) 139
Figure 9-5 Method spike replicates: recovery range (%) 139
Figure 9-6 Continuous calibration verification sample (CCV) plots 143
Figure A-l Measured and predicted weights of vacuum cleaner bags over
time from Day 1 A-19
Figure A-2 Measured and predicted weights of vacuum cleaner bags over
time from Day 2 A-22
Figure A-3 Standard deviation of one vacuum cleaner bag weight
measurements as a function of equilibration time A-24
Figure A-4 Standard deviation of one cassette weight measurement as a
function of equilibration time A-25
Figure A-5 Weight increase of vacuum bag during fiber preconditioning A-27
Xlll
-------
FIGURES (Continued)
Figure A-6 Histogram and time series plot of recovery measurements for
TaskS A-29
Figure A-7 Time series plot of dust applied and dust removed from the
carpet sample in Tasks 3 and 4 A-32
Figure A-8 Estimated coefficient of variation of the weight change of a bag
from vacuum cleaner A as a function of the weight change A-37
Figure A-9 Dust recovery versus vacuuming effort for dust deposited on the
carpet and dust deposited and ground into the carpet A-39
Figure A-10 Dust recovery versus vacuuming effort for dust deposited with
and without grind-in compared to fixed removal per
vacuuming A-40
Figure A-ll Weight of dust recovered from the vacuum cleaner bag as a
function of the weight of the dust deposited A-43
Figure B-l Five minute weight gain due to fibers versus cumulative
vacuuming time B-3
Figure B-2 Five minute weight gain due to fiber as a function of
cumulative vacuuming time B-6
Figure B-3 Average preconditioning dust recovery by vacuum and
substrate, with 95% confidence intervals B-17
xiv
-------
EXECUTIVE SUMMARY
Introduction
This project to expand the knowledge on household dust testing methods was
undertaken by the U.S. Environmental Protection Agency (EPA) as part of a major
national effort to address the public health issue of childhood lead poisoning. The
effort was given impetus by the CDC's statement on lead poisoning, which reduced
the level of concern for blood lead levels from 25 micrograms/deciliter ((ig/dl) to 10
H-g/dl. It has also been given impetus by the passage of the "Residential Lead-Based
Paint Hazard Reduction Act of 1992," also known as "Title X." In response to Title
X, EPA is proceeding towards the development of health-based standards for house
dust lead levels. To do this, appropriate methods for sampling house dust are
needed. As part of this effort, numerous questions about house dust sampling have
risen. This study was designed to address some of these questions.
This project was undertaken by the EPA Office of Pollution Prevention and Toxics
(OPPT) to evaluate house dust sampling methods and to assess the efficacy of typical
household vacuuming on removing leaded dust from residential surfaces. Dust-
lead sampling results from the National Survey of Lead-Based Paint in Housing
(HUD National Survey) are reexamined, based on new information collected in this
study about the performance of the dust sampler used during that survey.
Lead-contaminated house dust is considered one of the most significant sources of
childhood lead exposure in the United States. Millions of children live in dwellings
with hazardous dust-lead levels and routinely put dust-laden fingers, toys, and
other objects into their mouths. Although the potential hazards of house dust are
well recognized, it is currently unknown which dust sampling procedures and
methods are best for assessing residential lead hazards. It is also unknown how
effectively typical vacuuming lowers dust lead levels in the home.
The primary reason for the lack of scientific consensus on the best method to sample
house dust is that several recent studies, including this one, have shown that differ-
ent samplers give different results under the same circumstances. For example,
studies have shown that the amount of dust picked up by one sampling method
may be either considerably greater or considerably less than that dust picked up by a
second method. Therefore, any conclusions about the level of lead hazard posed by
dust may differ depending on the sampling method used. The situation is further
complicated by the fact that the previous studies designed to evaluate the perfor-
mance of house dust samplers are not directly comparable. Since interpretable
house dust lead measurements are needed by the Federal government to draft
guidelines to address lead hazards in housing, a standardized laboratory procedure
to characterize samplers was needed. Such a procedure was developed for use in
this study.
xv
-------
The test procedures developed for this study proved easy to implement. It is
recommended that they be duplicated by other researchers testing the performance
of house dust collection devices. By using the same standardized test procedures, a
baseline can be established for samplers and vacuum cleaners and the results from
future evaluation studies can then be compared to the baseline. In this way, infor-
mation from many studies can be combined to make the most appropriate decisions
on how to address lead hazards in housing.
Two standardized laboratory testing procedures were developed for this study. The
first procedure was designed to characterize the performance of house dust
samplers. The second was designed to evaluate how well commercially available
vacuum cleaners collect dust from various surface types. Three vacuum sampling
methods and one wipe sampling method were tested by the first procedure. These
methods included the "Farfel modified" High Volume Small Surface Sampler used
in the Baltimore Repair and Maintenance study (called the BRM sampler is this
report), the Comprehensive Abatement Performance Study (CAPS) cyclone sampler,
the Blue Nozzle sampler, and the Department of Housing and Urban
Development's (HUD) wipe sampling method. All of these sampling methods
have been used in previous EPA/OPPT studies. The second procedure was used to
characterize four commercially available household vacuum cleaners ranging in
price from $120 to $800. The most expensive vacuum cleaner was equipped with a
high efficiency particulate air (HEPA) filter. The protocols for both testing proce-
dures included using real house dust sieved into six particle size classes ranging
from 0 to 2,000 microns in size. The dust was applied to five substrates commonly
encountered inside a residence: tile, wood flooring, linoleum (sheet vinyl), uphol-
stery, and carpet.
Results and Conclusions
Test Dust Characteristics: The test dust used in this study was obtained from
volunteers who donated full vacuum cleaner bags of house dust. Bags were
collected from homes within two age groups: older homes built before 1963 when
lead paint was common and newer homes built after 1982, several years after lead
was banned from household paint. As expected, the dust from older dwellings had
a higher lead concentration than the dust from newer homes, with mean lead
concentrations of 474 parts per million (ppm) and 61 ppm, respectively, for all dust
particles smaller than 2,000 microns. The two groups of house dust contained
roughly the same proportions of total dust, by weight, in each particle size class.
However, the distribution of lead concentration by particle size class was dramati-
cally different between the two age groups. This result was unexpected and has not
been demonstrated in previous studies.
Most studies that have examined lead in house dust by particle size suggest that lead
concentrations in dust increase as particle size decreases. In this study, the lead
concentration in dust collected from newer homes follows the expected inverse
xvi
-------
relationship with particle size, but the lead concentrations in dust from the older
homes did not exhibit the same relationship. Lead concentrations in the dust
collected from pre-1963 homes remained relatively stable across particle sizes. One
possible explanation for this observation is that house dust from pre-1963 homes
likely contains some deteriorated lead-based paint, while dust from newer homes
does not. If deteriorated paint dust particles are larger and more variable in size
than other lead sources, such as soil or street dust, then the inverse relationship
between lead and particle size may disappear in the dust contaminated by deterio-
rated lead-based paint.
Samplers: The performances of one wipe sampling method and three vacuum
sampling methods were evaluated in this study. The vacuum samplers were tested
for total dust recovery (total dust cannot be measured by wipes) and all four
samplers were tested for lead recovery. Recovery is the amount of dust or lead
collected from the substrate as a percentage of the amount deposited on the
substrate. Tests were differentiated by substrate, by the nominal lead concentration
of the dust applied to the substrate (high and low lead concentration dust from older
and newer homes, respectively), by the dust loading levels (100 and 400 mg/sq ft.),
and by the dust particle size.
The results indicate that the BRM and the CAPS cyclone produced the highest dust
recoveries across all substrates and particle sizes. The recovery differences between
the two cyclone devices were not significant. The Blue Nozzle sampler had the
lowest recoveries, which were statistically significantly lower than for the cyclone
samplers. These results agree with findings from previous studies that indicate that
the Blue Nozzle sampler has lower dust recovery than other tested methods. The
average dust recoveries for the BRM, CAPS, and Blue Nozzle samplers were 89%,
84%, and 30%, respectively. The results also suggest that the measurements from
BRM and CAPS cyclone samplers are more precise than those from the Blue Nozzle
sampler.
The pattern of lead recovery across samplers was similar to dust recovery. In order
of decreasing lead recovery, the lead recoveries of the BRM, CAPS cyclone, wipe, and
Blue Nozzle samplers are 81%, 72%, 63%, and 26%, respectively. The lead recovery
for the BRM, CAPS, and wipe sampler are all significantly greater than for the Blue
Nozzle sampler. Differences in the recovery among different substrates were not
statistically significant.
The best methods for measuring lead in house dust depend on many factors, two of
which are dust particle size and substrate. It is clear from this study and others that
the selection of the sampling method is a critical factor. The differences among
samplers have particular application to the interpretation of health-based standards
and on the results from the HUD National Survey which are discussed below.
Commercial Vacuum Cleaners: Commercially available vacuum cleaners with
beater bar attachments for carpets were tested for total dust and lead pickup capabili-
xvn
-------
ties. The same test dust and substrates used for the samplers were also used for the
vacuum cleaners. For the vacuum cleaner tests, the dust loading in mg/sq ft was
the same as for samplers, but the size of the test area was larger so the amount of
dust applied was greater than for the sampler tests.
The overall ability of the vacuum cleaners to collect dust was, as expected, greatest
for the hard substrates and poorest for carpets. The average dust recovery ranged
from 76% on carpets with ground-in dust to 93% on wood substrates. The average
recovery for a particular substrate also varied among vacuum cleaners.
While it was simple to weigh the total dust collected in vacuum cleaner bags and
thus measure dust recovery, measurement of lead recovery proved difficult because
it was not possible to remove all of the dust from the vacuum cleaner bag for lead
analysis. It also was not feasible to measure the lead in the dust without removing
the dust from the bag. Based on the procedure developed for this study, which
analyzed only the portion of dust that could be shaken out of the bag, the overall
average lead recovery was 103%. The lead recoveries varied greatly and depended
on the combination of vacuum cleaner and substrate used in each test.
The vacuum cleaner tests also assessed the effect of vacuuming effort on dust
recovery. On all substrates, most of the dust applied to the substrate was recovered
within the first 40 seconds of vacuuming (over an area of 6.8 square feet). Although
additional vacuuming collected more dust, the effect of that effort was significant
only for carpets with ground-in dust.
The results of this study show that a highly rated vacuum cleaner with a beater bar
attachment will pick up at least three-quarters of the loose dust present on a variety
of surfaces within a moderate vacuuming time. The amount of additional dust
picked up depends on many factors, such as the vacuum cleaner design and whether
or not the dust is ground into the surface. The study suggests that lead recovery is
similar to the dust recovery. Because this is a laboratory study, no information is
available on how quickly dust accumulates in the home or whether acceptable
levels of dust lead can be maintained with regular vacuuming. It is clear that
vacuuming removes dust and leaded dust from the vacuumed surfaces, thereby
reducing the total amount of lead which might pose a risk to young children. It is
also clear from previous studies that lower lead loadings are correlated with lower
blood lead levels in children. Even though vacuuming removes leaded dust which
might be ingested by a child, however, we cannot definitely say that routine
vacuuming will reduce leaded dust in a way that will result in reduced blood lead
levels.
Tests were also conducted on the exhaust from the vacuum cleaners. The results
showed that 0.02% or less of the dust sucked into the vacuum cleaner hose passed
through the vacuum cleaner bag. The smallest dust particle size (<53 fim) was used
to represent a worst case situation. The exhaust from the vacuum cleaner equipped
with the HEPA filter had lower dust concentration than the ambient air. Although
XVlll
-------
these results indicate that very little of the dust passed through the vacuum cleaner
bag in the four tested vacuum cleaners, more research is required to determine
whether this result can be extended to other models and old vacuum cleaners found
in many homes.
Effect of Sampling Method on Estimates from the HUD National Survey of Lead-
Based Paint: The Blue Nozzle sampling method was used in the HUD National
Survey of Lead-Based Paint to estimate the number of priority homes in the U.S.
with children under seven years old. Priority homes are classified as private
dwelling units containing lead-based paint, with either non-intact paint present or
dust loading levels exceeding the HUD post-abatement clearance guidelines. Based
on the relative recoveries of the different samplers tested in this study, the number
of priority homes which would have been identified if the wipe sampling method
had been used in the HUD National Survey was calculated. HUD recommends that
wipe sampling be used for post-abatement clearance testing.
The number of priority homes with children under seven was reported as 3.8
million in the 1990 HUD Comprehensive and Workable Plan to Congress. This
number was later revised to 4.0 million after new information was included on the
performance of the x-ray fluorescence (XRF) instruments used to detect the presence
of lead-based paint. Based on the results of this study, the number of priority homes
with children under seven would be 4.6 million if wipe sampling techniques had
been used during the HUD National Survey.
Additional Questions
Some of the issues and questions raised by this study which have yet to be answered
include the following:
• In dust from older homes, the lead concentration was found to be
similar for all dust particle sizes except the largest size which had a
higher lead concentration. This relationship was based on dust
composited from vacuum cleaner bags from seventeen homes.
Additional studies of dust collected from individual homes may
provide information on the extent to which this relationship can be
generalized to all older homes.
• If it is determined that vacuuming can reduce the lead hazard from
floor dust without increasing the hazard from other sources, another
question to answer is: what vacuuming frequency is necessary to
adequately control dust and lead loading?
• For the vacuum cleaners, roughly 2% to 5% of the dust deposited on
the substrate was not accounted for. This dust may have been caught
in parts of the vacuum cleaner other than the bag, become airborne,
xix
-------
been deposited on surfaces other than the vacuumed area, or been
caught in the substrate so as not to be removable with extensive
vacuuming. Where is this dust and might it pose a threat to children?
• The extent to which the vacuum cleaner and its exhaust disturb dust,
making it airborne and creating a temporary lead hazard, has yet to be
determined. How much dust is kicked up by routine vacuuming? Is it
hazardous to young children? How soon does the airborne dust
resettle, and how soon after vacuuming are airborne dust and lead
levels safe for children? Does the vacuuming and/or exhaust cause the
uncollected dust to move to areas which provide an increased or
decreased lead risk to children?
These questions provide direction for future research.
xx
-------
1 INTRODUCTION
This project to expand the knowledge on household dust testing methods has been
undertaken by the U.S. Environmental Protection Agency (EPA) as part of a major
national effort to address the Public Health issue of lead poisoning. The EPA, the
U.S. Department of Housing and Urban Development (HUD), the Centers for
Disease Control (CDC), the Occupational Safety and Health Administration (OSHA),
and numerous other Federal, state, municipal, county, industry, and private
agencies have been mobilized in an effort to reduce the preventable occurrence of
lead poisoning, particularly in children. This effort has been given impetus by both
the CDC's statement on lead poisoning, which reduced the level of concern for
blood lead levels from 25 micrograms/deciliter (fig/dl) to 10 [ig/dl, and by the
passage of the "Residential Lead-Based Paint Hazard Reduction Act of 1992," also
known as "Title X."
The EPA is currently developing health-based standards for house dust lead levels
and approving methods for sampling house dust. As part of this work, numerous
questions about the sampling of house dust have arisen. Three important questions
are:
1. What are the best methods of measuring lead in house dust?
2. What levels of dust lead can be maintained by a typical homeowner
using regular vacuuming?
3. Can a homeowner be assured that the vacuuming process does not
create an airborne lead hazard? Or, stated another way, how much
leaded dust passes through household vacuum cleaner bags under
normal use?
These important questions lead to the following more specific questions:
4. How do different scientific field sampling devices perform under
various field conditions?
5. What factors affect household vacuum cleaner performance?
6. Do household vacuum cleaners perform about the same in the labora-
tory as they do in the home?
7. How fast does dust accumulate in the home?
8. How effective is regular cleaning in the home?
-------
This study addresses aspects of the these questions through a series of laboratory
experiments. The results of this study, other studies, and future field work should,
when combined, provide answers to these questions.
1.1 Purpose of the Project
Lead-contaminated house dust is considered one of the most significant sources of
childhood lead poisoning in the United States. Until recently, little was known
about the public health significance of house dust. Furthermore, little was known
about how to measure dust-lead levels in the home, how to relate sampling results
to actual health risks, or how to safely clean dust from residential surfaces. While
the significance of house dust is still not fully understood, recent advancements in
our state of knowledge have been made by the EPA, other government researchers,
and the private sector. These advancements include an increased understanding of
house dust characteristics, the realization that different samplers give different
results under the same circumstances, that different commercial household
vacuum cleaners are not equal in their dust-pickup capabilities, and that previous
studies designed to evaluate the performance of samplers or vacuum cleaners are
not necessarily comparable. Interpretable house dust lead measurements and safe,
reliable dust-cleanup methods are needed for the Federal government to draft
guidelines to address lead hazards in housing, to develop a standardized approach to
characterize house dust samplers, and to evaluate vacuum cleaners. These are
important objectives of the current study.
During a previous research study, the Comprehensive Abatement Performance
Study (CAPS), conducted by the EPA's Office of Pollution Prevention and Toxics
(OPPT), differences in results between wiped and vacuumed samples of house dust
were noted. Because of these differences, EPA was concerned over making policy
decisions based solely on dust sampling results. The purpose of the current task is to
answer some of the questions that have been raised concerning sampling house
dust. Of special concern is the vacuum sampling method used in the National
Survey of Lead-Based Paint in Housing (also known as the HUD National Survey)
and the resulting lead levels measured in the dust.
This project characterizes the performance of three vacuum and one wipe sampling
method used in previous OPPT studies. The characterization was accomplished by
measuring the recovery of the vacuums and wipes using several different particle
sizes of dust. The project results should improve interpretations and comparisons
across studies that used different means for collecting household dust.
The project also initiated research on the collection of dust and lead dust particles by
household vacuum cleaners available to homeowners and renters. It is anticipated
that, as residential lead hazards become even more widely publicized, homeowners
and renters will rely on vacuum cleaners to minimize the lead dust hazard in their
homes. The main purpose of this study is to identify factors which are important in
-------
determining the dust and lead pickup efficiency of household vacuum cleaners
including collecting data to evaluate how well several commercially available
vacuum cleaners collect different size dust particles from different surface types.
A secondary purpose is to assess the amount of dust exhausted into the air while
dust is being vacuumed. The Federal government has concerns that routine
vacuuming of highly lead-contaminated dust may create unseen health hazards by
polluting the air with lead particles. Lead abatement specialists use vacuum
cleaners equipped with a special high efficiency particulate air (HEPA) filter to clean
up lead-contaminated dust. The HEPA filters prevent fine lead particles from
escaping the vacuum cleaner through the exhaust and, thus, prevent a potential
airborne lead hazard. While vacuum cleaners fitted with HEPA filters are available,
they usually are expensive and not readily accessible to the general public, although
the situation is improving. This project measured the recovery and exhaust
emissions of lead dust in a laboratory setting by four different vacuum cleaners
currently available for household use. One of these vacuum cleaners was equipped
with a HEPA filter.
1.2 Overview of the Report
The rest of this report is devoted to the presentation of background information,
study objectives, and methods and results. The report is divided into two volumes.
Volume I presents the background, methods, and study results. For readers inter-
ested in the specific sampling and analysis procedures, or those interested in repli-
cating the procedures, Volume II contains the appendices from the Quality
Assurance Project Plan (QAPjP) which describe the sampling and analysis proce-
dures. The following list provides a brief description of the contents of each section
in this report.
Volume I: Objectives, Methods, and Results
Section 1 Provides a basic introduction to current issues in the sampling of
dust and dust lead and an overview of the report.
Section 2 Reviews background material and related studies on the house dust
sampling methods selected for this study.
Section 3 Describes the objectives for the laboratory evaluation of dust and
dust lead recoveries for samplers and vacuum cleaners, including
the data quality objectives.
Section 4 Presents the study design and sample collection procedures (specific
protocols are in Volume n).
-------
Section 5
Summarizes the laboratory analysis procedures (specific protocols
are in Volume II).
Section 6
Section 7
Section 8
Section 9
Appendix A
Appendix B
Appendix C
Appendix D
Appendix E
Presents the results of the study as they relate to the objectives
presented in Section 3.1.
Discusses the study results in light of the unanswered questions
about dust sampling, including a discussion of possible adjustments
to the results from the HUD National Survey.
Discusses the details of the data analysis, including data entry, data
processing, and statistical analysis procedures and measurement
precision.
Describes the quality assurance aspects of the study, including the
system audit, performance audits, data audit, and data assessment.
Presents the analysis and results from the Pilot Study.
Presents and summarizes the preconditioning data.
Lists the sieved dust test data from this study.
Lists the sampler test data from this shady.
Lists the vacuum cleaner test data from this study.
-------
2 BACKGROUND
The purpose of this section is to provide background information that will help the
reader to understand the objectives of the study. Important concepts that recur
throughout the document are explained below. Section 2.1 provides an overview of
house dust and the importance of dust particle size. Section 2.2 introduces the
sampling methods evaluated in this study and gives a brief history of the use of
these sampling methods in previous OPPT studies. Vacuum cleaners that are
available to the public are also discussed. Finally, Section 2.3 reviews previous
sampling method comparison studies, both in the laboratory and in the field.
2.1 Importance of House Dust Particle Size
House dust is a complex mixture of particles and fibers that accumulates on residen-
tial surfaces. All house dust contains lead particles. The amount depends on the
extent of lead contamination from sources such as soil tracked into the residence or
deteriorated lead-based paint. A significant portion of dust consists of fine particles,
which may be the most biologically important fraction of the dust. Fine dust parti-
cles stick to a child's hands more readily than do larger dust particles and it is
hypothesized that they are more likely to be swallowed during normal hand-to-
mouth activity.1/2'3 However, this has not been demonstrated by empirical
evidence nor has it been extensively studied.
Fine dust has further biological importance in that lead absorption into the body via
the gastrointestinal tract is inversely related to particle size.4 The smaller the lead
particle, the more efficiently it is absorbed into the body. Although it is not known
if children are exposed primarily to fine dust particles, because fine particles adhere
to the skin more readily than coarse particles, house dust sampling and cleaning
regimes that are efficient in removing fine particles may be the most efficacious.
The efficiency with which dust is collected from a surface during sampling or
cleanup may depend on particle size. Thus, it may be important to know how well a
particular sampler or vacuum cleaner collects various sizes of dust particles in order
to evaluate its performance. As mentioned in Section 1, the current study is
designed to evaluate the dust and lead pickup efficiencies for samplers and vacuum
cleaners using different particle sizes. The purpose is to create a reproducible
1Que Hee, S.S., Peace, B., Clark, C.S., Boyle, J.R., Bornshein, R.L., and Hammond, P.B.: Evolution of
Efficient Methods to Sample Lead Sources, Such as House Dust and Hand Dust, in the Homes of
Children. Environmental Research. 38:77-95(1985).
2Duggan, M.J., Inskip, M.J., Rundle, S.A., and Moorcroft, J.S.: Lead in Playground Dust and on the Hands
of School Children. The Science of the Total Environment. 44: 66-79 (1985).
3Driver, J.H., Konz, J.J., and Whitmyre, G.K.: Soil Adherence to Human Skin. Bulletin of
Environmental Contamination Toxicology 43: 814-820 (1989).
4Barltrop, D. and Meek, F.: Effect of Particle Size on the Lead Absorption. From the Gutmann Archives
of Environmental Health. 280-285 (July/August, 1979).
-------
baseline of performance characteristics, by particle size, to compare how well
different collection methods remove a wide range of particle sizes from surfaces.
2.2 House Dust Sampling and Collection Methods
Two fundamentally different types of dust sampling methods, wipe and the
vacuum methods, are available to sample house dust. Researchers have developed
more than 15 variations of the two methods. When used side by side, the different
variations typically give different results. This makes it difficult for the Federal
government to project national estimates or to develop health-based standards for
lead in house dust, since the lead level measured by a single dust sample is affected
by the sampling method employed. It is therefore important to understand how
different sampling techniques compare with each other before the results from
different studies can be interpreted.
When comparing the dust sampling methods, it is important to understand the
relationship among lead concentration, total dust, and lead loading. Lead concentra-
tion is a measure of the proportion or fraction of the dust which is lead, on a weight
basis. Total dust refers to the amount of dust on the surface. When these two
measures are multiplied together, the product is lead loading, the amount of lead on
the surface.
Lead loading is expressed by the Federal government as micrograms of lead per
square foot of surface (fig/ft2). Lead Concentration Qig/g) x Total Dust (g/ft2) = Lead
Loading (jig/ft2). Common wipe sampling techniques measure lead loading. The
total dust weight collected on the wipe is very small compared to the total weight of
the wipe material itself. Therefore, wipe sampling cannot measure lead concentra-
tion because there is no way to accurately weigh the total dust collected from the
sampled surface. The only way to measure dust-lead concentration is to use a
vacuum sampling technique. Most vacuum samplers collect the dust sample in a
way that allows a quantitative weight measure of the dust collected from the surface,
thus allowing analysis of the lead concentration in the collected dust.
2.2.1 Wipe and Vacuum Sampling Methods Used in this Study
The 1990 HUD Interim Guidelines for Public and Indian Housing describe the HUD
wipe sampling method used in this study. It is the most commonly used residential
wipe sampling method in the United States. This technique uses premoistened baby
towelettes and is designed for hard, relatively smooth surfaces.
The three vacuum sampling methods used in this study were the Blue Nozzle, the
CAPS cyclone, and the BRM. The Blue Nozzle method was developed in 1989 by
MRI for the HUD-sponsored National Survey of Lead-Based Paint in Housing. This
method was developed because other methods available at the time were not
-------
sufficient for the requirements of the HUD National Survey. Namely, a vacuuming
method was needed that could measure both lead concentration and loading and
allow sampling areas to be covered in only a few passes to reduce sampling time.
The Blue Nozzle sampler consists of a laboratory rotary vane pump connected to a
0.8 urn mixed-cellulose ester membrane filter cassette via thick-walled 3/8" Tygon
tubing. The cassette is used open faced and a specially designed angle-cut Teflon
nozzle, 4" long x 2" wide, fits over the cassette with O-rings to seal it. The Blue
Nozzle name was coined for the color of the nozzle.
In 1991, MRI developed the CAPS cyclone, a portable, AC-powered particle separator
sampler (similar to a cyclone) from standard PVC pipe and pipe fittings and a
commercially available hand-held vacuum. It was designed to be an inexpensive
vacuum sampler made from materials commonly found in hardware stores. This
sampler was characterized by the MRI Engineering Study to Explore Improvements
in Vacuum Dust Collection and used in the EPA Comprehensive Abatement
Performance Study (CAPS), both described below.
Shortly thereafter, Farfel at Baltimore's Kennedy Krieger Institute modified a
cyclone house dust sampler originally developed for EPA's Office of Research and
Development (ORD) in Research Triangle Park, North Carolina. Farfel used the
same cyclone developed for ORD, called the HVS3 cyclone, but used the portable
hand-held vacuum that MRI used for the CAPS cyclone, instead of the less-portable
upright vacuum used by ORD. Rigid PVC and, after further modifications, flexible
tubing was attached to the cyclone to allow small areas to be vacuumed. This
sampler was developed for the EPA Lead-Based Paint Abatement Repair and
Maintenance Study (R&M Study) conducted in Baltimore, discussed below, and is
referred to as the BRM sampler.
2.2.2 History of Dust Sampling Methods Used by OPPT
The Office of Pollution Prevention and Toxics (OPPT) has considerable experience
collecting house dust to measure residential lead levels. Some of the current
popular methods were used during, or first developed for, OPPT research studies.
This section gives a brief history of the use in previous OPPT studies of the samplers
tested in the current study.
The EPA, OPPT worked collaboratively with HUD during the HUD National Survey
of Lead-Based Paint conducted in 1989-1990. The purpose of HUD's national survey
was to better estimate the extent of lead paint hazards in the nation's housing stock.
House dust was collected in hundreds of homes, nationwide, with the Blue Nozzle
method developed specifically for this study. The Blue Nozzle sampler was also
used in the EPA, OPPT R&M pilot study, which was conducted in six Baltimore
dwellings. Currently the BRM sampler is being used in the full EPA R&M study.
-------
The EPA, OPPT Comprehensive Abatement Performance Pilot Study (CAPPS) was
designed in part to assess the performance of sampling and analysis methods and to
compare the Blue Nozzle with the HUD wipe sampling protocol planned for the
full Comprehensive Abatement Performance Study (CAPS). Because the CAPPS
study showed that the Blue Nozzle method performed poorly compared to the wipe
sampling, EPA contracted with MRI to conduct an engineering study to explore
improvements in vacuum dust collection. During this study, MRI developed the
CAPS cyclone for use in the CAPS study. The HUD wipe method was used during
both the CAPS study and CAPPS study.
2.2.3 Commercial Vacuum Cleaners
As mentioned in Section 1, the main purpose of the vacuum cleaner element of this
study is to evaluate how well several commercially available vacuum cleaners
collect different size dust particles from different surface types. A secondary purpose
is to assess the amount of dust exhausted into the air while dust is being vacuumed.
Four commercially available vacuum cleaners with beater bar attachments (i.e.,
"power nozzles") were selected for testing, ranging in price from $120 to $800. Each
tested vacuum cleaner is described below. They are not identified by the manufac-
turer's name, but by the letters A, B, C, or D.
Vacuum cleaners A, B, and C are canister-type vacuums and D is an upright
vacuum. Vacuum cleaner A is a top-of-the-line non-HEPA canister vacuum
cleaner ($400) and is widely available. Vacuum cleaner B is a lower-cost canister
vacuum that cost $300. Vacuum cleaner C is the only household vacuum tested in
the current study that is equipped with a HEPA filter. It represents a relatively
expensive vacuum ($800) and is new on the market. Finally, vacuum cleaner D is a
popular upright vacuum with a price of $120.
2.3 Previous Studies that Compared Samplers and Vacuum Cleaners
in the Laboratory and Field
Sampling method and vacuum cleaner characterization studies are important to
assess both the dust-pickup performance on different surfaces and how the perfor-
mance differs among devices. The following studies show the variety of procedures
used to test sampling methods. The test methods described in the next sections are
well designed, but they differ significantly. Researchers do not yet agree on the best
reference materials to characterize dust sampling methods. The procedures
designed for the current study were developed after careful review of the previous
work done in this area. The aim was to develop standard reference materials to be
used on standardized surfaces in a standardized manner to facilitate comparability
between studies in the future.
8
-------
2.3.1 Laboratory Comparison Studies: Real versus Artificial House
Dust
Several researchers have characterized house dust sampling systems in the labora-
tory with artificial house dust made from sand, soil, talc, corn starch, and paint
chips. The advantages of creating a well-defined artificial dust include the ability to
control outside variability in experiments and to obtain good measures of the
relative differences between sampling techniques on the substrates on which the
dust is placed. However, the downside of these experiments is that artificial house
dust may not represent dust found in homes. Dust found in homes is oily and
sticky and has other characteristics that cannot be duplicated with artificial dust.
Unfortunately, house dust must be collected first in the home to be used as a test
dust. This initial collection process may bias the dust particle size distribution
toward particles that are more easily collected. Thus, laboratory tests performed
with these dusts may overestimate the sampler's capability in the field. However,
although using dust collected from homes may introduce some limitations in
interpreting the results from laboratory performance tests, the limitations imposed
by using artificial dust were deemed to be much greater. For this reason, house dust
was used in the current study.
One of the first and frequently cited laboratory comparison studies was conducted by
Que Hee et al. (1985). He collected house dust in several houses with vacuum
cleaners containing standard vacuum cleaner bags. The portion of the dust from
these bags that passed through a 149 Jim sieve was retained as loose test house dust
and was used to determine sampling collection efficiency of a dust sampling method
Que Hee designed. Dust weights of 10, 20, 30, 40, 50, and 100 mg were placed as
evenly as possible on different surfaces and vacuumed up to measure the overall
collection efficiency of the sampler. Further tests were conducted with other house
dust sieved into the following six particle size classes: <44, 44-149, 149-177, 177-246,
246-392, and 392-833 fim. These additional tests determined the sampler collection
efficiency for different particle size classes on a variety of surfaces.
The EPA, ORD (1989)5 evaluated the High Volume Surface Sampler (HVS2, prede-
cessor of the HVS3) sampler for its own use using a modified American Society of
Testing and Materials (ASTM) Method F608-79 (1987, developed by the Hoover
Company, North Canton, Ohio for household vacuum cleaners). The ASTM
method called for an artificial test dust of 90 percent sand and 10 percent talc spread
on and embedded into a test carpet by dragging a large, smooth weight across the
surface. The EPA modified the test dust to better match the reported composition of
house dusts. The new mixture was 45 percent sand, 45 percent talc, 9.5 percent food-
grade cornstarch, and 0.5 percent technical-grade graphite. The cornstarch and
5United States Environmental Protection Agency. Project Summary: Development of a High Volume
Surface Sampler for Pesticides in Floor Dust, by J.W. Roberts and M.G. Ruby. EPA/600/s4-88/036.
January 1989
-------
graphite particles were found to be more than 99 percent less than 75 Jim, while the
particle size of the test sand was:
20 percent >300 Jim
70 percent between 300 and 150 p.m
2 percent between 150 and 106 urn
7 percent between 106 and 75 ^im, and
1 percent < 75 \im.
On a subsequent sampling system, Research Triangle Institute (1990) modified the
artificial dust to consist of 10 percent talc and 90 percent fine sand (<150 \im). The
same sampler was retested by Roberts et al. (1991) with "real" house dust collected
from carpets in six houses with an upright Hoover convertible vacuum cleaner
with an beater bar. The collected dust was removed from the vacuum cleaner bags,
mixed, and sieved to <150 Jim, similar to Que Hee's approach. Approximately
15.9 g/m2 of the dust was added to carpets using the ASTM method, and sampler
collection efficiency was then determined.
The EPA, OPPT-sponsored MRI Engineering Study to Explore Improvements in
Vacuum Dust Collection, mentioned previously, was designed to test samplers
using artificial dust. Three different particle size classes were prepared in the labora-
tory: < 250 ^im, 250 Jim to 2,000 um, and > 2,000 urn. The artificial dust consisted of
dirt, sand, and paint chips and was applied to a surface by hand as evenly as possible
over the one-foot square inscribed area of the surface. Each sampling test consisted
of vacuuming a one-square foot area on wood floor, linoleum, concrete, carpet, or a
window sill. Dust was not ground into the carpets. The authors' interpretation of
the results showed the Blue Nozzle to be the least efficient sampler for dust
sampling. The CAPS cyclone sampler achieved the best results.
Farfel (1993) used artificial dusts to characterize various house dust samplers,
including the Blue Nozzle, the CAPS cyclone, and the BRM. Three different dusts
were used: (1) a "large-diameter" dust (250-2,000 p.m) made from dried sand and soil;
(2) an "intermediate diameter" dust (38-149 ^m) made from NIST Standard
Reference Material #2704 (a soil standard); and (3) a "small diameter" dust (<44 p.m)
made from talc. Farfel's data showed that the BRM exhibited less bias across all of
the particle size classes than the other samplers.
Lioy et al. (1993)6 used two types of dust to characterize a wipe sampling device:
Arizona road dust with particle sizes less than 80 ^im (39%, < 5 fim; 18%, 5-10 Jim;
16%, 10-20 \im; 18%, 20-40 Jim; 9%, 40-80 urn) and an all-purpose potting soil,
composed of organics and sand, which was dried and sieved to provide a particle
size of less than 75 |im. The authors state that the sieving removed a large percent-
6Lioy/ P.J., Wainman, T., and Weisel, C: A Wipe Sampler for the Quantitative Measurement of Dust on
Smooth Surfaces: Laboratory Performance Studies. Journal of Exposure Analysis and Environmental
Epidemiology. 3(3): 315-330 (1993).
10
-------
age of the sand. They used a deposition chamber to load the test dust uniformly
onto different surface types. Actual house dust was not used in the resuspension
experiments because hair and other materials would clog the generator and inhibit
uniform deposition in the chamber.
2.3.2 Laboratory Comparison Studies: Test Surfaces and Collection
Efficiency
The test surface is related to efficiency of dust collection. The type of surface
sampled directly affects the amount of total dust collected from the surface.
Furthermore, different sampling techniques collect different amounts of dust from a
surface that has the same dust loading. The difference is due to different collection
efficiencies of the samplers. When evaluating devices, it is important to use several
different test surfaces to fully characterize and compare the samplers. If meaningful
comparisons among studies are desired, the same types of surfaces must be used by
different researchers conducting separate laboratory comparison studies.
2.3.3 Field Comparison Studies
Field comparison studies are important because they bring an element of reality that
cannot be duplicated in the laboratory. While it is not possible to obtain "true" dust
collection efficiencies in the field, it is possible to measure relative collection
efficiencies between sampling devices using side-by-side samples. It is important to
follow up on findings observed in the laboratory to determine if they hold up in the
"real world." The EPA, OPPT has conducted several field sampling method
comparison studies, which were briefly described in Section 2.2. They are discussed
in more detail below. A recent field comparison study conducted by the National
Center for Lead-Safe Housing (NCLSH) is also presented.
The EPA, OPPT R&M pilot study collected side-by-side wipe and vacuum dust
samples. The results showed that side-by-side wipe and vacuum floor dust samples
were highly correlated (r=0.84; p < 0.001; n=68). However, findings also revealed
that wipe lead loadings were 3.4 to 5.6 times higher than those observed by the Blue
Nozzle method.
The EPA, OPPT Comprehensive Abatement Performance Pilot Study (CAPPS)
collected two side-by-side floor samples using the Blue Nozzle vacuum and the
HUD wipe sampling method. The wipe sampling procedures showed lead loadings
(jig/ft2) for floor samples to be approximately 10 times higher, and lead loadings for
window stool samples to be approximately 5 times higher, than samples collected by
the Blue Nozzle method. For the EPA, OPPT Comprehensive Abatement
Performance Study (CAPS), side-by-side vacuum/wipe samples were not statistically
different. Unlike the pilot study, the CAPS study used the CAPS cyclone sampler.
The estimate of vacuum/wipe ratio was 1.42 with a confidence interval of 0.78 to
11
-------
2.60. The difference between the two methods appeared to increase with the
roughness of the substrate. It was also found that, on the average, side-by-side
vacuum measures were significantly more variable than side-by-side wipe
measures.
The NCLSH recently funded two studies: (1) a pilot study to field test five different
sampling methods, side by side and (2) a correlational study to assess the relation-
ships between settled lead dust and blood lead levels in children, using three
methods chosen from the first study. The first study was conducted by the
University of Cincinnati. The sampling methods used included the University of
Cincinnati method (a vacuum sampler made from common industrial hygiene
sampling materials), Farfel's BRM sampler, the HUD wipe method, Farfel's wipe
method, and the LWW wipe sampling method (a specially designed wipe sampling
device, capable of reporting both lead loading and concentration). Cincinnati
collected five side-by-side samples in 20 homes, in three rooms per home and two
samples per room.
The second NCLSH study includes quantifying the relationships among a wide
range of settled dust levels and blood lead levels. Methods include using side-by-
side vacuum and wipe sampling on floors, window sills, and window wells in a
minimum of three rooms per dwelling unit, including the child's bedroom and the
principal play area. The results for both studies are pending.
12
-------
3 STUDY OBJECTIVES
Many measures exist to determine the effectiveness of dust removal methods. One
such measure is "recovery" or the percentage of dust collected from a surface by
weight. Characteristics of dust and surface that may affect recovery include the size
and source of the dust particles, the type of surface on which the dust lies, and
whether it is ground into the surface. Characteristics of dust removal devices that
affect recovery include the amount of suction (or face velocity), the efficiency at
capturing dust particles, and the type of "head" that contacts the surface. The objec-
tives of this study are to examine the ability of several dust removal devices to
recover both dust and lead from five preselected surfaces under a variety of
conditions.
3.1 Questions to be Answered with the Data
This study focuses on two types of dust removal devices as discussed in the previous
sections: scientific field sampling devices and household vacuum cleaners.
Throughout this document, the term "sampler" refers to a device that is appropriate
for measuring dust and lead levels over small areas for scientific purposes. The four
sampling methods tested in this study are specifically designed for this purpose. The
term "vacuum cleaner" refers to a consumer product designed to vacuum in the
home. The following six study objectives are concerned with examining the differ-
ences between and within these two types of dust removal devices.
(1) For household dust collected in vacuum cleaner bags, estimate the
percentage of dust and the lead concentration for various dust particle
size classes.
(2) For selected samplers, estimate dust recovery and lead recovery for
various substrates and dust particle size classes.
(3) For selected vacuum cleaners, estimate dust recovery and lead recovery
for various substrates and dust particle size classes.
(4) For selected vacuum cleaners, estimate how dust recovery and lead
recovery change with cleaning effort.
(5) For selected vacuum cleaners, estimate how exhaust dust levels change
over time as dust enters a new vacuum cleaner bag.
(6) For all laboratory experiments, estimate sampling and measurement
errors.
Section 8 presents the study results that address these specific objectives.
13
-------
3.1.1 Samplers
Four samplers were used in this study: 1) the HUD wipe sampling method which
uses premoistened baby wipes (Wipes); 2) the BRM sampler (Baltimore Repair and
Maintenance Study modified HVS3 Cyclone sampler, BRM); 3) the CAPS Cyclone
sampler (CAPS); and 4) the Blue Nozzle sampler. These samplers were selected
because they were used in prior studies conducted by the EPA.
The objectives for samplers were designed to provide information to answer the
following questions:
• For the two fundamental methods of sampling house dust for scientific
purposes (wipe and vacuum), which method is more precise and how
can the relationship between them be characterized?
• What are the variances attributable to person-to-person differences
among sampling technicians and sample-to-sample variation by one
technician when taking wipe samples and vacuum samples?
Although not specified in the original objectives, the study was designed to use
different substrates and dust in different particle size classes in order to answer the
following questions:
• What role does the substrate play in the sampling of dust by both wipes
and vacuum samplers?
• Do vacuums and wipes show a preferential or uniform pickup of the
various particle size classes of house dust? If there is preferential
pickup, what are the recoveries of wipes and vacuums for the different
size classes?
Finally, can the study results be used to decide if it is necessary to adjust the HUD
National Survey vacuum data and, if so, how might it be adjusted? Section 9
discusses the study results in light of these questions.
3.1.2 Vacuum Cleaners
Four vacuum cleaners were used in this study: 1) an inexpensive canister model
(vacuum cleaner A), 2) a highly rated model without a HEPA filter (vacuum cleaner
B), 3) a highly rated canister model with a HEPA filter (vacuum cleaner C), and 4) a
popular upright model (vacuum cleaner D). These vacuum cleaners were selected
because they are commercially available, fairly popular, and/or cover a range of
vacuum cleaner characteristics, based on preliminary information. All were
equipped with a beater bar attachment.
14
-------
Collection Efficiency
As mentioned earlier, the vacuum cleaner element of this study was included to
evaluate the collection efficiency of the four vacuum cleaners. Two collection
efficiency objectives of the vacuum cleaner tests are to estimate the dust recovery
and lead recovery for various substrates and dust particle size classes and to estimate
how the dust recovery changes with cleaning effort.
Exhaust Tests
Concern has been expressed that small dust particles, possibly those with the greatest
lead hazard, will be expelled in vacuum cleaner exhaust, thereby reducing the effec-
tiveness of vacuuming for controlling leaded dust. This is not an issue for vacuum
cleaners that have a highly efficient HEPA filter in the exhaust stream, but it may be
for those without a HEPA filter. While the experiments performed in the collection
efficiency study will provide estimates of overall recovery rates, they will not
provide estimates of the amount of exhausted dust. Therefore, a separate experi-
ment was conducted to estimate the amount of dust that is expelled in vacuum
cleaner exhaust as dust enters a new vacuum cleaner bag.
3.2 Data Quality Objectives
The primary objectives of this study are to estimate dust and lead recovery for
samplers and household vacuum cleaners using dust of different particle sizes.
Because the collection of dust bags was based on voluntary procedures rather than a
probability sample, no data quality objectives have been established for determining
the distribution of dust mass by particle size class. The data quality objectives for the
recovery measurements are to estimate:
(1) Overall percent recovery of dust for a vacuum or wipe method across
all tests with a 95 percent confidence interval of +/- 8 percent
(2) Average percent dust recovery for a vacuum or wipe method on each
substrate with a 95 percent confidence interval of +/- 15 percent
(3) Average percent dust recovery for a vacuum or wipe method on each
combination of substrate and particle size class with a 95 percent
confidence interval of +/- 30 percent
These data quality objectives were established based on a consideration of what
could be achieved with the available resources and what precision was acceptable to
EPA. No specific data quality objectives for estimates of lead recovery were estab-
15
-------
lished because relevant estimates of precision were not available at the time the
study was designed. Since the objective of measuring the exhaust dust levels is to
identify relative changes over time and to determine if the measurements can be
made reliably, no data quality objectives have been established for these measure-
ments.
16
-------
STUDY DESIGN AND SAMPLE COLLECTION PROCEDURES
4.1 Study Design
This study required performing laboratory tests on four vacuum cleaners and four
samplers to determine their dust and lead (Pb) pickup efficiency (i.e., recovery). The
study design included tests of several factors, listed in Table 4-1, on the dust and lead
recovery. The tests were performed according to the study design previously
discussed in the QAPjP/ except for some changes that were made based on informa-
tion obtained from the pilot study.s Some changes were also made in precondition-
ing substrates, necessitated by difficulties in achieving the desired limits on weight
gain in vacuuming carpet and upholstery, as explained later in this section.
Vacuum cleaners A, B, and C are canister model units with all attachments includ-
ing powered beater bars (i.e., "power nozzles") for use on rugs (see Section 2.2.3).
Vacuum cleaner D is an upright model that uses a larger bag for dust collection. For
this model, the dust collection bag is on the discharge side of the blower, rather than
on the suction side as in the canister models. See Figures 4-1, 4-2, 4-3, and 4-4 for
photos of each vacuum cleaner. The attachments shown are those which came with
the vacuum cleaner. Only the attachments for floors (with the beater bar) and
upholstery were used in the study. When performing the tests, the canister vacuum
cleaners were placed on the floor beside the platform with the substrate.
The four samplers used are commonly referred to as baby wipes (wipes), BRM
sampler, CAPS sampler, and Blue Nozzle sampler. Photos of each sampler are
shown in Figures 4-5, 4-6, 4-7, and 4-8, and each is described in detail in Volume II of
this report.
Reference dust used in the tests was obtained from normal household vacuum
cleaner bags, as discussed in Section 4.2. There were two groups of dust: high lead
dust (dust from older homes built before 1963) and low lead dust (dust from newer
homes built after 1982). At the time of the study design, it was assumed that the
dust from older homes would have higher lead concentrations than the dust from
newer homes. Therefore, the dusts from older and newer homes are said to have
high and low nominal lead concentrations. Dust in each group was sieved into the
six particle size classes (Table 4-1) that were used in the study. Dust of selected
particle size classes was applied to substrates using two dust loadings, 100 mg/ft2 and
400mg/ft2
Since the personnel operating the vacuum cleaners and samplers could influence
the results, two different teams were used in carrying out the tests. These are
7 Quality Assurance Project Plan for the Wipe and Vacuum Study, dated September 24,1993 (EPA Task
Manager Dr. Ben Lim).
8 See Appendix A for "Pilot Test Results for the Wipe and Vacuum Study."
17
-------
referred to as Team 1 and Team 2; each team performed specific tests as directed in
the test sequence provided in this section.
The procedures for testing dust recovery followed the ASTM F608-89 method to the
extent that it was consistent with the objectives of the study. The ASTM method is
described in Appendix C of Volume II.
The substrate types for the study were chosen to include a variety of surface charac-
teristics commonly present in homes. The specific examples of each substrate were
selected based on usage reported by retailers. The selected substrates were
commonly used and available. Figures 4-9 to 4-13 present photos of each of the six
different substrates used in this study:
• Vinyl tile
• Carpet
• Carpet with ground-in dust (dust ground in following ASTM proce-
dure)
• Sheet vinyl (or linoleum)
• Wood
• Upholstery
To avoid cross-contamination due to the use of dust with two nominal lead concen-
trations and the two dust loadings, four separate sections of each of the six substrates
were required. Each of the four sections was labeled as follows:
• Low lead, low loading
• High lead, low loading
• Low lead, high loading
• High lead, high loading
Each substrate section measured 72 x 27 in., and the vacuum cleaner test area used
within that area was 54 x 18 in. (i.e., 6.75 ftz) per ASTM F608-89. The test area used
for the sampler tests was 12 x 12 in. (i.e., 1.00 ft?). A 6-in. high platform (72 x 29 in.)
was used to support substrates for the tests, as shown in Figure 4-14.
18
-------
Table 4-1 Factors affecting dust and lead recovery
Factor
VACUUM CLEANERS
Substrate
Dust particle size class
Dust loading
Nominal lead (Pb) concentration
Team
Vacuum cleaner
SAMPLERS
Substrate
Dust particle size class
Dust loading
Nominal lead (Pb) concentration
Team
Sampler
Type/levels
Carpet, carpet with grind-in, linoleum,
wood, upholstery, vinyl tile
<53 \im; 53 to 106 [im; 106 to 150 \Lm; 150
to 212 ^im; 212 to 250 Jim; 250 to 2,000 urn
100 mg/ft2, 400 mg/ft2
Low, high
1,2
Models A, B, C, and D
Carpet, carpet with grind-in, linoleum,
wood, upholstery
<53 [im; 53 to 106 Jim; 106 to 150 |im; 150
to 212 Jim; 212 to 250 Jim; 250 to 2,000 Jim
100 mg/ft2, 400 mg/ft2
Low, high
1,2
Baby wipes, BRM, CAPS, Blue Nozzle
19
-------
Figure 4-1 Vacuum cleaner A
20
-------
Figure 4-2 Vacuum cleaner B
21
-------
22
-------
Figure 4-4 Vacuum cleaner D
23
-------
Figure 4-5 Wipe sampling
24
-------
eg
2
fD
O
N
CA)
CU
3
"
Cn
-------
-------
N)
Ojf
f*
00
n
>
a
(/)
su
3
"
-------
Figure 4-9 Tile substrate
28
-------
Figure 4-10 Linoleum (sheet vinyl) substrate
29
-------
Figure 4-11 Wood flooring substrate
30
-------
Figure 4-12 Upholstery substrate
31
-------
Figure 4-13 Carpet substrate
32
-------
Figure 4-14 Platform for substrates
33
-------
Carpet used for the tests, including carpet with ground-in dust, was a commonly
used tufted cut pile type made of 100% staple nylon. A 3/8 inch thick foam pad was
used underneath each carpet section. For the tests, the carpet was clamped to the
platform at each of the four corners. The upholstery was 100% cotton, with a weight
of 2.19 Ibs. per linear yard 54 inches wide. The weave was "textures," a box weave
with warp and filler yarns. A layer of 1 /2 inch thick foam padding was used under-
neath each upholstery section. Upholstery substrates were stretched and clamped to
the platform along both ends to prevent "rippling" of the surface when vacuumed.
New wood flooring purchased for this study consisted of 3-inch wide x 3/8-inch
thick tongue and groove flooring available prestained and prewaxed. The flooring
was glued onto a piece of 3/4 inch-thick plywood.
Tile substrates consisted of 12 x 12-in. squares of self-adhering vinyl tiles applied to a
piece of 3/4 inch-thick plywood. The tile had a slight surface texture. The linoleum
(i.e., sheet vinyl) substrate was glued onto a piece of 3/4 inch-thick plywood. This
linoleum had a smooth surface.
The following list shows the sections which describe the sample collection proce-
dures, including the preparatory steps, along with the dates when the work was
carried out.
4.2 Description of House Dust Used in Study (7/27/93 to 8/19/93)
4.3 Fiber Preconditioning of Carpet and Upholstery (7/30/93 to 8/18/93)
4.4 Dust Preconditioning of All Substrates (8/23/93 to 8/25/93)
4.6 Vacuum Cleaner Tests (8/26/93 to 9/9/93)
4.7 Sampler Tests (9/10/93 to 9/15/93)
4.8 Vacuum Cleaner Exhaust Emission Testing (7/27/93 to 7/29/93)
NOTE: The pilot study, which included the vacuum cleaner exhaust emission
testing, was carried out from 6/22/93 to 7/29/93. The pilot study report can be found
in Appendix A.
4.2 Description of House Dust Used in Study
Vacuum cleaner dust bags from household vacuum cleaners were collected from
homes within one of two age groups: built before 1963 (older homes) or built after
1982 (newer homes). Seventeen bags were collected from older homes and 20 from
newer homes. The bags were donated by employees of EPA, Westat, and MRL
Other than stratifying homes by age, there was no control over the selection of
homes or the collection of dust within the homes. The bags were then sent to
Neutron Products in Maryland for sterilization.
34
-------
Following sterilization, the dust bags from both older homes and newer homes
were separately sieved into six particle size classes:
< 53 \im
53-106 Jim
106-150 urn
150-212 jim
212-250 \im
250-2,000 urn
All material above 2,000 microns was weighed and then discarded. The dust size
classes were selected to be similar to the size classes used in other studies and to
minimize the quantity of dust required for the study by using all of the dust under
2,000 microns. All sieving was performed according to Appendix B in Volume II,
"Protocol for Sieving Household Dust."
The size distribution across the six dust particle size classes for the dust from older
homes and newer homes is summarized in Section 6.2.1 of this report. Samples of
dust in each particle size class were analyzed for lead, and those results are summa-
rized in Section 6.2.2.
In addition to taking samples of the sieved dust for the initial lead analysis, samples
of each dust size were taken for lead analysis weekly during the vacuum cleaner and
sampler tests. These dust samples were obtained by the same procedure used for
distributing dust onto a substrate, except that the dust was distributed onto a sheet of
plastic instead of a substrate. Dust deposited on the plastic was transferred into the
sample bottle for analysis. Lead results for the initial analysis of sieved dust and the
weekly samples are included in Appendix C.
4.3 Fiber Preconditioning of Carpet and Upholstery
Prior to any vacuum cleaner or sampler tests, the carpet and upholstery substrate
sections were preconditioned, first by vacuuming the carpet and upholstery sections
to remove fibers, and second by applying and vacuuming dust several times. The
fiber preconditioning is described in this section. The dust preconditioning is
described in Section 4.4.
The new carpet and upholstery substrate sections were preconditioned by several
vacuumings (without applying dust) using all four vacuum cleaners. This was
done to minimize the weight of fibers picked up during subsequent vacuum cleaner
and sampler tests because the fibers could affect the measurement of dust recovery
data as well as the lead concentration.
Fiber preconditioning of carpet and upholstery substrates was carried out in accor-
dance with the test sequence attached to the "Protocol for Conditioning Carpet and
35
-------
Other Substrates" in Appendix D of Volume II. The preconditioning protocol speci-
fied repeated 5-minute vacuumings for a total of 20 vacuumings, or until the weight
gain was 20 mg or less for four consecutive vacuumings. The preconditioning of
carpet involved many more vacuumings than originally anticipated because, in
many cases, the weight gain after 20 vacuumings exceeded the 20-mg limit. There
was also a problem with the vacuum cleaner bag changing weight with time, even
without vacuuming, possibly due to changes in temperature and humidity. Thus, it
was not clear how much of the weight gain problem when vacuuming carpet was
due either to characteristics of the vacuum bags, or to pickup of material from the
substrates, or both.
Because the weight problem was evident in the first two preconditioning tests (Tests
1 and 2), several procedural changes were made in an effort to correct the problem.
The main procedural changes that were subsequently used in the preconditioning
tests were:
• The bag was cooled for 2 minutes in the room's vent duct after use.
• The bag was brushed with an anti-static brush, placed in a plastic bag,
and put on the balance. The weight reading was taken 1 minute after
the bag was removed from cooling (total of 3 minutes after removal
from vacuum cleaner). The bag remained in the plastic bag until
needed again.
Fiber preconditioning of the four sections of upholstery (Tests 9 to 12) gave similar
(and unexpected) results in that the incremental weight gain exceeded 20 mg after
many vacuumings. For this cotton upholstery material, it was clear that the weight
gain was primarily due to the material. Cotton fibers could easily be seen inside the
vacuum cleaner bags. Copies of the data from these preconditioning tests are given
in Appendix B of this report.
The limit of 20 mg of fiber per 5 minutes of vacuuming was set, based on the pilot
study results, as a level which could be achieved and which would have a negligible
effect on the recovery measurements. This target level was not achieved for all of
the carpet and upholstery samples used in the full study. Nevertheless, the effect of
fiber release on the weight of dust recovered from the substrates was small. In
addition, the analysis included a correction for fiber release and dust carryover.
4.4 Dust Preconditioning of all Substrates
Dust preconditioning of all six substrate materials was carried out according to the
test sequence attached to the "Protocol for Conditioning Carpet and Other
Substrates" in Appendix D of Volume II. For each section of substrate, this proce-
dure involved several applications of dust of different particle sizes, two different
teams, and vacuuming for 40 seconds using a different vacuum cleaner each time.
36
-------
For each of the six types of substrates, four sections of each type were necessary,
because the tests included two different dust loadings (100 and 400 mg/ft2) and two
different types of dust (i.e., dust from older homes-high lead, and dust from newer
homes-low lead). Therefore, the four sections of each substrate were identified for
specific dust loadings and lead concentrations as:
• Low lead, low loading
• Low lead, high loading
• High lead, low loading
• High lead, high loading
Each substrate section was used for the appropriate dust preconditioning tests in
accordance with the design sequence. The final design sequence for the dust
preconditioning and data from these tests are given in Appendix B of this report.
4.5 Statistical Design
The study was designed to estimate main effects for operator, dust loading, nominal
dust concentration, dust particle size, substrate, sampler or vacuum cleaner, and
interactions between sampler and both substrate and dust particle size. In the
original design, each combination of dust particle size and substrate shown in Table
4-2 was to be tested using each sampler.
The original experimental design was modified as a result of the pilot tests and, after
beginning the full study, in response to budget pressures. After the tests for the full
study began, it was necessary to cut back on the number of tests to stay within the
budget for the project. The redesign of the study was performed quickly and
consisted of specifying a fraction of the tests from the original design. In the
redesign, the tile substrate was eliminated from further testing, and not all samplers
were tested on each combination of substrate and dust particle size shown in Table
4-2.
The order of the tests was randomized in such a way that both operators could
perform tests at the same time and the chances of both operator needing either the
same substrate sample or the same sampler (or vacuum cleaner) at the same time
were minimized.
37
-------
Table 4-2 Combinations of substrate and dust particle class tested in the study
Substrate
Vinyl Tile (textured)
(Tile was used only in
the original design)
Sheet Vinyl /Linoleum
(Smooth)
Wood
Upholstery
Carpet
Carpet with ground-in
dust
Dust Particle Size Class
<53 \im
Tested
Tested
Tested
Tested,
not
using
wipes
Tested,
not
using
wipes
Tested,
not
using
wipes
53-106
jim
Tested
Tested
Tested
Tested,
not
using
wipes
Tested,
not
using
wipes
Tested,
not
using
wipes
106-150
p.m
Not
tested
Tested
Tested
Not
tested
Tested,
not
using
wipes
Not
tested
150-212
Hm
Tested
Tested
Tested
Tested,
not
using
wipes
Tested,
not
using
wipes
Tested,
not
using
wipes
212-250
um
Tested
Tested
Tested
Tested,
not
using
wipes
Tested,
not
using
wipes
Tested,
not
using
wipes
250-2,000
(im
Not
tested
Tested
Tested
Not
tested
Tested,
not
using
wipes
Not
tested
38
-------
4.6 Vacuum Cleaner Tests
Vacuum cleaner testing was carried out according to a specific test sequence. Four
vacuum cleaners were tested for dust recovery and lead recovery.
Vacuum cleaners
Model A - Canister model with beater bar and without HEPA filter
Model B - Canister model with beater bar and without HEPA filter
Model C - Canister model with beater bar and with HEPA filter
Model D - Upright with beater bar and without HEPA filter
An original test sequence consisted of tests on 240 combinations of substrate, dust
particle size class/ dust loading, lead concentration, team, and vacuum cleaner. The
original test sequence was revised during the course of the work as a result of
budgetary limitations. The revised test sequences (85 vacuum cleaner tests) are
shown in Tables 4-3 and 4-4. The revision of the original test pattern was guided by
the preliminary results from the pilot tests and the dust preconditioning tests. In
the revised test pattern, no tests with tile substrate, other than the initial tests, were
to be performed. Also, fewer dust particle sizes were to be tested for each combina-
tion of substrate and vacuum cleaner. The revised test design designated the
sequence for carrying out the tests and stipulated the parameters for each test, such
as:
• Test Number
• Substrate
• Particle Size Class of Dust
• Dust Loading (100 or 400 mg/ft2)
• Lead Concentration of Dust (High or Low)
• Team (Team 1 or Team 2)
• Vacuum Cleaner (A, B, C, or D)
39
-------
Table 4-3 Test sequence for
Substrate Particle Size
Class
Linoleum
Linoleum
Linoleum
Linoleum
Wood
Wood
Wood
Wood
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Upholstery
Upholstery
Upholstery
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Upholstery
Upholstery
Carpet
Linoleum
Linoleum
Linoleum
Carpet
Carpet
Carpet w Grind-in
Linoleum
Linoleum
Linoleum
Upholstery
Upholstery
Carpet
Carpet
Carpet w Grind-in
Wood
Wood
Wood
Wood
Wood
53-106
<53
212-250
106-150
150-212
106-150
150-212
<53
53-106
53-106
53-106
53-106
<53
212-250
<53
<53
212-250
212-250
212-250
<53
53-106
53-106
106-150
212-250
106-150
150-212
<53
212-250
53-106
<53
53-106
250-2000
150-212
150-212
250-2000
150-212
150-212
212-250
106-150
250-2000
53-106
<53
vacuum cleaner tests by team 1
Dust Loading Nominal Vacuum
Lead Cone
400 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
100 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
400 mg/sq ft
400 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
400 mg/sq ft
100 mg/sq ft
100 mg/sq ft
400 mg/sq ft
100 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
100 mg/sq ft
400 mg/sq ft
100 mg/sq ft
400 mg/sq ft
100 mg/sq ft
High
High
High
Low
High
Low
High
Low
Low
Low
Low
Low
Low
Low
High
High
High
High
High
High
High
High
High
High
Low
Low
Low
Low
High
High
High
High
Low
Low
High
High
Low
Low
Low
High
Low
Low
A
D
C
B
A
B
C
D
D
A
C
B
C
D
B
C
B
A
B
D
D
A
A
D
C
B
A
A
C
B
C
B
B
A
B
D
B
B
D
A
B
C
Original
Number
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1020
1085
1088
1052
1101
1104
1030
1032
1046
1026
1027
1056
1015
1083
1069
1075
1076
1095
1042
1044
1065
1119
1038
1079
1080
1111
1058
1077
Revised
Number
1-28
1-10
1-1
1-2
1-3
1-4
1-5
1-6
1-7
1-8
1-9
1-11
1-12
1-13
1-14
1-15
1-16
1-17
1-18
1-19
1-20
1-21
1-22
1-23
1-24
1-25
1-26
1-27
1-29
1-30
40
-------
Table 4-4 Test sequence for vacuum cleaner tests by
Substrate Particle Size Dust Loading Nominal
Class Lead Cone
Tile
Tile
Tile
Tile
Carpet
Carpet
Carpet
Carpet
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Linoleum
Linoleum
Carpet w Grind-in
Carpet w Grind-in
Linoleum
Linoleum
Linoleum
Carpet
Upholstery
Wood
Wood
Wood
Carpet w Grind-in
Carpet w Grind-in
Wood
Wood
Wood
Linoleum
Linoleum
Linoleum
Carpet w Grind-in
Carpet w Grind-in
Carpet
Carpet
Upholstery
Upholstery
Upholstery
Upholstery
Carpet
Carpet
Carpet
150-212
212-250
150-212
<53
212-250
<53
<53
212-250
212-250
212-250
<53
<53
53-106
150-212
212-250
<53
250-2000
<53
53-106
53-106
53-106
212-250
53-106
<53
150-212
150-212
106-150
150-212
250-2000
212-250
150-212
106-150
53-106
53-106
212-250
<53
<53
212-250
212-250
150-212
106-150
150-212
250-2000
100 mg/sq ft
400 mg/sq ft
100 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
100 mg/sq ft
100 mg/sq ft
400 mg/sq ft
400 mg/sq ft
100 mg/sq ft
400 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
400 mg/sq ft
100 mg/sq ft
400 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
400 mg/sq ft
400 mg/sq ft
100 mg/sq ft
400 mg/sq ft
100 mg/sq ft
100 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
100 mg/sq ft
400 mg/sq ft
100 mg/sq ft
100 mg/sq ft
Low
High
Low
High
High
High
High
High
Low
Low
Low
Low
Low
High
Low
Low
Low
Low
Low
High
Low
High
High
High
High
High
High
Low
Low
Low
High
High
Low
Low
High
High
Low
Low
Low
High
Low
Low
Low
team 2
Vacuum
C
B
A
D
C
D
A
B
C
B
D
A
A
B
D
C
C
A
D
C
C
C
A
D
A
C
B
A
D
A
D
A
D
B
D
B
D
C
A
D
C
B
C
Original
Number
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2098
2099
2033
2048
2061
2106
2025
2065
2068
2095
2109
2112
2018
2020
2052
2064
2085
2088
2043
2044
2058
2060
2089
2091
2101
2078
2071
2022
2074
Revised
Number
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
2-10
2-11
2-12
2-13
2-14
2-15
2-16
2-17
2-18
2-19
2-20
2-21
2-22
2-23
2-24
2-25
2-26
2-27
2-28
2-29
2-30
41
-------
Each test involved the following procedure:
• Tare weigh vacuum cleaner bag (after run free for 40 sec)
• Vacuum substrate for 40 sec and weigh bag
• Apply dust, vacuum 40 sec, weigh bag
• Apply dust, vacuum 40 sec, weigh bag
• Apply dust, vacuum 40 sec, weigh bag
• Vacuum substrate and weigh bag
• Vacuum substrate and weigh bag
• Vacuum substrate and weigh bag
• Recover dust from bag, weigh and submit for lead analysis
As shown above, each test involved three applications of dust, followed by vacuum-
ing and weighing, and then three additional vacuumings and weighings. The dust
from the bags was recovered after the last vacuuming by holding the bag upside
down with the dust inlet opening positioned over a wide-mouth sample bottle.
When the bag was tapped, part of the dust in the bag fell into the sample bottle; thus
only part of the dust in the bag was recovered for lead analysis.
Application of dust onto the substrates was begun by weighing the required amount
of dust into a small beaker. The application technique involved pouring the dust
from the beaker onto the appropriate size sieve screen while tapping the sieve as it
was moved around above the substrate. This technique provided the most even
distribution of dust onto the substrate, but some small amount of dust always
remained on the sieve. The weight of the dust applied, therefore, was determined
by weighing the sieve and beaker together, before and after application.
For carpet with grind-in, the dust was applied and then ground in using the ASTM
method described in Appendix C of Volume II. An example of the data for one
vacuum cleaner test is provided in Appendix E in this volume, along with a
printout of the database for all the vacuum cleaner sampling tests.
4.7 Sampler Tests
Sampler tests were similar to the vacuum cleaner tests except that the test area was
only one square foot. Therefore, the weight of dust applied was less than that
applied in vacuum cleaner tests since the dust loading used was the same for
42
-------
vacuum cleaner and sampler tests (100 and 400 mg/ft2). Only one application of
dust, rather than three, was used in sampler tests. As with the vacuum cleaner tests,
the original number of sampler tests was reduced from 161 to 52 due to budgetary
constraints.
The three vacuum samplers were tested for dust recovery and all four were tested
for lead recovery. As mentioned previously, the four samplers were:
• Wipes
• BRM sampler (BRM)
CAPS sampler (CAPS)
• Blue Nozzle sampler
As for the vacuum cleaner tests, the test design designated the sequence for carrying
out the tests and stipulated the parameters for each test, such as :
• Test Number
• Substrate
• Dust Particle Size Class
• Dust Loading (100 or 400 mg/ft2)
• Lead Concentration of Dusts (High or Low)
• Team (Team 1 or Team 2)
• Sampler (wipe, BRM, CAPS, or Blue Nozzle)
• Square Number to Be Used (1, 2, 3, or 4). The "square number" to be
used in each test referred to four one square foot squares marked on
each section of substrate.
Prior to the first sampler test on any substrate section, the entire test area (54 x 18 in)
was vacuumed for 40 seconds with vacuum cleaner A. This procedure helped to
minimize the effect of any dust that might remain from the previous vacuum
cleaner tests. The same substrates were used for both the vacuum cleaner and
sampler tests, with all the vacuum cleaner tests being done first.
Following the last sampler test on any substrate section, the entire test area was
vacuumed for 120 seconds with vacuum cleaner A. This was done to determine the
weight of material picked up by the vacuum cleaner after the sampler tests. That is,
43
-------
vacuuming was done for 120 seconds to simulate the three final 40-second vacuum-
ings done in each vacuum cleaner test.
Considering the above, each sampler test involved the following procedure:
• Only if first square is to be used (i.e., square 1):
• Reweigh bag (vacuum cleaner A)
• Vacuum entire substrate for 40 seconds with vacuum cleaner A
• Reweigh bag
• Deposit dust in specified square (i.e., square 1, 2,3, or 4)
• Use specified sampler to sample dust
• Weigh the dust collected by the sampler (except wipes)
• Prepare the dust sample for analysis
• If last square is to be used (i.e., square 4 for all substrates except carpet;
last square for carpet is square 3 since wipes are not done on carpet):
• Tare weigh bag (vacuum cleaner A)
• Vacuum entire substrate for 120 seconds with vacuum cleaner A
• Reweigh bag
• Vacuum dust from wand and brush of vacuum cleaner A (no
weighing)
Application of dust onto the 1-ft2 test area was done using the same procedure
described for vacuum cleaner tests. Grind-in, when specified, was done as per
Appendix C of Volume n, but only over the one foot square test area.
Dust samples from the samplers were recovered for lead analysis. For wipes, the
entire wipe was submitted to the lab. Dust was recovered from the BRM and CAPS
sampler using the procedures described in the appendices in Volume II. For the
Blue Nozzle sampler, the entire filter cartridge was transferred so that analysts could
remove the filter for digestion and analysis.
The sampler tests were carried out in accordance with the test sequence shown in
Tables 4-5 and 4-6. Tests with individual samplers were done using the procedures
in Volume II.
44
-------
Table 4-5 Test sequence for sampler tests by team 1
Substrate Particle Size Dust Loading Nominal Vacuum Square
Class Lead
Cone
Carpet
Upholstery
Upholstery
Upholstery
Upholstery
Wood
Wood
Carpet
Carpet
Carpet w Grind-in
Carpet w Grind-in
Wood
Linoleum
Wood
Wood
Upholstery
Linoleum
Linoleum
Carpet w Grind-in
Carpet w Grind-in
Linoleum
Linoleum
Linoleum
Wood
53-106
<53
<53
212-250
150-212
106-150
53-106
150-212
250-2000
53-106
150-212
150-212
106-150
212-250
<53
53-106
250-2000
53-106
212-250
<53
<53
212-250
150-212
250-2000
400 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
400 mg/sq ft
100 mg/sq ft
Low
High
High
High
Low
Low
Low
High
High
BRM
CAPS cyclone
Blue Nozzle
Blue Nozzle
Blue Nozzle
BRM
Blue Nozzle
BRM
Blue Nozzle
High CAPS cyclone
Low
High
Low
Low
Low
High
High
High
High
High
High
High
Low
High
Blue Nozzle
CAPS cyclone
CAPS cyclone
Blue Nozzle
CAPS cyclone
BRM
Blue Nozzle
CAPS cyclone
Blue Nozzle
BRM
Blue Nozzle
BRM
Blue Nozzle
Baby Wipe
4
2
3
4
3
2
3
1
2
3
1
4
2
1
2
3
1
2
2
3
3
4
4
4
Revised
Number
3-1
3-2
3-3
3-4
3-5
3-6
3-7
3-8
3-9
3-10
3-11
3-12
3-13
3-14
3-15
3-16
3-17
3-18
3-19
3-20
3-21
3-22
3-23
3-24
45
-------
Table 4-6 Test sequence for sampler tests by team 2
Substrate
Particle Size Dust Loading
Class
Nominal
Lead
Cone
Vacuum Square
Revised
Number
Linoleum
Linoleum
Carpet
Carpet
Carpet
Carpet w Grind-in
Carpet w Grind-in
Linoleum
Upholstery
Upholstery
Upholstery
Carpet w Grind-in
Carpet w Grind-in
Wood
Linoleum
Carpet w Grind-in
Linoleum
Linoleum
Carpet
Carpet
Wood
Wood
Upholstery
Wood
Wood
Wood
Carpet
212-250
<53
53-106
<53
212-250
<53
212-250
106-150
53-106
<53
212-250
53-106
53-106
150-212
150-212
150-212
250-2000
53-106
250-2000
150-212
<53
212-250
150-212
53-106
106-150
250-2000
106-150
400 mg/sq ft
400 mg/sq ft
100 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
400 mg/sq ft
100 mg/sq ft
400 mg/sq ft
400 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
400 mg/sq ft
400 mg/sq ft
100 mg/sq ft
100 mg/sq ft
100 mg/sq ft
400 mg/sq ft
400 mg/sq ft
Low
Low
High
High
High
Low
Low
High
Low
Low
Low
Low
Low
Low
High
High
Low
Low
Low
Low
High
High
High
High
High
Low
Low
Baby Wipe
Baby Wipe
CAPS cyclone
Blue Nozzle
BRM
CAPS cyclone
BRM
Baby Wipe
CAPS cyclone
BRM
CAPS cyclone
Blue Nozzle
BRM
Baby Wipe
BRM
CAPS cyclone
CAPS cyclone
BRM
CAPS cyclone
Blue Nozzle
BRM
CAPS cyclone
BRM
Baby Wipe
Blue Nozzle
BRM
CAPS cyclone
1
2
1
3
4
1
2
4
4
1
2
1
2
2
1
1
2
3
1
2
2
3
3
2
3
2
3
4-1
4-2
4-3
4-4
4-5
4-6
4-7
4-8
4-9
4-10
4-11
4-12
4-13
4-14
4-15
4-16
4-17
4-18
4-19
4-20
4-21
4-22
4-23
4-24
4-25
4-26
4-27
46
-------
An example of the sampling data for one test is provided in Appendix D of this
volume, along with a printout of the sampling database for all the sampler tests.
4.8 Vacuum Cleaner Exhaust Emission Testing
A series of tests were performed with all four vacuum cleaners to measure exhaust
dust concentrations. These tests, carried out during the pilot study, are documented
in their entirety in the final report for the pilot study (see Appendix A) and results
are summarized in Section 8 of this report. The procedures used in the tests are
given in Volume II, Appendix O.
47
-------
48
-------
5 LABORATORY ANALYSIS PROCEDURES
Dust samples collected in this study were digested with nitric acid (HNOs) and then
analyzed for lead by Inductively Coupled Plasma (ICP) or Graphite Furnace Atomic
Absorption (GFAA). The following samples were collected:
• Sieved dust samples
• Dust samples recovered from vacuum cleaner tests
• Dust samples recovered from sampler tests (including wipes and filter
cassettes from Blue Nozzle sampler)
Wipe samples were digested using the procedure in Appendix J of Volume H All
other samples were digested using the procedure in Appendix K of Volume n. The
digests were all analyzed by ICP per Appendix L of Volume n, except that the filters
from the Blue Nozzle sampler were analyzed by GFAA. Also, two dust samples and
one wipe sample were analyzed by GFAA because the ICP results were below 0.1
Hg/mL.
5.1 Lead Analysis of Sieved Dust
At the onset of this study, the dust to be used during testing was sieved and compos-
ited into six particle size classes (see Section 4.2), separately for newer and older
homes. Duplicate samples were taken from each the six size categories and were
analyzed for lead content (i.e., initial analysis).
In addition to the initial analysis of the sieved dust samples, samples that simulated
application of dust onto a substrate were taken each week for lead analysis.
The analytical results from all these sieved dust analyses are provided in Appendix
C of this report.
5.2 Lead Analysis of Dust Samples from Vacuum Cleaner and Sampler
Tests
All dust samples obtained from the sampler and vacuum cleaner tests were
analyzed for lead per the digestion and analysis procedures noted above. The analyt-
ical results are summarized in Appendices D and E of this report.
49
-------
50
-------
6 RESULTS
6.1 Summary of Results from the Pilot and Preconditioning Data
6.1.1 Pilot Test Results
The pilot tests were conducted to answer questions which would help improve the
study design. The tests provided data on fiber collection, dust recovery, and factors
which affect the test procedures. Details of the pilot study are presented in Appendix
A. The pilot tests consisted of five tasks. Results from Tasks 1 and 2 of the pilot
study, those that affect only the test procedures for the full study, are discussed in the
appendix. The results for the dust emission tests (Task 5) are summarized in
Section 6.4.4. Other results from the pilot study which are relevant to the objectives
of the full study are summarized here. Only vacuum cleaner A was used in the
pilot tests.
In Tasks 3 and 4 of the pilot study, the estimated dust recovery for vacuum cleaner
A was 84% (with 95% confidence interval from 80% to 87%) on carpets and 79%
(with 95% confidence interval from 74% to 85%) on carpets with ground-in dust.
Most of this dust is recovered in the first 40 seconds of vacuuming. For dust
deposited on carpets, 80% is recovered in the first 30 seconds of vacuuming, 4% is
recovered in successive vacuumings, and 16% is either caught in the carpet or lost.
For dust deposited and ground into carpets, 68% is recovered in the first 30 seconds
of vacuuming, 12% is recovered in successive vacuumings, and 20% is either caught
in the carpet or lost. In the emission tests, 0.02% or less of the dust was found to
pass through the vacuum cleaner bags.
From these figures it can be seen that roughly 16% to 20% of the dust deposited onto
carpets is not accounted for. Common sense suggests that this dust might be (1) in
the carpet and very difficult to remove with vacuuming, (2) below the carpet,
having passed through the carpet, (3) in the air, (4) scattered around the testing
room, perhaps onto parts of the carpet which were not in the vacuumed area, as a
result of disturbance while depositing the dust, grinding the dust into the carpet, or
vacuuming, or (5) caught in parts of the vacuum other than the bag. These prelim-
inary results are consistent with the results from the full study.
The precision of the dust recovery measurements was better than anticipated during
the preparation of the QAPjP. Therefore, the study as originally designed could
have achieved the data quality objectives. Due to subsequent budget considerations,
the number of tests planned for the full study was reduced. With this reduction it
was anticipated that the original data quality objectives would still be achieved based
on the precision attained in the pilot tests.
51
-------
6.1.2 Preconditioning Results
The fiber and dust preconditioning steps prepared the substrate samples to be used
in the sampler and vacuum cleaner tests. All four vacuum cleaners were used for
preconditioning the substrates. Successive vacuumings were used to remove loose
fibers from the carpet and upholstery substrates. The data suggest that the weight
gain due to fibers can be substantially reduced with 30 minutes of vacuuming.
However, the vacuum cleaners continue to pick up additional fibers after as much
as four hours of vacuuming. Therefore, the analysis of the sampler and vacuum
cleaner data included factors to account for fibers.
The average dust recovery achieved in the dust preconditioning ranged from 67%
on carpet and upholstery, using vacuum cleaner C, to 98% on carpet and upholstery
using vacuum cleaner A. Recovery on smooth substrates, wood, tile, and linoleum,
was similar for all vacuum cleaners and averaged 94%. The precision of the dust
recovery measurements depended on the substrate. Across all substrates, the pooled
standard deviation is 17%. This is greater than the value of 10% that was assumed
for the study redesign and suggested by the pilot study results.
6.2 Test Dust Characteristics
6.2.1 Dust Recovery by Particle Size Class for Older and Newer Homes
The dust in this study came from donated vacuum cleaner bags which were used in
either older homes built before 1963, or newer homes built after 1982. The dust was
removed from the vacuum cleaner bags and sieved into the seven dust particle size
classes shown in Table 6-1. Dust from homes in the same age class and the same
particle size class was physically mixed and placed in a plastic bag. For dust from
both older and newer homes, Table 6-1 shows the weight of dust in each particle size
class as a percentage of the weight of all dust removed from the bags.
The distribution of dust among the particle size classes is very similar for the
samples collected from older homes and newer homes. Most of the dust was found
in the smallest and largest size classes. The percent of dust in different size classes
depends on the definition of the size class boundaries. The selection of the sieve
sizes was based on what sieve sizes were available and size classes used in other
studies. Figure 6-1 shows histograms of dust weight by size class using a continuous
scale for the dust size. A log scale for the dust size was selected for the histograms
because the distribution was more symmetric. Figure 6-1 shows the distribution of
the dust weight by size in a manner which is relatively independent of the bound-
aries of the size classes. However, in order to plot the histogram, the lower end~ofj
the smallest size class (<53 fim) and the upper end of the largest size class (>2,000
\Lm) had to be specified. These limits were arbitrarily set at 10 Jim and 10,000 fim
respectively. Changing these limits does not greatly affect the shape of the distribu-
tions.
52
-------
Table 6-1 Percent of dust in each particle size class, for older and newer homes
Dust from
newer homes
(built after 1982)
Dust from
older homes
(built before 1963)
Dust particle
size (jim)
<53
53-106
106-150
150-212
212-250
250-2,000
>2,000
Total
<53
53-106
106-150
150-212
212-250
250-2,000
>2,000
Total
Dust weight
(grams)
1,052
967
566
470
231
1,645
5,492
10,424
1,398
987
462
484
202
1,623
5,438
10,594
Percent of total
10.1%
9.3%
5.4%
4.5%
2.2%
15.8%
52.7%
100.0%
13.2%
9.3%
4.4%
4.6%
1.9%
15.3%
51.3%
100.0%
53
-------
.SP
"3
Ol
10
Dust Composite from Newer Homes
100 t>fc "• 1,000
Particle Size (microns)
•ft
o
o
10
Dust Composite from Older Homes
100 - 1,000
Particle Size (microns)
10,000
Figure 6-1 Histogram of relative dust weight by dust particle size for composite
dust samples from newer and older homes
54
-------
6.2.2 Lead Concentration by Particle Size Class for Older and Newer
Homes
Samples of the dust from the six smallest particle size classes used in the study and
from the two ages of homes were analyzed for lead concentration. Before the study
began, duplicate grab samples of dust were taken from each bag after mixing the dust
within the bag. These grab samples were then analyzed for their lead content. As
the study progressed, samples of dust were collected periodically to measure the lead
concentration in the dust actually deposited onto the substrates and to determine if
the concentration changed over time due to settling or stratification in the bags of
dust.
For the dust collected for this study (not involving a statistical sample of homes),
the dust lead concentrations in dust from the older homes were significantly greater
than those from newer homes. For dust from older homes, the lead concentration
was similar for all size classes of dust except for the largest size which had the
highest lead concentration. For dust from newer homes, lead concentrations were
highest in the smallest dust particle classes. Figure 6-2 and Table 6-2 show the
geometric mean lead concentrations and 95% confidence intervals for the twelve
bags of dust used in the study. The spacing along the horizontal axis of Figure 6-2 is
corresponds to using a log scale.
The dust lead concentration in the dust at the beginning of the study was combined
with the measurements on the weight of dust to determine the amount of lead by
dust particle size. Figure 6-3 shows the distribution of lead by particle size for dust
from newer and older homes. The vertical scale is the same for both plots in Figure
6-3. For dust from newer homes, most of the lead is concentrated in particles with
sizes below 250 Jim. The lead in dust from older homes is distributed among all
sizes of dust. For particle sizes less than 2,000 Jim, the dust lead concentration in the
dust from homes built after 1982 is 61 |ig/g and, for homes built before 1963, is 474
6.3 Samplers
The sampler tests involved depositing a known amount of dust over a one-square
foot area of the substrate, using the sampler to recover the dust following standard
procedures for each sampler, and determining the weight of dust (gravimetric data)
and amount of lead recovered. These measurements were used to calculate the dust
recovery, lead recovery, and ratio of the lead concentration in the dust collected by
the sampler and the lead concentration in the dust deposited on the substrate. The
four samplers studied were the CAPS cyclone, BRM, Blue Nozzle, and baby wipes.
The wipes were not tested on upholstery, carpet, or carpet with ground-in dust
substrates. Also, only lead recovery could be -determined for wipes.
55
-------
2,500 -
2,000 -
u
§ 1,500
T3
rt
en
9
Q
9
1,000 -
500
6
o
0)
O
0
Mean-41
95% Confidence
Interval
Old Homes
New Homes
8
v
$
s
in
3
o
3 S
* «?
O CM
in IH
§
Dust Particle Size (microns)
Figure 6-2 Geometric mean dust lead concentration by dust particle size, with
approximate 95% confidence interval
Table 6-2
Geometric mean dust lead concentration (p.g/g) by dust particle size,
with approximate 95% confidence*intervals
Dust Particle Size (urn)
<53
53-106
106-150
150-212
212-250
250-2,000
Geometric mean dust lead concentration (ue/g)
Newer Homes
110 (104 to 117)
131 (123 to 140)
48 (44 to 53)
34 (30 to 37)
32 (28 to 38)
21 (11 to 40)
Older Homes
374 (359 to 390)
457 (435 to 480)
383 (360 to 407)
405 (377 to 436)
424 (380 to 474)
1,136 (586 to 2,204)
56
-------
c
3
o
6
(0
Jj
I
4*
JS
*«
10
Dust Composite from Newer Homes
100 1,000
Partcle Size (microns)
10,000
c
i
1
•3
10
Dust Composite from Older Homes
100 1,000
Particle Size (microns)
10,000
Figure 6-3 Histogram of relative lead weight by dust particle size for composite
dust samples from newer and older homes
57
-------
The study included sampler tests using each sampler, each substrate, and dust from
each dust particle size class. However, not all combinations of these factors were
tested. As a result, the statistical results are based on mathematical models. The
estimates from the models (called least square means) are presented in this section.
In the modeling, the effects of dust loading, nominal dust concentration (dust from
newer or older homes), operator, substrate, sampler, and dust particle size class were
tested along with tests of interactions, in particular, differences in sampler recovery
with different dust particle sizes, and different substrates. In some cases regression
weights were used to adjust for differences in measurement variance. This section
discusses the estimates for only those factors which are statistically significant at the
5% level. A more complete discussion of the statistical models is presented in
Section 8.
6.3.1 Sampler Dust Recovery
The sampler dust recovery is the weight of dust collected by the sampler as a
percentage of the weight of dust deposited on the substrate. Based on a weighted
analysis, the statistically significant predictors of dust recovery are the sampler type
(p < 0.0001) and the combination of sampler and dust particle size ( p = 0.038).
The average dust recovery for each sampler (with 95% confidence interval) is 30%
(14% to 47%) for the Blue Nozzle sampler, 84% (79% to 89%) for the CAPS cyclone
sampler, and 89% (82% to 96%) for the BRM sampler. These average sampler dust
recoveries and the associated 95% confidence intervals are shown in Figure 6-4 and
Table 6-3. The recovery estimate is shown as a dark circle in the figure. The vertical
line through the circle shows the range of the 95% confidence interval on the
estimated recovery. The standard deviations of the dust recovery measurements for
the Blue Nozzle, CAPS cyclone, and BRM samplers are 29%, 12%, and 9%, respec-
tively.
For each sampler, the dust recovery depends on the dust particle size, as shown in
Figure 6-5 and Table 6-4. The plotting position for the dust particle size classes on
the horizontal axis of Figure 6-5 is equivalent to using a log scale. The dust recovery
for the Blue Nozzle sampler decreases as the particle size increases. The dust
recovery for the CAPS cyclone and BRM sampler increases slightly or remains
constant as the dust particle size increases. The estimated average dust recovery is
the recovery for dust which has equal proportions of dust from each of the six dust
particle size classes. In any situation, the dust recovery will vary, particularly for the
Blue Nozzle sampler, depending on the relative proportion of dust in each dust
particle size class.
58
-------
recovery
CO
3
-a
1*
Cfl
100% 1
90% -
80% -
70% -
60% .
50% .
40% •
30% •
20% .
10% .
0%.
1
*
4>
Blue Nozzle CAPS BRM
Sampler
Mean-< i
NA
Wipe
95% Confidence
Interval
Figure 6-4 Sampler dust recovery by sampler, with 95% confidence intervals,
averaged across all substrates
Table 6-3 Average sampler dust recovery by sampler, with 95% confidence
intervals
Sampler
Blue Nozzle
CAPS
BRM
Wipe
Dust recovery
30%
84%
89%
95% confidence
interval
14% to 47%
79% to 89%
82% to 96%
Not applicable for dust recovery
59
-------
o
u
en
Dust particle size class
Figure 6-5 Sampler dust recovery by sampler and dust particle size
60
-------
Table 6-4 Sampler dust recovery by sampler and dust particle size class, with 95%
confidence intervals
Sampler
Blue Nozzle
CAPS
BRM
Dust particle size
(microns)
<53
53-106
106-150
150-212
212-250
250-2,000
<53
53-106
106-150
150-212
212-250
250-2,000
<53
53-106
106-150
150-212
212-250
250-2,000
lead^recovery
53%
49%
43%
21%
14%
2%
61%
78%
87%
86%
94%
98%
83%.
91%
94%
98%
77%
92%
95% confidence
interval
20% to 86%
9% to 90%
/ 100%
/ -8% to 49%
-19% to 47%
\ -38°yto43%
^"BtfXo to 72%
69% to 88%
74% to 101%
72% to 99%
81% to 107%
85% to 111%
69% to 97%
78% to 103%
70% to 119%
84% to 112%
63% to 91%
68% to 116%
61
-------
The dust recovery differences among different substrates are not statistically signifi-
cant. Although differences may exist, the differences are small enough that they
cannot be adequate assessed from the data. Therefore, the estimated dust recovery
for combinations of substrates and samplers is the dust recovery for the sampler
used to collect the dust, shown in Table 6-3. Dust recovery could not be determined
for the wipe method.
6.3.2 Sampler Lead Recovery
Lead Recovery
The sampler lead recovery is the weight of lead collected by the sampler as a
percentage of the weight of the lead deposited on the substrate. The significant
predictors of sampler lead recovery are the sampler type (p < 0.0001), dust particle
size class (p = 0.0033), and dust loading (p = 0.035). The samplers, in order of decreas-
ing lead recovery, are the BRM, CAPS cyclone, baby wipe, and Blue Nozzle sampler.
Lead recovery decreased as the dust particle sizes increased. The measurement
standard deviation, pooled across all tests, is 21%.
Figures 6-6, 6-7, and 6-8 and Table 6-5 show the average lead recovery and the associ-
ated 95% confidence interval by sampler, dust particle size class, and dust loading.
The recovery estimate is shown as a dark circle in the figure. The vertical line
through the circle shows the range of the 95% confidence interval on the estimated
recovery. Figure 6-6 also shows in gray the predicted average recovery by sampler
and dust loading. Figure 6-7 shows in gray the average lead recovery by dust particle
size class and sampler. For the smaller dust particle sizes, the lead recovery of the
BRM, CAPS cyclone, and Wipe samplers is close to 100%. By contrast, the lead
recovery for the Blue Nozzle sampler is significantly lower. The lead recovery
estimates for the vacuum samplers include measurements on carpet and upholstery
substrates which were not used with the wipe sampling method. Because the
substrate is not a significant predictor of sampler lead recovery, wipe recovery can be
compared with the lead recovery of the vacuum samplers without having to correct
for the different substrates used for different samplers.
The average lead recovery for each sampler (with 95% confidence interval) is 26%
(15% to 38%) for the Blue Nozzle sampler, 72% (60% to 84%) for the CAPS cyclone
sampler, 81% (70% to 93%) for the BRM sampler, and 63% (43% to 83%) for the wipe
sampler. The estimated average lead recovery is the average lead recovery for dust
which has equal proportions of dust from each of the six dust particle size classes. In
any situation, the lead recovery will vary depending on the relative proportion of
dust in each dust particle size class.
62
-------
100%
90%
fr 80% -
1 70%
60%
50%
40%
1* 30%
I 20%
10%
0%
2
JU
h4
V
mg/sq ft
Blue
Nozzle
CAPS
BRM
Wipe
Sampler
Average
95% Confidence
Interval
Average by dust loading
Figure 6-6 Average sampler lead recovery by dust particle size class, with 95%
confidence intervals, and by dust particle size class and dust loading
63
-------
>
u
2
T3
«J
o»
120% i
110% -
100%
90%
80%
70%
60% .
50% .
40% •
30%
20% i
10%
0%
.BRM
<53
53-
106
106-
150
150-
212
212-
250
250-
2,000
Dust particle size class (microns)
Average
95% Confidence
Interval
Average by sampler
Figure 6-7 Average sampler lead recovery by dust particle size class, with 95%
confidence intervals, and by sampler and dust particle size class
64
-------
fr
OI
recovi
•d
~
"Si
i
W3
100% -i
JLWV/ /O
90% .
80% .
70% •
60% •
50% •
40% .
30% .
20% .
10% •
n% .
, 1
41
100 mg/sq ft 400 mg/sq ft
Dust loading
Average <
95% Confidence
Interval
Figure 6-8 Average sampler lead recovery by dust loading, with 95% confidence
intervals
65
-------
Table 6-5 Average sampler lead recovery by sampler, dust particle size class, and
dust loading
Sampler
Dust particle size
Dust loading
Blue Nozzle
CAPS
BRM
Wipe
<53
53-106
106-150
150-212
212-250
250-2,000
100 mg/sq ft
400 mg/sq ft
Average lead
recovery
26%
72%
81%
63%
67%
72%
80%
59%
57%
28%
54%
67%
95% confidence
interval
15% to 38%
60% to 84%
70% to 93%
43% to 83%
53% to 81%
58% to 86%
58% to 102%
45% to 73%
41% to 74%
10% to 46%
44% to 64%
58% to 77%
66
-------
The lead recovery differences among substrates are not statistically significant.
Although differences may exist, the differences are small enough that they cannot be
adequate assessed from the data. Therefore, the estimated lead recovery for combi-
nations of substrates and samplers is the lead recovery for the sampler used to
collect the dust, shown in Table 6-5.
Concentration Ratio
The sampler concentration ratio is the ratio of the lead concentration in the dust
sample to the lead concentration in the dust deposited on the surface. The ratio
depends on one factor which is very significant and several factors which are
marginally statistically significant. The most significant predictor of the concentra-
tion ratio is the dust particle size class (p < 0.0001). The lead concentration ratio is
close to 1.0 for the smaller dust particle sizes and decreases as particle size increases.
The determination of the significance of other factors depends on the model chosen.
Based on the final model, the results suggest that the lead concentration ratio for
samplers is lower for the Blue Nozzle sampler than for the BRM and the CAPS
sampler. It is also lower for dust from older homes with higher lead concentrations
than for dust from newer homes, and higher on carpet and upholstery substrates
than on wood, sheet vinyl, and carpets with ground-in dust. The predicted lead
concentration ratio averaged across the tests using the three samplers is shown in
Figure 6-9 and Table 6-6. The lead concentration ratio cannot be determined for the
wipe method.
6.4 Commercial Vacuum Cleaners
The study included sampler tests using each vacuum cleaner, each substrate, and
dust from each dust particle size class. However, because not all combinations of
these factors were tested, the statistical results are based on mathematical models.
The estimates from the models (called least square means) are presented in this
section. In the modeling, the effects of dust loading, nominal dust concentration
(dust from newer or older homes), operator, substrate, vacuum cleaner, and dust
particle size class were tested along with tests of interactions, in particular, differ-
ences in vacuum cleaner recovery with different dust particle sizes, and on different
substrates. This section discusses the estimates for only those factors which are
statistically significant at the 5% level. A more complete discussion of the statistical
models is presented in Section 8.
67
-------
1.4
1.2
5
l"
2 0.8
0.6
§
0.2
0.0
CAPS
<53
53-
106
106-
150
150-
212
212-
250
250-
2,000
Dust particle size (microns)
Average <
95% Confidence
Interval
Average by sampler
Figure 6-9 Sampler concentration ratio by dust particle size, with 95% confidence
interval
Table 6-6 Sampler concentration ratio by dust particle size, with 95% confidence
interval
Dust particle size
<53
53-106
106-150
150-212
212-250
250-2,000
Average concentration
ratio
1.01
0.91
1.01
0.96
0.73
0.26
95% confidence interval
0.87 to 1.14
0.79 to 1.04
0.77 to 1.26
0.81 to 1.11
0.55 to 0.91
0.70 to 0.45
68
-------
6.4.1 Dust Recovery
In the vacuum cleaner tests, the substrates were vacuumed for 40 seconds before
depositing dust (vacuuming 1). Then dust was deposited on the substrate three
times, each time followed by 40 seconds of vacuuming (vacuumings 2, 3, and 4). An
additional three vacuumings of 40 seconds each were used to collect residual dust
(vacuumings 5, 6, and 7).
For this analysis, the dust recovery for vacuum cleaners is defined as that portion of
the dust deposited on the substrate which was subsequently collected in vacuumings
2 through 6. The estimates of dust recovery include a correction for dust from
sources other than the dust deposited, such as fibers and carryover dust from other
tests. The equation for calculating dust recovery is discussed in Section 8.2.5.
The substrate being vacuumed and the choice of the vacuum cleaner are significant
predictors of dust recovery. Figures 6-10 and 6-11 show the predicted average
vacuum cleaner dust recovery and the associated 95% confidence interval by
substrate and by vacuum cleaner. The recovery estimate is shown as a dark circle in
the figures. The vertical line through the circle shows the range of the 95% confi-
dence interval on the estimated recovery. The recovery was highest on wood,
upholstery, and vinyl substrates. It was lowest on carpet with ground-in dust and
next lowest on the carpet substrate. Differences among vacuum cleaners were small,
but significant. The averages and 95% confidence intervals are also shown in
Table 6-7.
The dust recovery can also be defined in other ways. Estimates based on the two
following alternate definitions of dust recovery are also shown in Figure 6-10:
• The weight of dust collected in all seven vacuumings as a percentage of
the weight of dust deposited in the three deposits. This weight approx-
imates the recovery which might be achieved after many more
vacuumings.
• The weight of dust collected on the first vacuuming after the first dust
deposit (corrected for any fibers or carryover) as a percentage of the
weight of the dust deposited in the first deposit.
These estimates represent the extreme recoveries which might be calculated using
different definitions of recovery. These high and low estimated average recoveries
for each substrate are shown in Figure 6-10 as dashes to the right of the confidence
intervals.
69
-------
JT
>
0
u
Ol
t-l
to
Q
.LV/V /o •
90% -
80% .
70% -
60% .
50% -
40% -
30% -
20% -
10% -
0%
|: *= «• i-
: \ <-
•
^ .M x ^ ^ _ w
1 '? ^ * o I"6 S
ll1 a i " ii 1
r j 60 » £3 K ^
Substrate
Average with 95%
confidence interval
Range of average using alternate
definitions of dust recovery
Figure 6-10 Predicted average vacuum cleaner dust recovery for tested substrates
with 95% percent confidence intervals and average dust recovery using
alternate definitions of dust recovery
70
-------
fr
>
0
u
£
-------
Table 6-7 Predicted average vacuum cleaner dust recovery for tested substrates
and vacuum cleaners, with 95% confidence intervals.
Dust recovery
Substrate
Carpet with ground-in dust
Carpet
Upholstery
Wood
Sheet vinyl (linoleum)
Vinyl tile
76%
79%
90%
93%
92%
90%
Vacuum
A
B
C
D
95%
92%
85%
84%
95 confidence
interval
70% to 82%
75% to 83%
86% to 94%
90% to 96%
89% to 95%
83% to 97%
92% to 97%
89% to 95%
80% to 89%
80% to 88%
72
-------
The estimated dust recovery depends on the definition used for recovery. However,
the same general pattern of relative recovery among substrates is seen with any of
the definitions considered. Regardless of the definition used, at least 5% of the dust
is not recovered or accounted for even on the sheet vinyl substrate on which little
dust was expected to accumulate. This dust may have been caught in other parts of
the vacuum, such as the wand, hose, and other internal parts. However, the study
provides no data from which to determine the final destination of the dust not
collected in the vacuuming bag. For carpet substrates, a higher proportion of the
dust is not recovered and unaccounted for. Experience suggests that some dust may
stay in the carpet or pass through the carpet to the floor underneath.
6.4.2 Vacuum Cleaner Lead Recovery
Lead Recovery
In the vacuum cleaner tests, the lead recovery was determined by multiplying the
lead concentration in the dust which could be shaken from the vacuum cleaner bag
by the weight of the dust recovered and dividing by the weight of lead applied to the
substrate. On the average, 26% of the dust collected in the vacuum cleaner bag was
removed in this shaking procedure. Because the lead concentration in the vacuum
cleaner bag dust may differ from that in the dust shaken from the bag, the lead
recovery results for vacuum cleaners must be qualified.
The vacuum cleaner lead recoveries were estimated after correcting for dust
removal from the vacuum cleaner bags. The statistical analysis suggests that the
observed lead concentration in the dust removed from the vacuum cleaner bag
depends on the percentage of dust removed from the bag. The results of this
analysis were somewhat inconsistent, suggesting that the effect of dust removal
efficiency depended on whether the dust came from newer or older homes. The
statistical results provide a correction for different dust removal amounts, such that
comparing relative lead recovery among vacuum cleaners or substrates does not
depend on the dust removal from the vacuum cleaner bag.
The statistical results do not allow a correction to the overall vacuum cleaner lead
recovery estimates for effects associated with dust removal efficiency. The interac-
tion of both (1) vacuum cleaner and substrate and (2) nominal lead concentration
(dust from older or newer homes) and percentage of dust removed from the bag
were statistically significant. The average vacuum cleaner lead recoveries and
associated 95% confidence intervals by vacuum cleaner and substrate are shown in
Figure 6-12 and Table 6-8. The lead recoveries by vacuum cleaner and substrate are
similar whether or not the dust removal is included in the model. However, the
differences are not significant if the dust removal is not in the model.
73
-------
250% -r-
200% --
150% --
o
100% --
50% --
0%
"§-§
•Vacuum A
•Vacuum B
•Vacuum C
•Vacuum D
95% Conf.
Interval
O
O
I
I
U
-
1
Figure 6-12 Average vacuum cleaner lead recovery by vacuum cleaner and
substrate, with 95% confidence intervals
74
-------
Table 6-8 Average vacuum cleaner lead recovery by vacuum cleaner and
substrate, with 95% confidence intervals
Substrate Vacuum 1
cleaner
Carpet with ground-in dust A
B
C
D
Carpet A
B
C
D
Upholstery A
B
C
D
Wood A
B
C
D
Sheet vinyl (linoleum) A
B
C
D
Vinyl tile A
B
C
D
-ead recovery 95 confidence interval
109% 141% to 78%
119% 146% to 92%
56% 83% to 28%
109% 136% to 82%
108% 129% to 86%
103% 127% to 79%
65% 87% to 43%
110% 135% to 86%
136% 167% to 104%
146% 177% to 115%
124% 155% to 92%
82% 113% to 51%
108% 136% to 81%
86% 117% to 55%
109% 141% to 78%
83% 110% to 56%
123% 147% to 98%
85% 109% to 60%
86% 116% to 55%
102% 128% to 75%
109% 171% to 46%
92% 150% to 35%
146% 204% to 87%
56% 115% to -2%
75
-------
The average vacuum cleaner lead recovery (after removing two outliers) was 103%,
suggesting that more lead was recovered than was deposited on the substrate. A
likely explanation for the high lead recovery is higher removal of leaded than non-
leaded dust from the vacuum cleaner bag. Thus, the estimated recovery shown in
Figure 6-12 may consistently overestimate the actual lead recovery. However, the
amount of the overestimation cannot be determined from the study data.
Concentration Ratio
The concentration ratio is the ratio of the lead concentration in the dust removed
from the vacuum cleaner bag to the lead concentration in the dust applied. The
average concentration ratio across all tests was 1.12. Thus, the measured lead
concentration in the vacuum cleaner bag dust was greater than in the dust deposited
on the substrate. This result suggests that the lead concentration measurement in
the dust removed from vacuum cleaner bags tends to overestimate the lead concen-
tration of floor dust. Since a new bag was used for each test, the relationship
between the lead concentration in dust that might be removed from a previously
used or partially full vacuum cleaner bag and the lead concentration in the floor
dust was not tested in this study.
6.4.3 Effect of Cleaning Effort
The measurements of dust collected for each 40 seconds of vacuuming allow an
assessment of the effectiveness of vacuuming for collecting dust as a function of the
time spent vacuuming. For each substrate, Figure 6-13 shows the average weight of
dust recovered for each of the seven vacuumings which made up the vacuum
cleaner tests. The plots in Figure 6-13 are scaled so that the vertical scale measures
percent recovery for vacuumings 2, 3, and 4. High recoveries are found for these
vacuumings, each of which immediately followed the deposition of dust. The dust
recovered in vacuumings 5, 6, and 7 represents what remained on the substrate
from previous depositions. The dust collected in vacuuming 1 includes dust from
previous tests. For carpets and upholstery, the weight of dust collected includes
fibers.
For the sheet vinyl, vinyl tile, and wood, all of which are smooth substrates, essen-
tially all of the dust was collected in the first 40 seconds of vacuuming after the dust
deposit. For the upholstery, there is some evidence of a small amount of carryover
from the fourth to subsequent vacuumings.
For carpets, the dust recovery for vacuumings 2, 3, and 4 was lower than for other
substrates, with a higher recovery for the other vacuumings, indicating dust
carryover beyond the first 40 seconds of vacuuming.
76
-------
4567
Carpet with
ground-in dust
Vinyl tile
1234567
Upholstery
-20%
234567
Vacuum number
-20% J
1 2
4567
Vacuum number
Vacuum number
1 = vacuuming before depositing dust
2 = vacuuming after first dust deposit
3 = vacuuming after second dust deposit
4 = first vacuuming after third dust deposit
5 = second vacuuming after third dust deposit
6 = third vacuuming after third dust deposit
7 = fourth vacuuming after third dust deposit
Figure 6-13 Dust recovery versus vacuuming effort for six substrates
77
-------
6.4.4 Exhaust Emissions
One of the objectives of the pilot study (Task 5) was to determine the amount of dust
expelled through the vacuum cleaner bags. The procedures used for this task
measured the exhaust emissions from vacuum cleaners by placing five grams of
dust on a turntable and feeding the dust into the inlet of each vacuum cleaner at a
rate of one gram per minute. Each vacuum cleaner was placed in a sealed enclosure
and exhaust emissions from the vacuum cleaners were expelled through the only
duct in the enclosure. The dust emissions were measured in ng/m3 and recorded on
a strip chart recorder. A pitot tube was used to determine the total gas flow rate in
the duct, so that the dust emission rate (jig/min) and total emissions (jig) could be
calculated.
Exhaust emission levels were calculated from both the strip charts and readings
taken at one-minute intervals. For all but vacuum cleaner C, emissions were
higher while dust was being injected into the vacuum cleaner than before injection
began. The exhaust emission levels peaked generally during the fourth minute of
injection. For vacuum cleaner C (with a HEPA filter) the dust concentration in the
exhaust was below the ambient level even when dust was being injected. The
exhaust emissions from all four vacuum cleaners were lower than expected; an
average of 0.01% and at most 0.02% of the dust placed on the turntable was expelled
as exhaust. No lead measurements were done for the exhaust emissions because of
the time required to collect a sufficient amount of dust from the exhaust for
analysis.
In the initial design, two exhaust emissions tests were planned for the pilot study
and 12 more were planned for the full study. However, it was determined that all
twelve could easily be done in the pilot study, so no exhaust emission tests were
performed in the full study. The results from the exhaust emission tests are shown
in Table 6-9 and the complete documentation and results of the exhaust emissions
tests can be found in Appendix A.
78
-------
Table 6-9 Average vacuum cleaner exhaust dust concentrations by vacuum
cleaner
Vacuum
cleaner
A
A
A
A
B
B
B
C
C
C
D
D
D
Bag
1
2
3
4
1
2
3
1
2
3
1
2
3
Ambient
air levels
0.004
0.009
0.006
0.004
0.013
0.012
0.012
0.017
0.012
Dust expelled (mg/m^) as
exhaust before, during, and
after injection
Before
0.090
0.057
0.053
0.060
0.018
0.011
0.009
0.004
0.006
0.004
0.031
0.019
0.013
During
0.092
0.061
0.037
0.070
0.028
0.018
0.014
0.003
0.003
0.003
0.158
0.091
0.093
After
0.065
0.041
0.033
0.051
0.015
0.011
0.010
0.003
0.003
0.003
0.020
0.015
0.013
Dust not captured as a
percent of dust place on
turntable
Dust not
captured in
the bag
5.0%
2.4%
2.4%
3.7%
4.0%
2.8%
2.4%
4.7%
2.7%
2.4%
7.6%
4.9%
84.0%
Dust from
exhaust
emissions
0.020%
0.013%
0.008%
0.015%
0.006%
0.004%
0.003%
0.001%
0.001%
0.001%
0.021%
0.011%
0.012%
Average of the measurements above for each vacuum cleaner
A
B
C
D
0.006
0.010
0.014
0.065
0.013
0.005
0.021
0.065
0.020
0.003
0.114
0.048
0.012
0.003
0.016
3.4%
3.0%
3.3%
6.3% *
0.014%
0.005%
0.001%
0.015%
* Average excluding the outlier of 84
79
-------
6.5 Sampling and Measurement Error
Some recovery measurement variation is contributed by variation in the sample
collection, sample preparation, and lead analysis procedures. However, most of the
variation in the measurements is due to differences among tests using the same
dust source and dust particle size class, dust loading, substrate, and sampler or
vacuum cleaner. The standard deviation of an individual lead recovery measure-
ment was 21% for the samplers and 27% for the vacuum cleaners. The coefficient of
variation (standard deviation divided by the mean) of the lead recovery measure-
ments was 36% for the samplers and 26% for the vacuum cleaner tests. Although
there were significant differences between the operators performing the tests in the
amount of dust removed from the vacuum cleaner bags for lead analysis, there were
no other differences associated with the operators.
There are many possible sources of error including, among others, differences
between substrate samples, operators, vacuum cleaner bags, temperature and
humidity, and spatial variation in the dust deposited on the substrate. The data
from the study can provide no insight as to which of these the possible sources of
error might be most important and how the unexplained variation could be
reduced.
80
-------
DISCUSSION OF RESULTS
There are currently no standardized laboratory methods to assess how well samplers
collect house dust and dust lead or how well household vacuum cleaners clean
surfaces contaminated with leaded house dust.9 The lack of a standardized
sampling method necessitated that one be developed for this study.
The final procedure developed for this study used house dust sieved into specified
particle size classes. The dust was applied to standard substrates commonly encoun-
tered inside a residence. Substrate preconditioning steps were used to ensure that
no test was biased from previous tests. The test procedures proved easy to imple-
ment and can be easily duplicated by other researchers testing house dust collection
devices. By using the same test procedures, a baseline can be established for
samplers and various leaded dust evaluation studies can be compared. New collec-
tion devices that enter the market can also be quickly evaluated and compared to the
baseline.
This section highlights some of the laboratory test results presented in Section 6.
Section 7.1 discusses characteristics of the test dust used in this study. Sections 7.2
and 7.3 summarize the results for the samplers and the household vacuum
cleaners/ respectively. Relationships between these findings and other studies are
presented where applicable.
7.1 Test Dust Characteristics
As noted previously, the test dust used in this study was obtained from volunteers
who donated vacuum cleaner bags full of normal house dust Bags were collected
from homes within two age groups, older homes built before 1963 and newer homes
built after 1982. Dust collected from volunteers whose homes were built between
1963 and 1982 was not used. The dust from homes within each age group was
sterilized and then sieved into seven dust particle size classes. The weight of the
sieved dust and the dust lead concentration for the six smaller particle size classes
used in the study are reported in Section 6.
The findings show that the two groups of house dust, from older and newer homes,
contained roughly the same proportion of total dust, by weight, in each particle size
class. Also, as predicted during the design phase of this study, the dust from the
older homes was more lead-contaminated than the dust from the newer homes.
The mean dust lead concentrations were roughly 474 Hg/g and 61 Hg/g for the older
and newer homes, respectively. However, the distribution of lead concentration by
9The ASTM has published standard method F609-79 to evaluate the carpet-embedded dirt removal
effectiveness of household vacuum cleaners. However, this method uses artificial dust and was not
designed to examine cleaning effectiveness on surfaces contaminated with leaded house dust
81
-------
particle size class was dramatically different for the two age groups. This result was
unexpected and has not been demonstrated by previous studies.
Most studies that have examined lead in house dust by particle size class suggest that
lead concentrations in dust increase as particle size decreases. This phenomenon is
well documented with numerous references for soil, street dust, and house dust. In
the current study, the lead concentration in dust collected from newer homes
follows the expected inverse relationship with particle size, but the lead concentra-
tions in dust from the older homes did not exhibit the same relationship. Lead
concentrations in the dust from older homes remained relatively stable across
particle size classes, except for the largest size class which had the highest lead
concentration.
This study and others suggest that the observed differences in lead concentration by
particle size for older and newer homes may be explained by two common sources
of lead contamination in residential environments, namely lead-contaminated soil
and deteriorated lead-based paint. Since houses built after 1982 are unlikely to be
painted with lead-based paint, the dust lead in these houses must come from soil,
street dust, or other external sources. Since numerous studies show that soil and
street dust exhibit the inverse relationship rule for lead and particle size class, it
follows that lead in dust from newer homes should exhibit the same inverse
relationship. Dust-lead contamination in houses built before 1963 likely results
from deteriorated lead-based paint in addition to external sources of lead. If deterio-
rated paint dust particles are larger and more variable in size than tracked-in or
wind blown soil and street dust, then the inverse relationship between lead and
particle size may disappear in the dust contaminated by lead-based paint.
The suggestion that higher lead concentrations may be found in larger dust particles
and the results from the study that the samplers have lower lead recovery associated
with larger dust particles has implications for future studies. If the common belief
that finer particles are more adherent to children's hands and more readily absorbed
is correct, it may be reasonable to ignore larger particles when sampling.
Alternatively, if larger particles contribute significant amounts of lead to children
then sampling methods which collect both small and larger dust particles would be
preferred.
7.2 Samplers
The performances of one wipe and three vacuum samplers were evaluated in this
study. The vacuum samplers were tested for total dust recovery (total dust cannot
be measured by wipes) and all samplers were tested for lead recovery. Tests were
differentiated by substrate, by the nominal lead concentration of the dust applied to
the substrate (high and low lead concentration dust from older and newer homes,
respectively), by the dust loading levels (100 and 400 mg/sq ft.), and by the dust
particle size.
82
-------
The dust recovery is the weight of dust collected by the sampler as a percentage of
the dust applied to the substrate before sampling. For the dust recovery tests, the
results from the study indicate that the BRM and CAPS cyclone produced the
highest recoveries across all substrates and particle size classes. The recovery differ-
ence between the two cyclone devices was not significant. The Blue Nozzle sampler
had the lowest recoveries, statistically significantly lower than for the cyclone
samplers. These results agree with findings from previous studies that indicate that
the Blue Nozzle sampler has lower dust recovery than other tested methods. The
dust recovery for the Blue Nozzle sampler decreases as the particle size increases.
Conversely, the dust recovery for the CAPS cyclone and BRM sampler increases
slightly or remains constant as the dust particle size increases.
Sampling precision is a very important factor when sampling house dust. The
results from this study suggest that the BRM and the CAPS cyclones are more
precise sampling methods than the Blue Nozzle sampler. This is evident in the
standard deviations of the dust recovery measurements for the BRM, CAPS, and the
Blue Nozzle samplers, which were 9%, 12%, and 29%, respectively.
The samplers, in order of decreasing lead recovery across all substrates and particle
size classes, were the BRM, CAPS cyclone, wipe, and Blue Nozzle sampler. The lead
recovery of the Blue Nozzle sampler was significantly lower than for the other
samplers tested. Average lead recovery across all sampling devices was assessed as a
function of particle size class. The lead recovery remains relatively stable for the
fine particle sizes and drops off with the largest particle size class. The lead recovery
calculation uses the lead concentration measured in the test dust. Any error in the
measured lead concentration in the test dust will affect the recovery estimate. The
lead concentration measurement for the dust with the largest particle size class has
the largest sampling and measurement error. As a result, the apparent drop in lead
recovery for dust with the largest particle size class may be due in part to error in the
associated dust lead concentration.
The ratio of the lead concentration in the sampled dust to that in the dust applied to
the substrate was also examined for the vacuum samplers. When the samplers are
pooled, the concentration ratio is dose to 1.0 for dust particle sizes less than 212 pm,
but drops off sharply with the larger particle sizes. This implies that of the larger
particles, the vacuum samplers are selectively collecting a higher percentage of the
non-lead particles than of the lead particles.
7.3 Commercial Vacuum Cleaners
Commercially available vacuum cleaners with beater bar attachments were tested
for total dust and lead pickup capabilities. The same test dust and substrates used for
the samplers were used for the vacuum cleaners. For the vacuum cleaner tests, the
83
-------
dust loading in mg/sq ft was the same as for samplers, but the size of the test area
was larger so that the amount of dust applied was greater.
The dust recovery is the weight of dust collected in the vacuum cleaner bag as a
percentage of the dust applied to the substrate during the test. The vacuum cleaner
tests involved measurements of the dust collected in seven successive vacuumings
of 40 seconds each. Dust was applied to the substrates before the second, third, and
fourth vacuuming. Under these test conditions, the dust recovery could be calcu-
lated several ways. One such way could be to calculate the weight of dust collected as
a percentage of the dust deposited immediately before the vacuuming. This method
is most comparable to the definition of recovery used for the samplers.
Alternatively, recovery could be defined as the weight of dust collected in all seven
vacuumings as a percentage of the dust deposited in the three deposits. This defini-
tion would provide higher recovery estimates. For the results presented below, the
following intermediate definition of recovery is used: the weight of dust collected in
the second through sixth vacuumings as a percentage of the weight deposited in the
three dust deposits, after correcting for possible fibers of dust carried over from other
tests.
The dust recovery performance of the vacuum cleaners was, as expected, highest for
the hard substrates and lowest for carpets. The average recovery ranged from 76%
on carpets with ground-in dust to 93% on wood substrates, and the average varied
among vacuum cleaners. Differences among vacuum cleaners were small though
statistically significant.
Measurement of lead recovery for the vacuum cleaners, which are not designed for
making lead measurements, proved difficult. It is not possible to remove all of the
dust from the vacuum cleaner bag for testing and it is difficult to measure the lead
in the dust without removing the dust from the bag. The procedure used in this
study required shaking the dust from the vacuum cleaner bag into a laboratory
bottle for subsequent lead analysis. The results can be biased if the dust removed
from the bag is not similar in lead concentration to the dust left behind in the bag.
The statistical analysis corrected to the extent possible for different amounts of dust
removed from the vacuum cleaner bags.
Overall average lead recovery was 103%. However, the vacuum cleaner lead
recovery depended on the combination of vacuum cleaner and substrate used in the
test. Average recovery for various combinations of vacuum cleaner and substrate
ranged from 56% to 146%. The consistent recoveries over 100% suggest that the dust
removed from the vacuum cleaner bags may have had higher lead concentrations
than dust remaining in the bags, although there is no data from the study to directly
support this conclusion. Although the high lead recovery estimates suggest that the
true lead recovery for tested vacuum cleaners is high, the amount of bias associated
with the test procedures cannot be assessed directly. The CAPS cyclone and BRM
samplers both operate in a manner similar to the vacuum cleaners except that the
vacuum cleaners had a beater bar and used a filter rather than a cyclone to remove
84
-------
the dust from the airstream. The vacuum cleaner tests involved more effort
vacuuming than did the sampler tests. Due to both of these factors, the expected
lead recovery for the vacuum cleaners would be greater than that for the BRM and
CAPS cyclone vacuum samplers, for which the lead recovery was 81% and 72%,
respectively.
The ratio of the lead concentration in the dust removed from the vacuum cleaner
bag, to the lead concentration in the dust deposited on the substrate, averaged across
all vacuum cleaner tests, is 1.12. In other words, dust from the vacuum cleaner bag
provides an estimate of floor dust lead concentrations that is biased by about 12%.
This result suggests that procedures which use dust from vacuum cleaner bags to
assess possible lead contamination problems will produce lead concentrations which
are somewhat higher than the actual concentration in the dust. However, this study
only used new vacuum cleaner bags. This conclusion might not apply to full bags.
The analysis of vacuuming effort versus dust recovery (in both the pilot tests and
the vacuum cleaner tests) indicates that 80% or more of the dust which has been
recently deposited is collected within the first 40 seconds of vacuuming, even when
the dust has been ground in. Of the remaining dust, most is collected in the next
few minutes of vacuuming. Some dust (at least 5%) may remain in the carpet,
upholstery or substrate, or parts of the vacuum cleaner and may not be collected in
the bag or otherwise accounted for. With the caveat that not all of the dust
deposited on the substrate is accounted for in the data, of the dust collected, almost
all is collected in the first 40 seconds of vacuuming.
In the exhaust emissions test originally planned for the study but conducted only in
the pilot tests, only 0.01% on average, and at most 0.02%, of the dust collected by the
vacuum was emitted in the exhaust. The dust emissions test was performed with
dust of the smallest dust particle size class, believed to be the size most likely to pass
through the vacuum bag into the exhaust. Based on these tests, the fraction of dust
which passed through the vacuum bag is very small. For tests on the vacuum with
a HEPA filter, the exhaust had lower dust concentrations than the ambient air.
One question not answered by this study is the extent to which the vacuum cleaner
exhaust kicks up dust on the floor and thereby increases the airborne lead concentra-
tion. In this study, the vacuum cleaner exhaust may have disturbed some of the
dust deposited on the substrates and thus account for some of the dust not otherwise
collected in the vacuum cleaner bag. However, the quantity of dust disturbed by the
exhaust is likely to be very small because the canister type vacuum cleaners were
located on the floor, six inches below the substrate testing surface, and the upright
model was well above the substrate surface.
85
-------
7.4 Effect of Sampling Method on Estimates from the National Survey
of Lead-Based Paint in Housing (HUD National Survey)
The dust samples in the HUD National Survey were collected using the Blue
Nozzle vacuum sampler. The results were used to estimate the number of priority
homes nationally, that is the number of private dwelling units with lead-based
paint (LBP), and either non-intact paint or dust loading exceeding the HUD guide-
lines. Priority housing is further classified as having or not having children under
age seven.
The HUD guidelines apply to clearance sampling after renovation and assume that
wipe samples are used. The results of this and other studies suggest that the dust
and lead recovery of the Blue Nozzle sampler is significantly below that of other
samplers, including the wipe sampler. If the wipe or another sampler had been
used in the HUD National Survey, how would the number of priority homes
change?
The number of priority homes with children under age seven was reported as 3.8
million in the Comprehensive and Workable Plan (CWP).10 In subsequent revision
of the survey results to account for the calibration of the x-ray fluorescence (XRF)
equipment and the incomplete sampling of rooms, this number was increased to 4.0
million. See the Report on the National Survey of Lead-Based Paint in Housing,
Appendix II, Table 2-8 for details. Using the wipe, BRM, or CAPS cyclone sampler,
all with higher recovery than the Blue Nozzle sampler, the estimated number of
priority homes with children under seven would be greater than 4.0 million.
Figure 7-1 shows the estimated number of priority homes with children under
seven as a function of the recovery of a selected sampler relative to that of the Blue
Nozzle sampler. The jagged shape of the curve is due to the discrete nature of
survey results. Homes are classified as either having or not having dust over the
HUD limits. A home cannot be classified as half over the limit and half under.
The average lead recovery of the Blue Nozzle and wipe samplers is 26% and 63%
respectively, and thus the wipe sampler collects about 2.4 times as much lead as does
the Blue Nozzle sampler. If we assume that, on all surfaces, the Blue Nozzle
sampler consistently recovers 42% of the lead that would be collected using a wipe
sampler (i.e. the lead loading for clearance is 2.4 times the measured loading), the
revised number of priority homes, determined by analyzing the survey data, would
be 4.6 million instead of the 4.0 million based on the Blue Nozzle sampler.
10U.S. Department of Housing and Urban Development, Office of Policy Development and Research
(1990), Comprehensive and Workable Plan for the Abatement of Lead-Based Paint in Privately owned
Housing: Report to Congress.
86
-------
8,000 -
7,000 -
-------
7.5 Additional Questions
As with most research studies, some questions are left only partially addressed and
others are generated as the findings from the study are analyzed. Some original
questions that are partially addressed include the relationship of lead and dust
recovery to dust particle size in combination with the substrate, sampler or vacuum
cleaner. When the study was redesigned due to budget constraints, the number of
tests was reduced, limiting the researcher's ability to identify the effect of interac-
tions between dust particle size and substrate or sampler on the lead or dust
recovery. In addition, the precision of some of the measurements (particularly lead
recovery ) was lower than was assumed in the planning stages. Thus, more data
collected in a similar manner can be used to provide additional and or more precise
information.
The study results have also suggested additional questions which were related to the
study objectives but not anticipated in the design. The primary questions involve
the location of the dust which was not collected in the vacuum cleaner bag and not
seen as carryover from previous tests. Additional information on the location of
the dust can be obtained from efforts to collect dust from the vicinity of the
substrate, on unvacuumed areas of the substrate, below the substrate for carpet and
upholstery substrates, in the air, and in the internal parts of the vacuum. A related
question is: would the unaccounted-for dust pose a threat to children? Other
questions are: what is the lead concentration in the dust which is easily removed
from the vacuum cleaner bags compared to the dust which remains in the bags?
How can the lead in vacuum cleaner bags be measured in an unbiased manner?
Still other questions are those which the study was not designed to answer but
which are important for addressing the overall objectives of the research effort.
These questions are discussed in the following paragraphs.
In dust from homes built before 1963, the lead concentration was found to be similar
for all dust particle size classes. This relationship was based on dust composited
from vacuum cleaner bags from many homes. Additional studies of dust collected
from individual homes can provide information on the extent to which this
conclusion can be generalized to all older homes. The relationship between dust
particle size and lead concentration may vary among homes depending on the age
of the home, presence of children and pets, or other factors. Any differences might
affect the risk to young children and, therefore; the choice of sampler for assessing
the risk.
EPA has recommended that household vacuum cleaners not be used to clean up
lead containing dust after renovation, in part due to concern about small particles
passing though the vacuum cleaner bag that then may produce an airborne dust
lead hazard. In this study, almost no dust passed through the vacuum cleaner bags,
However, only new vacuum cleaners and new bags were tested. Whether the
conclusion that very little dust passes through the vacuum cleaner bags can be
-------
extended to full vacuum cleaner bags and older models of vacuum cleaners has yet
to be determined.
The extent to which the vacuum cleaner exhaust disturbs dust, making it airborne
and creating a temporary lead hazard, has yet to be determined. How much dust
(and lead) is lifted into the air by the vacuum cleaner exhaust in typical home use?
How soon does the airborne dust resettle, and how soon after vacuuming are
airborne dust and lead levels safe for children? Of the dust which is not collected by
the vacuum cleaner bag, does the vacuuming and/or exhaust cause the residual
dust to move to areas which provide an increased or decreased lead risk to children?
Vacuuming, particularly with a new vacuum with a beater bar attachment, may
bring dust from deep in the carpet to the carpet surface where it is more easily
available to children, thus increasing the lead hazard. Questions to be answered
include: under what conditions can vacuuming be effective in reducing the lead
hazard? The type and condition of the substrate, type of vacuum cleaner and, for
carpets and similar substrates, the amount of accumulated lead will likely affect the
answer to this question.
If it can be determined whether and how vacuuming can reduce the lead hazard
from floor dust without increasing the hazard from other sources, other questions
to answer are: how quickly does dust accumulate? What vacuuming frequency is
necessary to control dust and lead loading? How much dust and lead do children
ingest from a freshly vacuumed floor, representing the minimum exposure that can
be achieved with vacuuming?
7.6 Final Comments
The questions first posed to motivate this evaluation of sampler and vacuum
cleaners include:
1. What are the best methods of measuring lead in house dust?
2. What levels of dust lead can be maintained by a typical homeowner
using regular vacuuming?
3. Can a homeowner be assured that the vacuuming process does not
create an airborne lead hazard? Or, stated another way, how much
leaded dust passes through normal household vacuum cleaner bags
used over an extended period of time?
Although complete answers to these questions require more research, this study
provides the following preliminary answers to these questions.
89
-------
1. The best methods of measuring lead in house dust vary by the situa-
tion and depend on many factors, such as the cost, ease of use, relative
recovery, and study objectives. This study provides information only
on relative recovery. Overall, the BRM, CAPS, and wipe methods
have similar recoveries and precision. More information is required to
determine which method is preferred in any one situation. This study
provides some information to help select the preferred sampling
method. It is clear, for instance, from this and other studies that the
selection of the sampling method does make a difference, with the
Blue Nozzle collecting less dust and dust lead than other sampling
methods tested. The differences have particular application to inter-
pretation of the results from the HUD National Survey (see Section 7.4)
and to the selection of sampling procedures for clearance testing.
2. The results of this study show that a highly rated vacuum cleaner with
a beater bar attachment will pick up at least three-quarters of the loose
dust present on a variety of surfaces with a moderate vacuuming time.
How much more dust is picked up depends on many factors, such as
the vacuum cleaner design and whether the dust is ground into the
surface.
The study suggests that lead recovery would be similar to the dust
recovery. This study provides no information on how quickly dust
accumulates and the levels of dust lead which could be maintained
with regular vacuuming. While it is clear that vacuuming removes
dust and leaded dust from the vacuumed surfaces, thus reducing the
total amount of lead which might pose a risk to young children, it has
yet to be determined if routine vacuuming will reduce leaded dust in a
way which will result in reduced blood lead levels.
3. This study shows that, for the four vacuums tested (one of which had a
HEPA filter) very little dust passes through the vacuum cleaner bag.
Further studies are required to determine if this result can be extended
to other vacuum designs and to older used vacuum cleaners. Aside
from the vacuum cleaner itself, the vacuum cleaner exhaust and the
vacuuming process can disturb dust in the room, increasing airborne
dust lead levels and possibly creating an airborne lead hazard.
90
-------
8 DATA PROCESSING AND STATISTICAL ANALYSIS PROCEDURES
This section describes, in detail, the data processing and statistical analysis proce-
dures used to derive the results presented in Section 6.
8.1 Data Entry and Data Processing Procedures
The data was supplied to Westat by MRI on paper and, for the lead analysis results,
as text files on computer diskettes. Westat entered the data into computer files,
identified outliers or possible errors, and verified the computer files against the
original data submissions. Outliers were reviewed by MRI and verified as correct or
corrected if possible. As part of the process, Westat had discussions with MRI and
visited MRI to make sure the Westat personnel understood the test procedures and
how each data element was generated. This helped to assure that the statistical
procedures were appropriate for the data. The data entry and verification proce-
dures were different for the gravimetrics (weight measurements) and lead analysis
data, as described below.
The first processing step for the gravimetrics data in this study required entering the
data onto spreadsheets whose layout was similar to the actual data sheet. Once the
data were entered, the spreadsheet files were converted to an ASCII data file and
transmitted to EPA's National Computer Center (NCC). A SAS data file was created
from the ASCII file and the data were printed in the format of the original data
sheets. The printed data sheets were then compared to the original data sheets and
any errors were corrected in both the NCC and spreadsheet files. With the correc-
tions in place, the spreadsheets files were again converted into ASCII files and the
corrections in both the NCC and newly created ASCII files were reverified. The
process of comparing the data in the SAS file to the data on the original sheets
checks for both data entry errors and data processing errors. Figure 8-1 shows the
processing steps required to prepare the final gravimetric data files.
The text files generated by MRI containing the lead analysis reports were edited to
remove introductory text material and then assimilated into one spreadsheet file
containing all the lead analysis information. Several variables were combined or
modified to make the subsequent analysis simpler. The spreadsheet file was
converted to a dBase file and then to a SAS file. The data in the SAS file were then
converted to a text file with the same format as the original text file. The text file
prepared from the SAS file was electronically compared to the text files generated by
MRI. Any errors were corrected in the spreadsheet and SAS files. The corrected
SAS file was converted to an ASCII file and sent to NCC where the ASCH file was
converted into a SAS file. Figure 8-2 shows the processing steps required to prepare
the final data file for the lead analysis data.
91
-------
Enter the data into a
spreadsheet on a PC
Convert the data to an
ASCH data file
1
Transmit the data to NCC,
create a SAS data file
Print the data in the format of
the original data sheets
Compare the original data
to the formatted data from
the database
Correct errors in the NCC
files
1
Correct errors in the PC
spreadsheet
Print the data in the format of
the original data sheets
Verify that the corrections
were made
Verify that the corrections
were made
Convert the data to an
ASCH data file
»
»
Check for outliers and
internally inconsistent data
Figure 8-1 Flow chart for data entry and verification of gravimetrics data
92
-------
Start with six text files, each
containing the analysis report
for one instrument batch
Read each text file into a
PC spreadsheet
Combine six spreadsheets
into one
Rearrange and modify the
spreadsheet data to eliminate
duplicate information
Create a PC-SAS data file
from the spreadsheet file
Prepare a SAS program to compare
the data in the PC-SAS file to the
data from the original text files
Run the program to verify
the data in the PC-SAS file
''Are there any
unexplained
X^ifferences?/^
No
Prepare a text file from the
PC-SAS file
Send the text file to NCC
Create a SAS file at NCC
Check for outliers and
internally inconsistent data
Document differences
between the SAS data files
and the text reports from MRI
Correct the PC-SAS
file or the
verification program
Correct and verify
any errors in the
spreadsheet file
Figure 8-2 Flow chart for data entry and verification of lead analysis data
93
-------
8.2 Statistical Analysis Procedures
8.2.1 Overview of the Statistical Analysis Procedures
The statistical procedures used to analyze the data were chosen to be appropriate for
the purpose of the analysis, the experimental design, and the characteristics of the
data. In general, regression models (including analysis of covariance models) were
used to analyze the data. The general procedures that formed the basis for most of
the analyses are described in this section. The specific statistical procedures used for
individual analyses may have differed somewhat from the general procedures,
depending on characteristics of the data and the purpose of the analysis.
Modifications of the general procedures are discussed in the following sections
which discuss individual analyses.
The general approach for fitting a regression model used the following steps:
(1) Starting with the basic model, identify the preliminary model, remove
outliers as necessary.
(2) Determine if regression weights are needed to equalize the measure-
ment variance across observations and, if so, calculate the regression
weights.
(3) Fit the final model.
(4) Check that observations removed as outliers are outliers based on the
final model and regression weights, check residuals for heteroscedas-
ticity, approximate normality, check for serial correlation or other
possible problems.
(5) Refit if necessary.
Choosing the model
The original design was designed to estimate main effects for operator, dust loading,
nominal dust concentration, dust particle size, substrate, sampler or vacuum
cleaner, and interactions between sampler and both substrate and dust particle size.
Each combination of dust particle size and substrate shown in Table 4-2 was to be
tested using each sampler.
There were four sections of each substrate, one for each combination of dust loading
(100 and 400 mg/sq ft) and nominal lead concentration (low and high, correspond-
ing to dust from newer and older homes). In the experimental design, the same
substrate sample was used for all tests using the same combination of dust loading
94
-------
and dust lead concentration. Since independent substrate samples were not used for
each test, the individual tests are nested within the substrate sample. The following
paragraphs describe how the analysis reflected this nested design.
The original experimental design was modified as a result of the pilot tests and, after
beginning the full study, in response to budget pressures. As a result of the time
required to precondition the carpet substrate in the pilot study, the final design was
modified to require the same substrate sample for all tests with the same dust
loading and dust lead concentration. This modification created a nested design.
After the tests for the full study began, it was necessary to cut back on the number of
tests to stay within the budget for the project. The redesign of the study was
performed quickly and consisted of specifying a fraction of the tests from the original
design. In the redesign, the tile substrate was eliminated from further testing, and
not all samplers were tested on each combination of substrate and dust particle size
shown in Figure 4-15.
The original design assumed a new substrate sample for each test. Due to an
oversight, the original design was not modified to reflect the nested design which
was adopted as a result of the pilot tests. The incorrect assumption of independent
substrate samples was also reflected in the redesign. Thus, the redesign did not
reflect the nested design which had actually being adopted. Although the redesign
created a roughly balanced experimental design for the main factors and important
interactions, these terms were not balanced with respect to the interaction of sub-
strate, dust loading, and nominal dust concentration corresponding to the substrate
samples in the nested design. As a result, some of the independent variables in the
full model are correlated, resulting in less power than was originally intended. In
most cases, the effect of the correlations appears to be small.
In a preliminary analysis, the results of the study were analyzed as a nested design
on the assumption that differences between substrate samples were significant. In
fact, the analysis results showed that the differences among substrate samples were
not close to statistically significant in any of the analyses and that the estimated
variance component among substrate samples was often negative. Therefore, for
the final analysis presented in this report, the nested nature of the design was
assumed to be insignificant and nesting was not included in the statistical model.
The statistical analysis started with a basic model which reflected all the factors in
the experimental design. The basic model had terms for:
• A full factorial model of substrate, dust loading, and dust lead concen-
tration (corresponding to the individual substrate samples in the
nested design).
• A quadratic model for the log of dust particle size, used to test for non-
linear differences associated with the log of the dust particle size.
95
-------
• An interaction between sampler (or vacuum cleaner) and substrate, the
two factors expected to most affect the recovery.
• An interaction between the log of dust particle size and both substrate
and sampler (or vacuum cleaner).
• An interaction between the log of dust particle size and nominal dust
lead concentration. This term was used only in models for lead recov-
ery and lead concentration ratio for which the recovery depended on
the measured lead concentration in the dust. This in turn was a func-
tion of the interaction of dust particle size and dust lead concentration.
For the basic model, the logs of 35, 75, 126, 178, 230, and 707 were used to approxi-
mate the median particle size in the size classes <53, 53-106,106-150,150-212, 212-250,
and 250-2,000 microns (M-m), respectively. This basic model was applied to the dust
recovery, lead recovery, and lead concentration ratio. The lead concentration ratio is
the ratio of the lead concentration in the dust removed from the sampler or
vacuum cleaner bag to the lead concentration in the dust deposited on the substrate.
Factors that were not statistically significant were eliminated from the basic model
in a step-wise (manual) manner to obtain a parsimonious model for the factors
which affect the response variable. As the least significant factors were removed,
the degrees of freedom for estimating measurement error increased. When there
were enough degrees of freedom and, if the terms involving the log of the dust
particle size class were close to significant, the continuous variable, log of the dust
particle size, was replaced by the dust size class variable. The model obtained after
additional stepwise elimination of statistically insignificant factors is referred to as
the preliminary model.
As described below, the residuals from the preliminary model were used to deter-
mine if there was significant heterogeneity in the measurement variances
(heteroscedasticity). If so, regression weights were calculated and used to identify the
final model.
Starting with the preliminary model and any regression weights, the final model
was identified by considering the effect of changes in the preliminary model on the
parameter estimates and p-values. The changes that were considered included
adding terms, removing terms, using or not using regression weights, including or
removing outliers, and using transformations of the response variables. The objec-
tive when identifying the final model was to understand how the assumptions
affected the statistical results and to identify one model which reasonably summa-
rized relationships among the data. The presentation of the statistical results
includes both a presentation of the final model and a description of how the results
are sensitive to the assumptions.
96
-------
The statistical methods test for differences among the different levels of a factor,
such as differences among samplers or dust particle size classes. If significant differ-
ences are found, the pattern of those differences is described in the text. Formal
multiple comparison procedures to compare pairs of levels (for example, to compare
two samplers) were not performed. Occasionally, in the description of the response
patterns, differences are designated as significant based on the following conserva-
tive procedure: if two 95% confidence intervals do not overlap, the means are
assumed to be significantly different. Similarly, differences are designated as not
significant based on the following conservative procedure: if the 95% confidence
interval for either mean overlaps the other mean, the means are not significantly
different.
Before the design was scaled back, four vacuum cleaner tests were performed using
the tile substrate. The small number of measurements on tile provided little
information on the correct model for the data. In order to have a more balanced
design, the vacuum cleaner tests on tile were excluded from the analysis while
identifying the factors in the final model. These tile measurements were then
included in order to calculate estimates for the final model. Thus, the model which
fit the data from tests using other substrates was assumed to fit the tests using tile.
In some cases, terms were temporarily added to the final model to test for possible
carryover (serial correlation associated with either the substrate or the sampler or
vacuum cleaner), instrument batch or calibration effects, or other effects. Log and
power transformations were considered to normalize the residuals. In most cases,
transformations were not needed and therefore not used. Residuals were analyzed
to verify that error variance did not vary significantly among classes of observations
and that the distribution of the residuals was roughly normal.
Identifying outliers
The extreme studentized residual (ESR) is used to identify residuals which are
associated with outlying observations. The extreme studentized residual is the
maximum absolute values of the studentized residual. The studentized residual is
an optional output from many regression programs. The ESR test assumes that the
residuals have a normal distribution. The critical values for the ESR, shown in
Table 8-1, depend on the number of observations. They also depend on the model
used to obtain the residuals and the criteria for defining the significance level for the
test. The values in Table 8-1 are appropriate when fitting just a mean.
Consideration of additional factors in the model would have slightly decreased the
critical values shown in Table 8-1. Therefore, use of the values in Table 8-1 repre-
sents a conservative test; the true probability of deciding that the most extreme
observation is an outlier is less than the nominal 5%. For numbers of observations
not shown in Table 8-1, interpolation was used.
97
-------
Table 8-1 Critical values for the extreme studentized residual (5% level)11
Number of observations
20
24
30
35
40
60
120
240
480 ,
960 1
Critical value
2.777
2.861
2.958
3.02a
3.08 «
3.23 a
3.47 a
3.67 a
3.86 «
4.03 a v
aApproximate values predicted, using nonlinear regression, from
theoretical values based on a normal distribution with known
mean and standard deviation.
11For a mean model, the extreme studentized residual (ESR) has the following relationship to the
extreme studentized deviate (BSD) and the maximum normed residual (MNR):
Critical values for the maximum normed residual are presented in Snedecor, G. W., and Cochran, W. G.,
1980. Statistical Methods, Seventh Edition, Iowa University Press, Ames, Iowa
98
-------
Outliers were removed from the analysis in order to identify the final model. They
were then included to determine if, based on the final model, they would still be
classified as outliers.
Determining regression weights
Regression methods assume that the errors, after applying any regression weights,
have constant variance across all observations. Appropriate regression weights
improve the estimates and their confidence intervals. These weights are propor-
tional to the inverse of the error variance, in this case, the sampling variance plus
measurement variance.
One method for identifying classes of observations that have different variance,
suggested by Levene,12 uses analysis of variance or regression on the absolute values
of the model residuals. In this study, a modification and refinement of this basic
approach was used. Regression analysis was performed on the following function of
the studentized residuals:
Vi = ln(0.05 + ri2)
where ri is the studentized residual for the ith observation i. The square of the
studentized residual has a chi-squared distribution with one degree of freedom
assuming constant variance. The constant 0.05 makes the log transformed values
approximately normally distributed. This constant is generally small relative to ri2
which has a mean of 1.0. The standard deviation of V is roughly constant even
when heteroscedasticity exists. A regression model, referred to as the variance
model, with terms for the main effects, is then fit to V. To account for the fact that
the studentized residuals are not independent, weighted regression, using regres-
sion weights equal to 1-hi, was used to fit the variance model, where hi is the diago-
nal element of the hat matrix and 1-hi is proportional to the variance of the
residual. The values hi are an optional output from many regression programs. In
addition, for testing significance of factors in the variance model, the residual
degrees of freedom for error in the variance model is the error degrees of freedom
from the preliminary model minus the number of parameters fit to V. Note that
the sum of (1-hj) is the error degrees of freedom from the preliminary model.
Simulations indicate that this approach performs well in maintaining the false
positive rate under constant variance and reasonable power to detect differences in
variance when they exist. This procedure for calculating regression weights has the
advantage that it is relatively simple, can be used for complex models, and provides
reasonable estimates of the regression weights.
12Levene, H. 1960. In Contributions to Probability and Statistics. Stanford Univ. Press, Stanford,
Calif., p. 278
99
-------
Because the regression weights need only be proportional to the inverse of the error
variance, the predicted values (Pred(V)) from the regression on V can be used to
calculate regression weights (Wgt) for the final model using the following formulas:
Wtemp = l/(exp(Pred(V)) - 0.05).
Wgt = Wtemp/Mean(Wtemp).
These weights are standardized for convenience so that the sum of the weights
equals the number of observations. A second iteration can be used to refine the
weights by using the residuals from the weighted regression to calculate a second set
of weights which are multiplied by the first. These refined weights, when calcu-
lated, were also standardized.
The following steps were used to test for non-constant measurement variance:
(1) Save the studentized residuals and hi (the diagonal of the hat matrix)
from the preliminary regression.
(2) Calculate V and the regression weights for the variance model.
(3) Fit a parsimonious model to V to test for non-constant variance,
assuming that the possible factors in the variance model are the same
factors as in the preliminary model. Sometimes other factors were also
considered.
(4) If the overall F statistic for the variance model is significant (based on
the adjusted error degrees of freedom), assume that heteroscedasticity
exists, save the predicted values, and calculate the regression weights
for the final model.
A parsimonious model for variance is required because the degrees of freedom for
estimating variance effects is reduced by the model degrees of freedom in the model
from which the residuals were obtained. There may not be many degrees of
freedom left to fit a complicated model. A parsimonious model is also reasonable
compared to the generally accepted procedure of assuming constant measurement
variance.
A note on the results
The theoretical values for the recovery range from 0% to 100%. Some confidence
intervals, as well as some of the measurements, extend beyond the theoretical limits
of 0% and 100% recovery. The measurements can be outside this range due to varia-
tion in the testing and measurement process. The confidence intervals apply to the
true mean for the measurement process and not to the true mean for the actual
100
-------
process. Although the recovery estimates should be unbiased in a statistical sense,
and the confidence intervals should fall within the 0% to 100% range as more data is
collected, some readers may be uncomfortable with estimates which appear to be
illogical. That the interval exceeds 100% does not indicate that an incorrect method
was used to calculate the confidence interval. Significant effort, possibly including
simulations, would be required to calculate and justify alternate confidence
intervals which are restricted to the range of 0% to 100%. For that reason, the
modeling was not performed.
8.2.2 Statistical Analysis of Sieved Dust Lead Concentration
The dust in this study came from donated vacuum cleaner bags which were used in
either older homes, built before 1963, or newer homes, built after 1982. The dust
from the vacuum cleaner bags was removed and sieved into the following six
particle size classes: less than 53 Jim, 53 to 106 ^im, 106 to 150 Jim, 150 to 212 um, 212
to 250 Jim, and 250 to 2,000 fim. Particles with sizes greater than 2,000 Jim were
discarded. Dust from homes in the same age class and in the same particle size class
was physically mixed and placed in a plastic bag.
Before the study began, duplicate grab samples of dust were selected from each bag of
dust after mixing the dust and were then analyzed for lead. As the study progressed,
samples of dust were collected periodically to measure the lead concentration in the
dust actually deposited onto the substrates and to determine if the concentration
changed over time due to settling or stratification in the bags of dust.
As described below, a weighted analysis was used to analyze this data because a
preliminary analysis suggested that the variance of the measurements varied
considerably among dust particle size classes. The variance of lead concentration
measurements can be estimated either from the data or by using a theoretical model.
The theoretical model provides insight into the expected patterns of variance as a
function of dust particle size and was used to calculate the regression weights for
analysis of the dust data. The following discussion describes the assumption behind
the theoretical model.
The dust particles are assumed to be of two types, leaded particles which have lead
associated with some non-lead material and non-leaded particles which contain no
lead. Within a dust particle size class, the dust particles are assumed to be the same
size. The lead concentration in the leaded particles is assumed to be the same for all
such particles. The process of sampling dust for analysis is assumed to be a random
selection of dust particles, some with lead and some without. Each type of particle is
assumed to have the same chance of selection. The number of particles in a sample
can be determined from the weight of the dust compared to the average weight of a
dust particle. The proportion of the leaded particles in the sample will have a
binomial distribution. The lead concentration in the dust sample will depend on
the number of leaded particles and the amount of lead in each particle. The relative
101
-------
variance of the lead concentration in the dust sample can be determined from the
binomial relative variance. These relationships are described in more detail in the
following equations.
If the weight of lead in the leaded particles is assumed to be R times the weight of
the non-lead particles and the proportion (P) of leaded particles is small, then P is
approximately:
P = C/(R* 1,000,000 Oig/g))
where C is the lead concentration (jig/g) in the dust sample. For these calculations,
R is assumed to be one-half, that is, the weight of lead in the leaded particles are
assumed to be half the weight of the non-lead particles. Because the density of lead
is much greater than 1.5 times the density of most dust components, setting R = 0.5
is equivalent to assuming that the lead is associated with paint constituents or other
relatively light material.
The relative variance (square of the coefficient of variation) of P is:
Relvar(P) = (1-P) / (nP)
where n is the number of particles in the sample to be analyzed in the lab. The
number of particles in the sample is the weight of the sample divided by the weight
of a particle, which is the volume of the particle multiplied by the density. The
volume (V) of a spherical particle in cubic centimeters is:
V=[0.81d/(10,000)]3
where d is the diameter in microns. The assumed diameter of the particles is 35, 75,
126,178, 230, 707 for the particle size classes <53, 53 to 106,106 to 150,150 to 212, 212
to 250, and 250 to 2,000 microns, respectively. These diameters are approximately
the geometric mean between the largest and smallest size within the size class.
Assuming that the density of the dust particles is 1.0 gm/cc (the density of water),
the number of particles in the sample is:
n = Wt / V
where Wt is the weight of the sample in grams. The estimated number of dust
particles is greater than 3,500 for all of the sieved dust samples. With this number of
particles, very little of the variation in the weight of the dust collected is associated
with the particle size.
When analyzing the natural log of the lead concentration measurements, the
regression weights should be roughly proportional to the inverse of the relative
variance of the measurements. The relative variance of the measurements is equal
102
-------
to the sum of the relative variance due to sampling and the relative variances due
to preparation and measurement. In the following formula for the regression
weights (Wgt), the relative variance associated with preparation and measurement
is based on the analysis of precision presented in Section 8.2.6.
Wgt =
\2
0.0183+ ,U-U184 +(0.0376)2
InstResp)
In this formula, InstResp is the ICP instrument response. All sieved dust samples
were analyzed using ICP.
One outlier was identified in the preliminary analysis. Based on the final regression
weights, this outlier was 4.55 standard deviations from the mean (test number 641,
using the smallest dust particle size). The outlier was removed when calculating
the geometric means and the final regression weights.
The regression weights range from .18 to 460. The ratio of the largest to smallest
weight is 2,557 to 1. The wide range in the regression weights suggests that the
variance of the lead concentrations varies considerably among dust samples of
different dust particle size classes and different lead concentrations.
Figure 8-3 shows the relative variance, expressed as a coefficient of variation (CV), as
predicted by the model for the regression weights, of the lead measurements for
each dust bag, averaged across all bags. According to the model, the measurements
for the coarse dust samples are much more variable than for other dust samples.
Similar results may apply to the lead recovery measurements from the sampler and
vacuum cleaner recovery tests. Figure 8-3 also shows the coefficient of variation for
the measurements from each bag of dust. In general, the observed CV's are close to
those predicted by the model.
These regression weights were used to fit a model and estimate the geometric mean
lead concentration for each bag of dust (one for each dust particle size class and age
of house) and the associated confidence interval, and to determine if there was a
trend in the lead concentrations over time.
If the assumptions used to calculate the regression weights and remove outliers are
correct, the mean square error in the final model would be 1.0. The mean square
error for the final model is 1.09. This value is very close to 1.0 considering the
approximations which were used to derive the regression weights, and it indicates
that the relative variance from the model may slight underestimate the true
relative variance for the data. Since the confidence intervals will be correct if the
relative regression weights are correct, even if the regression weights are consis-
tently biased, the regression weights appear to provide a reasonable basis for calculat-
ing the confidence intervals.
103
-------
1.4 -r-
en
3
T3
•a
8!
V
• •N
01
tn
§
5
1.2 --
0.8 --
8
I
m
1
0.6 --
^ 0.4 --
Predicted (older
homes)
Predicted
(newer homes)
CV (older
homes)
CV (newer
homes)
0.2 --
<53
53-
106
106-
150
150-
212
212-
250
250-
2,000
Dust particle size class (microns)
Figure 8-3 Relative standard deviation of lead concentration measurements for
sieved dust by dust particle size class and age of home, as predicted by
the theoretical model
104
-------
The lead concentration was modeled as a function of time (days since the first
samples were taken on September 17,1993) using regression. The weighted analysis
showed that the trend in the lead concentration with time varied significantly
among bags of dust (p=0.016). Therefore, a separate regression was fit to the data
from each bag. The slopes generally increase slightly, with the greatest increase in
concentration over time in the dust with the largest particle size class. Changes in
the dust characteristics can occur if handling of the dust bags causes the lead particles
to separate slightly from the non-lead particles. For example, if the lead particles
tend to move on top of the non-lead particles, the lead concentration at the top of
the bag, where the dust is removed for the tests, may have a higher lead concentra-
tion than samples taken from the bottom of the bag. The equation for the lead
concentration was used to predict the geometric mean concentration of lead in the
dust used in the tests, which was used to calculate lead recovery.
Statistical tests were performed to determine if the measured lead concentration
differed between the initial grab samples and the subsequent samples in which the
dust sample was collected by passing the dust through a sieve, as in the recovery
tests. No differences in the lead concentration were found from the way the sample
was collected. Statistical tests were also performed to test for differences between
preparation or instrument batches. These terms were not significant. Tests for
homogeneity of variance were just significant at the 5% level, suggesting that the
regression weights did not completely model the variance of the measurements.
However, there was no apparent pattern in the variances which would indicate a
possible change to the model. The dust lead concentrations by dust particle size class
for dust from new and old homes are presented and discussed in Section 6.2.
8.2.3 Statistical Analysis of Gravimetric and Lead Analysis Data for
Samplers
The sampler tests involved depositing a known amount of dust over a one square
foot area of the substrate, using the sampler to recover the dust following standard
protocols for each sampler, and determining the weight of dust recovered
(gravimetric data) and amount of lead recovered. These measurements were used
to calculate the dust recovery, the lead recovery, and the ratio of the lead concentra-
tion in the dust collected by the sampler to the lead concentration in the dust
deposited on the substrate. Four samplers were studied, the CAPS cyclone, BRM,
Blue Nozzle samplers, and baby wipes. The wipes were not tested on upholstery,
carpet, or carpet with ground-in dust. Total dust recovery was also not measured for
the wipes.
A procedural error was made on one test (3-12). This test was repeated as test 3-25.
The data for test 3-12 would be considered outliers and were not used in the analysis.
105
-------
Gravimetric Data
The sampler dust recovery is the weight of dust collected by the sampler as a
percentage of the weight of dust deposited on the substrate. In the final model for
sampler dust recovery, the sampler type was a highly significant predictor of
sampler dust recovery (p < 0.0001).13 The combination of sampler type and dust
particle size was significant (p = 0.038). An analysis of the residuals showed that the
measurement variance was significantly related to the sampler type (p = 0.029) and,
therefore, a weighted analysis was used. Similar results were obtained when regres-
sion weights were not used and when transformations were considered. Whether
one observation (test 4-23) could be considered to be an outlier depended on the
weights used. No observations were excluded from the final analysis. The term for
an interaction of sampler and substrate was not close to statistically significant. The
dust recovery estimates for samplers are discussed in Section 6.3.1.
The half-width of the confidence intervals for the average dust recovery for the Blue
Nozzle, BRM, and CAPS cyclone sampling methods are 16%, 7%, and 5%, respec-
tively. Except for the Blue Nozzle sampler, these confidence intervals meet the data
quality objectives of 8% for these estimates. The confidence interval half-width for
dust recovery using a selected sampler and substrate, averaged across substrates,
ranges from 15% for the CAPS, to 20% for the BRM sampler, and 45% for the Blue
Nozzle sampler. Of these half-widths, only the value for the CAPS meets the data
quality objective of 15%. The standard deviation of the dust recovery measure-
ments for the Blue Nozzle, BRM, and CAPS cyclone samplers are 28%, 12%, and 9%,
respectively.
An additional data quality objective of ±30% for recoveries for a combination of
substrate, sampler and dust particle size (corresponding to individual measure-
ments, if made) is met by the CAPS cyclone and BRM samplers but not by the Blue
Nozzle sampler. Most of the data quality objectives were achieved for the BRM and
CAPS cyclone samplers. The dust recovery measurements for the Blue Nozzle
sampler were more variable than for the other samplers, in part due to the low
recovery. Therefore, the Blue Nozzle sampler did not achieve any of the data
quality objectives. Note that the number of sampler tests was determined from the
precision of the dust recovery measurements for vacuum cleaners. No correspond-
ing precision data from which to estimate sample size were available for the
samplers.
Lead Data
For one test (test 4-4, using the Blue Nozzle sampler on carpet), a comment on the
data sheet indicates that the reported lead measurement may be low by 8% due to
13p-values indicate the probability that differences as large as those observed could be due to chance
alone. Generally, p-values less than 0.05 indicate statistical significance.
106
-------
over-dilution. Thus, the data for this sample was used in the statistical analysis after
increasing the lead amount and lead concentration by 8%.
Measurements from two of the 51 sampler tests were identified as outliers and were
removed from the analysis. These tests are summarized in the following table. The
ESR values indicate the number of standard deviations separating the observation
from its expected value after excluding more extreme outliers. After removing
these outliers, there were 49 measurements for the analysis.
Dust Concentration
Test No. Sampler Substrate Lead recovery recovery ratio
3-13
4-1
CAPS
Wipe
Linoleum
Linoleum
2.36 (ESR
2.05 (ESR
= 3.65)
= 3.52)
0.87
NA
2.72
NA
The lead recovery is calculated as the weight of lead in the sample divided by the
weight of dust deposited on the substrate and by the lead concentration in the dust.
The statistically significant terms in the final model for the lead recovery are the
sampler (p < 0.0001), dust loading (p = 0.035), and dust particle size (p = 0.0033). The
significance level of the interaction of sampler and substrate depended on whether
weights were used. An analysis of the residuals suggested that the measurement
variance was related to the interaction of dust loading and nominal dust concentra-
tion, but the significance of variance differences also depended on the model used.
Because the pattern of the observed variances by dust loading and nominal dust
concentration was difficult to interpret based on physical considerations, the final
model did not use regression weights. In an analysis of transformed data, using
ln(Dust Recovery + 0.5), only the sampler type and dust particle size class were statis-
tically significant. It was decided not to include terms for the interaction of sampler
and substrate in the final model. With the outliers included in the final model, the
dust loading is not statistically significant. The measurement standard deviation is
21%. The sampler lead recovery estimates are presented in Section 6.3.2.
Concentration Ratio
The concentration ratio is the ratio of the lead concentration in the dust sample
collected by the sampler to the lead concentration in the dust deposited on the
surface.
Three of the 42 observations were removed for the analysis as outliers. The studen-
tized residuals were estimated using a model with the one factor, sampler type,
which was statistically significant in all regressions. The following table describes
the outliers which were removed from the analysis. The ESR values in the follow-
107
-------
ing table indicate the number of standard deviations separating the observation
from its expected value after excluding more extreme outliers.
Test
No.
3-13
4-2
4-7
Sampler
CAPS
Blue Nozzle
BRM
Substrate
Linoleum
Carpet
Carpet
Lead
recovery
2.36
0.11
1.31
Dust
recovery
0.87
0.04
0.66
Concentration
ratio
2.72 (ESR = 3.32)
2.44 (ESR = 3.76)
1.99 (ESR = 4.07)
The concentration ratio appears to depend on many marginally statistically signifi-
cant factors. In the final model, the significant factors for predicting the sampler
concentration ratio are: dust particle size (p < 0.0001), sampler (p = 0.0193), nominal
lead concentration (p = 0.0368), and substrate (p = 0.0476). The most significant
predictor of the concentration ratio is the dust size. The determination of the signif-
icance of other factors depends on the terms chosen for the model. No regression
weights were needed or used to equalize the measurement error. If the three
outliers discussed above are included in the model, and insignificant terms are
removed, the only significant predictor of the lead concentration ratio is the dust
particle size class. The standard deviation of one concentration measurement is
15%. The sampler lead concentration ratio estimates are presented in Section 6.3.2.
8.2.4 Statistical Analysis of Gravimetric and Lead Analysis Data for
Vacuum Cleaners
For the vacuum cleaner tests, the procedure of vacuuming the substrate for 40
seconds and measuring the associated weight change in the vacuum cleaner bag was
repeated seven times. Dust was deposited onto the substrate before the second,
third, and fourth vacuumings. After the seventh vacuuming, dust was shaken
from the vacuum cleaner bag and analyzed for lead. These measurements were
used to calculate the dust recovery, lead recovery, and ratio of the lead concentration
in the dust collected by the vacuum cleaner to the lead concentration in the dust
deposited on the substrate.
Gravimetric Data
Statistical analysis of the dust recovery (gravimetric) data for vacuum cleaners
requires a definition of dust recovery and how to correct for accumulated dust from
previous tests (carryover) and for carpet or upholstery fibers picked up in each test.
108
-------
The vacuum cleaner tests used the following steps, also described in Section 4:
(1) Vacuum the substrate for 40 seconds and measure the combined
weight of fibers and dust collected (vacuuming number 1).
(2) Deposit a measured amount of dust.
(3) Vacuum for 40 seconds and measure the combined weight of fibers and
dust collected (vacuuming number 2).
(4) Repeat steps (2) and (3) two more times (vacuuming number 3 and 4).
(5) Vacuum for 40 seconds and measure the combined weight of fibers and
dust collected (vacuuming number 5).
(6) Repeat step (5) two more times (vacuuming number 6 and 7).
(7) Shake dust from the vacuum cleaner bag into a container for lead
measurement.
(8) Weigh the dust removed from the vacuum cleaner bag.
(9) Determine the lead content in the dust removed from the vacuum
cleaner bag.
The analysis in Section 8.2.5, on the quantity of dust collected versus vacuuming
effort, suggests that the weight of dust collected in the first vacuuming, before any
dust is deposited, may not be a good estimate of the fibers and dust carryover affect-
ing the subsequent vacuuming. The weight of dust collected on the last few
vacuumings is often less than that collected on the first vacuuming. Assuming that
(1) most of the dust deposited is collected in the second through sixth vacuumings,
(2) the dust collected on the seventh vacuuming represents fibers and carryover, and
(3) the same amount of dust from fibers and carryover affects vacuumings two
through six, the following definition of dust recovery is used for the statistical
analysis:
Dust recovery = (weight2 + weights + weight4 + weights + weight6 - 5 x weight?)
(depositl + deposit2 + deposits)
where: weightN = the change in weight of the vacuum bag in the Nth vacuuming,
and depositN = the weight of dust deposited in the Nth deposit.
The statistically significant terms in the final model are the substrate (p < 0.0001) and
the vacuum cleaner (p < 0.0001). Based on an analysis of the residuals, the
measurement variance decreased as the predicted dust recovery increased (with
minimum variance for dust recoveries close to 100%). The statistical conclusions
109
-------
are unchanged and the estimates are similar when no regression weights are used.
The results of the model are presented in Section 6.4.1. There was no evidence of
significant serial correlation.
The pooled standard deviation of one dust recovery measurement was 9.2%, much
better than required to meet the data quality objectives for the individual measure-
ments and the averages for vacuum cleaner recovery. The half-width of the confi-
dence intervals for dust recovery on combinations of substrate and vacuum cleaner
(based on a model with this interaction tern included) ranges from 4% to 18%,
compared to the data quality objective of 15%. Overall, the dust recovery measure-
ments for the vacuum cleaner tests achieved or nearly achieved the associated data
quality objectives.
To determine the effect of alternate definitions of dust recovery, the final model for
dust recovery (as defined above) was also fit using the following two alternate
definitions for dust recovery:
HDR= (weightl+weight2+weight3+weight4+weight5+weight6+weight7-7 x Fibers)
(depositl + deposit2 + deposits)
LDR = (weight2 - weight?)
(depositl)
where Fibers = the estimated 40-second uptake of fibers based on the dust precondi-
tioning data (see Appendix B).
HDR is the total weight of dust collected, after correcting for fibers, divided by the
weight of dust deposited. This represents an upper estimate of dust recovery corre-
sponding to extensive vacuuming, under the assumption that the dust in the carpet
has reached an equilibrium such that, on average, the dust carryover from the
previous test is the same as the dust carryover to the next test.
LDR is the recovery for the first deposit of dust, corrected (approximately) for both
fibers and dust carryover. Since the recovery for the first deposit of dust is generally
lower than for later deposits, possibly due to less carryover, this represents a lower
estimate of dust recovery. It corresponds to recovery based on a minimal amount of
vacuuming (40 seconds).
The average HDR and LDR estimates for substrates are shown in Figure 6-8. The
conclusions about which factors are significant predictors of vacuum cleaner dust
recovery are similar when modeling the dust recovery, HDR, or LDR except that
differences among vacuum cleaner are not statistically significant when modeling
the HDR.
110
-------
Lead Data
The lead recovery is the quantity of lead collected in the vacuum cleaner bag as a
percentage of the lead deposited on the substrate. Because lead analysis of the entire
vacuum cleaner bag and dust contained in it would be very difficult and would
require a correction for lead in the bag itself (making the measurement imprecise),
only that portion of the dust which was removed from the bag was analyzed.
Assuming that the dust removed from the bag is representative of the dust in the
bag, the dust recovery can be calculated as:
Lead recovery = weight of dust collected * lead concentration in dust removed from the bag
weight of dust deposited * dust lead concentration
The experimental procedures, with only one lead analysis per test, did not provide
information to correct for possible lead carryover from test to test. Therefore, statis-
tical analysis was used to identify possible lead carryover and, if necessary, to correct
the estimates for carryover.
The values shown in the following table were removed as outliers and were very
different than comparable measurements (ESR > 9.0 for both measurements). These
outliers affect only the lead recovery and concentration ratio analyses. One possible
explanation for the outliers, which cannot be checked, is that the lab technicians
chose the wrong dust bag by mistake (either using dust from older homes rather
than newer homes or using dust of a different size). MRI has checked these values
and finds no known explanation for the unusual results.
Test Vacuum Nominal lead Dust Concentration Lead
number cleaner Substrate concentration particle Team ratio recovery
size
1027(1-12)
1079(1-25)
C
B
Linoleum
Wood
Low
Low
106-150
212-250
1
1
9.8
5.1
9.4
4.8
As discussed below, the identification of the final model and the factors which affect
vacuum cleaner lead recovery depend on which other factors are in the model. The
only significant factor in the final model for vacuum cleaner lead recovery was the
choice of vacuum cleaner (p = 0.043). Differences in measurement variance among
tests with dust from older and newer homes were just statistically significant at the
0.05 level (p = 0.04). However, the identification of which factors affect the
measurement variance depends on the model fit to the lead recovery data. Because
the statistical results were insensitive to the use of weights in the model, and the
variance differences were only marginally significant, no weights were used in the
final model.
The average lead recovery across all tests was 103%, greater than the theoretical
maximum of 100%. The difference between the average of 103% and the theoretical
111
-------
maximum is not statistically significant, so the difference may be due to random
uncontrolled factors. However, given that the dust recovery averages 85% and thus
not all of the dust is collected, it is reasonable to assume that not all of the lead is
collected and that the true recovery is less than 100%. If this is true, the difference
between the estimated average recovery of 103% and the true vacuum cleaner lead
recovery may be due to factors other than chance.
Several possible explanations have been put forth to explain the high lead recovery,
including lead carryover between tests, higher vacuum cleaner recovery of leaded
dust than non-leaded dust, lead release from substrate samples (particularly carpets),
and differential recovery of leaded and non-leaded dust from the vacuum cleaner
bags. Of these explanations, the differential recovery of leaded and non-leaded dust
from the vacuum cleaner bags provides the most likely explanation. On average,
only 26% of the dust in the vacuum cleaner bags was removed for lead analysis,
leaving much of the dust in the bags. Leaded particles may be more easily shaken
from the bag.
Regression was used to test if lead carry-over, or differential removal of dust from
the vacuum cleaner bags, might explain the high lead recoveries. The weight of
dust removed from the bag as a percentage of weight of dust collected by the bag,
called the dust removal, was added to the model to determine if dust removal was
related to lead recovery. Terms were also added to assess serial correlation related to
successive tests within substrates, and to successive tests using the same vacuum
cleaner, and to assess trends over time.
None of these terms were statistically significant. However, a separate analysis of
the dust removal data showed that the dust removal depended on the dust particle
size. This finding may also be relevant to the lead recovery since the relationships
between dust particle size and dust lead concentration were different between dust
from newer homes and from older homes (see Section 6.2). It was therefore decided
to add a term for interaction between the nominal lead concentration (dust from
older or newer homes) and dust removal. This interaction term was highly signifi-
cant (p=0.0021). In addition, an interaction term between substrate and vacuum
cleaner was also significant (p = 0.0228). The relationship between dust recovery and
the vacuum cleaner and substrate tested is shown in Figure 6-12. The standard
deviation of a single vacuum cleaner lead recovery measurement is 27%. The
vacuum cleaner lead recovery estimates are discussed in Section 6.4.2.
If the leaded dust tends to be shaken out of the vacuum cleaner bag easier than the
non-leaded dust, then the initial dust shaken from the bag will have a higher lead
concentration than the dust remaining in the bag. Additional efforts to remove
dust from the bag may remove dust with a lower lead concentration than the dust
initially removed. In this case, one would expect the lead concentration in the
removed dust to decrease with increasing effort to remove the dust from the
vacuum cleaner bag. If (1) the lead recovery based on the lead concentration in the
initial dust removed from the bag was 103%, (2) the true lead recovery was 85%, and
112
-------
(3) the relationship between dust removal and lead recovery was linear such that
the lead recovery was estimated to be 85% when the dust removal was 100%, then
the expected slope relating the lead recovery to dust recovery would be -0.18
(ie., (103-85)/100).
By this simple argument, the expected parameter estimate for the dust removal
would be negative and roughly -0.18. While other more complex models might
suggest other values, this value provides a guide to evaluate the regression results.
For dust from newer homes with low dust lead concentration, the slope parameter
is -1.14. This is in the expected direction and somewhat larger in magnitude than
expected, although its confidence interval is large, from -0.21 to -2.07. For dust from
older homes with high dust lead concentration, the slope parameter is 0.77. This is
not in the expected direction, but its confidence interval is also large, from -0.06 to
1.60. These parameters are difficult to interpret because of the large differences
between the parameters for dust from older and newer homes and the large magni-
tude of the estimated slopes.
A model for the dust removal indicates that many factors affect it, including the
following:
• Dust loading (p < 0.0001) - - more dust, as a percentage of the dust in the
bag, was removed for tests with low dust loading than for tests with
high dust loading. It was more difficult to get enough dust for lead
analysis from the tests with low dust loading and, therefore, perhaps
more effort was used.
• Nominal dust lead concentration (p = 0.0039) - - more dust was
removed in tests using dust from older homes (high lead concentra-
tion) than newer homes. This result is consistent with the assumption
that the leaded dust is easier to remove than the non-leaded dust
• The combination of substrate and vacuum cleaner (p = 0.0471) - - this
may in part reflect the varying difficulty in removing dust from the
differently constructed and shaped vacuum cleaner bags.
• Dust particle size class (p < 0.0001) - - the dust removal efficiency
increased as the dust particle size increased.
• The combination of dust loading on the substrate and dust particle size
class (p = 0.0001) - - the pattern of dust removal from the vacuum
cleaner bag as a function of dust loading and dust particle size is diffi-
cult to interpret. For each dust particle size class, high dust removal for
tests with high dust loading is associated with low dust removal for
tests with low dust loading. Similarly, low dust removal for tests with
113
-------
high dust loading is associated with high dust removal for tests with
low dust loading.
• Operator (p = 0.0034) - - one vacuum cleaner operator removed, on
average, 70% more dust from the bags than the other operator.
If the dust recovery depends on the dust removal and the dust removal depends on
the factors above, then the observed dust recovery may appear to depend on the
factors above through the dust removal. The regressions which include dust
removal as an independent variable provide a correction for differential dust
removal. However, the results are difficult to interpret and do not explain the high
lead recovery estimates, particularly for the dust with high lead concentration,
which is of most concern.
On the average, 26% of the lead deposited on the substrate is removed from the
vacuum cleaner bag.14 The remaining 74% of the deposited lead is in (1) the
vacuum cleaner bag, (2) the substrate, (3) other parts of the vacuum cleaner, or (4)
other areas of the test room or expelled into the air. The estimates of lead recovery
are uncertain due to uncertainty in the dust removal. At a minimum however, it is
possible to say that average lead recovery is greater than 26%, based on the worst case
assumption that all of the leaded dust is removed for lead analysis, leaving only
non-leaded dust in the bag. Because of the difficulty in interpreting the coefficients
in the model for vacuum cleaner lead recovery (which included the dust removal
interaction with nominal lead concentration), tentative conclusions are as follows:
• Vacuum cleaner lead recovery may depend on the combination of
vacuum cleaner and substrate tested.
• Measured vacuum cleaner recoveries average about 103%, but, the
measurements are difficult to interpret because of the methods
employed and conflicting statistical results.
Concentration Ratio
The concentration ratio is the ratio of the lead concentration in the dust removed
from the vacuum cleaner bag to the lead concentration in the dust applied. Two
outliers were removed for the analysis, the same two outliers removed from the
lead recovery analysis. In the final model, there were no significant predictors of the
lead concentration ratio. The mean and standard deviation of the concentration
14The average of both (1) the amount of dust removed from the bag for analysis as a percentage of the
dust in the bag and (2) the amount of lead removed from the bag as a percentage of the amount of lead
deposited on the substrate (after removing two outliers) are the same value, 26%, when rounded to two
significant figures.
114
-------
ratio measurements are 112% and 27%, respectively. The vacuum cleaner lead
concentration ratio results are presented in Section 6.4.2.
8.2.5 Statistical Analysis of Vacuuming Effort Data
Data were obtained on the quantity of dust collected by the vacuum (Le. the increase
in weight of the vacuum cleaner bag) before depositing dust (vacuuming 1), after
each of three dust deposits (vacuumings 2, 3, and 4), and in three subsequent
vacuumings (vacuumings 5, 6, and 7). All vacuumings were for 40 seconds. This
data can be used to determine whether the dust deposited on the substrate is
collected in the first 40-second vacuuming or whether additional vacuumings are
required to remove the dust.
For the statistical analysis, the dust recoveries for vacuumings 2, 3, and 4 were
defined as the ratio of the weight of dust collected in the vacuum bag to the weight
of dust deposited just prior to the vacuuming. If there is no dust carryover, the
average dust recovery for the three deposits are the same as the overall dust
recovery. The weights of all dust deposits with the same nominal dust loading are
quite similar.
For the statistical analysis, the dust recoveries for vacuumings 1, 5, 6, and 7 were
defined as the ratio of the weight of dust collected in the vacuum bag to the average
weight of dust deposited in the three deposits. By scaling the weight increases by the
average dust deposit, the weight of dust collected on vacuumings 1, 5, 6, and 7 are
put onto the same scale as the dust recovery for vacuumings 2, 3, and 4. Not
counting the effect of fibers, the sum of the dust recovery, as defined here, over all
vacuumings divided by weight of dust deposited in the three deposits, estimates the
overall dust recovery.
The statistical analysis was used to describe the pattern of dust recovery versus
vacuuming effort for each substrate only after correcting for factors which affect the
dust recovery. The full analysis of dust recovery is presented in Section 8.2.4.
Because the descriptive nature of the analysis, the model ignored nested effects for
the multiple measurements within a test and possible serial correlation of
measurements within a test.
Because the measurement variance appeared to vary among observations and to
affect the selection of the model, preliminary regression weights were used to
identify a preliminary model from which the final regression weights were deter-
mined, using the following steps:
(1) Remove apparent outliers and get a preliminary fit to the data.
(2) Analyze the residuals to determine preliminary regression weights.
115
-------
(3) Model the data using a weighted analysis to identify a preliminary final
model.
(4) Determine the final regression weights.
(5) Based on the final regression weights, formally identify and remove
outliers and fit the final model. One term was removed from the
preliminary final model to obtain the final model.
The model for the final regression weights predicts the measurement standard
deviation as a function of dust loading (dust recovery is more variable based on
smaller dust loadings), nominal lead concentration, dust particle size class, substrate
(measurements on carpet are more variable than on other substrates), and the
number of the vacuuming within the test (vacuumings 2, 3, and 4, with higher
recoveries, have more variable measurements).
Four observations were identified as outliers and were removed from the analysis.
The test number and vacuuming number of the outliers are: vacuuming 1 of test
2033 on linoleum, vacuuming 1 of test 2020 on wood, vacuuming 7 of test 1013 on
carpet, and vacuuming 1 of test 2006 on carpet. All of these outliers were more than
five standard deviations from their estimated mean. The weighted residuals were
used to identify outliers in the final model. Inclusion or exclusion of the outliers
made very little difference in the estimates.
The final model for dust recovery versus vacuuming effort had terms for interac-
tions between the vacuuming number and both dust loading and substrate, the
interaction between vacuum cleaner and substrate, and a term for dust particle size
class. The interaction between substrate and vacuuming number accounted for
most of the prediction sum of squares. The predicted least square means for the
substrate and vacuuming number interaction are shown in Figure 6-13 The impli-
cations of these results for the vacuuming effort are discussed in Section 6.4.2.
The amount of material (dust and fibers) collected in the last three vacuumings is
often less than that collected in the first vacuuming before any dust is deposited.
Thus, the weight of material collected in the first vacuuming appears to provide a
poor estimate of the effect of fibers and dust on subsequent vacuumings. Because of
this, it was decided to use the weight of dust collected in the last vacuuming to
correct for fibers and dust carryover in the analysis of dust recovery, as discussed in
Section 8.2.4.
116
-------
8.2.6 Statistical Analysis of Sampling and Measurement Precision
Lead Measurement Precision
The measurement error (or variation) is the difference between the observed
measurement and the true value being measured. It can be described mathemati-
cally as the sum of several independent sources of error, called components of
variance. The variance of the measurement error is the sum of the variance of the
components contributing to the error. The error components for measurements of
lead are shown in Table 8-2.
Some of the samples were analyzed using the GFAA analysis and others using the
ICP analysis. Because all GFAA samples were analyzed in the same batch, it is not
possible to estimate the variance of the preparation batch and instrument batch
components for the GFAA method. The results for the ICP analyses and the GFAA
analyses are presented separately.
Variance Components for ICP Samples
Because all the samples from one preparation batch were generally analyzed in the
same instrument batch, the measurement errors for the preparation and instrument
batch components cannot be estimated independently. Similarly, sample variation
within a preparation batch and within an instrument batch cannot be estimated
independently. The samples sent for lead analysis can be divided into the different
types shown in Table 8-3, Although these samples can be used to estimate different
components, estimates of individual components are difficult to determine and
compare because different types of samples may have slightly different factors
contributing to each component. Table 8-3 shows the components which can be
estimated from measurements on different types of samples.
The analysis of variance components assumed that (1) the instrument batch and
preparation batch components were confounded and could not be estimated sepa-
rately and (2) the variance of the components depended on the concentration being
measured but were similar for samples with the same lead concentration. The
differences among batches are expected to affect all samples with similar lead
concentration in a similar way, regardless of the sample type. This relationship is
maintained in the estimates when all samples with similar concentration are
analyzed together.
The model fit to the data had a term for sample type and a random effect term for
the instrument batch. A separate analysis was performed for each group of samples
with the similar lead concentration. In the analysis of the samples with zero lead
concentration, one apparent outlier was removed from the interference check
standards.
117
-------
Table 8-2 Variance components for lead measurements
Variance component
Source
Test conditions
Variation in the lead recovery among
replications of the test conditions
Sampling
Variation among the possible samples of
dust, only one of which was collected for
analysis.
Preparation batch
Variation in the procedures and
reagents among sample preparation
batches, only one of which was used to
prepare the digestate for analysis.
Preparation
Within-preparation batch variation in
the lead concentration in the digestate
among the possible digestate beakers.
Instrument batch
Variation in the instrument condition
and calibration among instrument
batches.
Measurement
Variation in the measured lead concen-
tration due to variation (assumed to be
random) in the instrument's measure-
ment process.
118
-------
Table 8-3 ICP variance components which can be estimated from each type of
sample
Sample type
Instrument
calibration blanks
Interference check
Method blank
Field blank
Detection limit
Interference check
standards with lead
Spiked samples
Continuing
calibration
verification
Independent
calibration
verification
High-calibration
standard
Standard reference
material
Lead
(Hg/mL)
0
0
0
0
.1
1
4
10
10
20
Depends
on
dilution
Components which can be
measured
[Instrument batch]
[measurement]
[Instrument batch]
[measurement]
[Instrument batch]
[measurement]
[Instrument batch +
preparation batch]
[preparation +
measurement + sample]
[Instrument batch]
[measurement]
[Instrument batch]
[measurement]
[Instrument batch]
[measurement]
[Instrument batch]
[measurement]
[Instrument batch +
measurement]
[Instrument batch +
measurement]
[Instrument batch +
preparation batch]
[preparation +
measurement + sample]
Comments
Low-calibration
standard
Interference check
standards with no
lead
Blank samples
prepared for
digestion
Blank prepared
during the
experiment and
sent for analysis
Standard with
concentration near
the detection limit
Interference check
standards with lead
Spiked samples
with wipes
excluded from the
analysis
Mid-calibration
standard
Independently
prepared mid-
calibration
standard
High-calibration
standard
Variance is affected
by interferences
119
-------
The analysis output provided estimates of the average measurement for each batch
and the variance of the measurements within a batch and the variance of the batch
averages. The sum of the within batch and batch average variance components is
the variance of one independent measurement associated with the laboratory
analysis. For discussion and presentation, these variances are expressed as standard
deviations (the square root of the variance) which have the same units as the
measurements.
Figure 8-4 shows a plot of the standard deviation of the within batch measurement
component, the instrument batch component, and the sum of these two variance
components. The measurement component could not be estimated for the high
calibration standards because there was only one such measurement per instrument
batch.
Both the standard deviation of the within batch measurement component and the
standard deviation of the batch component increase roughly linearly with the lead
concentration in the sample. This linear relationship between standard deviation
and concentration is typical for many laboratory concentration measurements and
generally applies to all of the variance components. The within instrument batch
variance is consistently less than the between batch component of variance. The
variance for the spiked sample measurements is greater than predicted by the trend
for the other samples. Only the spiked, method blank, and field blank samples
passed through the preparation step. The method blank and field blank samples
have zero lead, assuming no contamination. For these samples, the components
associated with the preparation step would be small because the lead concentration
was small. The method and field blank samples exhibited no greater variance than
the calibration blank and interference check samples that also had no lead. On the
other hand, the lead concentration in the spiked samples may have been affected by
the preparation step. The increased variance for the spiked samples most likely
represents the contribution of the preparation and preparation batch components.
A simple regression line was fit to predict the standard deviation of the combined
between batch and measurement components. The prediction line is shown as a
dotted line in Figure 8-4. The method spike samples were not used to fit the regres-
sion. However, the difference between the variance for the method spike samples
and the predicted measurement error using the regression line was used to estimate
the standard deviation of the error associated with the preparation step. This
estimated coefficient of variation was 3.76%. The predicted standard deviation of a
single independent measurement, represented by the regression line, was used to
estimate the coefficient of variation of the lead measurements as a function of
instrument response.
120
-------
0.6 -i
*g 0.5
e
o
1
•3
en
rt
tn
(0
§
I
0.4 -
0.3 -
to
g 0.2
§4 0.1
I
u
V
0 -
-0.1
•2
(0
I
CO
1
E
•§
(0
u
•a
jg , Predicted
*
Batch*
measurement
vy
Instrument batch
Measurement
.x
(0
^H
oa
•4 0 4 8 12 16 20
Lead concentration in the standard or spike sample (ug/mL)
i
24
Figure 8-4 Standard deviation of the variance components as a function of lead
concentration in the instrument sample after any dilution
121
-------
Figure 8-5 shows the predicted coefficient of variation associated with lead
measurements as a function of the instrument response and a histogram of the
observed instrument response for the dust samples in the study. For higher
instrument responses, the coefficient of variation for lead measurement is about 2%
and is roughly independent of the instrument response. For lower instrument
responses, the coefficient of variation can be considerably greater than 2% and can
approach 20% for samples with instrument response at 0.1 jig/mL (the response
below which GFAA analysis was used). Many of the samples had associated
instrument responses below 1 ug/mL with corresponding coefficient of variations
above 4%.
The measurement bias (average measured - known lead concentration) for each
batch as a function of the known lead concentration is shown in Figure 8-6. For
samples with higher instrument response, the bias associated with each instrument
batch is roughly constant. For samples with lower instrument responses, the bias
varies considerably among batches and for different instrument responses. The
relatively high bias for larger instrument responses for instrument batch E12023B is
reflected in the instrument drift shown in Figure 9-6. The differences among the
batch averages are statistically significant (p<.01 for all but the high calibration
standards).
Figure 8-6 suggests that, for most samples, the bias in the lead measurements is less
than 10% compared to the calibration standards. For the analysis of lead recovery, a
correction for the bias in the lead measurements may be possible by including a term
for an interaction between instrument response and instrument batch. However,
due to the relatively large expected magnitude of the sampling error, such a correc-
tion may not be useful.
Variance Components for GFAA Samples
The GFAA QC and calibration samples provide some information on the magni-
tude of the variance components for lead concentration based on the GFAA
method. Since all GFAA samples were analyzed in the same batch, only the within
batch variance can be estimated. As with the ICP measurements, the measurement
variance appears to increase with instrument response (after removing one outlier
from the calibration blanks). The coefficient of variation of the continuing calibra-
tion standards is 2.2%, suggesting that the measurement variation for the GFAA
methods, as measured by the coefficient of variation, is similar to that for the ICP
method.
122
-------
20%
18% -
16% -
0
Predicted cv
Vacuum
cleaner
Sampler
1 Sieved dust
2 4 6 8 10
Instrument response (ug/mL)
100
90
80
•a
01
I
70 T&
60 jj
S
s s
50 i? B-
40
20
10
0
12
v
"3
6
u
30
Ol
Figure 8-5 Predicted coefficient of variation (cv) associated with lead
measurement as a function of the instrument response and histogram
of the observed instrument response for the dust samples in the study
123
-------
a
30% -i
25%
20% -
15% '
g 10% -
a,
-------
Variance Component Associated with Dust Collection and Sampling
The coefficient of variation of one lead recovery measurement on a dust sample can
be estimated from the measurements on standard reference materials and on sieved
dust. The coefficients of variation of measurements on sieved dust are discussed in
Section 8.2.2 and range from about 5% to 20%, except on dust in the largest dust size
class. The coefficients of variation of measurements on the standard reference
materials are discussed below.
Due to the small number of standard reference material samples, separate estimates
for the variance components were not calculated. The coefficient of variation of the
recoveries provides a measure of the combined effect of the variance components.
For the samples analyzed using ICP, for SRM 1646 (with the lower lead concentra-
tion and particular problems with interferences), the coefficient of variation of the
recoveries is 10.5%. For SRM 2704, the coefficient of variation is 8.6%.
For the purpose of estimating the coefficient of variation associated with the dust
sampling component, the following assumptions are used: (1) the coefficient of
variation of lead measurement is 3% (see Figure 8-5), (2) the coefficient of variation
due to batch differences is 3.8%, and (3) the coefficient of variation of a lead
measurement in a sample of dust is 11% (compared to estimates of 8.6%, 10.5%, and
5% to 20%). Using these assumptions, the coefficient of variation of the sampling
component is:
V.112-0.032 - .0382 = .099 = 9.9%
Thus, most of the variation in the lead measurement on a sample of dust is associ-
ated with the sampling of the dust.
Variance Component Associated with Test Conditions
The coefficient of variation of lead recovery measurements between tests conducted
under the same conditions can be estimated from the error variance from the statis-
tical models for lead recovery. The pooled standard deviation of the sampler lead
recovery measurements is 17%. The average lead recovery for the BRM, CAPS
cyclone, and wipe samplers is 69%, giving a coefficient of variation of 25%. For the
vacuum cleaner lead recovery estimates, the coefficient of variation of the lead
recovery measurements was about 26% (standard deviation of 27% divided by a
mean of 103%). Because this is much greater than the roughly 9.9% associated with
the dust sampling, most of the variation in the measurements is due to differences
between tests.
125
-------
126
-------
9 QUALITY ASSURANCE
An independent evaluation of the sample collection and analysis activities on this
work assignment was performed by the MRI Quality Assurance Officer (QAO) for
the program. The evaluation included a system audit, performance audit, data
audit, and data assessment. An explanation of each type of audit or review is given
below, along with a discussion of the audit results. Also, Westat audited the data
entry and statistical analysis procedures, as discussed in this section.
9.1 System Audit
The system audit performed by MRI for this work assignment was a qualitative
examination of the vacuuming and analytical systems. Since the activities were
significantly different for each system, a separate inspection was performed on each
system. The results of the system audits are given below.
9.1.1 Vacuuming Task
A system audit on the vacuuming task was conducted on August 4, 1993. The areas
inspected during this audit were the facility, equipment, and documentation. The
facility was found to be adequate for the task. The equipment necessary for the
activities was either in the facility or on order. No systematic problems were
observed with the facility or the equipment.
Vacuum and wipe protocols were followed as per the QAPjP and no discrepancies or
problems were found. Forms and laboratory notebooks were used for the documen-
tation of work on this work assignment.
9.1.2 Analytical Task
The system audit of the analytical task was conducted on August 13,1993. The areas
inspected during this audit were personnel qualifications, sample control, sample
preparation techniques (on samples similar to those being analyzed for this work
assignment), and Standard Operating Procedures. No systematic problems were
observed during this audit
9.2 Performance Audits
For the analytical activities, two Performance Evaluation Samples (PESs) were
prepared for each analytical preparation batch. The PESs were prepared by the
project sample custodian using National Institute of Standards and Technology
(MIST) Standard Reference Material (SRM). The two SRMs used for the dust PES
127
-------
material were Estuarine Sediment (SRM 1646) with a lead level of 28.2 |ig/g and
Buffalo River Sediment (SRM 2704) with a lead level of 161
9.2.1 Performance Evaluation Sample Results
The individual results for the PESs used in the sample batches associated with this
work assignment are given in Table 9-1. As noted in the table, the recovery of the
PES did not meet the original DQOs for SRM 1646 in sample preparation batch No.
502. That is the recovery was below the lower control limit of 75%. This situation
was investigated before proceeding with the analysis of the remaining sample
batches.
The standard reference materials, NIST SRM 1646 and 2704, have been analyzed by
ICP as blind PESs on several program tasks, and the recovery results have been
control charted since late June 1991. A two year history of SRM 1646 recovery results
is shown in Figure 9-1 (ICP sequence Nos. 1 through 62, covering the period from
June 27, 1991, to September 8, 1993). In this figure, the results pertaining to the
current work assignment are shown as full bullets and all other results are shown as
hollow bullets. Figure 9-1 includes the results from two batches analyzed after batch
No. 502 which do not pertain to this work assignment. Each data point represents a
single recovery result.
Figure 9-1 shows an obvious change in the recovery pattern of SRM 1646 following
ICP sequence No. 45. Prior to that date (August 6, 1992), difficulties were encoun-
tered in obtaining acceptable recoveries for SRM 1646. This material has a low lead
concentration (28.2 M-g/g) combined with high levels of other metals such as iron
(iron:lead ratio exceeds 1,000:1), which causes interferences and necessitates further
dilution of the samples. To correct for non-lead interferences, the analyst would
perform serial dilutions of the digests. This in turn would result in lead levels for
the blind SRM 1646 that were either below or within a few multiples of the instru-
mental detection limit, thus producing variable and sporadically poor recoveries.
Starting with ICP sequence No. 46 (May 3,1993), action was taken to correct for non-
lead interferences. This was achieved by (1) establishing a consistent serial dilution
pattern of samples for both high levels of lead and high levels of spectral interfer-
ences and (2) by raising the interference check standard from 200 to 250 ng/mL. This
resulted in more consistent but lower recoveries for lead in SRM 1646 PESs, as
reflected in Figure 9-2. This figure shows recovery results for ICP sequence Nos. 46
through 108, covering the period May 3,1993, through January 28,1994.
128
-------
Table 9-1 Percent recoveries of blind performance evaluation samples
ICP
Sequence
No.
56
60
85
86
87
95
96
NA3
Preparation
Batch No.
501
502
503
504
505
506
507
508
SRM 16461
Concentration (M-g/g)
Certified
28.2
28.2
28.2
28.2
28.2
28.2
28.2
28.2
Found
22.07
18.94
20.12
23.30
24.10
23.62
22.07
21.09
Recovery
(%)
78.28
67.153
71.36
82.63
85.46
83.77
78.27
74.78
SRM 27042
Concentration
Certified
161
161
161
161
161
161
161
161
Found
152.7
140.5
148.1
150.1
152.0
153.4
147.6
152.8
Recovery
(%)
94.87
87.27
91.96
93.25
94.39
95.26
91.65
94.91
1 SRM 1646 accuracy DQOs for batch Nos. 501 and 502: target value is 100%±20%
(warning limits) and ±25% (control limits)
2 SRM 2704 accuracy DQOs for all batches: target value is 100% ±20 % (warning limits)
and ±25% (control limits)
3 Preparation batch No. 508 was analyzed using GFAA spectroscopy instead of ICP
a % recovery does not meet DQO of 75%
129
-------
I"
o*
B>
•1
¥
§
8
1
HISTORICAL RECOVERY CONTROL CHART
180
150 -
120 -
90 -
60 -
30
SRM—1 640
28.2
= Data points for WA-11
I i i i i I i i i i I i i i i I i i i i I i i t i I
0 5 10 15 20 25 30 35 40 45 50 55 60
ICP Sequence No.
-------
OQ
sr
8
I
I
r
HISTORICAL RECOVERY CONTROL CHART
SRM-1646 28.2 jig/g
180
150
120
90
60
30
UCL.
LTTL
tCL
= Data points for WA-11
. ... I .... I .... I . I .. I , I ., I ... I I , ... I .... I
45 50 55 60
65 70 75 80 85 90 95 100 105
ICP Sequence No.
-------
In the investigation of the low recovery for SRM 1646 in batch 502 (67.15%), it was
recognized that this change toward consistent, but lower, recoveries was obtained
after ICP sequence No. 46. It also was recognized that the original control limits of
100%±25%, which were arbitrarily selected, needed to be statistically determined.
Therefore, revised control limits were statistically determined using the 16 results
available at the time (beginning with ICP sequence No. 46 through ICP sequence No.
62) as given in Table 9-2. The mean value for these data was a recovery of 84% with
a standard deviation of 10. Based on this information, the warning limits and
control limits were statistically specified as:
Control limits = Mean ± 3 standard deviations = 84 ± 30%
Warning limits = Mean ± 1.96 standard deviations = 84 ± 20%
These statistically-based control limits were approved for this task and a QAPjP
amendment record was prepared, dated October 29,1993. After receiving approval,
work resumed on the analysis of subsequent sample batches, and the results were all
within the revised control limits. Moreover, these results, along with results for
SRM 1646 on other sample batches not associated with this work assignment and
included in Figure 9-2, support the validity of the revised control limits and their
use for the intended purpose of identifying problems in sample analysis.
Based on the information discussed above and shown in Figure 9-2, it is clear that
the results for SRM 1646 on this work assignment were not a problem and were in
fact "in control," including the result for batch No. 502.
The historical control chart for SRM 2704 recovery results is shown in Figure 9-3 for
completeness. This chart covers the period from June 27, 1991, to January 28, 1994,
and reflects the fact that all results are within the control limits. Since the blind
performance samples consisted of both SRM 1646 and SRM 2704 and both SRMs
were prepared and analyzed during the same period, the results obtained from SRM
2704 show an analytical system that is in control with no systematic errors. These
results also show that the problems with SRM 1646 were in the nature of the SRM
rather than in the analytical system.
9.3 Data Audit
The data audit is a qualitative and quantitative evaluation of the documentation
and procedures associated with the measurements to verify that the resulting data
are of known and acceptable quality.
For this work assignment, two types of data were submitted for audit. The first type
was primarily weight data obtained during the dust application and vacuuming
activities. The second type was analytical data used to evaluate the lead concentra-
tions in the sieved dust as applied to substrates and the dust recovered from the
vacuum cleaners and samplers.
132
-------
Table 9-2 SRM 1646: percent recovery of 16 blind control samples
ICP Sequence
No.a
46
47
48
50
51
52
53
54
55
56
57
58
59
60
61
62
Analysis Date
05/03/93
05/03/93
07/22/93
07/28/93
08/03/93
08/14/93
08/14/93
08/16/93
08/16/93
08/23/93
08/28/93
08/28/93
09/01/93
09/02/93
09/08/93
09/08/93
Preparation
Batch No.
XDV
XDT
601
804
809
801
802
803
807
501b
814
821
817
502b
811
812
Recovery
(%)
76.24
91.00
93.47
71.71
78.04
80.47
90.46
76.53
72.21
78.28
87.02
80.94
97.71
67.15
97.88
96.96
Number of samples = 16
Mean recovery = 84
Standard deviation = 10
Coefficient of variation (%) = 12
a ICP sequence No. 49 results were not available at the time
corrective action was taken
b Batch associated with present work assignment
133
-------
I
I
8
C/l
I
HISTORICAL RECOVERY CONTROL CHART
SRM-2704 161 >ig/g
140
130 -
120
110
100
90 ~
80 ~
70 ~
60
Data points for WA.-11
i .. i , i i ... i .... i .
10 20 30 40 50 60 70
ICP Sequence No.
80
90 100 110
-------
9.3.1 Vacuum Weight Data
Five data audits were conducted on dust weight data. These audits evaluated
approximately 25% of the data collection and processing systems for the vacuum
cleaner emission study, the carpet pre-conditioning study, and all the vacuum
cleaner and sampler recovery data sets. The balances used for the weighing activi-
ties were serviced within the past year. Each balance was verified to be calibrated
against weights traceable to NIST Standards and was operating properly. During
each weighing session, check weights were used to verify that the response of the
balance was accurate before the weighing of the samples (i.e., each day). No system-
atic errors were detected in the data audited, and any random errors found in the
data were corrected prior to release of the data.
Of the 1,598 test weight results, a total of 595 (37.2% weights randomly selected
within batches) were audited. Two random errors (one sample misidentification
and on arithmetic error) were found and corrected. The estimated error rate, before
correction, was 0.34% with lower and upper 95% confidence limits of 0.03% and
1.21%, respectively, based on a Poisson distribution. Adjusting for the two correc-
tions made and the fact that one batch (360 test weights) was 100% checked and
error-free, the average error rate for dust weight test results is estimated at 0.14%
with a 95% confidence interval of 0% to 0.81%.
9.3.2 Analytical Data
The data audits were conducted on approximately 20% of the analytical data by
personnel assigned to the QA Unit under the supervision of the QAO. The
analytical data generated for this work assignment were audited to assure quality
and reliability. The quality of the analytical data was evaluated using blind PESs
prepared from NIST standard reference materials and internal quality control
samples prepared by the analyst. The data obtained from these samples were
evaluated against the DQOs and the measurement objectives for the analytical
process as presented in the QAPjP and its appendices. Audits of the analytical data
showed no systematic errors in the data measurement process. These data were
found to be in compliance with the DQOs and measurement objectives, with the
exception of the recovery (accuracy) data from three of the low-level, blind perfor-
mance evaluation samples (NIST SRM 1646), as was discussed in Section 9.2 of this
report.
Of the 222 analytical test results, 41 (18.5%) were randomly selected from within the
sample batches and were audited. One (1) random error that involved a dilution
factor calculation where the spreadsheet had been changed to accommodate a larger
dilution volume was found and corrected. The estimated error rate, before correc-
tion, was 2.4% with lower and upper 95% confidence limits of 0.24% and 13.7%,
respectively, based on a Poisson distribution. Adjusting for the one correction, the
135
-------
average error rate for the analytical test results is estimated at 2.0% with a 95%
confidence interval of 0% to 13.2%.
9.4 Data Assessment
All analytical data were reviewed to verify that all study requirements were met.
Various sets of data were compared to the DQOs stated in the QAPjP. Where
necessary, corrective actions were taken and documented in work assignment
records. The following subsections document quality control results pertaining to
sample preparation, instrument calibration, and data processing and statistical
analysis procedures.
9.4.1 Sample Preparation QC Data
Potential laboratory contamination was assessed by the use of digestion blanks.
These blanks were included in each sample preparation batch at the ratio of one
blank for every 20 samples, with a minimum of one per batch. The DQO for the
measured value of a digestion blank was set at 10 times the instrumental detection
limit. The digestion blank results for each batch are shown in Table 9-3.
For all but two preparation batches, the levels found in the method digestion blanks
(last column of Table 9-3) were below their respective calculated sample detection
limit. The levels in one blank of batch No. 505 and in the three blanks of batch No.
508 were above their respective calculated sample detection limit. However, all but
one of these blank levels were below 10 times the calculated sample detection limit
(i.e., the DQO). One blank level in batch 508 was above the DQO- A cassette from
the lot of cassettes that were in the collection laboratory was used as a blank for batch
508. Although the levels of lead found were above the limit of detection, they were
within the range of lead levels found in blank cassettes from previous studies.
To evaluate the accuracy and precision of the laboratory analytical procedures,
replicate spike QC samples were included in each sample preparation batch at the
ratio of two replicate spike samples for every 20 samples, with a minimum of two
per batch. Percent recoveries were calculated for each spike sample. From these
results, the range of duplicate percent recoveries was calculated as the difference
between the highest and lowest recovery in each batch. All percent recoveries met
the DQOs as stated in the QAPjP: lower and upper control limits of 75% and 125%,
respectively. The ranges of replicate percent recoveries were all below the upper
control limit of 20%. All recovery statistics are shown in Table 9-4 and in Figures 9-4
and 9-5, including the associated DQOs.
136
-------
Table 9-3 Method digestion blank results
Analytical
batch No.
E08233B
E09023B
E11053A
El 1083 A
E12023B
V12073A
Preparation
batch No.
501
501
502
502
503
503
504
504
505
505
506
506
507
507
508
508
508
Sample
type
Bottle
Bottle
Bottle
Bottle
Bottle
Bottle
Wipe
Cassette
Instrumental
detection
limit
(Hg/mL)
0.0129
0.0408
0.0331
0.0240
0.0240
0.0184
0.0184
0.4718 ug/L
Digestion
volume
(mL)
25
25
25
25
25
25
100
0.025 L
Calculated
sample
detection
limit (ug)a
0.323
1.020
0.828
0.600
0.600
0.460
1.840
0.0118
Data
quality
objective**
(M8)
3.23
10.20
8.28
6.00
6.00
4.60
18.40
0.118
Value
found
(M8»
< 0.32
<0.32
< 1.02
< 1.02
< 0.83
<0.83
<0.60
< 0.60
0.85
< 0.60
<0.46
<0.46
< 1.84
< 1.84
0.08
0.07
0.17
a Sample detection limit (fig) = instrument detection limit (ug/mL) x digestion volume (mL)
b DQO: Total |ig found is to be less than 10 times the sample detection limit (jig)
137
-------
Table 9-4 Method spike replicate results
Analytical
batch No.
E08233B
E09023B
E11053A
E11083A
E11083A
E12023B
E12023B
V12073A
Preparation
batch No.
501
502
503
504
505
506
507
508
Sample
type
Bottle
Bottle
Bottle
Bottle
Bottle
Bottle
Wipe
Cassette
Method spike
recovery (%)
104.58
104.40
109.65
109.73
101.73
96.83
101.93
98.41
101.10
98.98
98.58
102.93
104.54
103.18
103.35
97.86
103.08
98.46
98.88
98.90
109.95
102.66
102.22
111.96
99.34
96.48
98.40
% 103.70
Range1 of
replicate %
recoveries
5.33
5.10
4.35
6.68
4.62
9.74
2.86
5.30
1 Range of replicate % recoveries = highest - lowest % replicate recovery
DQOs: Upper warning limit = 15%; upper control limit = 20%
138
-------
IJV1
120
110
fr
K 100
f
a- 90
80
70
UCL
a
-a a
° 0 g B ° B
g a 0 D =0
501 I 503 I 505 I 507 I 509
502 504 506 508
Sample Preparation Botch No.
Figure 9-4 Method
26
24
22
20
fr 1B
5 16
u
K ' 14
K
% 12
e
? 10
«
6
4
2
n
spike replicates: recovery (%)
-
-
a
a
D n Q
0 °
a
501 I 503 I 505 I 507
502 504 506 508
Sample Preparation Batch No,
Figure 9-5 Method spike replicates: recovery range (%)
139
-------
9.4.2 Instrumental Analysis QC Data
A series of instrumental QC samples were analyzed with each analytical batch to
determine the performance of the instrumental measurements independently of
sample preparation. The following summarizes the results from these QC samples.
Initial calibration blanks (ICBs) were analyzed using one per run at the beginning of
each run. The ICBs and continuing calibration blanks (CCBs) were prepared by the
analyst on the day of analysis, using the same acid matrix that was used for sample
detection at the instrument. All measured values were below their respective
instrumental detection limits.
CCBs were used to verify blank response and freedom from carry-over. These
blanks were analyzed after each continuing calibration verification (CCV). The DQO
for these blanks was identical to that for ICBs. Of the 65 CCB samples run, 60 had
levels below their respective instrumental detection limits. One CCB sample in
each of batch Nos. 501, 504, and 508, and two in batch No. 505, were above their
respective instrumental detection limits. However, all measured CCB values were
below 10 times the instrumental detection limit; therefore, all CCB values met the
DQO.
Initial calibration verification samples (ICVs) were analyzed once per run following
calibration. These samples were analyzed to verify proper instrumental calibration
prior to the start of the analytical batch, and were from alternate stock standards
than those used in the original calibration. All ICV sample values met the DQO of
±10% of the known value. The results are shown in Table 9-5.
CCV samples were analyzed using one sample during or after calibration, after each
set of 10 samples, and at the end of the analytical run. These samples were analyzed
to monitor instrumental drift, utilizing the original mid-point calibration standard.
The measured values of these samples were all within ±10% of their respective
initial values. The CCV sample results are summarized in Table 9-6 and plotted in
Figure 9-6.
140
-------
Table 9-5 Initial calibration verification sample (ICV) results
Analytical
batch No.
E08233B
E09023B
E11053A
E11083A
E12023B
V12073A
Sample type
Bottle
Bottle
Bottle
Bottle
Bottle and wipe
Cassette
Concentration (|ig/g)
Known
10
10
10
10
10
20
Found
10.123
10.042
10.062
10.193
10.277
20.040
Recovery1(%)
101.2
100.4
100.6
101.9
102.8
100.2
1 DQO: Found value to be within ±10% of known value
141
-------
Table 9-6 Continuous calibration verification (CCV) sample results
Analytical
batch No.
E08233B
E09023B
E11053A
E11083A
E12023B
V12073A
Preparation
batch No.
501
502
503
504,505
506,507
508
Known value
(Ug/mL)
10
10
10
10
10
10
10
10
10
10
10
10
10
.10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
20
20
20
20
20
20
20
20
Found value
(U£/mL)
10.111
10.286
10.127
10208
10.260
9.9916
10.130
9.9681
10.057
10.178
10.173
10.307
10.401
10.317
10.387
10.417
10.204
10210
10.214
10.327
10.318
10.484
10.486
10.356
10.440
10.538
10.545
10.486
10.511
10.631
10.711
10.827
10.732
20.080
19.230
18.780
19210
19.380
19.070
18.760
19.540
Recovery
(%)
101.1
102.9
101.3
102.1
102.6
99.9
101.3
99.7
100.6
101.8
101.7
103.0
104.0
103.2
103.9
104.2
102.0
102.1
102.1
103.3
103.2
104.8
104.9
103.6
104.4
105.4
105.5
104.9
105.1
106.3
107.1
108.3
107.3
100.4
96.2
93.9
96.1
96.9
95.4
93.8
97.7
Instrument
drift (%)1
1.7
0.2
1.0
1.5
1.4
-0.2
0.7
1.9
1.3
2.2
1.4
2.1
2.4
0.1
0.1
1.2
1.1
2.7
2.8
1.5
2.3
3.3
-0.6
-0.3
0.8
1.6
2.7
1.8
-4.2
-6.5
-4.3
-3.5
-5.0
-6.6
-2.7
1 DQO: Instrument drift to be within ±10%
142
-------
•fl
CCV—Instrument Drift
r
I
TJ
sr
^
K
^•^
£
0
"c
V
In strum
10
14
12
10
8
6
4
2
0
-2
-4
-6
-8
-10
-12
-14
1 ft
f-
-
i n
UL
—
i
—
—
P^ D _
~ D j-p D D D^^ DTJ DD D D
n n m
J ULJ DM
n
D
"
D D
-
• i
LL
-
-
E08233B I E11053A I E12023B I
E09023B E11083A V12073A
Analytical Batch No.
-------
9.4.3 Statistical Analysis QC Results
Quality control of the statistical analysis was achieved through two analyses of the
data by the same analyst and peer review by another statistician. The data were
initially analyzed to prepare preliminary results for EPA's review and to determine
the most appropriate analytical procedures. After correcting a few minor errors in
the data files (identified during the preliminary analysis and the final verification
steps in the preparation of the data files), the analysis files were again prepared from
the revised data files and the final analyses were performed. The programming for
the final statistical analysis was independent of the programming used in the initial
analysis. Statistical procedures used in the first analysis were carefully reviewed
before being used in the second analysis. A Macintosh PC based statistical analysis
program called JMP from the SAS Institute was used to analyze the data. The two
analyses were separated by a period of about a month. Where applicable, the results
from the different models were compared to identify features of the data which were
not apparent from the primary analysis. In addition, both the statistical procedures
used to analyze the data and the results from the statistical analyses were reviewed
by a second statistician.
144
-------
APPENDIX A: PILOT TESTS RESULTS FOR THE WIPE AND VACUUM
STUDY
Pilot Tests for EPA's Wipe and Vacuum Study were conducted to test some of the
procedures proposed for the full Wipe and Vacuum Study and to provide informa-
tion for improving the design of the study. The objectives of the full Wipe and
Vacuum Study are broad in scope: first evaluation of two kinds of dust collection
methods (samplers and household vacuum cleaners) with multiple examples of
each method; second estimation of lead recovery, dust recovery, and amount of dust
expelled through the exhaust, with recovery assessments for multiple substrates,
multiple amounts of dust, multiple particle size classes, and multiple dust lead
concentrations. The results of the pilot tests were used to help select the solutions to
the most important design problems encountered. This appendix describes the
individual tests which together comprise the Wipe and Vacuum Pilot Study and the
results of those tests.
A1.0 INTRODUCTION
A draft study design document for a full laboratory study was prepared by Westat,
Inc. and reviewed by MRI and EPA for inclusion in the Quality Assurance Project
Plan (QAPjP)15 for the Wipe And Vacuum Study. During the development of the
test design document, several uncertainties were identified regarding testing of
vacuum cleaners. The first major uncertainty concerned the amount of dust that
should be used on the substrates (e.g., carpet) for testing the dust and lead pickup
efficiency of vacuum cleaners. Concerns about accuracy in determining dust
removal efficiency raised questions about variability of weighing new bags used in
vacuum cleaners (i.e., precision of tare weights). A related question concerned how
much preconditioning of new carpets would be necessary so that the weight of
carpet fibers picked up in the tests would be insignificant compared with weight of
dust. A final question concerned whether new carpet should be used for each test
(after preconditioning) or whether the same substrate could be used in several tests,
without significant "carryover" from test to test.
In another area, a concern was the possibility that a significant portion of the dust
used on the substrate might adhere to the brush and wand of the vacuum cleaner. If
this were true, some dust would actually have been removed from the substrate but
not included in the weight change of the bag.
Uncertainty about the amount of dust that should be used was also associated with
determining lead removal efficiency. Lead in the dust could be determined either by
digesting the entire vacuum bag or by removing dust from the bag for digestion and
15Quality Assurance Project Plan for the Wipe and Vacuum Study. EPA Contract nos. 68-DO-0137 (Task
3-55) and 68-D3-0011 (Task 1-07). July 21,1993.
A-l
-------
analysis. The amount of lead in the bag and captured dust would be affected by the
lead content of the bag itself (blank level) if the amount of dust was small and could
not be removed from the bag. Initial tests carried out on two blank bags indicated
that, when including the bag in the lead analysis, the amount of dust used would
likely have to be as large as 5 grams or more. It was therefore important to
determine if a sufficient quantity of dust could be removed from vacuum cleaner
bags.
The mechanics of carrying out the tests required a method of securing substrates in
some suitable way for performing the vacuuming tests. A method described in
ASTM F608-89 was a likely option, but the apparatus needed to be built and success-
fully used prior to the full laboratory testing.
Finally, there was uncertainty as to how dust emissions from the vacuum cleaners
could be measured (i.e., dust that passes through the bag and is exhausted out of the
vacuum cleaner). This uncertainty raised several questions about how dust could be
fed evenly into the vacuum cleaners for such tests, and how the emissions could be
sampled isokinetically for the emission measurements.
A1.1 Objectives
Having no information to answer the questions presented above, MRI, with the
help of Westat, prepared a work plan for the pilot test16. The work plan was
reviewed and approved by EPA, after which the pilot tests were performed by MRI.
The tests were designed to determine dust quantities needed for the full study, to
determine what preconditioning procedures were necessary, and to optimize the
sampling protocols discussed in the QAPjP. The various tests were organized into
the following five tasks:
Task 1 Determine the Stability of Tare Weights for New, Clean Vacuum
Cleaner Bags
Task 2 Demonstrate a Method of Securing Carpet and Upholstery
Substrates for Testing Vacuum Cleaners
Task 3 Determine if Preconditioning Procedures Would Allow Use of
New Carpet in the Tests, and Determine if the Reuse of the
Same Substrate for Each Series of Tests is Feasible
Task 4 Determine the Amount of Dust Needed for the Tests
16Revised Work Plan for Pilot Tests. Wipe and Vacuum Study. EPA Contract No. 68-DO-0137. Work
Assignment No. 55. MRI Project No. 9802-a(55). June 14,1993
A-2
-------
Task 5 Develop and Demonstrate a Method for Measuring Exhaust
Emissions from Vacuum cleaners
The detailed description of work performed in each of these tasks is provided in the
study design presented in Section A2.
The following five sections of this report cover the study design and procedures
(Section 2), the data collection (Section 3), the statistical data analysis (Section 4), and
the discussion of the test results (Section 5).
A-3
-------
A2.0 STUDY DESIGN AND PROCEDURES
The pilot tests were performed using four vacuum cleaners on two types of
substrates, carpet and upholstery. The vacuum cleaners were: three canister types,
subsequently denoted as A, B, and C and one upright, denoted as D. The following
accessories were included with the purchased vacuum cleaners:
Vacuum Cleaner Code
and Type
Vacuum Cleaner Accessories
Vacuum cleaner A
(canister) without
HEPA filter
Beater bar for use on rugs
Upholstery attachment (including a brush)
Hard surface attachment (including a brush)
Vacuum cleaner B
(canister) without
HEPA filter
Beater bar for use on rugs. (Power nozzle used on
hard surfaces with beater bar stopped.)
Upholstery attachment (with a brush)
Vacuum cleaner C
(canister) with HEPA
filter
Beater bar for use on rugs
Upholstery attachment (no brush included)
Hard surface attachment (including a brush)
Vacuum cleaner D
(upright) without
HEPA filter
Hose connection for upholstery attachment
(including a brush) and hard surface attachment.
(Insertion of hose stops beater bar and diverts
suction to hose.)
The exhaust emission tests were conducted using each of the vacuum cleaners and
dust which had passed through a 53 micron mesh. For all of the other tests,
Vacuum cleaner A was used with dust which had passed through a 250 micron
mesh.
The steps followed in carrying out the work for each task listed in Section Al.l are
described in the following sections. In some cases, the test procedures were
modified as the pilot tests progressed, or additional tests were performed which
were not outlined in the original work plan. The following sections describe the
tasks as they were conducted. If changes to the procedures in the original work plan
were made during the progress of the pilot tests, they are noted in the description of
each task.
A-4
-------
A2.1 TASK 1— Determine Stability of Tare Weights for New Clean
Vacuum Cleaner Bags
Data on stability of the tare weight for new vacuum cleaner bags were observed by
two different types of tests done on two different days. For Day I, each of the four
types of vacuum cleaner bags was weighed 10 times with at least half-hour intervals
between weighings. On Day 2, the bag used on Day 1 was inserted in the vacuum
cleaner which was run for 1 minute (without actually vacuuming a surface). The
bag was then removed and reweighed and the process repeated 10 times using the
same bag. This procedure was carried out for all four vacuum cleaners. The specific
procedure used on Day 1 and Day 2 is given below.
Dayl
Day 2
Weigh four different new vacuum cleaner bags 10 times each
throughout the day, with at least half-hour intervals between
weighings. Use one bag for each of the four different brands of vacuum
cleaners.
Record the weights, relative humidity (RH) and temperature (T) in the
lab, and the time when weighings are made.
Run the vacuum cleaner for 5 minutes (without actually vacuuming a
surface) with an old bag in place to purge loose dust.
Discard the bag.
Reweigh the bag from Day 1.
Record RH and T and the time when weighed.
Insert the tared bag into the vacuum cleaner and run the unit for 1 min
(without actually vacuuming a surface).
Remove the bag and weigh it.
Repeat this vacuum procedure and weigh sequence 10 times for the
same bag.
Repeat this sequence of 10 weighings for each of four brands of vacuum
cleaner bags.
A-5
-------
Extra Test
Results from the Day 2 tests showed that the weight of Bag D was increasing
over time as it sat on the scale. Therefore, an extra test was done as
follows:
• Run vacuum cleaner for 40 seconds with new bag.
• Repeat above three times, 1 minute between, using same bag.
• Run vacuum cleaner 40 seconds.
• Immediately weigh bag, and record weight every 1 minute for 10
minutes.
• Repeat all the above once (total of two times).
• Repeat all the above for vacuum cleaners B, C, and D.
Because the sampling cassettes used for some of the samplers might also show
weight changes over time, the following tests were also performed.
• Obtain a cassette which has been acclimated to room conditions for >24
hours (with the plugs removed).
Weigh the cassette and record the weight every minute for 5 minutes.
• Remove the top half of the cassette and install the bottom half in the
Blue Nozzle sampler.
• Run the sampler for 120 seconds.
• Remove the cassette and reinstall the top half.
• Weigh the cassette and record the weight every minute for 5 minutes.
• Repeat running the sampler and weighing the cassette once.
• Repeat all of the above with another cassette once.
A-6
-------
A2.2 TASK 2—Demonstrate Method of Securing Carpet and
Upholstery Substrates for Testing Vacuum Cleaners
This task required construction of a 6-in high rectangular table designed to support
and secure carpet, upholstery, and other substrates for vacuuming, in accordance
with an ASTM method. The procedure for carrying out this task was:
• Construct a 6-in high rectangular table (1.83 m x 0.69 m) with the top
made from 3/4-in thick exterior grade plywood (per ASTM Method
F608-89). Provide a mechanism for securing carpet section, or
upholstery section, at the corners of the table. The carpet section will
include the pad underneath, and the upholstery will include a 1/2-in
thick foam pad underneath.
• Determine suitability of table for vacuum cleaner tests by vacuuming a
carpet section with the upright vacuum cleaner and with one of the
canister vacuum cleaners. Similarly, vacuum an upholstery section
with the same two vacuum cleaners using the proper attachments.
• Revise system for securing carpet or upholstery, if necessary, after
approval by project leader.
A2.3 TASK 3—Determine if Preconditioning Procedures are Feasible
for Using New Carpet in the Laboratory Tests, and Determine if
use of the Same Carpet for Each Series of Tests is Feasible
Task 3 consisted of a two-step process using vacuum cleaner A only. Step 1, precon-
ditioning of carpet without dust, involved vacuuming the carpet section for 5
minutes followed by weighing the bag, and repeating this process 10 times. In step 2,
which used the carpet section from Step 1, two different amounts of dust were
applied to the carpet, with each amount being applied and followed by three
30-second vacuumings. Three sets of tests were done for each of the two amounts of
dust. These three tests differed in that the vacuuming of the wand and brush was
done in three different ways:
• After completing the entire test.
• After each set of three vacuumings within each test.
• After each vacuuming.
The multiple parts of Step 2 provided important data on the effect of vacuuming the
wand and brush and on dust pickup efficiency, after the carpet had already previ-
ously been used (i.e., dust applied and vacuumed up).
A-7
-------
The procedures used in Step 1 and in all six parts of Step 2 (a to f) are listed below.
Stepl
• Vacuum the entire area of the secured carpet for 5 min.
• Discard bag.
• Measure RH and T.
• Place a new tared bag in the vacuum cleaner; vacuum carpet again for
5 minutes.
• Reweigh bag.
• Repeat the 5-minutes vacuuming and weigh process 10 times using the
same bag.
• Determine incremental weight gain for each vacuuming.
Step 2 (using the preconditioned carpet from Step 1)
2a. Tests using 0.678 g of dust:
• Vacuum the wand and brush before the initial test; discard bag.
• Place new tared bag in vacuum cleaner.
• Put 0.678 g of sieved dust (e.g., < 250 Jim) on the test area (100 mg/ft2).
• Vacuum for half a minute.
• Weigh the vacuum bag.
• Repeat the last two steps (vacuum and weigh) three times using the
same bag.
• Repeat the last five steps (new bag, deposit dust, vacuum and weight
three times) three times.
• After completing all the above tests and the last weighing, vacuum the
wand and brush and reweigh the bag.
A-8
-------
2b. Tests using 2.71 g of dust:
• Vacuum the wand and brush before the initial test; discard bag.
• Place new tared bag in vacuum cleaner.
• Put 2.71 g of sieved dust on the test are (400 mg/ft2).
• Vacuum for half a minute.
• Weigh the vacuum bag.
• Repeat the last two steps (vacuum and weigh) three times using the
same bag.
• Repeat the last five steps (new bag, deposit dust, vacuum and weight
three times) three times.
• After completing all the above tests and the last weighing, vacuum the
wand and brush and reweigh the bag.
2c. Tests using 0.678 g of dust (same as 2a, except more frequent vacuuming of
wand and brush):
• Vacuum the wand and brush before the initial test; discard bag.
• Place new tared bag in vacuum cleaner.
• Put 0.678 g of sieved dust on the test area (100 mg/ft2).
• Vacuum for half a minute.
• Weigh the vacuum bag.
• Repeat the last two steps (vacuum and weigh) three times using the
same bag.
• After the third vacuuming and weighing, vacuum the wand and brush
and reweigh the bag.
• Repeat the last six steps (new bag, deposit dust, vacuum and weight
three times, vacuum wand) three times.
A-9
-------
2d. Tests using 2.71 g of dust (same as 2b, except more frequent vacuuming of
wand and brush):
• Vacuum the wand and brush before the initial test; discard bag.
• Place new tared bag in vacuum cleaner.
• Put 2.71 g of sieved dust on the test area (400 mg/ft2).
• Vacuum for half a minute.
• Weigh the vacuum bag.
• Repeat the last two steps (vacuum and weigh) three times using the
same bag.
• After the third vacuuming and weighing, vacuum the wand and brush
and reweigh the bag.
• Repeat the last six steps (new bag, deposit dust, vacuum and weight
three times, vacuum wand) three times.
2e. Tests with 0.678 g of dust, including vacuuming of wand and brush with each
vacuuming of the carpet:
• Vacuum the wand and brush before the initial test; discard bag.
• Put new tared bag in vacuum cleaner.
• Put 0.678 g of sieved dust on the test area (100 mg/ft2).
• Vacuum for half a minute.
• Use the vacuum hose to vacuum dust from the wand and brush.
• Weigh the vacuum bag.
• Repeat the above sequence (vacuum carpet, vacuum wand and brush,
weigh bag) three times using the same bag.
• Repeat the last six steps (new bag, deposit dust, vacuum-weight-
vacuum wand three times,) three times.
A-10
-------
2f. Tests with 2.71 g of dust, including vacuuming of wand and brush with each
vacuuming of the carpet:
• Vacuum the wand and brush before the initial test; discard bag.
• Put new tared bag in vacuum cleaner.
• Put 2.71 g of sieved dust on the test area (400 mg/ft2).
• Vacuum for half a minute.
• Use the vacuum hose to vacuum dust from the wand and brush.
• Weigh the vacuum bag.
• Repeat the above sequence (vacuum carpet, vacuum wand and brush,
weigh bag) three times, using the same bag.
• Repeat the last six steps (new bag, deposit dust, vacuum-weight-
vacuum wand three times,) three times.
One of the three sets of vacuumings described in step 2a showed a noticeably lower
dust pickup efficiency, and was done by a different operator. To test if this lower
recovery was associated with differences between operators, an extra test was
performed in which the entire Step 2a was repeated twice, using a different operator
for each repetition.
A2.4 TASK 4—Determine the Amount of Dust Needed for the Tests
Tests carried out in Task 4 used the same carpet sample and the same vacuum
cleaner (A) as in Task 3. Task 4 required applying and embedding two different
amounts of dust. The dust was applied either in 10 separate applications and
vacuumed after each application, or once followed by 10 vacuumings (e.g., 0.678 g
applied and vacuumed 10 times, or 6.78 g applied once and vacuumed 10 times).
Two of the four tests done in Task 4 required determining of the weight of dust that
could be recovered from the vacuum cleaner bag after the test had been completed.
This was important, since at least 100 mg needed to be recovered for lead analysis. If
that amount could not be recovered, it would be very difficult to determine the lead
content of the dust collected by the vacuum cleaner. It was anticipated that the larger
amount would yield recovery of 100 mg even if the smaller amount did not.
According to the original work plan and the data sheets in Section A6.0, the specific
procedures for Task 4 involving carpet samples and vacuuming of dust from the
wand and brush depend on the results from Task 3. Based on the preliminary
A-ll
-------
analysis of the Task 3 results (see Section A4.3), the Task 4 tests used the same carpet
sample for all tests and the wand and brush were not vacuumed as part of the test.
The procedures used in carrying out the tests for Task 4 are as follows.
a. Application of 0.678 g of sieved dust to test area, 10 times:
• Put new tared bag in vacuum cleaner. Record RH and T.
• Apply 0.678 g of dust and embed. Brush any dust that sticks to
embedding tool back onto the carpet. If it appears that significant
amounts of dust are lost or cannot be brushed back onto carpet, contact
project leader before proceeding.
• Vacuum carpet for half a minute.
• Weigh bag.
• Repeat 10 times using the same bag (adding 0.678 g of dust each time,
for a total of 6.78 g).
• Remove bag from vacuum cleaner and make sure that bag inlet is wide
open (cut away any sealing flaps if necessary). Place opening of bag
over top of tared beaker and tap on outside of bag to dislodge dust into
beaker. Determine weight of dust recovered, that could be used for lead
analysis. Observe dust to determine if fibers from bag are present in the
sample.
• Repeat the entire process in Step a once.
b. Application of 6.78 g of sieved dust to test area, once:
• Put new tared bag in vacuum cleaner. Record RH and T.
• Apply 6.78 g of dust to test area and embed.
• Vacuum carpet for half a minute.
• Weigh bag.
• Vacuum surface again, using the same bag, without adding any dust to
the test area.
• Weigh bag.
• Repeat the vacuum and weigh process for a total of 10 times.
A-12
-------
• Repeat the entire process in Step b once.
c. Application of 2.71 g of sieved dust to test area, 10 times:
• Repeat the entire process described in Step a using 2.71 g of dust rather
than 0.678 g.
d. Application of 27.1 g of sieved dust to test area—once:
• Repeat the entire process described in Step b using 27.1 g of dust rather
than 6.78 g.
Note: After completing Steps a through d of Task 4, Step a was repeated twice with
three applications of dust. The three dust applications were followed by either three
vacuumings after last application of dust or by seven vacuumings, as suggested by
Westat.
A2.5 TASK 5—Develop and Demonstrate Method for Measuring
Exhaust Emissions from Vacuum Cleaners
Work on this task involved fabricating a system to feed a specific quantity of dust (5
g) into the inlet of a vacuum cleaner over a specific period of time (5 min). It also
involved fabrication of a sealed enclosure, suitable for all vacuum cleaners to be
tested (including upright), so that only the suction tube extended outside the
enclosure. The enclosure was built so that all exhaust emissions discharged through
one duct. The diameter of the exhaust duct was designed so that isokinetic sampling
could be carried out near the center of the duct, with the sample directed to a
particulate concentration monitor measuring dust emissions in jig/m3- A pitot tube
was used to determine the total gas flow rate in the duct, so that the dust emission
rate (ng/min) and total emissions (jig) could be calculated.
The vacuum cleaner enclosure and dust feed system used for Task 5 are described in
Appendix O of Volume n of this report. The enclosure and feed system were used
to carry out three replicate tests for each of the four vacuum cleaners. The dust used
in tasks 1 through 4 was sieved to obtain dust which passed through a 53 micron
sieve for use in Task 5.
A procedure for conducting the vacuum cleaner exhaust emission tests as part of the
full study had been prepared for the QAPjP. These procedures, described in
Appendix O of Volume n, were used in the pilot tests with only minor changes.
Throughout each test the concentration of particulate in the exhaust duct was
continuously monitored and recorded on a strip chart recorder. The particulate
concentration monitor is based on the detection of near-forward scattered electro-
A-13
-------
magnetic radiation in the near-infrared (940 nm). The monitor was Model RAM-1
purchased from Monitoring Instruments for the Environment Inc. (MIE) in
Billerica, Massachusetts.
The test procedures specified the following steps:
• Turn on the particulate monitor and strip chart recorder. Mark the
date, time, and run number on the strip chart. Also identify each of the
following steps on the strip chart, and record the time.
• Turn on the vacuum cleaner. Run for 1 minute.
• Turn on the turntable and lower the vacuum cleaner nozzle until it
nearly touches the turntable.
• Continue running for 5 minutes, thereby removing all of the dust
from the turntable (i.e., one revolution).
• Continue running for 1 minute, then stop the test.
• Remove the bag from the vacuum cleaner; wait 5 minutes, then record
the weight of the bag.
• Repeat the test three times for each vacuum cleaner.
The original plan specified repeating the test twice for one vacuum cleaner. When
the pilot test was performed, it was easy to test all of the vacuum cleaners, following
the procedures which were planned for the full study.
A-14
-------
A3.0 DATA COLLECTION
All pilot tests were conducted at MRI's laboratory according to the work plan for this
study and the laboratory procedures previously described for Tasks 1 through 5.
A3.1 Pilot Test Data Collection
All weights (bags and dust samples) were determined using Mettler PM 1200 and PM
2500 balances which were checked for accuracy each morning using standard check
weights. Ambient relative humidity and temperature in the laboratory were
recorded for each test. Embedding of dust into carpet, when prescribed, was
performed according to the protocol for grinding dust into carpet (Appendix C of
Volume II). For each series of tests, data were recorded on forms developed for
these tests. The data for each task are reproduced in Sections A6.1 through A6.5 for
Tasks 1 through 5 data, respectively.
Several of the tests involved vacuuming a section of carpet without dust applica-
tion. Other tests involved applications of dust followed by vacuuming. One section
of carpet was used in all the tests, mounted with carpet pad underneath, on the 6-in
high table described previously in Task 2. The carpet section (1.83 m x 0.69 m) was
made of nylon, purchased locally.
The dust for the tests was obtained from vacuum cleaner bags collected by Westat
and MRI and then sterilized. Dust from these bags was mixed together. The portion
of dust which passed through the 250 micron sieve was used for Tasks 3 and 4. This
dust was resieved using a 53 micron sieve, to provide dust for the exhaust tests.
In order to apply dust as evenly as possible onto the carpet test area (1.37 m x 0.46 m),
the prescribed amount of dust was weighed in a beaker along with the 250-pin sieve.
Dust in the beaker was then poured onto the sieve over the test area, and the sieve
lightly tapped as it was moved around above the test area. Most, but not all, of the
dust passed through the sieve when using this method. Therefore, the sieve and
beaker were subsequently reweighed to determine, by difference, the weight of dust
that actually passed through the sieve onto the carpet test area.
A3.2 Quality Assurance Activities
The data sets from each task were audited for accuracy of weight data, balance
calibrations, calculations, etc. In addition, a systems audit was conducted during the
pilot study phase of this project.
A-15
-------
A4.0 STATISTICAL DATA ANALYSIS
The statistical analyses of the pilot test data were performed by Westat.
A4.1 Analysis of Task 1
Stability of Tare Weights For New Vacuum Cleaner Bags (Task 1, Day 1)
Four bags, one from each of the four vacuum cleaners, were weighed at half-hour
intervals. Measurements of temperature and relative humidity were recorded at
the time of weighing.
Prior to the tests, it was suspected that the weight of a vacuum cleaner bag would
change with a change in the relative humidity or temperature. Plots of the data
showed a change in the bag weights over time where the rate of change was greatest
in the beginning of the test. Although the trend in the weights might be associated
with the fluctuations in the room temperature or relative humidity at the time of
measurement, it might also result from the bag coming into equilibrium with the
surrounding laboratory environment during the test and after it was removed from
its storage area. If the rate at which the bag weight changes as it comes into equilib-
rium with the laboratory environment is proportional to the difference in the bag
weight and the equilibrium bag weight, the bag weight will follow a simple
exponential decay relationship. The difference between the bag weight at the
beginning of the test and that at reaching equilibrium may be due to differences in
temperature and relative humidity between the laboratory and the bag storage area.
In order to identify whether temperature, humidity, or approach to equilibrium
provides the best explanation of the weight changes, a model was fit to the data with
the terms for (1) a liner relationship between the bag weight and temperature, (2) a
linear relationship between the bag weight and relative humidity, and (3) an
exponential decay for the return to equilibrium. Nonlinear regression was used to
fit the model. The regression parameters were used to identify which factors were
most influential in determining the bag weight The root mean square error
estimates the standard deviation of one measurement. The equation to fit the data
was:
Bag weight = C + R*(Relative humidity) + TTemperature + D * (1 - exp(-(time)/M))
where:
Bag weight is measured in grams.
Relative humidity is measured in percent.
Temperature is measured in degrees Fahrenheit.
A-16
-------
Time is measured as the number of minutes from the time of the first measure-
ment.
C = a constant (the initial weight of the bag, in grams, at 0% relative humidity and 0
degrees Fahrenheit).
R = the change in the bag weight, in grams, associated with an increase in the
relative humidity of 1%.
T = the change in the bag weight, in grams, associated with an increase in the
temperature of one degree Fahrenheit.
D = the change in the bag weight, in grams, from the beginning of the test to until
equilibrium is reached.
M = the equilibration time of the bag, the time, in minutes, for the bag weight to
reach 69% of equilibrium.
This model was fit to the data for each vacuum bag tested. The parameter estimates
and the root mean square error are shown in Table A-l. Following the table, for
each vacuum cleaner bag, Figure A-l shows the weight measurements (using
circles), the predicted weights (using diamonds), and the predicted weight change
(trend) associated with the bag coming into equilibrium with the laboratory
environment. The differences between the trend and the model prediction are due
to changes in the temperature and relative humidity during the tests.
For all four vacuum bags, the estimated weight changes associated with an approach
to equilibrium were statistically significant. The estimated weight changes associ-
ated with changes in temperature were not statistically significant. The estimated
weight changes associated with changes in the relative humidity were statistically
significant for bags from vacuums B, C, and D.
The importance of trend, temperature, and relative humidity in determining the
precision of the bag weight depends on the changes in time, temperature, and
relative humidity which might be expected during the test. During the pilot tests
the temperature and humidity were fairly stable over short periods of time.
Assuming that fluctuation in temperature and relative humidity are similar during
the full tests to those during this pilot test, and assuming further that the bag
weights are close to equilibrium, the root mean square error measures the standard
deviation of a single weight measurement. These estimates are shown in Table A-l.
A-17
-------
Table A-l Regression estimates for predicting the weight of vacuum cleaner bags
as a function of time, relative humidity, and temperature for data
collected on Day 1.
Parameter
C Constant
R Change in weight with
change in relative
humidity (g/%RT) with
95% confidence
intervals
T Change in weight with
change in temperature
(g/°F) with 95%
confidence intervals
D Difference between
initial weight and
equilibrium weight (g)
M Equilibrium time in
minutes with 95%
confidence intervals
Root mean square error (g)
Vacuum
A
35.5
0.003
-.003 to 0.009
-.005
-.022 to 0.012
0.023
i
1 29.7
7.8 to 112.9
0.0034
B
42.0
.0.013
0.001 to 0.024
.003
-.409 to 0.362
0.048
43.0
12.7 to 145.7
0.0064
C
30.2
0.006
0.002 to 0.010
.003
0.12 to 0.012
0.014
109.3
23.6 to 505.8
0.0018
D
41.0
0.007
0.003 to 0.011
.003
-0.10 to 0.010
0.017
24.3
6.7 to 88.5
0.0023
Model: Bag weight = C + R * (Relative humidity) + T * Temperature + D * (1 - exp(-(time)/M))
Statistically significant results are shown in bold text
A-18
-------
Vacuum Cleaner A
Vacuum Cleaner B
11:00 12:00 13:00 14:00 15:00
Time of day
41.060 T
_ 41.050-
* 41.040
1> 41.030 I
| 41.020
$ 41.010
W 41.000 f
40.990
11:00 12:00 13:00 14:00
Time of day
15:00
I
30.485
30.480
30.475
I 30.470 i
60
« 30.465
30.460
Vacuum Cleaner C
11:00 12:00 13:00 14:00
Time of day
15:00
:00
Vacuum Cleaner D
12:00
13:00 14:00
Time of day
Data
Trend
15:00
Model
Figure A-l Measured and predicted weights of vacuum cleaner bags over time from Day 1
-------
Recovery measurements are based on the change between the initial and final
weight of the vacuum cleaner bag. The standard deviation for this change is 1.414
(sqrt(2)) times the standard deviation of one weight measurement. Further,
assuming that accurate weight change measurements can be achieved if the weight
change is 10 times the standard deviation of the weight change measurements, the
weight changes would need to be roughly 14 times the root mean square error
shown in Table A-l. For the largest root mean square error in Table A-l, 0.0064, the
standard deviation (i.e., 0.091 grams) could therefore be measured with acceptable
precision. The precision would be better for some vacuum cleaners than for others.
The estimates of the equilibration times have implications for the how long the bags
should sit before making weight measurements. These values will be discussed
later.
Stability of Tare Weights for Vacuum Cleaner Bags Placed in the Vacuum (Task 1,
Day 2)
During Day 2, our bags, one from each of the four vacuum cleaners, were weighed
10 successive times. Each weighing was separated by placing the bag into the
vacuum cleaner and running the vacuum for 1 minute. Measurements of tempera-
ture and relative humidity were recorded at the time of weighing. The model fit to
this data is the same as fit to the data from Task 1, Day 1. Table A-2 and Figure A-2
present the results from fitting the model to the data for each vacuum.
When weighing unused bags on Day 1, the weight increased over time as the bags
came into equilibrium However, when putting the bags into the vacuum and
running the vacuum, the weight decreased over time.
Based on the root mean square error, weight changes of 0.100 grams would have
acceptable precision for all the vacuum cleaners. This estimate is similar to that
from the measurements on Day 1.
The equilibrium time estimates for the data from Day 1 and Day 2 are similar. That
is, it takes about 30 minutes for the bag weight to go 69% of the way to its equilib-
rium weight. This time is similar whether or not the bag is placed in the vacuum
and run for a minute. Unless the laboratory, staff waits a long time for the bag
weight to come to equilibrium, the weight measurement will depend on the time at
which the weight is taken.
It was decided to standardize the time between removing the bag from the vacuum
and the weight measurement for subsequent tests to control the weighing error. For
Tasks 3 and 4, the time between vacuuming and weighing was 5 minutes, timed
with a stop watch.
A-20
-------
Table A-2 Regression estimates for predicting the weight of vacuum cleaner bags
as a function of time, relative humidity, and temperature for data
collected on Day 2.
Parameter
C Constant
R Change in weight with
change in relative
humidity (g/%RT) with
95% confidence
intervals
T Change in weight with
change in temperature
(g/°F)with95%
confidence intervals
D Difference between
initial weight and
equilibrium weight (g)
M Equilibrium time in
minutes with 95%
confidence intervals
Root mean square error (g)
Vacuum
A
40.1
0.034
-.096 to 0.029
-.039
-.108 to 0.029
-.109
30.3
16.3 to 44.2
0.0042
B
45.9
-.037
0.064 to 0.011
-0.39
-.064 to -.015
-.104
30.1
24.5 to 35.6
0.0027
C
29.9
0.003
-0.22 to 0.016
.006
-.013 to 0.025
-0.20
25.8
6.2 to 45.3
0.0020
D
49.5
-0.32
-.121 to 0.058
-.191
-.284 to -.097
-.226
16.8
13.7 to 20.0
0.0071
Model: Bag weight = C + R * (Relative humidity) + T * Temperature + D * (1 - exp(-(time)/M))
Statistically significant results are shown in bold text.
A-21
-------
Vacuum Cleaner A
Vacuum Cleaner B
34.75 T
34.7
-a 34.65
r 34.6
§34.55
| 34.5
6034.45
£ 34.4
34.35
34.3
13:
30
41.54
14:00
14:30 15:00
Time of day
Vacuum Cleaner C
15:30
16:00
13-30 14:00 14:30 15:00
Time of day
15:30
16:00
:30 14:00
14:30 15:00 15:30
Time of day
Vacuum Cleaner D
16:00
Data
Predicted
Figure A-2 Measured and predicted weights of vacuum cleaner bags over time from. Day 2
-------
Although the time between vacuuming and weighings was 5 minutes for Tasks 3
and 4, a question arose about the best time to use in the full study. To answer this
question, extra tests were performed to measure the weight change over time after
removing the bag from the vacuum. These tests were done using two bags for each
of the four vacuums. Similar tests were performed for the sampling cassettes used
for the Blue Nozzle vacuum sampler, testing two cassettes twice each. The results of
these extra tests are discussed below.
The exponential decay model: Bag weight - C + D * (1 - exp(-(time)/M)) was fit to the
measurements from the extra tests using nonlinear regression.
Waiting for the bags to reach equilibrium might add considerable time to the test
process. An alternative is to specify a fixed time between the removal of the bag
from the vacuum and weighing (call this the time after vacuuming), thus standard-
izing the weighing procedure. When specifying a fixed time after vacuuming for
the weighing, the precision of a weight measurement depends on the rate at which
the weight is changing and the precision with which the time after vacuuming can
be set. Assuming that the actual time between vacuuming and weighing varies
around the proscribed time and has a standard deviation of 5 seconds, the variance
of the weight is equal to the mean square error plus the error associated with timing
(the rate of weight change times the variance of the time after vacuuming). Using
these assumptions, the standard deviation of a weight measurement is shown in
Figure A-3 for each of the vacuum cleaner bags tested and in Figure A-4 for each of
the sampling cassette tests.
For both the sampling cassettes and the vacuum cleaner bags, the standard deviation
of one weight measurement decreases with increasing time between turning the
vacuum off and weighing the bag. However, except for the bags from vacuum
cleaner D, there is little improvement in precision beyond the first several minutes.
A4.2 Analysis of Task 2
No data was generated for analysis of Task 2.
A4.3 Analysis of Task 3
Task 3 consisted of (1) fiber preconditioning of a test piece of carpet to determine
how much effort was required to precondition the carpets and (2) tests to determine
if there was significant carryover of dust from test to test and how much dust
adhered to the vacuum cleaner wand and brush during vacuuming.
A-23
-------
4.5 -,-
60
S 4.0
-------
I
a*
i
§
0.20 -r-
0.18 --
0.16 --
0.12 -
-------
Fiber Preconditioning
For the fiber preconditioning tests, a piece of carpet was vacuumed 11 times, each
time for 5 minutes. The first vacuuming was used to warm the vacuum cleaner,
after which the weight increase in the bag was measured for each of the 10 vacuum-
ings. The weight of the bag was determined after the bag had been on the balance for
5 minutes. The weights were clearly increasing at the five-minute point. Therefore,
after the llth vacuuming, additional bag weights were obtained to determine the
rate of change in the bag weight with time, in the absence of vacuuming.
The weight increase of the vacuum bag due to fibers removed from the carpet for
each successive vacuuming is shown in Figure A-5. For this sample of carpet used
in the pilot studies, the additional fiber collected for each 5 minutes of vacuuming
decreased to below 0.500 grams after two vacuumings (or equivalently after 10
minutes of vacuuming) and below 20 milligrams after six vacuumings, or equiva-
lently, after 30 minutes of vacuuming. When developing the pilot study protocols,
it was decided that weight gains from fibers of more than 100 milligrams per minute
would be unacceptable and provide too much bias in the measurements. This level
of weight increase due to fibers was obtained in only 10 minutes of vacuuming.
However, the data also suggested that, with about 30 minutes of vacuuming, the
fibers vacuumed could be kept to a low level which would have little effect on the
recovery measurements and would vary little from test to test. Assuming that
similar amounts of fibers were vacuumed from other carpet samples, a target of 20
milligrams per 5 minutes of vacuuming was set for the fiber preconditioning.
The additional measurements of the weight of the vacuum bag collected after the
last vacuuming were analyzed. The equilibrium time was estimated to be 21
minutes, in the same general range as determined using the data from Task 1.
Dust Recovery
For the dust recovery pilot tests, either 0.678 grams or 2.71 grams (approximately) of
dust were applied to the carpet, after which the carpet was vacuumed three times for
40 seconds each time. The dust recovery was measured for each of the vacuumings.
A total of 18 tests were performed, six sets of three tests each. For the first two sets of
three tests, the amount of dust on the vacuum wand and brush was measured at the
end of the set. For the second two sets of three tests, the amount of dust on the
vacuum wand and brush was measured at the end of the test. For the last two sets
of three tests, the amount of dust on the vacuum wand and brush was measured
after each vacuuming of the carpet and thus the dust on the wand and the brush
was included into the estimate of dust vacuumed. Within each set of two tests, the
first test used nominally 0.678 grams of dust and the second used 2.71 grams of dust.
For each test, the recovery was calculated as the ratio of the amount of dust
deposited to the weight increase in the bag summed across the three vacuumings.
A-26
-------
u./ -
? °-6 -
e
«
O n c
•c 0.5 -
V
•S
J§ 04 -
^^" \/»TC
9
0
u
5 0.3 -
.S
V
g 0.2-
'!? 0.1 -
0 .
^•••M
I | Weigh*
gain
20mgper
5 minutes
1 — 1
_ •• «•! ' ' '^=~l ' ' l~=~l ^" 1
123456789 10
Number of 5 minute vacuumings
Figure A-5 Weight increase of vacuum bag during fiber preconditioning
A-27
-------
Figure A-6 shows both a histogram and a time series plot of the recovery measure-
ments. The time series plot shows the recovery for the tests in the order they were
conducted, going from top to bottom. Along the left axis a bar graph shows the
relative amounts of dust deposited for each test. The recovery measurement for the
second test is noticeably different from the other measurements and is classified as
an outlier by standard outlier tests. One possible explanation for this unusual
observation is that this test was performed by a different lab technician than the
other tests in Task 3. The recovery measurements for the first tests appear to be
more variable than for later tests. There may be an associated learning time in
which the technicians learned to perform the tests consistently. The analysis of the
data was performed both with and without the outlier.
Analysis of variance was used to identify factors which affect the recovery. The
factors considered included the test procedure for vacuuming the wand and brush,
the amount of dust deposited, and the interaction of these factors. Other terms were
included to test for carryover of dust from one test to the next and to test for trends
over time. No factors were statistically significant at the five percent level. This was
true whether or not the outlier was included in the analysis. Thus, there is no
evidence that the procedure for vacuuming the wand and brush affect the recovery
estimates. There is also no evidence of a trend across the three tests within each set
that might suggest that dust carries over from one test to another.
Since no factors appear to affect the recovery, the measurements are summarized
here by their mean and confidence intervals. The mean recovery over all tests is
85.3% with a 95 percent confidence interval from 82.1% to 88.5%. With the outlier
removed, the mean recovery is 86.7% with a confidence interval from 85.3% to
88.1%. In either case, the recovery is relatively high. This recovery on carpet is
expected to be lower than for all other substrates except carpet with ground in dust.
The standard deviation of the recovery measurements is 6.5% with the outlier
included and 2.7% it excluded. These precisions for measuring recovery are well
within the requirements for meeting the data quality objectives for the study.
Measurements of the dust collected on the wand and brush were obtained for two of
the test procedures. The average amount of dust removed from the wand and
brush by vacuuming, as a percentage of the dust deposited since the wand and brush
were last cleaned, was 0.19% when the wand and brush were cleaned after each set of
three tests and 0.44% when the wand and brush were cleaned after each test. The
largest measurement of dust vacuumed from the wand and brush was 0.71% of that
deposited.
A-28
-------
asurements
n ON NQ oo vo
i 4:
° 3-
Jg 2-
1 l'
50
1
2
3
4
5
6
s 7
1 9
% 10
1 11
H 12
13
14
15
16
17
18
1
1 , 1. Illll ,
% 60% 70% 80% 90% 100%
j 9
•
• Testl
~"| Procedure: vacuum wand and •
— i brush after every three tests 9
Zl •
•
• Test2
I Procedure: vacuum wand •
-~| and brush after every test 9
=3 •
•
•
Procedure: vacuum wand * Test 3
ZD and brush after each •
~] vacuuming within each test •
Pi . 1*1
50% 60% 70% 80% 90% 100%
Recovery measurements
D Relative dust amount • Recovery
i
Figure A-6 Histogram and time series plot of recovery measurements for Task 3
A-29
-------
The amount of dust removed from the wand and brush was consistently less than
1% of the dust deposited. Any differences in the recovery measurement due to the
procedure used to vacuum the wand and brush were too small to be identified in
the statistical analysis. Due to the time required to vacuum the wand and brush and
to the small amount of dust collecting on the wand and brush, it was decided not to
vacuum the wand and brush as a part of each test. In addition, due to the relatively
long time required to precondition each carpet sample (at least 10 minutes) and the
lack of evidence for carryover of dust from one test to another, it was decided to use
the same carpet sample for all tests in Task 4 rather than preconditioning a new
carpet sample for each test.
A4.4 Analysis of Task 4
In Task 4, tests with different amounts of dust applied in either 1 or 10 applications
were used to help determine how much dust was needed to get a measurable
amount in the vacuum bag and a usable amount from the bag for lead analysis.
These tests used the conditions with the lowest expected dust recovery, carpet with
ground-in dust. In the first two tests, the 0.678 grams or 2.71 grams of dust were
applied 10 times, followed each time by vacuuming for 30 seconds. In the second
two tests, 6.78 grams and 27.1 grams of dust were applied once, followed by 10 succes-
sive vacuumings. For each vacuuming, the weight increase in the vacuum cleaner
bag was determined. For each test, dust was removed from the vacuum cleaner bag
and weighed to determine if the amount of dust obtained was adequate for the
laboratory analysis of lead.
The preliminary results from Task 4 suggested that three dust applications followed
by three more vacuumings, would provide good estimates of recovery and enough
dust for measuring the lead concentration and would result in little carryover of
dust from test to test. An extra series of tests was conducted to determine if this
preliminary design would indeed meet these objectives.
The results from the first set of three tests in Task 3 suggested that the vacuum
cleaner operator could make a significant difference in the recovery measurement.
An extra test was performed repeating this first set of tests with each of two
operators. The dust was not ground in for these tests or for the other Task 3 tests.
Rather than analyze the data from Task 4 by itself, the data from Task 3, Task 4, and
the extra tests with three dust applications and the repeat of the Task 3 set of tests
were combined to create a complete history of the dust deposited and dust
vacuumed from the one carpet sample used in the pilot tests. The combined data
was analyzed to identify factors which affect recovery and to provide a model for
dust recovery that can be used as a basis for establishing the final design for the main
tests.
A-30
-------
Figure A-7 shows the amount of dust applied before and the amount of dust
recovered for each of the 168 half-minute vacuumings of the carpet after the
completion of fiber preconditioning. For each of the vacuumings, the vertical black
bar in Figure A-7 indicates the increase in the bag weight due to dust and fibers
removed from the carpet. If dust was applied to the carpet immediately before
vacuuming, the top of the white vertical bar indicates the weight of the dust applied.
In all but one case was greater than the weight of dust and fibers removed.
The model fit to the data assumes that a fixed percentage of the dust deposited is
picked up during the first 30 second vacuuming. Of the dust that remains on the
carpet, a different fixed percentage is picked up during the second vacuuming, and
so on for subsequent vacuumings. Some of the dust may not be removed by
vacuuming, and thus the dust amounts removed may equal less than the dust
amounts deposited. This general model can be fit to the data using regression. The
model can be written as:
Weight of dust removed = a + blXl + b2X2 + b3X3 + ...
Where the following process is assumed to have occurred:
(1) An amount of dust X3 is deposited.
(2) The carpet is vacuumed for 30 seconds.
(3) An amount of dust X2 is deposited.
(4) The carpet is vacuumed for 30 seconds.
(5) An amount of dust XI is deposited.
(6) The carpet is vacuumed for 30 seconds and the weight of the dust
caught in the bag is Y.
The model assumes that the vacuum bag captures a fraction bl of the dust XI, a
fraction b2 of the dust X2, a fraction b3 of the dust X3, a fraction of dust deposited
before the X3 deposit, and a small quantity a, which might be fibers or dust from
other sources.
A-31
-------
N>
30 -i
25 -
o
e 20 -
O
•d
o
2 10 -I
O
5 -
0
.1
removed C applied
Tasks
No grind-in
Task 4
With grind-in
I I I I I I I I
Extra tests
: With
: grind-in No grind-in
I I I • I I
* indicates outliers
Each of 168 half minute vacuumings, in sequential order
Figure A-7 Time series plot of dust applied and dust removed from the carpet samples in Tasks 3 and 4
-------
Additional terms were added to this model to account for the following effects:
• Different proportions of dust being picked up on the first vacuuming
when different densities of dust are deposited.
• Effects associated with temperature and relative humidity and changes
in temperature and relative humidity.
• Trends over time.
• Differences associated with the test protocols.
The final model fit to the data was:
Y = a + bii-Xu +bi2*X12 + b2*X2 + b3*X3 + b4*X4 + b5*X5 +
fcll-20*Xn-20 + gll*zll + g!2*z!2 + gl3*z!3 + g!4*z!4 + g2*z2 +
g3*z3 + g4*z4 + g5*z5 + g6-10*z6-10 + gll-20*zll-20 + 1*™ +
r2*Temp + r3*DRH + r4*DTemp + t*obs + pii=l/7 [1]
where:
Y = the change in the dust weight during 30 seconds of vacuuming.
a = a constant.
the amount of dust which was deposited at a loading of 100 mg/sq.ft.
prior to vacuuming
the proportion of dust which was deposited at a loading of 100 mg/sq.ft.
which was picked up on the first vacuuming.
Xl2 = tne amount of dust which was deposited at a loading of 400 mg/sq.ft.
prior to vacuuming
bj2 = the proportion of dust which was deposited at a loading of 400 mg/sq.ft.
which was picked up on the first vacuuming.
X2 = the amount of dust deposited two vacuumings prior to the end of the
present vacuuming. Similar definitions apply to X3, X4, etc.
b2 = the proportion of dust amount X2 which contributes to Y, similarly for
b3 and X3, b4, and X4 etc.
^6-10 - the total amount of dust deposited between the sixth and tenth
vacuuming prior to the present vacuuming. Similar definitions apply
toXll-20.
^6-10 = tne proportion of dust amount X6-10 which contributes to Y. Similar
definitions apply to bll-20.
A-33
-------
the proportion of dust deposited and ground in at 100 mg/sq.ft. picked
up on the first vacuuming.
the amount of dust deposited and ground in prior to vacuuming at a
loading of 100 mg/sq.ft.
g!2 = the proportion of dust deposited and ground in at 400 mg/sq.ft. picked
up on the first vacuuming.
Z\2 = the amount of dust deposited and ground in prior to vacuuming at a
loading of 400 mg/sq.ft.
813 = me proportion of dust deposited and ground in at 1,000 mg/sq.ft. picked
up on the first vacuuming.
Z]3 = the amount of dust deposited and ground in prior to vacuuming at a
loading of 1,000 mg/sq.ft.
614 = me proportion of dust deposited and ground in at 4,000 mg/sq.ft. picked
up on the first vacuuming.
Zj4 = the amount of dust deposited and ground in at a loading of 4,000
mg/sq.ft.
Z2 = the amount of dust deposited and ground in two vacuumings prior to
the end of the present vacuuming. Similar definitions apply to Z3, Z4,
etc.
g2 = the proportion of dust amount Z2 which contributes to Y, similarly for
g3 and Z3, g4, and Z4 etc.
RH = the relative humidity as a percent.
T\ = the effect of relative humidity on the weight gain measurement.
Temp = the temperature in degrees Fahrenheit.
T2 = the effect of temperature on the weight gain measurement.
DRH = the change in relative humidity from the previous to the current
weight gain measurement.
r3 = the effect of changes in relative humidity on the weight gain
measurement.
DTemp = the change in the temperature from the previous to the current weight
gain measurement.
T4 = the effect of changes in temperature on the weight gain measurement.
obs = the number of the vacuuming, 1 to 168, provides a measure of time.
t = the effect of time, as measured by the number of observations, on the
weight gain measurement.
Pi(l=l,7) = a classification variable used to indicate the test procedure used.
A-34
-------
The model has many terms which, in the end, turned out to be insignificant.
Because the objective of the modeling effort was to identify factors which affected
the weight gain measurement, rather than to fit a specific model or to identify a
parsimonious model for prediction, all terms which were initially thought impor-
tant were included.
A preliminary analysis indicated that the measurement variance was a function of
the weight of dust removed. Weighted regression was used to fit the data, where
the regression weights were proportional to the inverse of the estimated measure-
ment variance. The regression weights were determined based on the following
analysis. The log of the absolute value of the residuals is proportional to the log of
the standard deviation of the residuals. During the investigation of the variance of
the residuals, it was noted that the log of the absolute residuals were linearly related
to the log of the predicted values from the regression. This suggested that the
following model might be used:
Ln(Abs(residuals)+0.001) = c + d*Ln(predicted+0.001) [2]
The small value of 0.001 was used to make the distribution of the values closer to
normal and to reduce the influence of values which were very close to zero, perhaps
by chance. Using this model, the weights for regression were:
Wgt = l/(exp(c + d*Ln(predicted+0.001))A2 [3]
Because the predicted regression weights depend on the model fit to the data and the
regression weights used to fit the model, the following iterative procedures were
used to calculate the regression weights used in the final analysis:
(1) fitting the model (1) to the data using unweighted regression, saving the
residuals and predicted values.
(2) fitting model (2) to the residuals and predicted values using regression and
using the parameters to define the preliminary regression weights using
equation (3). Using the unweighted regression, some predicted values
were negative. In this case 0.02 rather than 0.001 was added to the
predicted values before taking the log.
(3) fitting the model (1) to the data using weighted regression, saving the
residuals and predicted values.
(4) identifying three outliers and fitting model (2) to the residual and
predicted values without using the outliers and using equation (3) to
calculate provisional regression weights.
(5) fitting the model (1) to the data without the outliers using weighted
regression, saving the residuals and predicted values.
A-35
-------
(6) fitting model (2) to the residual and predicted values without using the
outliers and calculating the final regression weights using equation (3)
where the predicted values in equation (3) come from the regression in
step (5) except for the outliers in which the predicted values from the
regression in step (3) are used.
Plots indicated that the procedure for calculating the regression weights had equal-
ized the variance of the weighted residuals. The regression weights varied substan-
tially, with the ratio of the largest to smallest weight being about 8,000. The large
variation in the regression weights indicates the importance of using weighted
regression.
The equation for the regression weights and the regression output was used to
calculate the coefficient of variation of one weight gain measurement as a function
of the magnitude of the weight gain in the vacuum cleaner bag during a 30-second
vacuuming. This relationship is shown in Figure A-8. Assuming that the amount
of dust applied can be measured with relatively little error, the coefficient of
variation of the recovery measurement is the same as for the weight gain in the bag.
Assuming that the weight gains from each of three successive vacuumings are
statistically independent, the coefficient of variation of the weight gain summed
across three vacuumings was also calculated and is plotted in Figure A-8.
The assumption that three successive weight gains are independent is probably not
true. However, there is no evidence from the data to support a lack of indepen-
dence. Nonetheless, the estimated coefficient of variation for weight gain from
three vacuumings should be considered approximate.
The model (equation (1)) provided a very good fit to the data, explaining over 99%
of the variance in the weight gain measurements. The only factors which were
statistically significant at the 5 percent level were those associated with the deposit of
dust and the deposit and grind-in of dust. No effects of temperature, relative
humidity, time, or test procedures were significant. After fitting the full model,
additional terms were added to determine if there were differences between
operators, if the residuals were significantly correlated, and if the change over time
might be represented better by a quadratic rather than a linear relationship. None of
these tests gave statistically significant results.
A-36
-------
0.3 -r
0.25 --
§
1
0.2 --
£ 0.15 --
g
»4
u
9
o>
a
0.1 -.
0.05 --
One 30 sec.
vacuuming
Three 30 sec.
vacuumings
(approximate)
0
0.5 1 1.5 2 2.5
Increase in vacuum bag weight (g)
Figure A-8 Estimated coefficient of variation of the weight change of a bag from
vacuum cleaner A as a function of the weight change
A-37
-------
Based on the parameter estimates from the model, the recovery, in each successive
30 seconds of vacuuming, of dust deposited, or deposited and ground in, at a loading
of 100 mg/sq.ft. is shown in Figure A-9. As can be seen in this figure, most of the
dust is recovered on the first vacuuming, with significantly reduced amounts
collected on each successive vacuuming. Compared to recovery for ground-in dust,
recovery for dust deposited on the carpet without any grind-in is greater for the first
vacuuming which picks up the loosest dust and significantly less for subsequent
vacuumings. The ground-in dust is more difficult to remove on the first vacuum-
ing, leaving more dust for removal on subsequent vacuumings.
Figure A-10 shows the recovery versus vacuuming effort using a log scale. The
shape of the curve suggests that the dust might be considered to have three
components, loose dust which is removed entirely in the first vacuuming, dust
which is gradually removed in successive vacuumings, and dust which is either not
removed using the vacuum cleaner or is otherwise lost from the carpet. The change
in the recovery with increasing vacuuming effort suggests that after the first
vacuuming, each successive vacuuming removes about half of the dust which can
be removed using the vacuum. This suggests that, for dust which is just deposited
on the carpet surface, roughly 16% is either not collected by the vacuum or is
otherwise lost, 4% is caught in the carpet (of which half is removed with each 30
seconds of vacuuming) and the remaining 80% is loose dust which is removed on
the first vacuuming. For dust which is ground in, these numbers are, 21% that is
either not collected or otherwise lost, 12% that is caught in the carpet (of which half
is removed with each 30 second vacuuming) and the remaining 68% that is loose
and is removed on the first vacuuming. Figure A-10 shows the slope of the
relationship between recovery and vacuuming effort which corresponds to
vacuuming 47% of the remaining dust which can be vacuumed on each successive
30 seconds of vacuuming.
The estimated cumulative dust recovery after many vacuumings is 84% (with 95
percent confidence interval from 80% to 87%) for dust deposited on the carpet and
79% (with 95 percent confidence interval from 74% to 85%) for dust ground into the
carpet. Although these recoveries are not statistically different, they suggest that
recovery of ground-in dust is lower than dust deposited without grind-in, consistent
with common sense.
Although differences in recovery after many vacuuming may not be statistically
significant, there are statistically significant differences in the measured recovery for
the first vacuuming as a function of the dust loading (i.e., weight of dust applied).
These recovery measurements are shown in Table A-3. In general, larger loadings
were correlated with higher recoveries on the first vacuuming.
A-38
-------
100% -T-
0
Deposit
Deposit with
grind-in
95% Confidence
interval
0
2345
Number of 30 second vacuumings
Figure A-9 Dust recovery versus vacuuming effort for dust deposited on the carpet
and dust deposited and ground into the carpet
A-39
-------
100.0% -r
o
tn
o
4-1
10.0% --
1.0% --
0.1%
0
47% removal
per vacuuming
Deposit, no
grind-in
Deposit with
grind-in
95% Confidence
interval
+
H 1 1 1 1
1 2 3 t 4 5 6
Number of 30 second vacuumings
Figure A-10 Dust recovery versus vacuuming effort for dust deposited with and
without grind-in compared to fixed removal per vacuuming
A-40
-------
Table A-3 Dust recovery on the first vacuuming as a function of the dust loading
Nominal dust loading
100 mg/sq.ft.
400 mg/sq.ft.
1,000 mg/sqJt
4,000 mg/sq.ft.
Dust deposited only
80.1%
82.5%
Dust deposited and
ground in
67.3%
66.7%
63.7%
72.6%
A-41
-------
The weight of dust recovered from the vacuum bags is shown in Figure A-ll, along
with the regression line relating the dust deposited to the dust collected from the
bag. Assuming that the detection limit for lead is 0.8 micrograms, the lead concen-
tration in the dust is 20 M-g/g, and the desired level in the sample is three times the
detection limit, then a dust sample with at least 0.120 grams of dust is required.
Using the regression relationship, a deposit of at least 1.25 grams of dust is required
to obtain 0.120 grams for the vacuum cleaner bag. This target weight of dust is also
shown in Figure A-ll.
The results of the analysis of Task 3 and 4 suggested that the carryover from test to
test is small, that two or three vacuumings will remove virtually all of the dust
which might be removed by vacuuming, and thus that using the same carpet
sample for successive tests was reasonable. The results also showed that roughly
three applications of dust would both shorten the test compared to the original
design and provide enough dust to measure the lead in the dust.
A4.5 Analysis of Task 5
In the exhaust emissions tests, Task 5, the dust concentration in the exhaust of the
vacuum cleaner was measured before, during, and after a known amount of dust
was picked up by the vacuum cleaner. The exhaust dust concentrations were
recorded as indicated by the output on the front of the instrument, at one-minute
intervals. The weight of dust captured in the vacuum bag was also measured. The
dust concentrations were converted to dust amounts to calculate the proportion of
the dust picked up by the vacuum which was in the exhaust.
To calculate the exhaust emissions from the vacuum cleaners, information includ-
ing the gas flow rate (in cubic feet per minute), the amount of dust applied to the
turntable (in grams), the exhaust emissions concentration (in mg per cubic meter,
recorded both on a strip chart and at one-minute intervals), and the initial and final
vacuum cleaner bag weights (in grams) are taken from the data forms for vacuum
cleaner emission tests. The strip chart and one-minute interval concentration
values are used to obtain exhaust emission estimates. The two methods, the
integration method using the strip chart, and trapezoid method using the one-
minute readings, produce somewhat different results. The conditions under which
one estimate is better than the other are discussed later.
A-42
-------
100 -,-
60
(0
I
10 --
i
CO
2 i --
*
o.i
• x
• Measurements
Regression
line
/ •
•Target amount
^"TP^T^T^j
10
Weight of dust deposited
100
Figure A-ll Weight of dust recovered from the vacuum cleaner bag as a function of
the weight of the dust deposited
A-43
-------
In the original plan for the pilot tests, the purpose of Task 5 was to evaluate the
possibility of measuring the lead content in the dust exhausted from the vacuum
cleaners. Preliminary measurements indicated that the exhaust levels were so low
that it would not be possible to sample enough dust to measure lead concentrations.
Therefore, dust measurements were not attempted. According to the original plan,
the purpose of the pilot exhaust tests was to evaluate the feasibility of taking the
exhaust measurements. At the time the tests were performed, it was convenient to
test all vacuums and, in effect, complete the tests planned for the full study. It had
been planned that dust passing through the 250 micron sieved would be used for the
pilot tests and that dust passing through the 53 micron sieve would be used for the
exhaust tests in the full study. By using dust which passed through the 53 micron
sieve for the pilot exhaust tests, it was not necessary to repeat the exhaust tests in the
full study.
The following information was used to calculate the exhaust emissions from the
vacuum cleaners: (1) the gas flow rate (in cubic feet per minute), (2) the amount of
dust applied to the turntable (in grams), (3) the exhaust emissions concentration (in
mg per cubic meter), recorded both on a strip chart and at one-minute intervals, and
(4) the initial and final vacuum cleaner bag weights (in grams) taken from the data
forms for vacuum cleaner emission tests. Two procedures were used to determine
the total amount of dust exhausted during the period before, during, and after the
dust pickup period: (1) integrating the exhaust dust concentrations recorded each
minute using approximating trapezoids and (2) integrating the continuous trace on
the strip chart recorder (integration method). The first method was more accurate
for the point in time when the emissions were recorded. The second allowed
estimation in periods where the emissions were fluctuating and the instantaneous
emission level was not representative of the average emissions level.
The integration method approximates the area under the strip chart curve by
totaling the number of squares under the curve. For each vacuum cleaner and
replicate, an area, in mg, associated with a small square on the strip chart is
calculated and multiplied by the number of small squares to estimate the exhaust
emissions. For example, vacuum cleaner A-replkate 1 had a flow rate of 75.03 cubic
feet per minute, or 2.1246 cubic meters per minute. The area (in mg) of one small
square, 0.05 minutes wide and 0.02 mg per cubic meter high, is then calculated as
0.02 mg/m3 * 2.1246 m3/min. * 0.05 min. = 0.00212 mg
For the first time interval, from 0 to 1 minutes, the area under the strip chart curve
consisted of 90 small squares. Therefore, the estimate of exhaust emissions for the
first time interval is 90 * 0.00212 mg = 0.191 mg. This method works better than the
trapezoid method for intervals covering larger areas and whose strip chart curves
cannot be accurately estimated with a straight line.
The trapezoid method assumes that the strip chart curve between time intervals can
be accurately estimated with a straight line. The area under the curve between two
A-44
-------
intervals, an estimate of exhaust emissions in that interval, can be calculated using
the trapezoid rule where the concentrations at each interval are the heights of the
trapezoids. For example, vacuum cleaner B replicate 1 had concentrations at the
second and third minutes of 0.037 and 0.034 mg per cubic meter, respectively. The
area of the resulting trapezoid is 0.0355 mg minutes per cubic meter. The area,
0.0355 mg minutes per cubic meter, multiplied by the gas flow rate, 2.269 cubic
meters per minute, equals 0.081 mg, the estimate of the exhaust emissions in the
first replication of vacuum cleaner B. This method works better than the integra-
tion method for intervals covering smaller areas and whose strip chart curves can
be accurately estimated by a straight line.
For each of the 13 tests (four tests for vacuum A and three tests each for vacuums B,
C, and D), the strip charts were divided into seven intervals (0-1 minutes, 1-2
minutes,..., 6-7 minutes) where the appropriate method for each interval and test
was used to estimate the exhaust emissions for that test in that interval. No one-
minute intervals were recorded for the first replication of vacuum cleaner A,
though, so estimates from the integration method were used. In addition, when
borderline cases arose, the exhaust emission estimates from the two methods were
similar. Thus, the potential errors resulting from choosing the inappropriate
method would be negligible.
Three questions are presented in the work plan for Task 5: How much do the
exhaust emissions change when dust is injected into each vacuum cleaner? The
exhaust emissions and dust not captured in the vacuum cleaner bags are what
percentage of total dust applied to the turntable for each vacuum cleaner? Where
are the peaks in the concentrations of exhaust emissions for the vacuum cleaners?
One speculation prior to the tests was that the exhaust emissions would peak early
in the test and subsequently decrease as the pores of the vacuum bag were plugged
by fine dust particles. The fluctuations in the emissions over time make testing of
this hypothesis difficult. When examining the strip charts, the larger exhaust dust
concentrations appeared in the beginning and then slowly tapered off, although
there were some unexplained late peaks for vacuum cleaner A. Although this
general trend could possibly be due to the fact that the larger dust particles were
clogging the bag, the late peaks are difficult to interpret. The emissions levels
fluctuated too much to make reasonable estimates of the rate at which the emissions
decreased after the peak. In general, evidence suggests that concentrations peak in
the first minute and decrease to near pre-injection levels during the last minute the
turntable is on. However, since the exhaust emissions were dose to ambient levels,
the concentration peaks are not likely to raise much concern.
The changes in the exhaust emissions resulting from injection of dust into the
vacuum cleaner differ significantly from vacuum cleaner to vacuum cleaner. For
example, the exhaust emissions from vacuum cleaners A and B while injecting dust
were slightly higher than exhaust emissions without injecting the dust. The
amount of exhaust emissions expelled from vacuum cleaner C was reduced by 25%
A-45
-------
while injecting dust into the vacuum cleaner. That is, it reduced dust levels in
ambient air, probably because it was equipped with a HEPA filter. Vacuum cleaner
D, the upright vacuum, had the largest average increase in exhaust emissions.
Exhaust emissions while injecting dust were six times larger than those when the
dust was not being injected. These average exhaust emission levels are shown in
Figure A-12 and Table A-4. The minute-by-minute averages are shown in Section
A6.5.
There are two possible ways to compute the amount of exhaust emissions. The first
and most straightforward is to measure the concentrations of dust in the exhaust as
measured by the dust emissions monitor. The second is to calculate the amount of
dust not captured in the bag. On average, 4% of the dust placed on the turntable was
not captured in the vacuum cleaner bag, with the percentages ranging from over 6%
for vacuum cleaner D to 3% for vacuum cleaners A, B, and C5 However, only about
0.01% of the dust placed on the turntable was expelled as exhaust emissions, with
the percentages ranging from 0.021% for vacuum cleaner D to less than 0.001% for
vacuum cleaner C. Across all vacuum cleaners, roughly 4% of the dust, that which
was not caught in the bag and not measured in the exhaust, has not been accounted
for. The missing dust may have adhered to the hose, the outside of the bag but
inside the vacuum cleaner (e.g., inside the machinery), or to the turntable.
D-3 is considered an outlier and is not used in these calculations
A-46
-------
Table A-4 Average exhaust concentrations for each vacuum cleaner exhaust test
Vacuum
cleaner
A
A
A
A
B
B
B
C
C
C
D
D
D
Bag
1
2
3
4
1
2
3
1
2
3
1
2
3
Ambient
air levels
0.004
0.009
0.006
0.004
0.013
0.012
0.012
0.017
0.012
Dust expelled (mg/m^) as
exhaust before, during and
after injection
Before
0.090
0.057
0.053
0.060
0.018
0.011
0.009
0.004
0.006
0.004
0.031
0.019
0.013
During
0.092
0.061
0.037
0.070
0.028
0.018
0.014
0.003
0.003
0.003
0.158
0.091
0.093
After
0.065
0.041
0.033
0.051
0.015
0.011
0.010
0.003
0.003
0.003
0.020
0.015
0.013
Dust as a percent of dust
place on turntable
Dust not
captured in
the bag
5.0%
2.4%
2.4%
3.7%
4.0%
2.8%
2.4%
4.7%
2.7%
2.4%
7.6%
4.9%
84.0%
Average for each vacuum cleaner
A
B
C
D
0.006
0.010
0.014
0.065
0.013
0.005
0.021
0.065
0.020
0.003
0.114
0.048
0.012
0.003
0.016
3.4%
3.0%
3.3%
6.3% *
Dust from
exhaust
emissions
0.020%
0.013%
0.008%
0.015%
0.006%
0.004%
0.003%
0.001%
0.001%
0.001%
0.021%
0.011%
0.012%
0.014%
0.005%
0.001%
0.015%
* Average excluding the outlier of 84%
A-47
-------
0.12 -r-
0.1 --
» 0-08
I
a
o
§
*-»
en
9
0.06 --
£ 0.04
0.02 --
0
Ambient
Vacuum
Before During
Time relative to dust injection
After
-0 D
Figure A-12 Average dust emissions before, during, and after injection of dust
A-48
-------
A5.0 DISCUSSION OF RESULTS
The results of the separate tasks which made up the pilot tests are summarized
below, along with associated changes for the tests in the full study.
Taskl
Task 1 showed that there appears to be a trend in the vacuum cleaner bag weights
consistent with the assumption that they are adjusting to the laboratory environ-
ment. Although there is evidence for a relative humidity effect, and possibly a
temperature effect, on the weight measurements, these effects were small compared
to the trend over time and were difficult to estimate because the temperature and
relative humidity stayed relatively constant during the tests.
The trend suggests that the time between stopping the vacuuming and weighing the
bag must be carefully controlled to minimize the measurement variance associated
with weighing an object whose weight is changing. An analysis of the data suggests
that adequate precision can be obtained by waiting three minutes between turning
off the vacuum and weighing the vacuum bag; little added precision is obtained by
waiting longer. Based on a preliminary analysis, the time between vacuuming and
weighing was set at five minutes for Tasks 3, 4, and 5. Following the more complete
analysis described here, and considering the time involved and flow of work in the
lab, MRI and Westat decided to set the time between vacuuming and weighing the
vacuum bag at three minutes in the full study.
The precision of the recovery measurements was better than originally anticipated
during the preparation of the QAPjP. Therefore, the study as originally designed
could easily achieve the data quality objectives. Due to subsequent budget considera-
tions, the number of tests planned for the full study was reduced. With the reduced
number of tests, it was anticipated that the original data quality objectives would be
achieved based on the precision attained in the pilot tests.
Task 2
In Task 2, the platform for holding the carpet samples performed well. The design
was not modified for the full study.
Task 3
The primary purpose of Task 3 was to assess the significance of carryover of dust
from one test to the next and to evaluate whether or not new carpet samples are
needed for each test.
In the first part of the task it was determined that the carpet could be vacuumed
such that few fibers would be picked up in each 40-second vacuuming and this
A-49
-------
might affect the recovery estimate. At the same time, it was determined that the
time to precondition each carpet was roughly a half hour.
The tests designed to identify significant carryover from test to test showed no
measurable carryover. Without carryover, recovery measurements can be made
using the same carpet sample for all, or many, tests. Use of the one carpet sample
for multiple tests also reduces the time required for each carpet test.
Task 4
The purpose of Task 4 was to determine how much dust was required. The tests
were set up also to determine if depositing all the dust at once gave results similar to
depositing the dust in multiple applications.
The analysis of the Task 4 data included the Task 3 data and allowed for estimation
of recovery as a function of vacuuming effort. The results of a regression analysis
indicated that the recovery on carpets was roughly 80% and that recovery was lower
when the dust was ground in than when the dust was simply deposited on the
carpet. Although there was some evidence of carryover, it was small after the third
30-second vacuuming. The dust removed from the vacuum bag for lead analysis,
when using three applications of dust at the low loading amount, was adequate for
the lead analysis.
Based on results from Task 4, for the full study it was decided to apply the dust in
three applications followed by three vacuumings without applying dust.
TaskS
The results of Task 5 indicated that the exhaust emissions from the vacuum
cleaners were low and that very little of the dust passed through the vacuum
cleaner bags. Because the tests specified for the full study were completed during the
pilot test, the exhaust tests will not be performed during the full study.
Other test procedure revisions
Due to continuing concern over the possibility that carpet fibers might affect the
measurements, particularly early in the study, that dust accumulated in the carpet
might affect later measurements, and that carryover might have some effect on the
results, it was decided to vacuum the carpet for each test prior to depositing the dust.
This initial vacuuming provides a measure of the dust or fibers which might bias
the recovery estimate. The results from this initial vacuuming will be used to
correct for any bias or carryover and make the statistical analysis simpler.
The decision to use the same carpet sample for many tests was further modified
such that for all substrates, four substrate samples were prepared. Each sample in
the full study is to be used for tests with one combination of dust loading and dust
A-50
-------
lead concentration. In effect, each substrate sample corresponds to a house with
either low or high dust lead concentrations and low and high dust loadings prior to
vacuuming. The use of the same substrate for all tests, with the associated dust lead
concentration and loading, corresponds roughly to the history of vacuuming in a
home where there is a sequence of dust depositions and vacuumings over time.
A-51
-------
A6.0 LABORATORY DATA
A6.1 TASK 1: Determine the stability of tare weights for new, clean vacuum
cleaner bags
Tables A-5, A-6, A-7 and A-8 consist of the data generated from Task 1. The description
of the data in each column of Table A-5 and A-6 is listed below:
Column Name Description
Time Number of minutes since the initial weighing
Bag A Weight Weight of the bag for vacuum cleaner A
Bag B Weight Weight of the bag for vacuum cleaner B
Bag C Weight Weight of the bag for vacuum cleaner C
Bag D Weight Weight of the bag for vacuum cleaner D
Cassette 1 Weight Weight of the first cassette used in the task
Cassette 2 Weight Weight of the second cassette used in the task
The description of the data in each column of Table A-7 and A-8 is listed below:
Date Date of the test
Time Time the test was performed
Relative Humidity Relative humidity at the beginning of the test
Temperature Temperature at the beginning of the test
Wt of Housevac A Bag Weight of the bag for vacuum cleaner A
Wt of Housevac B Bag Weight of the bag for vacuum cleaner B
Wt of Housevac C Bag Weight of the bag for vacuum cleaner C
Wt of Housevac D Bag Weight of the bag for vacuum cleaner D
Form No Form number - 1 signifies the vacuum cleaner bags were
weighed without running the vacuum cleaners and 2
signifies the vacuum cleaner bags were weighed after
running the vacuum cleaners for 1 minute
Bag Sequential number to distinguish between bags
Replication Replication of test on the same cassette
The procedure generating the data in table A-5 follows:
1) Run vacuum cleaner 40 sec with new bag
2) Repeat above 3 times with 1 minute between using same bag
3) Run vacuum cleaner for 40 sec
4) Immediately weigh bag, and record wt every minute for 10 min
A-52
-------
5) Repeat steps 1) to 4) once
6) Repeat steps 1) to 5) for each vacuum cleaner
The procedure generating the data in table A-6 follows:
1) Obtain cassette that has been acclimated to room for more than 24 hours
2) Weigh cassette - record wt every minute for 5 min
3) Remove top half of cassette; install blue nozzle sampler
4) Run sampler for 120 sec
5) Remove cassette; reinstall top half
6) Weigh cassette; record wt every minute for 5 min
7) Repeat items 3) to 6) one time
8) Repeat 1) to 7) with another cassette, once
The procedure generating the data in table A-7 follows:
1) Weigh each bag type 10 times, one half hour between
The procedure generating the data in table A-8
1) Insert bag
2) Run vacuum cleaner 5 min; discard bag
3) Weigh new bag and insert in vacuum cleaner
4) Run vacuum cleaner for 1 min, reweigh bag
5) Repeat run/weigh 10 times with same bag
A-53
-------
Table A-5: Tare weight of vacuum cleaner bags at one minute intervals for Task 1
Gravimetric Data
Time
(min)
0
1
2
3
4
5
6
7
8
9
10
0
1
2
3
4
5
6
7
8
9
10
Bag A
Weight (g)
37.547
37.544
37.544
37.542
37.542
37.541
37.541
37.541
37.539
37.540
37.539
36.755
36.754
36.752
36.752
36.752
36.752
36.752
36.752
36.752
36.751
36.750
BagB
Weight (g)
36.121
36.118
36.112
36.107
36.103
36.100
36.099
36.098
36.098
36.092
36.092
40.938
40.933
40.929
40.925
40.923
40.920
40.919
40.918
40.916
40.914
40.913
BagC
Weight (g)
30.309
30.303
30.303
30.302
30.300
30.301
30.301
30.301
30.302
30.300
30.300
30.221
30.218
30.219
30.221
30.221
30.222
30.224
30.226
30.226
30.227
30.228
BagD
Weight (g)
35.161
35.194
35.221
35.238
35.255
35.269
35.281
35.290
35.299
35.308
35.313
34.455
34.467
34.483
34.505
34.511
34.518
34.527
34.536
34.542
34.548
34.553
Bag
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
2
A-54
-------
Table A-6: Tare weight of sampling cassette at one minute intervals for Task 1
Gravimetric Data
Time
(min)
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
Replication
1
1
1
1
1
1
2
2
2
2
2
2
3
3
3
3
3
3
Cassette 1
Weight (g)
19.4825
19.4822
19.4821
19.4820
19.4820
19.4820
19.4852
19.4841
19.4837
19.4837
19.4831
19.4829
19.4840
19.4829
19.4823
19.4820
19.4817
19.4814
2 Weight
fe)
19.3281
19.3260
19.3253
19.3248
19.3245
19.3242
19.3222
19.3210
19.3205
19.3201
19.3198
19.3195
19.3211
19.3200
19.3194
19.3191
19.3189
19.3186
A-55
-------
Table A-7: Tare weighht of vacuum cleaner bags at 30 minute interrvals for Task 1
Gravimetric Data
Date
6/24/93
6/24/93
6/24/93
6/24/93
6/24/93
6/24/93
6/24/93
6/24/93
6/24/93
6/24/93
Relative
Time Humidity (%)
11:00
11:30
12:00
12:30
13:00
13:30
14:00
14:30
15:00
15:30
47.0
47.6
47.6
47.6
48.5
47.9
47.7
48.2
48.3
46.6
Weight of
Temperature Housevac
(F) ABag(g)
69.8
70.2
70.3
70.4
70.7
70.6
71.0
71.2
70.9
70.8
35.275
35.288
35.298
35.292
35.301
35.296
35.291
35.296
35.296
35.294
Weight of
Housevac
B Bag (g)
41.010
41.024
41.042
41.046
41.053
41.050
41.042
41.040
41.037
41.024
Weight of
Housevac
CBag(g)
30.462
30.467
30.472
30.470
30.481
30.477
30.477
30.479
30.480
30.472
Weight of
Housevac
DBag(g)
41.390
41.406
41.411
41.410
41.418
41.416
41.414
41.417
41.413
41.403
A-56
-------
Table A-8: Tare weight of vacuum cleaner bags after one minute of use, for Task 1
Gravimetric Data
Date
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
6/25/93
Time
13:32
13:44
13:59
14:11
14:28
14:43
14:56
15:10
15:28
15:43
15:57
13:35
13:47
14:01
14:14
14:31
14:45
14:58
15:12
15:30
15:45
15:59
13:37
13:49
14:03
14:17
14-33
14:47
15:00
15:14
15:32
15:47
16:02
13:39
13:51
14:05
14:21
14:35
14:49
15:03
15:16
15:36
15:50
16:05
(%)
40.2
39.9
39.8
39.6
39.7
39.5
39.4
39.2
39.3
39.4
39.4
40.2
40.0
39.7
39.6
39.6
39.6
39.5
39.3
39.2
39.5
39.3
40.2
39.9
39.8
39.8
39.5
39.6
39.5
39.4
39.4
39.6
39.3
40.3
39.8
39.7
39.8
39.6
39.7
39.4
39.4
39.4
39.5
39.3
(F)
70.8
71.0
71.0
71.2
71.1
71.3
71.4
71.6
71.6
71.6
71.6
70.8
71.1
71.2
71.1
71.2
71.2
71.3
71.5
71.6
71.5
71.6
70.8
71.1
71.1
71.1
71.2
71.2
71.3
71.5
71.5
71.5
71.5
70.8
71.1
71.1
71.0
71.1
71.2
71.3
71.4
71.4
71.4
71.5
Weight of Housevac Bag (g)
36.005
35.972
35.952
35.926
35.915
35.913
35.905
35.900
35.902
35.891
35.892
41.656
41.617
41.599
41.588
41.575
41.570
41.564
41.562
41.563
41.552
41.550
30.250
30.246
30.241
30.234
30.238
30.238
30.235
30.237
30.237
30.237
30.235
34.710
34.530
34.465
34.443
34.414
34.394
34.376
34.371
34.358
34345
34.340
Housevac
A
A
A
A
A
A
A
A
A
A
A
B
B
B
B
B
B
B
B
B
B
B
C
C
C
C
C
C
C
C
C
C
C
D
D
D
D
D
D
D
D
D
D
D
A-57
-------
A6.2 TASK 2: Demonstrate method of securing carpet and upholstery
substrates for testing vacuum cleaners
A 6-in high rectangular table( 1.83 m x 0.69 m) with the top of the table made from 3/4-
in thick exterior grade plywood (per ASTM Method F608-89) was constructed. A
mechanism was provided for securing carpet section s at the corners of the table and for
securing upholstery sections along the entire length at both ends. A pad was placed
underneath the carpet sections, and a 1/2-in thick foam pad was placed underneath the
upholstery sections.
The suitability of the table for vacuum cleaner tests was determined by vacuuming a
carpet section with the upright vacuum (vacuum cleaner D) and with one of the canister
vacuum cleaners with beater bar. Similarly, an upholstery section was vacuumed with
the same two vacuum cleaners using the proper upholstery attachment.
The table was tried out on June 30,1993. The method of securing substrates to the table
worked well for carpet sections, However, upholstery sections had to be clamped along
one end and then stretched tight before clamping onto the table at the opposite end.
Not doing so allowed the upholstery section to ripple up in front of the vacuum cleaner
nozzle. Two pieces of channel were cut for proper securing of upholstery.
A-58
-------
A6.3 TASK 3: Determine if preconditioning procedures are feasible for using
new carpet in the laboratory tests, and determine if the use of the same
substrate for each series of tests is feasible
Tables A-9 and A-10 consist of the data generated from Task 3. The description of the
data in each column of Table A-9 is listed below:
Column Name Description
Time Time the test was performed
Bag Wt Weight of the vacuum cleaner bag
RH Relative humidity at the time the bag weight was observed
Temp Temperature at the time the bag weight was observed
The description of the data in each column of Table A-10 is listed below:
Vac Run Order of the vacuuming
Increase Increase in the bag weight from the vacuuming
RH Relative humidity at the time the bag weight was observed
Temp Temperature at the time the bag weight was observed
Bag No Sequential number to distinguish between the bags
Measurement Type Type of dust retrieval used • 1 signifies the wand was
vacuumed after every ninth test/ 2 signifies the wand was
vacuumed after every third test/ and 3 signifies the wand
was vacuumed as a part of each test
Dust Amount Dust amount deposited - 1 signifies 100 mg/sq ft and 2
signifies 400 mg/sq ft
Dust Despot Amount of dust deposited on the substrate
Dust from Wand Amount of dust vacuumed from the wand (except for
measurement type 3)
Time Time the test was performed
The procedure for generating the data in table A-9
1) Vac entire carpet 5 min; discard bag
2) Weigh new bag/ vac for 5 min
3) Reweigh same bag; vac 5 min
4) Repeat reweigh/vac 10 times
The procedures for generating the data in table A-10
1) Vacuum wand and brush; discard bag
A-59
-------
2) Insert new tared bag
3) Apply dust; vac for 30 sec, wait 5 min, reweigh bag
4) Vac for 30 sec, wait 5 min, reweigh bag
5) Vac for 30 sec, wait 5 min, reweigh bag
6) Repeat items 2) through 5) above
7) Repeat items 2) through 5) again
8) Vacuum off wand and brush
9) Reweigh Bag
A-60
-------
Table A-9: Fiber conditioning, Task 3
Gravimetric Data
Time
15:55
16:06
16:19
16:32
16:44
16:56
17:10
17:21
17:32
17:43
17:54
18:01
18:11
18:22
18:35
33:01
Bag
Wtfe)
35.572
36.237
36.705
36.900
37.038
37.202
37.215
37.231
37.247
37.262
37285
37.316
37.350
37.372
37.384
37.433
RH (%)
52.6
52.4
52.2
52.0
51.6
51.6
51.1
50.9
50.9
50.8
50.7
50.3
50.2
50.0
49.8
43.9
Temp (F)
73.5
73.6
73.6
73.7
73.7
73.8
73.8
73.8
73.8
74.0
74.1
74.1
73.8
73.5
73.4
71.4
A-61
-------
Table A-10: Dust carry-over test for Task 3
Gravimetrics Data
Vac Increase
Run (g)
1st vac
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
1st vac
0.619
0.011
0.006
0.383
0.018
0.021
0.571
0.024
0.003
2.270
0.061
0.012
2.252
0.033
0.007
2.021
0.046
0.008
0.592
0.020
0.002
0.533
0.022
0.000
0.532
0.016
0.007
2.257
0.061
0.036
2.278
0.096
0.019
2.272
0.084
0.030
0.548
0.008
0.014
0.559
0.017
0.002
0.572
RH (%)
44.7
44.5
44.5
44.4
44.4
44.8
44.4
44.5
44.3
44.3
44.2
44.0
43.7
43.6
43.4
43.4
43.1
42.9
42.5
42.4
42.6
42.3
42.4
42.2
42.0
42.3
42.0
52.4
52.9
52.5
52.9
53.0
53.0
53.2
53.0
53.0
51.8
51.9
51.3
49.9
49.5
48.8
49.2
Temp
(F)
74.4
74.3
74.2
74.2
74.3
74.3
74.4
74.5
74.4
72.3
72.3
72.4
72.5
72.6
72.6
72.8
72.8
72.8
73.2
73.2
73-3
73.5
73.4
73.5
73.6
73.6
73.6
73.9
74.1
74.1
74.5
74.6
74.6
74.8
74.9
75.0
75.3
75.3
75.5
75.5
75.5
75.6
75.8
Bag Measurement Dust
No Type Amount
1
1
1
2
2
2
3
3
3
4
4
4
5
5
5
6
6
6
7
7
7
8
8
8
9
9
9
10
10
10
11
11
11
12
12
12
13
13
13
14
14
14
15
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
,2
2
2
1
1
1
1
1
1
1
Dust
Deposit (g)
0.726
0.000
0.000
0.685
0.000
0.000
0.682
0.000
0.000
2.687
0.000
0.000
2.724
0.000
0.000
2.618
0.000
0.000
0.677
0.000
0.000
0.661
0.000
0.000
0.671
0.000
0.000
2.675
0.000
0.000
2.690
0.000
0.000
2.729
0.000
0.000
0.657
0.000
0.000
0.665
0.000
0.000
0.671
Dust from
Wand (g)
0
0
0
0
0
0
0
0
0.02
0
0
0
0
0
0
0
0
0.04
0
0
0.004
0
0
0.017
0
0
0.017
0
0
0.018
0
0
0.03
0
0
0.033
0
0
0
0
0
0
0
Time
14:20
14:25
14:30
14:59
15:04
15:08
15:28
15:36
15:44
11:02
11:08
11:15
11:30
11:36
11:43
11:57
12:04
12:11
14:48
14:54
15:00
15:28
15:34
15:41
16:06
16:13
16:20
9:37
9:43
9:50
10:12
10:19
10:25
10:49
10:56
11:03
12:59
13:08
13:15
13:42
13:51
14:01
14:28
A-62
-------
Table A-10: Dust carry-over test for Task 3 (continued)
Gravimetrics Data
Vac Increase
Run (g)
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
1st vac
2nd vac
3rd vac
0.004
0.010
2.229
0.041
0.029
2.295
0.064
0.024
2.277
0.097
0.048
RH (%)
48.5
48.4
48.5
48.4
48.3
48.4
48.3
48.5
47.6
47.3
47.1
Temp
(F)
75.8
75.9
76.1
76.0
75.9
76.1
76.2
76.2
76.2
76.2
76.2
Bag Measurement Dust
No Type Amount
15
15
16
16
16
17
17
17
18
18
18
3
3
3
3
3
3
3
3
3
3
3
1
1
2
2
2
2
2
2
2
2
2
Dust
Deposit (g)
0.000
0.000
2.627
0.000
0.000
2.706
0.000
0.000
2.710
0.000
0.000
Dust from
Wand (g) Time
0
0
0
0
0
0
0
0
0
0
0
14:36
14:44
15:10
15:18
15:26
15:51
16:01
16:10
16:34
16:42
16:49
A-63
-------
A6.4
Determine the amount of dust needed for the tests
Table A-ll consists of the data generated from Task 4. The description of the data in
each column of Table A-ll is listed below:
Column Name
Run No
Date
Time
Diff
RH
Temp
Bag No
Meas Type
Dust Amount
Amount Deposit
Amount Ground-in
Amount Picked by Wand
Description
Order of the vacuuming within each series of tests
Date of the test
Time of the test
Increase in bag weight from the previous vacuuming
Relative humidity at the time the bag weight was
observed
Temperature at the time the bag weight was observed
Sequential number to distinguish between vacuum
cleaner bags
Procedure used to perform the test - a description of each
procedure is at the bottom of the page
Dust amount deposited - 1 signifies 100 mg/sq ft and 2
signifies 400 mg/sq ft
Amount of dust deposited on the substrate and not
groundin to the substrate
Amount of dust ground-in to the substrate
Amount of dust vacuumed from the wand (except for
measurement type 3)
The procedure for generating the data in Table A-ll with a measurement type of 4
follows:
1) Vacuum carpet 3 times, discard bag
2) Warm up vacuum cleaner for 30 seconds
3) Insert new tared bag
4) Apply dust and embed
5) Vacuum 30 sec; wait 5 minutes; reweigh bag
6) Repeat steps 4) and 5), a total of 10 times
7) Recover dust from bag, and weigh dust (not bag)
8) Repeat all above, one time
The procedure for generating the data in Table A-ll with a measurement type of 5
follows:
1) Vacuum carpet 3 times, discard bag
2) Warm up vacuum cleaner for 30 seconds
A-64
-------
3) Insert new tared bag
4) Apply dust and embed
5) Vacuum 30 sec; wait 5 minutes; reweigh bag
6) Repeat step 5), a total of 10 times
7) Repeat all above, one time
The procedure for generating the data in Table A-ll with a measurement type of 6
fr»11rrtA7c-
follows:
1) Vacuum carpet 3 times, discard bag
2) Warm up vacuum cleaner for 30 seconds
3) Insert new tared bag
4) Apply dust and embed
5) Vacuum 30 sec; wait 5 minutes; reweigh bag
6) Repeat steps 4) and 5), a total of 3 times
7) Vac and weigh 3 more times
8) Recover dust from bag, and weigh dust (not bag)
9) Repeat all above, one time
A-65
-------
Table A-ll: Testing different amounts of dust and vacuuming the wand for Task 4
Gravimetrics Data
Run
No
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
1
2
3
Date
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/9/93
7/12/93
7/12/93
7/12/93
7/12/93
7/12/93
7/12/93
7/12/93
7/12/93
7/12/93
7/12/93
7/12/93
7/12/93
7/12/93
Time
920
932
9:43
9-52
10:01
10:09
10:19
I(h31
10-39
10:48
13:01
13:11
13:20
13:29
1339
13:50
1359
1420
1431
14:40
15:49
1557
16:05
16:11
16:18
16-25
1634
16:41
16:48
1655
10:14
10:24
1035
11:02
11:11
11:21
1129
11:40
1152
12.-04
16:05
16:14
1621
Difffe)
0.480
0.488
0511
0539
0510
0.498
0509
0.468
0508
0503
0.460
0.499
0.493
0.435
0.466
0.466
0.498
0.433
0.498
0.484
1.741
1.862
1592
£028
2D60
2.492
2215
1.988
2j039
2231
1.838
1.923
1.912
1.974
2j012
2245
2319
Z019
2.118
2J023
4.419
0.616
0.232
RH
(%)
515
513
512
512
51.0
515
513
515
512
512
51.6
52.1
51.4
50.9
513
515
503
505
50.7
50.8
512
52J)
51.0
51.1
512
51.0
512
50.8
51.8
51.9
48.8
49.0
48.9
48.7
483
48.9
485
482
482
485
SIX)
55.0
56.0
Temp
(F) Bag No
75.8
75.8
76.0
76.1
76.1
762
762
763
76.2
76.4
76.8
76.8
76.8
76.7
77.0
77.0
77.1
77.4
775
77.7
78.0
78.1
77.9
77.9
77.9
78.0
78.0
78.1
78.1
78.0
75.1
752
75.4
75.9
75.9
76.1
76.1
762
763
76.4
76.9
77.0
77.0
19
19
19
19
19
19
19
19
19
19
20
20
20
20
20
20
20
20
20
20
21
21
21
21
21
21
21
21
21
21
22
22
22
22
22
22
22
22
22
22
23
23
23
Meas Dust
Type Amount
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4'
4
4
4
4
4
4
5
5
5
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
3
3
3
Amount
Deposit
(K)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Amount
Ground-
in fe)
0.674
0.643
0.649
0.641
0.653
0.648
0.626
0.663
0.631
0640
0.631
0.688
0.673
0.615
0.639
0.639
0.646
0.634
0.659
0.631
2570
2.622
1.835
2J22
2.753
3.482
2.664
2.629
2504
2.614
2732
1.861
2457
2-674
2594
2.606
2.710
2.671
2,629
2.618
6.662
0.000
0.000
Amount
Picked by
Wand(g)
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
A-66
-------
Table A-ll: Data from Task 4 (continued)
Gravimetrics Data
Run
No
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
Date
7/12/93
7/12/93
7/12/93
7/12/93
7/12/93
7/12/93
7/12/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/13/93
7/14/93
7/14/93
7/14/93
7/14/93
7/14/93
7/14/93
Time
16:31
1638
16:44
1651
1657
17:03
17:09
958
10:05
10:11
10:19
10:25
I
-------
Table A-ll: Data from Task 4 (continued)
Gravimetrics Data
Run
No
1
2
3
4
5
6
7
8
9
10
Date
7/14/93
7/14/93
7/14/93
7/14/93
7/14/93
7/14/93
7/14/93
7/14/93
7/14/93
7/14/93
Time
14:48
14:56
15:04
15:11
15:18
15:25
15:32
15:38
15:45
15:58
DiffQr)
0.503
0.533
0.573
0.081
0.055
0.034
0.032
0.030
0.021
0.026
RH
(%)
53.2
53.0
53.5
53.0
52.9
53.0
52.9
53.0
52.7
52.8
Temp
(F) Bag No
77.1
77.2
77.2
77.3
77.3
77.4
77.5
77.5
77.6
77.6
28
28
28
28
28
28
28
28
28
28
Meas Dust
Type Amount
6
6
6
6
6
6
6
6
6
6
1
1
1
1
1
1
1
1
1
1
Amount Amount
Deposit Ground-
(K) in (g)
0
0
0
0
0
0
0
0
0
0
0.657
0.669
0.675
0.000
0.000
0.000
0.000
0.000
0.000
0.000
Amount
Picked by
Wandfe)
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
A-68
-------
A6.5 Develop and demonstrate a method for measuring exhaust emissions
Table A-12 consists of the data generated from Task 5. The description of the data in
each column of Table A-12 is listed below:
Column Name
Vac
Rep
Date
BagWt
Net Dust Wt
Gas Flow Rate
Rate Selected
Nozzle Size
Time
Particulate Cone
Final Bag Wt
Description
Vacuum cleaner used in the test (either A, B, C or D)
Replication within vacuum cleaner
Date the test was performed
Initial weight of the vacuum cleaner bag
Amount of dust deposited on the turntable
Actual flow rate of the air leaving the pitot tube
Selected flow rate of the air leaving the pitot tube
Diameter of the nozzle
Time the test was performed
Concentration of the dust leaving the encased vacuum
cleaner
Vacuum cleaner bag weight after all replications had been
completed
A-69
-------
Table A-12: Exhaust emissions data for Task 5
Gravimetric Data
Vac
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
B
B
B
B
B
B
B
B
B
Rep
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
1
1
1
1
1
1
1
1
2
Date
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/27/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
BagWt
(B)
36.756
36.756
36.756
36.756
36.756
36.756
36.756
36.756
35.895
35.895
35.895
35.895
35.895
35.895
35.895
35.895
35.452
35.452
35.452
35.452
35.452
35.452
35.452
35.452
36.298
36.298
36.298
36.298
36.298
36.298
36.298
36.298
36.092
36.092
36.092
36.092
36.092
36.092
36.092
36.092
40.869
Net Dust Gas Flow
Wt (g) Rate (acrni)
4.9744
4.9744
4.9744
4.9744
4.9744
4.9744
4.9744
4.9744
5.1020
5.1020
5.1020
5.1020
5.1020
5.1020
5.1020
5.1020
5.0486
5.0486
5.0486
5.0486
5.0486
5.0486
5.0486
5.0486
4.9632
4.9632
4.9632
4.9632
4.9632
4.9632
4.9632
4.9632
5.0541
5.0541
5.0541
5.0541
5.0541
5.0541
5.0541
5.0541
5.0816
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
75.03
80.16
80.16
80.16
80.16
80.16
80.16
80.16
80.16
80.16
Rate Selected Nozzle
(L/min) Size (in)
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
J 0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
Time Particulate Cone Final Bag
(min) (mg/cu m) Wt (g)
0
1
2
3
4
5
6
7
0
1
2
3
4
5
6
7
0
1
2
3
4
5
6
7
0
1
2
3
4
5
6
7
0
1
2
3
4
5
6
7
0
0.021
0.066
.
.
.
'
0.080
0.042
0.038
0.036
0.037
0.076
0.020
0.041
0.078
0.040
0.035
0^32
0.033
0.035
0.032
0.035
0.049
0.128
0.171
0.062
0.055
0.045
0.043
0.060
0.015
0.018
0.037
0.034
0.030
0.020
0.016
0.014
0.011
41.481
41.481
41.481
41.481
41.481
41.481
41.481
41.481
40.877
40.877
40.877
40.877
40.877
40.877
40.877
40.877
40.381
40.381
40.381
40.381
40.381
40.381
40.381
40.381
41.077
41.077
41.077
41.077
41.077
41.077
41.077
41.077
40.944
40.944
40.944
40.944
40.944
40.944
40.944
40.944
45.812
A-70
-------
Table A-12: Exhaust emissions data for Task 5 (continued)
Gravimetric Data
Vac
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
D
D
Rep Date
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
1
1
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/28/93
7/29/93
7/29/93
BagWt
(g)
40.869
40.869
40.869
40.869
40.869
40.869
40.869
42.089
42.089
42.089
42.089
42.089
42.089
42.089
42.089
30.210
30.210
30.210
30.210
30.210
30.210
30.210
30.210
30.273
30.273
30.273
30.273
30.273
30.273
30.273
30.273
31.111
31.111
31.111
31.111
31.111
31.111
31.111
31.111
34.383
34.383
Net Dust
Wt(K)
5.0816
5.0816
5.0816
5.0816
5.0816
5.0816
5.0816
5.0012
5.0012
5.0012
5.0012
5.0012
5.0012
5.0012
5.0012
5.1301
5.1301
5.1301
5.1301
5.1301
5.1301
5.1301
5.1301
5.0403
5.0403
5.0403
5.0403
5.0403
5.0403
5.0403
5.0403
5.0578
5.0578
5.0578
5.0578
5.0578
5.0578
5.0578
5.0578
4.9777
4.9777
Gas Flow
Rate (acfm)
80.16
80.16
80.16
80.16
80.16
80.16
80.16
80.16
80.16
80.16
80.16
80.16
80.16
80.16
80.16
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
67.85
45.81
45.81
Rate Selected Nozzle
(L/min) Size (in)
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.240
0.297
0.297
Time Particulate Cone Final Bag
(min) (mg/cu m) Wt (g)
1
2
3
4
5
6
7
0
1
2
3
4
5
6
7
0
1
2
3
4
5
6
7
0
1
2
3
4
5
6
7
0
1
2
3
4
5
6
7
0
1
0.013
0.024
0.017
0.017
0.022
0.011
0.010
0.008
0.010
0.023
0.015
0.011
0.013
0.010
0.009
0.004
0.003
0.003
0.003
0.003
0.003
0.003
0.003
0.008
0.003
0.003
0.003
0.003
0.003
0.003
0.003
0.005
0.003
0.002
0.003
0.003
0.003
0.003
0.003
0.012
0.044
45.812
45.812
45.812
45.812
45.812
45.812
45.812
46.972
46.972
46.972
46.972
46.972
46.972
46.972
46.972
35.098
35.098
35.098
35.098
35.098
35.098
35.098
35.098
35.176
35.176
35.176
35.176
35.176
35.176
35.176
35.176
36.049
36.049
36.049
36.049
36.049
36.049
36.049
36.049
38.980
38.980
A-71
-------
Table A-12: Exhaust emissions data for Task 5 (continued)
Gravimetric Data
Vac
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
Rep
1
1
1
1
1
1
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
Date
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
7/29/93
BagWt
(K)
34383
34383
34.383
34383
34383
34.383
34.412
34.412
34.412
34.412
34.412
34.412
34.412
34.412
34593
34.593
34.593
34.593
34.593
34.593
34.593
34.593
Net Dust Gas Flow
Wt (g) Rate (acfm)
4.9777
4.9777
4.9777
4.9777
4.9777
4.9777
5.1199
5.1199
5.1199
5.1199
5.1199
5.1199
5.1199
5.1199
5.0122
5.0122
5.0122
5.0122
5.0122
5.0122
5.0122
5.0122
45.81
45.81
45.81
45.81
45.81
45.81
45.81
45.81
45.81
45.81
45.81
45.81
45.81
45.81
45.81
45.81
45.81
45.81
45.81
45.81
45.81
45.81
Rate Selected Nozzle
(L/min) Size (in)
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
0.297
0.297
0.297
0.297
0.297
0.297
0.297
0.297
0.297
0.297
0.297
0.297
0.297
0.297
0.297
0.297
0.297
0.297
0.297
0.297
0.297
0.297
Time Particulate Cone Final Bag
(min) (mg/cum) Wt(g)
2
3
4
5
6
7
0
1
2
3
4
5
6
7
0
1
2
3
4
5
6
7
0.266
0.152
0.210
0.136
0.021
0.022
0.018
0.020
0.183
0.080
0.098
0.068
0.015
0.016
0.013
0.017
0.180
0.128
0.093
0.170
0.013
0.013
38.980
38.980
38.980
38.980
38.980
38.980
39.282
39.282
39.282
39.282
39282
39.282
39.282
39.282
36.396
36.396
36.396
36.396
36.396
36.396
36.396
3o*39o i
A-72
-------
APPENDIX B: PRECONDITIONING DATA
B1 Fiber Preconditioning
Fiber preconditioning was performed to remove loose fibers from the carpet and
upholstery samples which might adversely affect the measurements of dust recovery
and lead concentration. The fiber preconditioning procedures are discussed in Section
4.3. Fiber preconditioning for carpets and upholstery were analyzed separately and are
discussed in the following two subsections.
B1.1 Fiber Preconditioning on Carpets
The preconditioning tests determined the increase in weight of the vacuum cleaner bags
when vacuuming the substrates for either 5 minute or, in a few cases, 40 seconds. The
fiber preconditioning was performed on 8 carpet samples using all four vacuum
cleaners. The data sheets identified each vacuum cleaner used and the substrate
section. The number of vacuumings varied among the substrate samples and in some
cases 40 second vacuumings were used to estimate fiber recovery under the conditions
used in the study tests. Therefore, the data were summarized by calculating the
cumulative weight gain for each five minute period of vacuuming (i.e. the period from 0
to 5 minutes, 5 to 10 minutes, 10 to 15, minutes of vacuuming, etc.). These estimates of
five-minute weight gain were analyzed as a function of cumulative vacuuming time
and other factors.
During the analysis, 5 outliers were identified, all using vacuum cleaner D. These
outliers include the following points:
Planned Substrate usage
Substrate Dust loading Nominal lead Vacuum Cumulative minutes
(mg/sqft) concentration cleaner of vacuuming
Carpet 400 High D 80
Carpet 400 High D 140
Carpet 400 High D 160
Carpet with 400 Low D 65
Ground-in dust
Carpet with 400 High D 20
Ground-in dust
Possible problems with the first reading of the day using vacuum cleaner D (the
upright) had been noted. However, even after correcting for possible effects associated
with the first vacuuming using vacuum cleaner D, there observations appeared to be
distinct outliers (Because the analysis of the preconditioning data was not central to the
study, to save time formal outlier tests were not performed).
B-l
-------
A preliminary model was fit to the data with terms for the interaction between vacuum
cleaner and substrate sample, minutes of vacuuming on the substrate sample, and
indicator to identify the first measurement of the day using vacuum cleaner D. The
least square estimates of the mean recovery versus cumulative vacuuming time are
shown in Figure B-l. The preliminary analysis clearly indicates that the five-minute
weight gain due to fibers from carpets decreases substantially within the first 20
seconds of vacuuming, after which it remains relatively constant over the next four
hours of vacuuming.
Due to the variability in the weight gain measurements in the first 20 minutes of
vacuuming among carpet samples and the change in 5-minute weight gain with time at
the beginning of the preconditioning, the weight gain measurements in the first 20
minutes of vacuuming (the first vacuuming with each vacuum cleaner) were excluded
from the final analysis of the fiber preconditioning data.
The final model fit a separate linear trend to the 5-minute weight gain as a function of
cumulative vacuuming time for each substrate sample. The final model also had terms
for differences among combinations of vacuum cleaner and substrate samples and the
first vacuuming of the day using vacuum cleaner D. A weighted analysis was used,
with the regression weights being a function of the substrate sample. All terms were
highly significant (p < 0.0001) except for the differences among slopes for the different
substrate samples (p = 0.0383). Weight gains were highest for vacuum cleaner D. There
was no evidence for serial correlation among the residuals.
In the full study, the vacuum cleaner tests use 40 second vacuumings rather than the 5-
minute vacuumings using in most of the fiber preconditioning. The final model was
used to predict the fiber uptake in 40 seconds of vacuuming which might be seen in the
full study by dividing the predicted 5-minute weight gain by 7.5. The predicted weight
gain due to fibers is shown in Table 8-1, broken down by substrate, dust loading and
nominal dust lead concentration (i.e., by individual substrate sample) and by the
vacuum cleaner used on the sample. These values were used as possible covariates in
the analysis of data from the vacuum cleaner and sampler tests.
B-2
-------
1.8 -r
o>
c
I 1.6
o
CO
o 1.4 --
CO
o
in
ww
3
•o
2
o
E
0
oc
1.2 --
1 - -
0.8 --
3 0.6 - -
CO
Q
O
S> 0.4
o
O)
1 0.2
\
•••. /'"^••
/ V
+
+
50 100 150
Total Minutes of Vacuuming
200
250
Figure B-l Five minute weight gain due to fibers versus cumulative vacuuming time
B-3
-------
B1.2 Fiber Preconditioning on Upholstery
The data provided by the preconditioning are the increase in weight of the vacuum
cleaner bags when vacuuming the upholstery samples for 5 minutes. The fiber
preconditioning was performed on 4 upholstery samples using all four vacuum
cleaners. The data included indicators for the vacuum cleaner and substrate sample.
Two samples were vacuumed for a total of 100 minutes and two for a total of 120
minutes. Three outliers, negative weight gains all of which were the from the first
vacuuming of the day using vacuum cleaner D, were removed for the preliminary
analysis. After determining that regression weights would improve the model, a fourth
outlier was identified based on the weighted analysis (from the second vacuuming of
the day using vacuum cleaner D). These outliers are summarized below.
Planned Substrate Usage
Substrate Dust Nominal lead Vacuum Cumulative minutes
loading concentration cleaner of vacuuming
(mg/sq ft)
Upholstery 100 High D 45
Upholstery 400 Low D 15
Upholstery 400 Low D 75
Upholstery 400 High D 80
Preliminary analysis, using two way analysis of variance, indicated that the weight gain
due to fibers depended on the cumulative vacuuming time and the vacuum cleaner
used. The predicted average five-minute weight gain (with its 95% confidence interval)
as a function of time is shown in Figure B-2 (using dark circles). Except for the last four
vacuumings, performed only on two of the four substrates, the weight gain appears to
follow a decreasing curve. No reason has been found to explain the apparent change
after 100 minutes of vacuuming. To provide balanced data for the analysis, only the
data for the first 100 minutes of vacuuming were used in the final analysis.
B-4
-------
Table B-l Predicted 40-second fiber uptake from carpets by substrate sample and
vacuum cleaner
Planned Substrate Usage
Substrate
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Carpet w Grind-in
Vacuum
cleaner
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
Dust loading
(mg/sq ft)
400
400
400
400
400
400
400
400
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
400
400
400
400
400
400
400
400
Nominal dust
lead
concentration
High
High
High
High
Low
Low
Low
Low
Low
Low
Low
Low
High
High
High
High
Low
Low
Low
Low
High
High
High
High
Low
Low
Low
Low
High
High
High
High
Predicted
weight of
fibers /40sec
(s)
0.006
0.001
0.002
0.005
0.003
-0.001
-0.004
0.013
0.002
-0.001
0.001
0.004
0.005
0.001
0.001
0.01
0.003
-0.002
0.001
0.004
0
0.001
0
0.002
0.01
0.004
0.003
0.008
-0.002
-0.001
-0.003
0.021
B-5
-------
0.40
0.35
O)
C
I
si
2 TO
O >
— «*-
O °
O «>
si
o> c
Si
•si
0.30
0.25
0.20
0.10
0.05 -
0.00
Based on data Based on data
from 4
upholstery1 upholstery"
samples
Nonlinear fit
Mean with 95%
confidence interval
from 2
samples
in o 10 o in
o
o
(0
§
03
9
5
B
C
A
• r>
oino *
T-T-CM
omoinomoiQO
Total Minutes of Vacuuming
Figure B-2 Five minute weight gain due to fiber as a function of cumulative
vacuuming time; predicted means (with 95% confidence intervals) and
trend (non-linear curve)
B-6
-------
Several relationships to describe the trend in the weight gain as a function of time were
considered. The following non-linear model incorporating an exponential decay and a
linear trend provided the best and most parsimonious description of the overall trend:
Weight gain = Z * exp(-R * Minutes-of-vacuuming) + constant + slope * Minutes-of-vacuuming
The predicted values from the non-linear model were used in regression to identify the
final model for the data. Due to concern for unequal measurement variance, regression
weights were determined using the procedures in Section 6.2. The measurement
variance was found to depend on the predicted weight gain, with larger variance
associated with larger predicted weight gain.
The final weighted model had terms for time, represented by the predicted non-linear
relationship, vacuum cleaner and sample. Because the predicted non-linear relationship
had four parameters, the degrees of freedom for error was slightly biased, however this
had little effect on the results, all terms were statistically significant at the 2% level or
better. The least square estimated of average five minute weight gain by vacuum are
shown on the right side of Figure B-2. The corresponding averages for substrates
samples were more similar, i.e., showed less variation, than for the vacuum cleaners.
On upholstery, vacuum cleaner D collected fewer fibers than other vacuum cleaners,
unlike for carpets. This difference is due in part to the differences in the beater bar
attachments used to vacuum carpets and the upholstery attachments used with the
upholstery samples. After the first half hour of vacuuming upholstery, the weight gain
due to fibers decreases slowly with increasing vacuuming
The predicted 40-seconds weight gain due to fibers was calculated for each of the
substrates and vacuum cleaners and is shown in Table B-2 These values were used as
possible covariates in the analysis of data from the vacuum cleaner and sampler tests.
The predicted values are based on the first 100 minutes of vacuuming and apply to the
vacuum cleaner and sampler tests assuming the measurements for cumulative
vacuuming times from 100 to 120 minutes do not represent the weight gain at later
times. Even if this assumption is not correct, the effect of the study results are expected
to be very small.
B-7
-------
Table B-2 Predicted 40-second fiber uptake from upholstery by substrate sample and
vacuum cleaner
Vacuum
cleaner
A
A
A
A
6
B
B
B
C
C
C
C
D
D
D
D
Dust loading
(mg/sq ft)
100
400
400
100
100
400
400
100
100
400
400
100
100
400
400
100
Nominal dust
lead
concentration
High
High
Low
Low
High
High
Low
Low
High
High
Low
Low
High
High
Low
Low
Predicted
Fibers /40sec
(s)
0.009
0.004
0.006
0.007
0.023
0.018
0.02
0.021
0.012
0.008
0.009
0.01
0.003
-0.002
0
0
B-8
-------
B1.3 Fiber Preconditioning Data
Table B-3 consists of the fiber preconditioning data used in the preconditioning analysis.
The description of the data in each column of Table B-3 is listed below:
Column name Description
Sample Sample number
Team Team responsible for performing the test
Grindin Whether the dust was ground in or not
Amount Amount of dust applied (100 or 400 mg/sq ft)
Pb Cone Nominal lead concentration (HIGH or LOW)
Wt Gain Increase (g) from the vacuuming
Housevac Vacuum cleaner used in the test
Date Date of the test
Vac Min Cumulative number of minutes vacuumed
Substrate Substrate used in the test
The procedure for generating the data in Table B-3 follows:
1) Put a new bag in each vacuum each morning
2) For new bag, run vacuum for 5 min, wait 3 min and record weight For
used bag, used last weight
3) Cycle through the vacuums in order A,B,C,D,A,B,C,D,A,B... (depending
on which vacuum cleaner was the initial vacuum cleaner)
4) Vacuum the substrate for 5 min, remove the bag, wait 3 min, weigh the
bag
5) Calculate the increase in weight since the last weighing of the same bag
6) Cyde through the vacuums until two successive vacuums collect less than
20 mg of dust in 5 min
B-9
-------
Table B-3: Fiber Preconditioning Data
Gravimetrics Data
Sample
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 1
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Team
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
Grindin
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
Amount
(mg/sq ft)
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
Pb Cone
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
LOW
LOW
LOW
LOW
LOW
LOW
WtGain
te)
1.809
0.147
0.050
0.429
0.134
0.080
0.056
0.050
0.115
0.044
0.020
0.138
0.097
0.039
0.010
-0.129
0.062
0.085
0.075
0.196
0.065
0.013
0.038
0.268
0.049
-0.062
-0.008
-0.348
-0.016
-0.002
0.009
0546
0.030
-0.005
0.007
0.184
-0.013
-0.013
0.033
0.070
0.058
-0.028
0.045
0.090
0.093
0.027
0.029
0.060
2.266
0.125
0.668
0.302
0.081
0.026
Housevac Date Vac Min
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
B
C
D
A
B
C
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
8/3/93
8/3/93
8/3/93
8/3/93
8/3/93
, 8/3/93
8/3/93
8/3/93
8/4/93
32723
32723
8/4/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
105
110
115
120
125
130
135
140
145
150
155
160
165
170
175
180
19033
195.33
20033
205.33
21033
21533
22033
22533
235.67
240.67
245.67
250.67
5
10
15
20
25
30
Substrate
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
B-10
-------
Table B-3: Fiber Preconditioning Data (continued)
Gravimetrics Data
Sample
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Sample 2
Samples
Sampled
Sampled
Sampled
Sampled
Sampled
Sample d
Sampled
Sampled
Sample d
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Team
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
Grindin
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
Amount
(mg/sq ft)
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
Pb Cone
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
WtGain fe)
0.471
0.170
0.117
0.035
0329
0.088
0.061
0.001
0.363
0.104
0.079
0.003
0313
0.132
0.052
-0.009
0.145
0.061
0.026
0.071
0.027
0.031
0.000
0.129
0.037
0.045
0.030
0.100
0.041
2.322
0378
0.088
0.021
0.023
0.059
0.000
0.009
0.061
0.056
0.018
0.014
0.005
0.056
4.002
0.018
0.019
0.025
0.020
-0.013
0.023
0.008
0.028
0.001
-0.014
Housevac
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
B
C
D
A
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
Date
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
7/30/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/6/93
8/9/93
8/9/93
8/9/9d
8/9/9d
8/9/93
8/9/9d
8/9/93
VacMin
35
40
45
50
55
60
65
70
75
80
85
90
95
100
105
110
115
120
125
130
135
140
145
150
155
16533
17033
17533
18033
5
10
15
20
25
30
35
40
49.67
54.67
59.67
64.67
69.67
74.67
79.67
84.67
89.67
94.67
99.67
104.67
109.67
114.67
119.67
124.67
129.67
Substrate
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
B-ll
-------
Table B-3: Fiber Preconditioning Data (continued)
Cravimetrics Data
Sample
Sample 3
SampleS
Sample 3
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Sample 4
Samples
SampleS
Samples
SampleS
Samples
SampleS
SampleS
SampleS
SampleS
SampleS
SampleS
SampleS
SampleS
SampleS
Samples
SampleS
SampleS
Samples
SampleS
SampleS
Sample 6
Sample 6
Sample 6
Team
2
2
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2
2
Grind in
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
Amount
(mg/sq ft)
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
PbConc
LOW
LOW
LOW
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
HIGH
HIGH
HIGH
WtGain fe)
0.046
0.060
-0.029
0.980
0563
0.008
0.045
0.152
0.088
0.010
0.005
0.125
0.048
0.001
0.021
0.095
0.042
0.007
0.023
0.097
0.055
0.055
0.027
0.011
0.024
-0.008
0.015
0.063
0.045
0.060
0.014
0.786
0.035
0.103
0.033
0.007
0.009
0.065
0.041
0.001
0.014
0.041
0.023
-0.029
0.020
0.010
0.017
0.009
0.025
0.027
0.034
1.092
0575
0.029
Housevac
D
A
B
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
C
D
A
Date
8/9/93
8/9/93
8/9/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/4/93
8/6/93
8/6/93
8/6/93
8/6/93
8/6/93
8/6/93
8/9/93
8/9/93
8/9/93
8/9/93
8/9/93
8/9/93
8/9/93
8/9/93
8/9/93
8/9/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/6/93
8/6/93
8/6/93
8/6/93
8/6/93
8/6/93
8/6/93
8/6/93
8/5/93
8/5/93
8/5/93
VacMin
134.67
139.67
144.67
5
10
15
20
25
30
35
40
47.67
52.67
57.67
62.67
67.67
72.67
77.67
82.67
87.67
92.67
97.67
102.67
107.67
112.67
117.67
122.67
127.67
132.67
137.67
142.67
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
5
10
15
Substrate
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
B-12
-------
Table B-3: Fiber Preconditioning Data (continued)
Gravimetric Data
Sample
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Sampled
Sample 6
Sample 6
Sampled
Sampled
Sampled
Sampled
Sample 7
Sample/
Sample 7
Sample 7
Sample 7
Sample 7
Sample 7
Sample?
Sample 7
Sample?
Sample?
Sample?
Sample?
Sample?
Sample?
Sample?
Sample?
Sample 7
Sample 7
Sample?
Samples
Samples
Samples
Samples
Samples
Samples
Samples
Samples
Samples
Samples
Samples
Samples
Samples
Samples
Samples
Samples
Samples
Team
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
Grindin
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
YES
Amount
(mg/sq ft)
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
PbConc
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
WtGain fe)
-0.023
0.023
0.037
0.020
0.011
0.019
0.044
0.013
0.033
0.006
-0.017
-0.017
0.011
0.007
0.038
0.016
-0.002
0.979
0.196
0.039
0.058
0.093
0.065
0.027
0.024
0.058
0.074
0.017
0.019
-0.104
0.142
0.092
0.019
0.047
0.043
0.013
0.034
2.000
0.087
0.034
0.323
O.OSd
0.075
-0.003
0.208
0.029
0.044
0.02d
0.222
0.002
O.Old
0.019
0.134
-0.035
Housevac
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
Date
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/5/93
8/6/93
8/6/93
8/6/93
8/6/93
8/d/93
8/6/93
8/6/93
8/6/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/11/93
8/11/93
8/11/93
8/11/93
8/11/93
8/11/93
8/11/93
8/11/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/10/93
8/11/93
8/11/93
8/11/93
8/11/93
8/11/93
VacMin
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
Substrate
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
Carpet
B-13
-------
Table B-3: Fiber Preconditioning Data (continued)
Gravimetrics Data
Sample Team
Samples
Samples
Samples
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 9
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 10
Sample 11
Sample 11
Sample 11
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2
1
Grindin
YES
YES
YES
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
Amount
(mg/sq ft)
400
400
400
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
400
400
400
Pb Cone
HIGH
HIGH
HIGH
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
WtGainte)
-0.033
-0.002
0.188
0.219
0.105
0.240
0.409
0.139
0.056
0.141
0.186
0.177
-0.007
0.064
0.151
0.087
0.077
0.100
0.161
0.094
0.047
0.070
0.182
0.118
0.095
0.058
0.240
0.170
0.272
0.333
0.170
0.138
0.181
0.222
0.172
-0.041
0.119
0.182
0.135
0.062
0.100
0.160
0.122
0.071
0.081
0.186
0.128
0.083
0.104
0.187
0.105
0.277
0370
0.195
Housevac Date VacMin Substrate
B
C
D
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
A
B
C
8/11/93
8/11/93
8/11/93
8/12/93
8/12/93
8/12/93
8/12/93
8/12/93
8/12/93
8/12/93
8/12/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/12/93
8/12/93
8/12/93
8/12/93
8/12/93
8/12/93
8/12/93
8/12/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/13/93
8/16/93
90 Carpet
95 Carpet
100 Carpet
5 Upholstery
10 Upholstery
15 Upholstery
20 Upholstery
25 Upholstery
30 Upholstery
35 Upholstery
40 Upholstery
45 Upholstery
50 Upholstery
55 Upholstery
60 Upholstery
65 Upholstery
70 Upholstery
75 Upholstery
80 Upholstery
85 Upholstery
90 Upholstery
95 Upholstery
100 Upholstery
105 Upholstery
110 Upholstery
115 Upholstery
120 Upholstery
5 Upholstery
10 Upholstery
15 Upholstery
20 Upholstery
25 Upholstery
30 Upholstery
35 Upholstery
40 Upholstery
45 Upholstery
50 Upholstery
55 Upholstery
60 Upholstery
65 Upholstery
70 Upholstery
75 Upholstery
80 Upholstery
85 Upholstery
90 Upholstery
95 Upholstery
100 Upholstery
105 Upholstery
110 Upholstery
115 Upholstery
120 Upholstery
5 Upholstery
10 Upholstery
15 Upholstery
B-14
-------
Table B-3: Fiber Preconditioning Data (continued)
Gravimetrics Data
Sample Team
Sample 11
Sample 11
Sample 11
Sample 11
Sample 11
Sample 11
Sample 11
Sample 11
Sample 11
Sample 11
Sample 11
Sample 11
Sample 11
Sample 11
Sample 11
Sample 11
Sample 11
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
Sample 12
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
Grindin
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
Amount
(me/sq ft)
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
Pb Cone
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
WtGain (g) Housevac Date VacMln Substrate
0.076
0.098
0.231
0.126
0.044
0.096
0.230
0.109
0.037
-0.011
0.172
0.064
0.142
0.043
0.105
0.096
-0.070
0596
0.147
-0.161
0.065
0.246
0.004
0.065
0.091
0.196
0.112
0.039
0.137
0.204
0.096
-0.148
0.014
0.169
0.076
0.000
0.063
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
B
C
D
A
8/16/93
8/16/93
8/16/93
8/16/93
8/16/93
8/16/93
8/16/93
8/16/93
8/16/93
8/16/93
8/16/93
8/17/93
8/17/93
8/17/93
8/17/93
8/17/93
8/17/93
8/13/93
8/13/93
8/16/93
8/16/93
8/16/93
8/16/93
8/16/93
8/16/93
8/16/93
8/16/93
8/16/93
8/16/93
8/16/93
8/16/93
8/17/93
8/17/93
8/17/93
8/17/93
8/17/93
8/17/93
20 Upholstery
25 Upholstery
30 Upholstery
35 Upholstery
40 Upholstery
45 Upholstery
50 Upholstery
55 Upholstery
60 Upholstery
65 Upholstery
70 Upholstery
75 Upholstery
80 Upholstery
85 Upholstery
90 Upholstery
95 Upholstery
100 Upholstery
5 Upholstery
10 Upholstery
15 Upholstery
20 Upholstery
25 Upholstery
30 Upholstery
35 Upholstery
40 Upholstery
45 Upholstery
50 Upholstery
55 Upholstery
60 Upholstery
65 Upholstery
70 Upholstery
75 Upholstery
80 Upholstery
85 Upholstery
90 Upholstery
95 Upholstery
100 Upholstery
B-15
-------
B2 Dust Preconditioning
Several vacuum cleaner and sampler tests were conducted using each substrate sample.
To make the test conditions more similar between the first and last test on a substrate,
dust was applied to each substrate and vacuumed off to simulate the previous tests.
The weight of dust vacuumed from the substrate was determined, from which the dust
recovery was calculated. The dust preconditioning procedures are discussed in Section
4.4. The dust preconditioning was done to all types of substrates and used all dust
particle sizes and all vacuums on each substrate sample. This as done prior to use of
substrates in any of the actual tests.
The dust preconditioning recovery data were analyzed separately for smooth substrates
(tile, linoleum, wood) and rough substrates (carpet, carpet with grind-in, and
upholstery). The initial model had factors for all two way interactions of substrate,
nominal dust lead concentration, and dust loading, interaction of vacuum cleaner and
dust loading, and all two way interactions of substrate, vacuum cleaner, and, for rough
substrates, particle size. Terms which were not significant at the 5% level were
eliminated from the model to determine the final model.
For dust preconditioning of smooth surfaces (tile, linoleum, wood), no factors were
significant predictors of dust recovery. For rough substrates (carpet, carpet with grind-
in, and upholstery), only the vacuum cleaner was a significant predictor of dust
recovery. Differences in measurement variance among vacuum cleaners for rough
substrates were not statistically significant. However, differences between rough and
smooth substrates were significantly different. Therefore, the mean dust recovery and
95% confidence interval were calculated separately for each vacuum cleaner on rough
substrates and for smooth substrates, without pooling the variance. Figure B-3 and
Table B-4 show the average dust recovery for each vacuum cleaner on rough substrates
and all vacuum cleaners on smooth substrates, with 95% confidence intervals. The
pooled standard deviation for dust recovery measurements is 17%, greater than the 10%
value assumed for the redesign of the study. Therefore, based on the dust
preconditioning results, the full study may not achieve its data quality objectives.
The results show high dust recovery on smooth substrates (averaging 94% recovery).
Recovery on rough substrates carpet and upholstery, depend on the vacuum cleaner
used. For the canister vacuum cleaners, the highest dust recovery is found on the least
expensive vacuum cleaner and the lowest dust recovery on the most expensive vacuum
cleaner.
B-16
-------
1.2 -T-
1 .-
0.8 --
I
I °-6 r
1
Q
0.4 --
0.2 --
0
If
g-si
f|
I
U
Su
^9
U
Substrate and Vacuum
1
.S 8
nJ ^Q
I
Figure B-3 Average preconditioning dust recovery by vacuum and substrate/ with
95% confidence intervals
B-17
-------
Table B-4 Average dust recovery by vacuum and substrate, with 95% confidence
intervals
Substrate
Carpet/Upholstery
Carpet/Upholstery
Carpet/Upholstery
Carpet/Upholstery
Wood/Tile/Linoleum
Vacuum
cleaner
A
B
C
D
All Vacs
Average dust
recovery
98%
83%
67%
86%
94%
95% confidence
interval
92% to 104%
73% to 94%
57% to 77%
78% to 94%
91% to 98%
B-18
-------
B2.1 Dust Preconditioning Data
Table B-5 consists of the dust preconditioning data used in the preconditioning analysis.
The description of the data in each column of Table B-5 is listed below:
Column Name Description
Substrate Substrate used in the test
Grind-in Whether the dust was ground in (applies only to carpet and
upholstery substrates)
Amount Amount of dust applied to substrate (100 or 400 mg/sq ft)
Pb Cone Nominal lead concentration (HIGH or LOW)
Date Date of the test
Time Time of the test
Test No Test number
Housevac Vacuum cleaner (either A, B, C, or D)
Dust Size Size of the dust
Dust Applied Amount of dust applied
Bag Weight Change Increase in the bag weight from vacuuming
The procedures generating the data in Table B-5 follow:
1) Use a new bag in each vacuum at the beginning of the day
2) Perform the tests according to the test sequence for the dust
preconditioning
3) Deposit dust. Determine the actual weight of dust deposited
4) Grind-in if applicable
5) Determine tare weight of each bag before each use
6) Run free for 40 sec, cool 2 min, brush and record weight after 1 more min
7) Vacuum for 40 sec with the vacuum indicated in the test sequence for dust
preconditioning
8) Record the time of the vacuuming
9) Reweigh the bag after 40 sec vac (cool 2 min, brush and record weight
after 1 more min)
10) Repeat tare-vac-reweigh using the vacuum cleaner and particle size
designated in test sequence, which utilizes the same substrate with the
same dust loading and lead cone.
11) Vacuum the wand and brush on all vacuum cleaners after completing all
tests on the substrate (no weighing)
B-19
-------
Table B-5: Dust Preconditioning Data
Gravimetrics Data
Substrate
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
UPHOL
UPHOL
UPHOL
UPHOL
UPHOL
UPHOL
UPHOL
UPHOL
UPHOL
UPHOL
UPHOL
UPHOL
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
LINO
LINO
LINO
LINO
WOOD
WOOD
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
TILE
TILE
TILE
TILE
WOOD
WOOD
CRPT
CRPT
Grind-in
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
YES
YES
YES
YES
YES
NO
NO
NO
NO
NO
NO
YES
YES
YES
YES
YES
YES
NO
NO
NO
NO
NO
NO
NO
NO
Amount
(mg/sq ft) Pb Cone
400
400
400
400
400
400
400
400
400
400
400
400
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
400
100
100
400
100
100
400
400
400
400
400
400
100
400
400
100
400
400
400
400
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
HIGH
HIGH
LOW
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
LOW
HIGH
LOW
HIGH
LOW
HIGH
HIGH
HIGH
Date
8/23/94
8/23/94
8/23/94
8/23/94
8/23/94
8/23/94
8/23/94
8/23/94
8/23/94
8/23/94
8/23/94
8/23/94
8/23/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/25/94
8/25/94
8/25/94
8/25/94
8/25/94
8/25/94
8/25/94
8/25/94
8/25/94
8/25/94
8/23/94
8/23/94
8/23/94
8/23/94
8/23/94
7/25/94
8/23/94
8/23/94
Time
9-56
10:21
1053
11:27
1151
12:12
12:30
1256
1339
14:40
15:15
1555
16:30
8:55
9:11
930
9:47
10:07
10:41
11:07
11:25
11:40
12:45
13:05
13:25
13:46
14:11
14:30
1455
15:15
1535
16:11
11:05
11:25
1157
12:20
12:48
13:45
1432
1453
15:10
15:25
9:49
10:28
11:02
1135
12:05
13:45
14:10
14:33
Test No
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
201
202
203
204
205
206
207
208
Dust Size Dust
Housevac (microns) Applied (g)
D
B
B
C
A
C
A
B
D
C
A
D
D
D
A
B
C
C
C
D
A
A
B
D
D
B
A
D
C
A
B
D
A
C
B
C
B
A
C
D
A
D
A
D
C
B
D
A
B
C
150-212
250-2000
53-106
106-150
212-250
<53
106-150
<53
250-2000
212-250
150-212
53-106
106-150
<53
150-212
212-250
53-106
250-2000
150-212
106-150
53-106
250-2000
<53
212-250
<53
106-150
250-2000
212-250
150-212
53-106
106-150
106-150
106-150
106-150
106-150
106-150
150-212
106-150
212-250
53-106
<53
250-2000
106-150
106-150
106-150
106-150
106-150
106-150
212-250
250-2000
2.643
2.743
2.678
2.796
2.737
2.608
2.752
2.611
2.860
2.802
2.661
2.763
0.700
0.590
0.743
0.720
0.696
0.702
0.723
0.725
0.663
0.659
0.628
0.722
0.660
0.667
0.741
0.733
0.710
0.727
2.783
0.741
0.740
2.710
0.654
0.737
2.785
2.937
2.744
2.807
Z730
2.901
0.661
2.672
2.868
0.711
2.705
2.730
2.729
1.904
Bag Weight
Change (g)
2.172
2.451
2.115
2.063
3.104
1.997
2.705
2.291
2.636
2528
2.640
2.155
0.650
0.419
0.748
0.769
0587
0.705
0.332
0.759
0.656
0.659
0.739
0576
0.477
0514
0.738
0520
0.340
0.901
2545
0.711
0.724
2513
0.633
0.678
1.965
2.105
0575
2.828
2.833
1.871
0.679
2.459
2.953
0.618
2502
2360
1565
1353
B-20
-------
Table B-5: Dust Preconditioning Data (continued)
Gravimetrics Data
Substrate
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
UPHOL
UPHOL
UPHOL
UPHOL
UPHOL
UPHOL
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
UPHOL
UPHOL
UPHOL
UPHOL
UPHOL
UPHOL
Grind-in
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
YES
YES
YES
YES
YES
NO
NO
NO
NO
NO
NO
YES
YES
YES
YES
YES
YES
NO
NO
NO
NO
NO
NO
Amount
(mg/sq ft)
400
400
400
400
100
100
100
100
100
100
400
400
400
400
400
400
100
100
100
100
100
100
100
100
100
100
100
100
400
400
400
400
400
400
Pb Cone
HIGH
HIGH
HIGH
HIGH
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
Date
8/23/94
8/23/94
8/23/94
8/23/94
8/23/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/24/94
8/25/94
8/25/94
8/25/94
8/25/94
8/25/94
8/25/94
8/25/94
8/25/94
8/25/94
Time
1455
15:14
1535
15:57
16:17
8:35
854
9:10
9:35
953
10:25
10:43
11:08
11:26
11:41
1233
13:12
13:32
13:40
14:01
14:25
14:43
14:48
15:33
1553
10:19
11:05
11:28
12:46
13:00
14:15
14:35
1450
15:05
Test No
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
Dust Size Dust
Housevac (microns) Applied (g)
C
B
D
A
C
A
D
D
A
B
D
A
C
B
C
B
B
D
D
A
A
C
C
C
B
A
D
B
D
B
A
A
C
B
53-106
106-150
<53
150-212
212-250
106-150
53-106
250-2000
<53
150-212
106-150
<53
250-2000
150-212
53-106
212-250
250-2000
106-150
150-212
53-106
212-250
<53
<53
106-150
53-106
212-250
150-212
250-2000
212-250
106-150
250-2000
53-106
150-212
<53
2.698
2.695
2.658
2.718
0.685
0.699
0.642
0.750
0.605
0.677
2.736
2.623
2.708
2.720
2.727
2.750
0.700
0.678
0.688
0.661
0.700
0.643
0.613
0.689
0.684
0.721
0.678
0.690
2.756
2.778
2.785
2.775
2.742
2.620
Bag Weight
Change (g)
2.021
0.803
3574
3.145
0.439
0.762
0527
0.647
0.455
0583
2.019
2.146
1.731
2.155
1.895
2.605
0.638
0.604
0575
0564
0.699
0362
0336
0394
0.721
0.716
0539
0561
2.640
2582
2.700
2500
2540
1.834
B-21
-------
B-22
-------
APPENDIX C: SIEVED DUST DATA
The data in the following tables is derived from both the gravimetrics and lead analysis
data. The two files were merged matching the lead analysis with the corresponding test
data for sieved dust and the values for relevant variables are reported. Table C-l
consists of the sieved dust data. The description of the data in each column of Table C-l
is listed below;
Column Name
Test
TestNos
Size
Date
Team
Dust Type
Dust Sample Weight
Run
Preparation Batch
Instrument Botch
Lead Comment
Instrument Response
Sample Weight
PB
Dust Lead Cone
Instrument
Description
Sample test number
Six sample test numbers sieved together
Size of the dust sampled
Date of the sampling
Team responsible for sampling
Type of dust (either from NEW or OLD home)
Amount of dust sent to lab for lead analysis
Lead analysis run number
Lead analysis preparation batch number
Lead analysis instrument batch number
Lead analysis comment number
Lead analysis instrument response
Weight of sample used in the lead analysis
Lead amount estimated from analysis
Lead concentration estimated from analysis
Type of instrument used in analysis (either ICP or GFAA)
C-l
-------
Table C-l: Sieved Dust Data
Gravimetrics Data
Test
601
602
603
604
605
606
621
622
623
624
625
626
607
608
609
610
611
612
627
628
629
630
631
632
641
642
643
652
647
648
649
644
645
646
650
651
667
668
661
662
663
664
665
666
669
670
671
672
681
682
TestNos
601-606
601-606
601-606
601-606
601-606
601-606
621-626
621-626
621-626
621-626
621-626
621-626
607-612
607-612
607-612
607-612
607-612
607-612
627-632
627-632
627-632
627-632
627-632
627-632
641-646
641-646
641-646
647-652
647-652
647-652
647-652
641-646
641-646
641-646
647-652
647-652
667-672
667-672
661-666
661-666
661-666
661-666
661 "666
661-666
667-672
667-672
667-672
667-672
681-686
681-686
Size
<53
53-106
106-150
150-212
212-250
250-2000
<53
53-106
106-150
150-212
212-250
250-2000
<53
53-106
106-150
150-212
212-250
250-2000
<53
53-106
106-150
150-212
212-250
250-2000
<53
53-106
106-150
250-2000
<53
53-106
106-150
150-212
212-250
250-2000
150-212
212-250
<53
53-106
<53
53-106
106-150
150-212
212-250
250-2000
106-150
150-212
212-250
250-2000
<53
53-106
Date
8/17/93
8/17/93
8/17/93
8/17/93
8/17/93
8/17/93
8/18/93
8/18/93
8/18/93
8/18/93
8/18/93
8/18/93
8/19/93
8/19/93
8/19/93
8/19/93
8/19/93
8/19/93
8/19/93
8/19/93
8/19/93
8/19/93
8/19/93
8/19/93
8/27/93
8/27/93
8/27/93
8/27/93
8/27/93
8/27/93
8/27/93
8/27/93
8/27/93
8/27/93
8/27/93
8/27/93
9/3/93
9/3/93
9/3/93
9/31/93
9/3/93
9/3/93
9/3/93
9/3/93
9/3/93
9/3/93
9/3/93
9/3/93
9/10/93
9/10/93
Operator
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
chambers
chambers
chambers
MOORE
MOORE
MOORE
MOORE
chambers
chambers
chambers
MOORE
MOORE
MOORE
MOORE
chambers
chambers
chambers
chambers
chambers
chambers
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
Dust Type
NEW
NEW
NEW
NEW
NEW
NEW
NEW
NEW
NEW
NEW
NEW
NEW
OLD
OLD
OLD
OLD
OLD
OLD
OLD
OLD
OLD
OLD
OLD
OLD
NEW
NEW
NEW
OLD
OLD
OLD
OLD
NEW
NEW
NEW
OLD
OLD
OLD
OLD
NEW
NEW
NEW
NEW
NEW
NEW
OLD
OLD
OLD
OLD
NEW
NEW
Dust Sample
Weight
0.778
0.709
0.777
0.783
0.699
0.761
0.724
0.739
0.789
0.788
0.729
0.719
0.731
0.722
0.730
0.735
0.797
0.715
0.743
0.740
0.767
0.749
0.792
0.762
0.639
0.497
0.688
0.734
0520
0.658
0.692
0.655
0.643
0.715
0.680
0.731
0.612
0.774
0.597
0.614
0.669
0.685
0.684
0.685
0.783
0.740
0.678
0.754
0592
0.624
C-2
-------
Table C-l: Sieved Dust Data (continued)
Lead Concentration Data
Run
31
32
33
34
40
41
42
43
44
45
46
47
48
49
58
59
60
61
62
63
64
65
66
67
49
58
59
61
62
63
64
65
66
67
74
75
40
41
42
43
44
45
46
47
48
49
58
59
125
126
Preparation
Batch
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
501
502
502
502
502
502
502
502
502
502
502
502
502
504
504
504
504
504
504
504
504
504
504
504
504
505
505
Instrument
Batch
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E08233B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
Lead
Comment
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Instrument
Response
05822
0.6798
0.6919
1.0915
0.1626
03242
0.6156
0.7221
0.8501
05144
0.2273
0.0746
2.1829
25710
2.4238
2.5491
23702
2.0726
2.1683
2.6315
1.8484
2.4412
2.0525
6.8177
03102
1.2123
0.6068
1.8463
15606
2.4905
2.0689
0.4202
03527
0.2951
1.9442
2.7349
1.8708
2.8396
0.7168
0.7011
0.6026
0.6562
05564
0.2671
25937
63087
65902
63124
0.6752
0.6791
Sample
Weight
0.778
0.709
0.777
0.783
0.699
0.761
0.724
0.739
0.789
0.788
0.729
0.719
0.731
0.722
0.730
0.735
0.797
0.715
0.743
0.740
0.767
0.749
0.792
0.762
0.639
0.497
0.688
0.734
0.520
0.658
0.692
0.655
0.643
0.715
0.680
0.731
0.612
0.774
0597
0.614
0.669
0.685
0.684
0.685
0.783
0.740
0.678
0.754
0592
0.624
PB
72.77
84.98
34.60
5457
2033
16.21
76.95
90.27
4250
25.72
28.41
9.32
272.86
32138
302.98
318.64
296.28
259.08
271.04
328.94
231.05
305.15
25656
1704.43
38.78
60.61
30.34
230.79
195.08
31131
258.61
21.01
17.63
14.76
243.02
341.86
233.85
354.95
71.68
70.11
30.13
32.81
27.82
1335
324.21
315.44
32951
85155
6752
8488
Pb Cone
Recover
93.540
119.857
44524
69.700
29.081
21.302
106.290
122.148
53.871
32.640
38.976
12.968
373.273
445.118
415.034
433520
371.738
362.343
364.788
444510
301.239
407.410
323.943
2236.778
60.681
121.962
44.096
314.424
375.144
473.119
373.717
32.078
27.422
20.637
357390
467.664
382.108
458592
120.074
114.184
45.035
47.896
40.675
19.493
414.064
426.264
486.003
1129377
114056
136.030
Instrument
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
C-3
-------
Table C-l: Sieved Dust Data
Cravimetrics Data
Test
683
684
685
686
687
688
701
702
703
704
705
712
706
689
690
691
692
707
708
709
710
711
TestNos
681-686
681-686
681-686
681-686
687-692
687-692
701-706
701-706
701-706
701-706
701-706
707-712
701-706
687-692
687-692
687-692
687-692
707-712
707-712
707-712
707-712
707-712
Size
106-150
150-212
212-250
250-2000
<53
53-106
<53
53-106
106-150
150-212
212-250
250-2000
250-2000
106-150
150-212
212-250
250-2000
<53
53-106
106-150
150-212
212-250
Date
9/10/93
9/10/93
9/10/93
9/10/93
9/10/93
9/10/93
9/16/93
9/16/93
9/16/93
9/16/93
9/16/93
9/16/93
9/16/93
9/10/93
9/10/93
9/10/93
9/10/93
9/16/93
9/16/93
9/16/93
9/16/93
9/16/93
Operator
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
chambers
chambers
chambers
chambers
chambers
MOORE
chambers
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
MOORE
Dust Type
NEW
NEW
NEW
NEW
OLD
OLD
NEW
NEW
NEW
NEW
NEW
OLD
NEW
OLD
OLD
OLD
OLD
OLD
OLD
OLD
OLD
OLD
Dust Sample
Weight
0.690
0.693
0.698
0.690
0.612
0.656
0.552
0.593
0.667
0.686
0.651
0.724
0.675
0.677
0.619
0.691
0.662
0560
0.666
0.675
0.679
0.682
C-4
-------
Table C-l: Sieved Dust Data (continued)
Lead Concentration Data
Run
127
128
129
130
131
132
66
67
73
74
75
76
77
158
159
160
161
79
80
81
82
91
Preparation
Batch
505
505
505
505
505
505
504
504
504
504
504
504
504
505
505
505
505
504
504
504
504
504
Instrument Lead
Batch Comment
El 1083 A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
2
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Instrument
Response
0.5109
03652
03397
0.2610
1.8299
2.3695
0.6500
0.9013
0.6470
0.4195
03411
8.5532
0.8184
2.3901
6.0027
5.3871
11.6840
2.1064
25017
2.4269
5.3365
6.4787
Sample
Weight
0.690
0.693
0.698
0.690
0.612
0.656
0.552
0593
0.667
0.686
0.651
0.724
0.675
0.677
0.619
0.691
0.662
0.560
0.666
0.675
0.679
0.682
PB
25.55
18.26
16.99
13.05
228.74
296.19
65.00
90.13
32.35
20.97
17.06
1069.15
40.92
298.76
300.14
26936
584.20
210.64
312.71
30336
266.83
323.94
PbConc
Recover Instrument
37.025
26349
24335
18.916
373.754
451.505
117.754
151.990
48501
30573
26.200
1476.727
60.623
441304
484.871
389.805
882.477
376.143
469538
449.426
392.968
474.978
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
ICP
C-5
-------
Entry Shaat
for
Pretest and Weekly Analysis of Staved Duct
Procedure
Collect sample of each dust size, from new homes, as follows:
Take approximately 0.678 0 from dust container
Deposit dust through sieve, onto plastic sheet, to simulate depositing of
dust on substrate
Determine weight of dust deposited
Transfer dust on plastic sheet into labeled and weighed sample bottle
Determine weight of dust sample in sample bottle
Repeat all the above for each particle size
Using new data sheet, repeat all the above for dust from older homes
Test No.
Date
Operator
Dust Type OusrVPg
(old or now homes)
Dust Applied (gm)
(Balance No. APflBfll )
Total Wt Rnal Wt Net Wt
Dust Sample (gm)
(Balance No. &MPAAL)
NetWt
> 53 am
T65T
AfPftN
3AMPP/r\J
Bar Code
Label
53-106 i/m
Bar Coda
Label
IPS-ISO am
Bar Code
Label
150-212 um
Bar Code
Label
212-250
Bar Code
Label
250-2000 am
Bar Code
Label
Samples Relinquished by:
Samples Received by:
Data:
Reviewed by.
Date
C-6
-------
APPENDIX D: SAMPLER DATA
The data in the following tables is derived from both the gravimetrics and lead analysis
data. The two files were merged matching the lead analysis with the corresponding test
data for samplers and the values for relevant variables are reported. Table D-l consists
of the sieved dust data. The description of the data in each column of Table D-l is listed
below:
Column Name
Test
Date
Team
Sampler
Substrate
Grindin
Amount
Nom Dust Lead Cone
Dust Size
Square
Time
Initial Gain
Dust Load 1
Sampler Collect
Final Collect
Dust Comment
Run
Preparation Batch
Instrument Batch
Instrument Response
Sample Weight
Lead Amount
Q
Dust Lead Cone
Lead Comment
Instrument
Description
Test number
Date of the test
Team responsible for performing the test
Sampler used in the test (either BN, CAPS, WIPE, or BRM)
Substrate used in the test
Whether or not the dust was ground in to the substrate
(applied only to carpet and upholstery)
Amount of dust applied (either 100 or 400 mg/sq ft)
Lead concentration deposited (either HIGH or LOW)
Size of the dust particles deposited on the carpet
Square of the substrate used in the test
Time of the test
Initial increase from vacuuming with no dust deposited
Amount of dust applied to the substrate
Increase in the cassette or sampler dust container weight
during the test
Increase in the weight of the cassette or sampler dust
container weight from the final vacuuming
Gravimetrics comment number
Lead analysis run number
Lead analysis preparation batch number
Lead analysis instrument batch number
Lead analysis instrument response
Weight of the sample analyzed
Lead amount estimated from analysis
Notifier (*) of whether the sample was below IDL
Lead concentration estimated from analysis
Lead analysis comment number
Instrument used in the analysis
D-l
-------
Table D-l: Sampler Data
Cravimetrics Data
Test Date Team Sampler Substrate
4-1 9/10/93
3-1 9/10/93
3-2 9/10/93
3-3 9/10/93
3-4 9/10/93
4-2 9/10/93
4-3 9/10/93
4-4 9/13/93
4-5 9/13/93
4-6 9/13/93
4-7 9/13/93
4-8 9/13/93
3-6 9/13/93
3-7 9/13/93
3-8 9/13/93
3-9 9/13/93
3-10 9/13/93
3-11 9/13/93
3-12 9/13/93
3-13 9/13/93
3-14 9/14/93
3-15 9/14/93
4-9 9/13/93
4-10 9/13/93
4-11 9/13/93
4-12 9/13/93
4-13 9/13/93
4-14 9/14/93
4-15 9/14/93
4-16 9/14/93
4-17 9/14/93
4-18 9/14/93
4-19 9/14/93
4-20 9/14/93
4-21 9/15/93
4-22 9/15/93
3-16 9/14/93
3-17 9/14/93
3-18 9/14/93
3-19 9/15/93
3-20 9/15/93
3-21 9/15/93
3-22 9/15/93
3-23 9/15/93
3-24 9/15/93
3-25 9/16/93
4-23 9/15/93
4-24 9/15/93
4-25 9/15/93
4-26 9/15/93
4-27 9/15/93
3-5 9/10/93
2
1
1
1
1
2
2
2
2
2
2
2
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
2
2
2
2
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
1
WIPE
HVS3
CAPS
BN
BN
WIPE
CAPS
BN
HVS3
CAPS
HVS3
WIPE
HVS3
BN
HVS3
BN
CAPS
BN
CAPS
CAPS
BN
CAPS
CAPS
HVS3
CAPS
BN
HVS3
WIPE
HVS3
CAPS
CAPS
HVS3
CAPS
BN
HVS3
CAPS
HVS3
BN
CAPS
BN
HVS3
BN
HVS3
BN
WIPE
CAPS
HVS3
WIPE
BN
HVS3
CAPS
BN
LINO
CRPT
UPHO
UPHO
UPHO
LINO
CRPT
CRPT
CRPT
CRPT
CRPT
LINO
WOOD
WOOD
CRPT
CRPT
CRPT
CRPT
WOOD
LINO
WOOD
WOOD
UPHO
UPHO
UPHO
CRPT
CRPT
WOOD
LINO
CRPT
LINO
LINO
CRPT
CRPT
WOOD
WOOD
UPHO
LINO
LINO
CRPT
CRPT
UNO
LINO
LINO
WOOD
WOOD
UPHO
WOOD
WOOD
WOOD
CRPT
UPHO
Amount
Crindin (mg/sq ft)
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
YES
NO
NO
NO
NO
NO
YES
YES
NO
NO
NO
NO
NO
NO
NO
YES
YES
NO
NO
YES
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
YES
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
400
400
100
100
100
400
100
400
400
400
400
400
400
400
400
400
400
400
400
100
100
100
100
400
400
100
100
100
100
100
100
100
100
100
400
400
400
400
400
100
100
100
100
400
100
400
100
100
100
400
400
400
Nom Dust
Lead Cone Dust Size
LOW
LOW
HIGH
HIGH
HIGH
LOW
HIGH
HIGH
HIGH
LOW
LOW
HIGH
LOW
LOW
HIGH
HIGH
HIGH
LOW
HIGH
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
HIGH
HIGH
LOW
LOW
LOW
LOW
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
LOW
HIGH
HIGH
HIGH
HIGH
HIGH
LOW
LOW
LOW
212-250
53-106
<53
<53
212-250
<53
53-106
<53
212-250
<53
212-250
106-150
106-150
53-106
150-212
250-2000
53-106
150-212
150-212
106-150
212-250
<53
53-106
<53
212-250
53-106
53-106
150-212
150-212
150-212
250-2000
53-106
250-2000
150-212
<53
212-250
53-106
250-2000
53-106
212-250
<53
<53
212-250
150-212
250-2000
150-212
150-212
53-106
106-150
250-2000
106-150
150-212
Square Time
1
4
2
3
4
2
1
3
4
1
2
4
2
3
1
2
3
1
4
2
1
2
4
1
2
1
2
2
1
1
2
3
1
2
2
3
3
1
2
2
3
3
4
4
4
4
3
2
3
2
3
3
10:24
10:40
11:30
12:45
13:00
10:40
12:15
8:48
9:08
10:13
10:30
11:21
9:10
9:45
10:10
10:40
11:30
13:15
14:05
15:20
9:10
10:50
13:20
14:15
14:38
15:39
16:04
8:45
9:29
10:54
12:45
14:28
14:52
8:44
9:21
11:10
13:10
15:00
9:20
9:45
10:30
11:05
13:00
13:35
9:15
10:20
11:10
12:05
12:40
13:16
16:30
Initial
Gain
fe)
0.014
0.734
0.011
0.104
1.404
0.493
0.017
0.072
0.093
1.304
0.065
0.010
-0.008
0.002
0.046
0.022
0.055
.
0.014
0.035
0.433
0.007
0.295
0.035
0.016
-0.004
0.045
-0.006
.
0.000
0.002
•0.006
0.042
0.069
.
0.005
0.127
-0.003
Dust
Load
1(K)
0.406
0.359
0.081
0.119
0.116
0.371
0.115
0.380
0.404
0.421
0.400
0.411
0.422
0.404
0.433
0.433
0.507
0.418
0.139
0.113
0.106
0.103
0.098
0.433
0.401
0.110
0.115
0.096
0.107
0.099
0.119
0.085
0.115
0.093
0.392
0.399
0.408
0.427
0.427
0.097
0.110
0.107
0.124
0.420
0.112
0.389
0.108
0.100
0.107
0.421
0.404
0.421
Sampler
Collect fe)
0.320
0.046
0.076
-0.001
0.098
0.195
0.358
0.272
0.263
0.398
0.352
0.306
0.008
0.404
-0.003
0.388
0.098
0.051
0.064
0.066
0.329
0.382
0.013
0.103
0.104
0.076
0.108
0.081
0.121
0.004
0.335
0.371
0.360
0.013
0.346
-0.005
0.096
0.047
0.095
0.206
.
0.369
0.135
0.046
0.388
0.354
0.127
D-2
-------
Table D-l: Sampler Data (continued)
Gravimetrics Data
Final Dust
Collect (g) Comment
0.218
0.123
0.039
0.069
0.672
0.386
0.024
0.016
0.535
0.717
0.345
0.015
0.021
.
0.042
0.03
0.124
0.087
0.013
-0.002
0.189
0
.
0.288
0.03
0.004
m
0.366
0.078
0.051
0.183
0.011
0.005
0.033
0.059
0.013
0.114
0.282
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
7
NA
NA
NA
NA
8
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
9
NA
10
NA
NA
NA
NA
NA
Lead Concentration Data
Preparation Instrument
Run Batch Batch
92
61
62
23
25
94
63
27
64
65
66
95
67
29
73
31
74
34
75
76
35
60
77
133
139
38
140
85
141
142
74
144
145
39
146
147
61
41
62
42
63
50
64
54
98
65
148
99
61
83
157
67
507
506
506
508
508
507
506
508
506
506
506
507
506
508
506
508
506
508
506
506
508
504
506
505
505
508
505
507
505
505
505
505
505
508
505
505
504
508
504
508
504
508
504
508
507
504
505
507
508
505
505
508
E12023B
E12023B
E12023B
V12073A
V12073A
E12023B
E12023B
V12073A
E12023B
E12023B
E12023B
E12023B
E12023B
V12073A
E12023B
V12073A
E12023B
V12073A
E12023B
E12023B
V12073A
E11083A
E12023B
El 1083 A
El 1083 A
V12073A
E11083A
V12073A
E11083A
E11083A
V12073A
El 1083 A
E11083A
V12073A
E11083A
E11083A
El 1083 A
V12073A
El 1083 A
V12073A
El 1083 A
V12073A
E11083A
V12073A
E12023B
El 1083 A
El 1083 A
E12023B
V12073A
V12073A
E11083A
V12073A
Instrument
Response
(ug/mL)
0.23360
1.13130
0.61650
0.05156
0.04411
0.40550
1.54290
0.02487
5.70260
0.70460
0.28650
1.51190
0.33940
0.03165
5.12140
0.07298
3.86600
0.01260
6.12420
0.45870
0.02710
0.30830
0.40000
0.72820
0.14320
0.04474
0.52900
0.04647
2.34540
0.88130
0.05411
0.35760
0.11310
0.01157
2.52040
2.91790
3.30770
0.06789
3.59200
0.00536
1.54170
0.54739
1.49570
0.27383
0.10430
4.96250
1.34810
0.39520
0.44710
0.18961
0.46600
0.16978
Sample Lead
Weight Amount
(ugj (UK) Q
0.320
0.046
0.076
-0.001
0.098
0.195
0.358
0.272
0.263
0.398
0.352
0.306
0.008
0.404
-0.003
0.388
0.098
0.051
0.064
0.066
0.329
0.382
0.013
0.103
0.104
0.076
0.108
0.081
0.121
0.004
0.335
0.371
0.360
0.013
0.346
-0.005
0.096
0.047
0.095
0.206
,
0.369
0.135
0.046
0.388
0.354
0.127
23.358
56.57
15.41
25.78
1.1 *
40.547
38.57
62.18
14157
35.23
14.33
151.19
16.97
39.56
128.04
1.82
193.3
0.32 *
153.11
11.47
0.68
7.71
10
36.41
7.16
1.12
13.23
1.162
58.64
22.03
1.35
8.94
183
0.29
126.02
145.9
165.39
1.7
179.6
0.13 •
38.54
13.68
37.39
6.85
10.432
124.06
33.7
39.524
11.18
4.74
23.3
4.24
Dust Lead
Cone Lead
(ug/g) Comment
176.766
335.033
339.211
393.597
318.846
398.226
129.528
54.470
42.641
112.393
418.415
228.063
478.465
394.601
117.013
13.284
120.434
151.500
110.673
18.741
86.038
128.405
563.798
289.885
12.525
110.380
23.357
71313
376.179
393.248
459.403
130.558
519.075
401.484
291.165
393.605
33.232
336.213
249.648
.
242.989
12.217
65.815
33.421
NA
NA
NA
NA
NA
NA
NA
5
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
7
NA
NA
6
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
6
NA
NA
Instrument
ICP
ICP
ICP
GFAA
GFAA
ICP
ICP
GFAA
ICP
ICP
ICP
ICP
ICP
GFAA
ICP
GFAA
ICP
GFAA
ICP
ICP
GFAA
ICP
ICP
ICP
ICP
GFAA
ICP
GFAA
ICP
ICP
GFAA
ICP
ICP
GFAA
ICP
ICP
ICP
GFAA
ICP
GFAA
ICP
GFAA
ICP
GFAA
ICP
ICP
ICP
ICP
GFAA
GFAA
ICP
GFAA
D-3
-------
Project 9802
WA*55
Data Entry Sheets
for
Sampler Tests
Test Identification
Sampler
Substrate
Grind-in
Dust Amount
PbConc
Test Sequence Number T£ST
Date
Operator
(Blue Nozzle, CAPS, HVS3 or WIPE)
UNOIeum, WOOD, UPHOtetery, CaRPeT)
No)
(100, 400 mg/ft2)
(Low. High)
Dust Size gtze (>53, 53-106, 106-150, 150-212, 212-250, 250-2000)
Team TfeAM (Number 1 or 2)
Square number SftoAgfe- (1, 2, 3 or 4) 1 -first, 3-last for carpet and upholstery, else 4-last)
Procedure
Perform the tests according to the sampler test sequence in Appendix Q, and procedures in Appendix E, F, G or H.
Housevac A will be used to vacuum the first square before sampler tests, or to vacuum the last square after sampler
tests.
If first square:
Tare weigh bag (run free for 40 seconds, cool 2 minutes, brush and record weight after 1 more minute)
Vac square for 40 seconds with Housevac A
Reweigh bag (cool 2 minutes, brush and record weight after 1 more minute)
Deposit dust in specified square and weigh the amount deposited (Grind-in dust if applicable)
Sample dust according to the appropriate protocol, weigh the dust collected (except for wipes)
Prepare the dust sample for analysis
If last square:
Tare weigh bag (run free for 120 seconds, cool 2 minutes, brush and record weight after 1 more minute)
Vac square for 120 seconds with Housevac A
Reweigh bag (cool 2 minutes, brush and record weight after 1 more minute)
Vacuum dust from wand and brush (no weighing)
Weight of Dust
(Balance *
Weight of Bag
(Balance #
Initial weight of bag
(if first or last square)
Vacuum and reweigh
bag (if first square)
Dust deposited
Dust collected by
sampler (exdu wipes)
Vacuum & reweigh bag 4.
(if last square)
Total Wt
am.
.0
.1
Final Wt.
gm.
NetWt
am.
Time
Weight
am.
Increase
om.
Bar Code
for Sample
BarCode
for Blank
NOTE: Submit one blank for each sampler, once each week
Sample relinquished by Reviewed by
Sample received by Date reviewed _
Date of transfer
D-4
-------
APPENDIX E: VACUUM CLEANER DATA
The data in the following tables is derived from both the gravimetrics and lead analysis
data. The two files were merged matching the lead analysis with the corresponding test
data for vacuum cleaners and the values for relevant variables are reported. Table E-l
consists of the sieved dust data. The description of the data in each column of Table E-l
is listed below:
Column Name
Test
Date
Team
Vac
Substrate
Grindin
Amount
Nom Dust Lead Cone
Dust Size
Time
Initial Gain
Dust Load 1
Gain - Load 1
Dust Load 2
Gain - Load 2
Dust Load 3
Gain - Load 3
No Dust - Gain 1
No Dust - Gain 2
No Dust - Gain 3
Dust Comnt
Run
Prep Batch
Instr Batch
Instr Resp
Sample Wgt
Lead Amount
Dust Lead Cone
Lead Comnt
Description
Test numbert
Date of test
Team responsible for test
Vacuum cleaner (either A, B, C, or D)
Substrate used in the test
Whether the dust was ground-in to the substrate (applies to
carpet and upholstery substrates only)
Amount of dust applied (either 100 or 400 mg/sq ft)
Lead concentration deposited (either HIGH or LOW)
Size of the dust applied
Time of the test
Initial increase from vacuuming with no dust deposited
Amount of first dust loading
Increase from first vacuuming
Amount of second dust loading
Increase from second vacuuming
Amount of third dust loading
Increase from third vacuuming
Increase from fourth vacuuming (no additional dust)
Increase from fifth vacuuming (no additional dust)
Increase from sixth vacuuming (no additional dust)
Gravimetrics comment number
Lead analysis run number
Lead analysis preparation batch number
Lead analysis instrument batch number
Lead analysis instrument response
Weight of sample used for lead analysis
Lead amount estimated by analysis
Lead concentration estimated by analysis
Lead analysis comment number
(Note: Lead concentrations for all vacuum cleaner tests were performed using the ICP
instrument)
E-l
-------
Table E-l: HousevacData
Gravimetrics Data
Test Date Team
1001 8/26/93
1002 8/26/93
1003 8/26/93
1004 8/26/93
1005 8/27/93
1006 8/27/93
2001 8/26/93
2002 8/26/93
2003 8/26/93
2004 8/26/93
2005 8/27/93
2006 8/27/93
2007 8/27/93
1007 8/27/93
2008 8/27/93
2009 8/30/93
1008 8/27/93
1009 8/30/93
1010 8/30/93
2010 8/30/93
2011 8/30/93
2012 8/30/93
2013 8/30/93
2014 8/30/93
1011 8/30/93
1012 8/30/93
1013 8/30/93
1014 8/30/93
1020 9/1/93
2098 9/1/93
2099 9/1/93
2033 9/1/93
2048 9/1/93
2061 9/1/93
2106 9/1/93
2025 9/2/93
2065 9/2/93
2068 9/2/93
1085 9/1/93
1088 9/1/93
1052 9/1/93
1101 9/1/93
1104 9/1/93
1030 9/2/93
1032 9/2/93
1046 9/2/93
1026 9/2/93
1027 9/3/93
2095 9/2/93
1
1
1
1
1
1
2
2
2
2
2
2
2
1
2
2
1
1
1
2
2
2
2
2
1
1
1
1
1
2
2
2
2
2
2
2
2
2
1
1
1
1
1
1
1
1
1
1
2
Vac
A
D
C
B
A
B
C
B
A
D
C
D
A
C
B
C
D
D
A
B
D
A
A
B
C
B
C
D
B
D
C
C
A
D
C
C
C
A
C
B
A
B
D
D
A
A
D
C
D
Substrate Grindin
LINO
LINO
LINO
LINO
WOOD
WOOD
TILE
TILE
TILE
TILE
CRPT
CRPT
CRPT
WOOD
CRPT
CRPT
WOOD
CRPT
CRPT
CRPT
CRPT
CRPT
LINO
LINO
CRPT
CRPT
CRPT
CRPT
UPHO
CRPT
CRPT
LINO
LINO
LINO
CRPT
UPHO
WOOD
WOOD
UPHO
UPHO
CRPT
CRPT
CRPT
UPHO
UPHO
CRPT
LINO
LINO
WOOD
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
NO
NO
NO
YES
YES
YES
NO
NO
NO
NO
NO
NO
NO
YES
YES
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
YES
YES
NO
NO
NO
NO
NO
NO
Amount
400
100
100
100
400
400
100
400
100
400
400
400
400
400
400
400
100
400
400
400
400
400
100
100
400
400
100
100
100
400
400
100
400
100
100
100
400
100
100
100
100
100
100
400
400
100
100
100
400
Pb Cone- Dust Size
Deposit (microns)
HIGH
HIGH
HIGH
LOW
HIGH
LOW
LOW
HIGH
LOW
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
LOW
LOW
LOW
LOW
LOW
LOW
LOW
LOW
HIGH
LOW
LOW
LOW
LOW
HIGH
LOW
LOW
LOW
LOW
LOW
HIGH
LOW
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
HIGH
LOW
HIGH
53-106
<53
212-250
106-150
150-212
106-150
150-212
212-250
150-212
<53
212-250
<53
<53
150-212
212-250
212-250
<53
53-106
53-106
212-250
<53
<53
53-106
150-212
53-106
53-106
<53
212-250
<53
212-250
<53
250-2000
<53
53-106
53-106
53-106
212-250
53-106
<53
212-250
212-250
212-250
<53
53-106
53-106
106-150
212-250
106-150
<53
Time
11:35
13:20
14:30
15:22
9:50
10:05
11:10
13:20
14:15
15:20
9:01
10:13
11:19
11:05
15:05
8:52
14:55
9:00
10:05
10:08
11:23
12:54
14:24
15:15
11:05
12:45
14:00
15:15
830
8:22
9:29
10:52
13:35
14:37
15:34
8:55
10:00
12:35
930
10:45
1330
14:30
15:30
9:00
10:00
11:00
12:00
930
11:00
Dust Gain - Dust Gain -
Initial Load 1 Load Load 2 Load 2
Gain fe) Ife) fe) (g)
0.000
-0.020
0.000
-0.001
0.128
0.008
0.042
0.001
0.087
0.067
0.153
1.101
0.232
0.045
0.043
0.219
-0.024
0.406
0.081
0.237
0.214
0.098
0.015
0.008
0.048
0.230
0.071
0.089
0.056
0.134
0.086
0.129
0.033
-0.026
0.077
0.040
0.032
0.036
0.084
0.035
0.216
0.034
0.035
-0.013
0.023
0.212
0.074
0.037
-0.044
2.720
0.600
0.708
0.688
2.735
2.716
0.697
2.871
0.697
2.717
2.801
2.648
2.630
2.710
2.778
2.729
0.542
2.656
2.833
2.781
2.640
2.695
0.698
0.734
2.677
2.645
0515
0.607
0.674
2.695
2.633
0.705
2.799
0.692
0.684
0.685
2.747
0.693
0.727
0.792
0.693
0.774
0.688
2.887
2.913
0.724
0.682
0.776
2.747
2.617
0.425
0.706
0.668
2.712
2.609
0.671
2.780
0.590
2.005
1.452
2.532
2.343
2.551
1.650
1.067
0.510
2.441
2.615
2.461
1.991
1.856
0.659
0.694
2.126
2.493
0.460
0.474
0.505
1.467
1.323
0.597
2504
0.646
0.488
0575
2.641
0.688
0545
0.795
0.681
0.385
0519
2.633
2.791
0.680
0.678
0.743
2352
2.898
0.639
0.692
0.698
2.747
2.750
0.715
2.727
0.747
2.761
2.732
2.680
2.703
2.756
2.672
2.715
0.673
2.818
2.741
2.745
2.872
2.750
0.679
0.702
2.820
2.736
0.684
0.750
0.730
2.666
2.748
0.680
2.808
0.696
0.697
0.678
2.778
0.676
0.702
0.743
0.761
0.671
0.603
2.910
2.937
0.665
0.742
0.800
2.727
2.774
0.510
0.684
0.670
2.715
2.677
0.696
2.653
0.726
1.889
1.486
2.190
2.200
2.652
1.895
1.842
0577
2.485
2564
2571
2.168
1.908
0.615
0.681
2.350
2534
0.436
0509
0591
1.894
1.604
0.665
2521
0.611
0513
0595
2.685
0.663
0.607
0.739
0.859
0.497
0558
2.676
2.843
0.671
0.703
0.745
2.419
E-2
-------
Table E-l: Housevac Data (continued)
Gravimetrics Data
Dust
Load 3
fe)
2.759
0.669
0.671
0.702
2.889
2.764
0.679
2.753
0.680
2.694
2.735
2.707
2.722
2.739
2.707
2.712
0.716
2.697
2.752
2.700
2.822
2.720
0.699
0.710
2.782
2.814
0.645
0.684
0.737
2.708
2.779
0.686
2.962
0.681
0.675
0.703
2.773
0.682
0.744
0.651
0.806
0.684
0.687
2.775
2.702
0.703
0.685
0.759
2.751
Gain - No Dust-
Load 3 No Oust- No Dust- Gain 3 Dust
(g) Gainl(g)Gain2fe) fe) Comnt
2.669
0.457
0.664
0.675
2.851
2.659
0.637
2.690
0.680
2.131
1.628
2.389
2.230
2.542
2.249
1556
0.528
2.518
2.542
2.667
2.121
1.978
0.651
0.662
2.294
2.596
0521
0.684
0583
1.828
1.621
0.671
2.680
0595
0.491
0599
2.696
0.662
0588
0.668
0.798
0517
0.492
2528
2596
0.662
0504
0.723
2.541
-0.009
-0.029
0.001
-0.007
0.022
-0.008
0.012
0.007
-0.001
0.017
0.332
-0.053
0.100
0.061
0.541
0.590
0.000
0.155
0.096
0.660
0.114
0.116
0.002
-0.014
0.142
0.099
0.031
0.104
0.060
0.673
0.252
-0.001
0.019
-0.005
0.046
0.026
0.004
0.009
0.196
0.015
0.098
0.136
-0.106
0.119
0.054
0.052
0.099
0.015
0.067
-0.026
0.004
-0.009
0.009
0.016
-0.031
0.021
-0.006
0.001
0.037
0.263
0.022
0.062
•0.021
0.351
0392
-0.059
0.013
0.039
0.354
0.075
0.076
0.010
-0.001
0.044
0.072
0.005
-0.028
0.050
0.470
0.087
0.001
0.015
-0.025
0.029
0.010
0.011
0.002
-0.126
0.026
0.039
0.069
0.120
0.008
0.031
0.081
0.035
0.007
0.018
-0.020
-0.005
0.009
-0.004
0.007
-0.013
0.001
-0.014
0.005
0.021
0.224
0.030
0.040
-0.009
0.276
0.331
-0.012
0.078
0.044
0.139
-0.003
0.069
-0.010
-0.007
0.047
0.057
0.238
0.087
0.029
0.267
0.069
0.002
-0.003
0.006
0.015
0.025
0.006
0.014
0.074
-0.001
0.036
0.016
0.012
0.013
0.023
0.016
0.056
0.001
-0.020
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
1
NA
NA
NA
NA
NA
2
NA
NA
NA
3
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Lead Concentration
Run
31
32
33
34
40
41
42
43
44
45
46
47
48
60
73
31
76
32
33
34
40
41
42
43
44
45
46
47
48
49
58
59
60
61
62
63
64
65
66
67
73
74
75
76
77
78
79
31
80
Prep
Batch
502
502
502
502
502
502
502
502
502
502
502
502
502
502
502
503
502
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
503
504
503
Data
Instr Lead Pb Conc-
Instr Resp Sample Amount Recover Lead
Batch (ug/mL) Wgt fe) (UK) (UK/K) Comnt
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E09023B
E11053A
E09023B
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
El 1053 A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11053A
E11083A
E11053A
2.6404
1.6785
3.1479
0.4322
4.3170
0.5809
0.4036
3.6742
0.3010
23356
2.4423
2.1723
3.0518
2.9575
3.0671
03930
1.1783
1.1763
0.9297
0.4234
0.6896
0.9632
1.9394
4.6537
0.9883
0.6321
0.5938
0.2385
2.3475
0.6899
1.0456
0.1948
0.8788
1.6534
1.9281
1.7470
5.7564
2.8758
15296
4.9871
2.9894
35664
3.6968
3.0388
3.3300
4.2495
3.8264
3.2515
2.2853
0576
0.134
0.482
0386
1.195
1.637
1.000
4.149
1.198
3.224
1.329
1.360
1.713
1556
1.672
1.460
0.231
0.378
0.772
1.902
1542
1.619
0.453
0548
0.946
0535
0.128
0518
0.159
1362
1.139
0.623
1.743
0.540
0576
0.409
1.680
0.622
0.188
0.978
0.981
0.446
0396
1.725
1.842
1.058
0.897
0.906
1.615
330.05
41.96
157.40
21.61
539.63
72.61
50.46
1837.10
37.62
1167.80
61057
543.07
762.95
73938
766.78
49.12
29.46
58.82
116.21
105.85
172.41
240.79
96.97
232.68
12353
79.01
14.85
29.81
58.69
86.23
130.70
9.74
219.71
82.67
241.01
8735
71955
359.47
76.48
62339
373.67
17832
184.84
759.70
83250
531.19
47830
406.44
57132
573.003
313.153
326546
55.986
451569
44.354
50.455
442.781
31.404
362.221
459.424
399.320
445.388
475.177
458.597
33.645
127522
155595
150528
55.652
111.806
148.728
214.062
424.608
130.584
147.680
115.980
57.553
369.104
63.313
114.750
15.633
126.050
153.093
418.424
213570
428.304
577.934
406.809
637.411
380.912
399.821
466.768
440.406
451.954
502.068
533.222
448.607
353.762
NA
NA
NA
NA
NA
NA
NA
1
NA
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
E-3
-------
Table E-l: Housevac Data (continued)
Gravimetrics Data
Test
2109
2112
2018
2020
1056
1.31
1.32
2052
2064
2085
2088
2043
2044
2058
1015
1083
1069
1075
1076
1095
1042
1044
1065
1119
1038
2060
2089
2091
2101
2078
1079
1080
1111
1058
1077
2071
2022
2074
Date Team
9/2/93
9/2/93
9/3/93
9/3/93
9/3/93
9/3/93
9/3/93
9/3/93
9/3/93
9/7/93
9/7/93
9/7/93
9/7/93
9/8/93
9/7/93
9/7/93
9/7/93
9/7/93
9/7/93
9/8/93
9/8/93
9/8/93
9/8/93
9/8/93
9/9/93
9/8/93
9/8/93
9/8/93
9/9/93
9/9/93
9/9/93
9/9/93
9/9/93
9/9/93
9/9/93
9/9/93
9/9/93
9/9/93
2
2
2
2
1
1
1
2
2
2
2
2
2
2
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
1
1
1
1
1
2
2
2
Vac
A
C
B
A
B
A
A
D
A
D
A
D
B
D
A
A
C
B
C
B
B
A
B
D
B
B
D
C
A
D
B
D
A
B
C
C
B
C
Substrate Grindin
CRPT
CRPT
WOOD
WOOD
LINO
CRPT
CRPT
WOOD
LINO
LINO
LINO
CRPT
CRPT
CRPT
CRPT
CRPT
CRPT
LINO
LINO
LINO
UPHO
UPHO
CRPT
CRPT
CRPT
CRPT
UPHO
UPHO
UPHO
UPHO
WOOD
WOOD
WOOD
WOOD
WOOD
CRPT
CRPT
CRPT
YES
YES
NO
NO
NO
NO
NO
NO
NO
NO
NO
YES
YES
NO
NO
NO
YES
NO
NO
NO
NO
NO
NO
NO
YES
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
Amount
100
100
100
100
400
100
400
400
400
100
400
100
100
400
100
100
400
100
400
400
400
400
400
400
400
400
400
400
400
100
100
400
100
400
100
400
100
100
PbConc- Dust Size
Deposit (microns)
HIGH
HIGH
HIGH
LOW
LOW
LOW
LOW
LOW
LOW
HIGH
HIGH
LOW
LOW
HIGH
LOW
LOW
HIGH
HIGH
HIGH
HIGH
LOW
LOW
HIGH
HIGH
LOW
HIGH
LOW
LOW
LOW
HIGH
LOW
LOW
HIGH
LOW
LOW
LOW
LOW
LOW
150-212
150-212
106-150
150-212
150-212
<53
212-250
250-2000
212-250
150-212
106-150
53-106
53-106
212-250
<53
212-250
53-106
<53
53-106
250-2000
150-212
150-212
250-2000
150-212
150-212
<53
<53
212-250
212-250
150-212
212-250
106-150
250-2000
53-106
<53
106-150
150-212
250-2000
Time
13:35
14:40
9:35
10:48
10-.30
12:30
13:30
12:42
15:12
12:11
13:08
14:07
15:10
1157
9:40
11:05
12:30
13:45
15:15
10:00
10:50
12:45
13:45
15:00
8:30
13:15
14:13
15:15
8:28
9:40
9:40
11:00
13:00
1350
15:00
11:08
12:45
1351
Dust Gain- Dust Gain-
Initial Loadl Load Load 2 Load 2
Gain fe) Ife) fe) fe)
0.086
0.024
0.006
0.138
0.014
0.071
0.047
0.019
0.007
-0.025
0.008
0.111
0.000
0.189
0.111
0.010
0.121
0.005
0.021
0.003
0.104
0.027
0.047
0.183
0.417
0.059
-0.042
0.092
0.027
-0.031
0.008
0.017
0.021
-0.006
0.033
0.043
0.023
0.043
0.689
0.697
0.674
0.692
2.649
2.831
0.633
2.727
2.700
0.688
2.771
0.685
0.694
2.804
0.704
0.727
2.583
0.738
2.700
2.818
2.703
2.846
2.806
2.676
2.636
2.703
2.684
2.756
2.764
0.696
0.664
2.752
0.794
2.466
0.864
2.816
0.670
0.716
0.661
0.244
0.632
0.685
2.559
2.290
0542
2556
2.635
0.655
2.686
0580
0523
1.874
0560
0596
1.424
0.655
2.482
2.768
2.668
Z760
2.247
1.959
2.019
2.416
2.099
2.641
2.711
0.652
0.600
2524
0.781
2.269
0.756
2.100
0.481
0.505
0.685
0.694
0.676
0.685
2.980
2.801
0.616
2.758
2.705
0.624
2.744
0.688
0.682
2.685
0.655
0.721
2.742
0.674
2.608
2.725
2.882
2.853
2.714
2.734
2.707
2.748
2.706
2.800
2.740
0.689
0.657
2.736
0.730
2.755
0.713
2.728
0.682
0.716
0.630
0.368
0.624
0.666
2.900
2.495
0.706
2.637
2.636
0588
2.637
0584
0521
1.809
0532
0.604
2.003
0.615
2.420
2.689
2.858
2.812
2.040
2.164
2315
2501
2.145
2.702
2.694
0.641
0.642
2.662
0.730
2525
0.653
2.180
0528
0.461
E-4
-------
Table E-l: Housevac Data (continued)
Gravimetric* Data
Dust Gain - No Dust-
Load 3 Load 3 No Dust- No Dust- Gain 3 Dust
te) fe) Gain Ife) Gain 2 (g) (g) Comnt
0.683
0.693
0.679
0.673
2.737
2.957
0.788
2.739
2.693
0.688
2.725
0.684
0.675
2.741
0.639
0.716
2.758
0.713
2.761
2.836
2.687
2.655
2.820
2.740
2.698
2.699
2.685
2.685
2.799
0.679
0.658
2.635
0.717
2.770
0.710
2.738
0.675
0.688
0.640
0.372
0.646
0.659
2.681
2.605
0.388
2.593
2.642
0.660
2.613
0562
0537
2.053
0.505
0.628
1.949
0.674
2.604
2.739
2.660
2.616
2357
2.211
2331
2.458
2.256
2.596
2.710
0.638
0.634
2.496
0.718
2.640
0.628
2.185
0564
0513
0.071
0.108
0.003
-0.003
-0.058
0.098
0.126
0.067
0.012
-0.027
-0.009
0.017
0.052
0313
0.038
0.099
0.470
0.010
0.011
-0.018
0.056
•0.001
0.144
0.430
0.655
0.152
0.009
0.061
0.008
0.048
-0.005
0.031
0.004
0.012
0.002
0.203
0.068
0.032
0.000
0.080
-0.023
0.005
-0.016
0.053
0.044
0.008
-0.006
-0.004
-0.006
-0.020
0.005
0.268
0.010
0.040
0.186
-0.023
-0.014
-0.006
0.043
0.041
0.069
0.205
0.216
0.038
-0.036
0.030
0.018
-0.057
0.001
0.016
0.021
0.027
0.017
0.085
0.043
0.018
0.009
0.043
0.004
-0.002
-0.016
-0.006
0.055
0.008
-0.011
0.007
-0.039
0.045
0.015
0.218
-0.003
0.023
0.110
-0.007
0.023
0.013
0.043
-0.009
0.025
0.185
0.178
0.067
0.012
0.040
0.035
0.027
-0.006
•0.008
0.002
0.014
0.002
0.103
0.033
0.008
NA
NA
NA
NA
NA
4
5
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
6
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Lead Concentration Data
Run
81
82
91
92
32
33
34
93
94
31
32
33
34
100
106
107
108
109
110
111
112
113
40
41
42
114
115
124
43
44
45
46
47
48
49
58
59
60
Prep
Batch
503
503
503
503
504
504
504
503
503
506
506
506
506
505
505
505
505
505
505
505
505
505
506
506
506
505
505
505
506
506
506
506
506
506
506
506
506
506
Instr Lead Pb Conc-
Instr Resp Sample Amount Recover Lead
Batch (ug/mL) Wgt fe) (UK) (ug/g) Comnt
E11053A
E11053A
E11053A
E11053A
E11083A
E11083A
E11083A
E11053A
E11053A
E12023B
E12023B
E12023B
E12023B
E11083A
El 1083 A
E11083A
E11083A
E11083A
E11083A
E11083A
E11083A
El 1083 A
E12023B
E12023B
E12023B
E11083A
E11083A
E11083A
E12023B
E12023B
E12023B
E12023B
E12023B
E12023B
E12023B
E12023B
E12023B
E12023B
3.8362
3.0607
35074
0.3705
0.3781
1.0397
0.2526
0.2814
0.4255
6.1292
5.2082
1.2641
15475
5.0513
0.8042
0.2661
3.3998
1.5299
3.8394
6.2411
0.8920
0.6304
10.2920
5.6419
0.4604
2.4624
0.6795
05141
05753
3.1784
1.0438
0.6510
10.6960
1.7098
0.7802
03791
0.2166
0.1295
0.887
0.603
0.517
0.543
0.825
2.079
0.975
1.805
1.415
0.625
1388
0.470
0.421
1552
0.252
1.108
2.065
0.219
1.987
2.087
1.801
1.440
1543
1.912
1.664
1397
1.483
1.830
1536
0.954
0.815
1.781
1.276
1.427
0.118
1.405
0567
0.474
47953
153.04
17537
1852
18.90
259.93
25.26
35.17
53.18
306.46
651.03
63.21
7738
631.41
40.21
26.61
849.95
7650
959.85
1560.28
11150
78.80
2573.00
705.24
5755
615.60
169.87
64.26
71.91
397.30
130.48
8138
1337.00
213.73
1950
4738
10.83
6.47
540.614
253.789
339.207
34.115
22.912
125.024
25.907
19.485
37.585
490.336
469.038
134.479
183.789
406.838
159569
24.015
411598
349.292
483.065
747.616
61.911
54.719
1667531
368.848
34587
440.659
114547
35.113
46.819
416.457
160.092
45.691
1047.806
149.772
165.294
33.724
19.102
13.659
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
E-5
-------
Protocol: Sampling Housevac Exhaust Emissions
Revision No. 1
Date: September 24, 1 993
Page 1 0 of 10
Entry Sheets
for
HOUMVBC Tests
Test Sequence Number
Date
Operator
Project 9802
WA No. 55
.sr
Test
Housevac
Substrate
Grind-in
(A, B, C or D)
(TILE, UNOIeum, WOOD, UPHOIstery, CaRPeT)
4 (Yes, No)
Dust Amount ANCUOT- 1100. 400 mo/ft3
Pb Cone PFtoMtat (Low, High)
Dust Size Srzfe- «53. 53-106, 106-150, 150-212, 212-250. 250-2000)
Team TEAH (number 1 or 2)
Procedure
Perform the tests according to the housevac test sequence in Appendix P, and vac procedure in Appendix I
Tare weigh new bag:
Run free 40 sec, cool 2 mm, brush and record weight after 1 more min
Vacuum for 40 sec before any dust deposit
Reweigh bag (cool 2 min, brush and record weight after 1 more min)
Deposit dust vacuum 40 sec, weigh the bag. Total of 3 times. (Grind-in after each dust deposit, if applicable)
Repeat vacuuming only (vacuum 40 sec, weigh the bag) 3 times
Shake dust from the bag, weigh, prepare for lead analysis
Vacuum dust from wand and brush (no weighing)
Weight of Dust
(Balance f feQLDusr )
Weight of Bag
(Balance*
TareWt
om.
Tare weight of bag
Vacuum and weigh
Add dust vac & weigh
Add dust vac & weigh
Add dust vac & weigh
Vacuum & weigh
Vacuum & weigh
Vacuum & weigh
Dust sent to lab
Net Wt
Time
Weight
orp.
Increase
VACWTZ
V/AtUrtM
V*C*/T5
UAfcTTtft
Bar Code
for Sample
Bar Code
for Blank
Submit one blank for each week
Sample relinquished by
Sample received by
Date of transfer
Reviewed by _
Date reviewed"
------- |