EPA/600/2-85/011
February 1985
MEASUREMENT OF
VOLATILE ORGANIC
COMPOUND CAPTURE EFFICIENCY
Final Report
by
D. B. Hunt
and
J. L. Randall
Radian Corporation
8501 Mo-Pac Boulevard
Austin, Texas 78766
EPA Contract No. 68-03-3038
Work Assignment 10
Project Officer
Ronald J. Turner
Alternate Technologies Division
Hazardous Waste Engineering Research Laboratory
Cincinnati, Ohio 45268
HAZARDOUS WASTE ENGINEERING RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U. S. ENVIRONMENTAL PROTECTION AGENCY
CINCINNATI, OHIO 45268
-------
NOTICE
This document has been
Protection Agency policy and
names or commercial products
dation for use.
reviewed in accordance with U.S. Environment
approved for publication. Mention of trade
does not constitute endorsement or recomir.en-
ii
-------
FOREWORD
When energy and material resources are extracted, processed, converted,
and used, the related pollutional impacts on our environment and even our
health often require that new and increasingly more efficient pollution control
methods be used. The Hazardous Waste Engineering Research Laboratory assists
in developing and demonstrating new and improved methodologies that will meet
these needs both efficiently and economically.
Determination of overall emission reduction efficiency and regulatory
compliance status of some new surface coating operations requires knowledge
of the capture efficiency of emissions. A feasible method for measurement of
capture efficiency has been proposed by the U.S. Environmental Protection
Agency. It requires total containment of emissions from process lines. In
many web coating plants such containment has almost become the norm. The
Agency would permit the installation of temporary enclosures for purposes of
conducting the performance test. If a method for capture efficiency could be
developed which does not require construction of a temporary enclosure, that
expense could be eliminated.
This report presents the results of an investigation to identify and
evaluate acceptable alternate measurement methods. The results should be
of interest to regulatory agencies and industrial sources involved in VOC
emissions control. Requests for further information regarding this study
should be directed to the project officer at the Hazardous Waste Engineering
Research Laboratory, Cincinnati.
David G. Stephan
Director
Hazardous Waste Engineering Research Laboratory
Cincinnati
iii
-------
ABSTRACT
This research program was initiated with the overall objective of
determining an acceptable alternate method for measuring capture efficiency
of volatile organic compounds at surface coating operations.
An information search was conducted, and alternate methods were
conceptualized and evaluated for feasibility and applicability. The
alternate approaches included liquid/ga.s-phase material balances, use of
tracer materials, modeling, and other indirect methods.
i Several methods were considered feasible, although the liquid/gas-phase
material balance approach was preferred,. Therefore, the remainder of the
study focused on testing of the 1iquid/gas-phase material balance method. In
the liquid/gas-phase material balance technique, the amount of solvent used
at the coater is compared with the measured amount of VOC in the capture
stream. j
Both laboratory and field tests were conducted to fully evaluate the
methodology. Laboratory tests proved encouraging and identified no major
problems with the methodology using an EPA Method 25A gas-phase analysis.
Laboratory material balance closures averaged 99.9% for pure solvents and
102.2% for solvent and solid blends. However, field capture efficiency test
results were unexplainably high. Liquid/gas-phase closures employing Method
25A analyses in the field test averaged' 106.7% when determined for 24 hour
periods and 103.0% when determined for batch periods.
¦ i
Capture efficiency determinations by the liquid/gas-phase method at
sites with more complex stream configurations or more non-steady state
operations are expected to be more variable.
A comparative evaluation of the alternatives considered and the existing
gas-phase enclosure method concluded that there is not a readily available
alternate method that will produce equal or better results.
i . :
! This report was submitted in fulfillment of Contract No. 68-03-3038 by
Radian Corporation under the sponsorship of the U.S. Environmental Protection
Agency. This report covers the period of February 1982 through August 1984.
iv
-------
CONTENTS
Foreword iii
Abstract iv
Figures vi
Tables vii
1. INTRODUCTION 1
1.1 Background 1
1.2 Technical Objective 2
1.3 Program Approach 2
2- SUMMARY OF RESULTS, CONCLUSIONS, AND RECOMMENDATIONS 3
2.1 Results and Conclusions of the Feasibility Evaluation ... 3
2.2 Results and Conclusions of the Laboratory Testing 4
2.3 Results and Conclusions of Field Testing 7
2.4 Recommendations 9
3. REVIEW AND SELECTION OF ALTERNATE MEASUREMENT CONCEPTS 11
3.1 Review of Material Balances 12
3.2 Review of Tracer Concepts 20
3.3 Review of Modeling Concepts 25
3.4 Review of Approximation or Assumption Concepts 26
4. LABORATORY TESTING 29
4.1 Experimental Approach 29
4.2 Experimental Work 33
4.3 Experimental Results and Discussion 43
4.4 Quality Assurance and Quality Control 55
5. FIELD TESTING 57
5.1 Test Site Description 57
5.2 Testing Approach 62
5.3 Test Methods and Procedures 67
5.4 Results and Discussion 71
5.5 Quality Assurance and Quality Control 89
REFERENCES 95
APPENDICES 97
v
-------
' FIGURE'S'
i
I
¦Number
!
! 1
2
3
A
5
6
Page i
Typical Coating Operation StreainiFlow and Emissions 13
Test Apparatus for Capture Efficiency Laboratory Study 35
Detailed Plant-Wide Flow Diagram! 59
Coating Area Flow Diagram . • . I 63 |
Simplified Plant-Wide VOC Flow Diagram 65 j
Range of Error for Means of Multiple Measurements of Coating ,
Line Liquid/Gas Closure • . . . J 90
Response Factor Variations with Compositions for a Byron 401 . • . 93
vi
-------
"TABLE'S"
Number > Page
1 LIQUID GAS-PHASE TESTING RESULTS! 15|
i 2 SUMMARY OF VOC CAPTURE EFFICIENCY SAMPLING/ANALYTICAL METHODS ~ . 37}
3 TEST MATRIX FOR FLOW RATE INTERACTION EFFECTS 38!
4 TEST MATRIX FOR MASS BALANCE EXPERIMENTS 40
5 TEST MATRIX AND RESULTS FOR FLOW jRATE INTERACTION EFFECTS .... 44
6 ANALYSIS OF GAS PHASE FLOW RATE TEST RESULTS 46
| 7 LIQUID/GAS MASS BALANCE CLOSURES' 48
| 8 ANALYSIS OF MASS BALANCE TEST RESULTS 49
9 VOLATILE CONTENT OF COATING MIXTURES 53|
I 10 COMPARISON OF RF FACTORS ON CLOSURES FOR MEK/TOLUENE MIXTURE
I EXPERIMENTS [ 54!
I 11 SUMMARY OF SAMPLING/ANALYTICAL METHODS 68"
12 LIQUID AND GAS MASS FLOW RATES AND CAPTURE EFFICIENCIES FOR
THE COATING ROOM | 72
13 CAPTURE EFFICIENCY AVERAGES BY DAY 76^
14 CAPTURE EFFICIENCY AVERAGES BY SLURRY BATCH ll\
15 GAS-PHASE MASS FLOW RATES AND MATERIAL BALANCES FOR THE SLA
SYSTEM | 79!
16 SLA GAS-PHASE MATERIAL BALANCES USING DAILY AVERAGING PERIODS . . 82|
17 SLA GAS-PHASE MATERIAL BALANCES USING SLURRY BATCH PERIODS .... 83'
18 LIQUID MASS INVENTORY DATA . . . j 85'
19 INPUT VALUES FOR MONTE CARLO SIMULATION OF COATING LINE !
LIQUID/GAS CLOSURE MEASUREMENTS \ 88;
20 BYRON 401 RESPONSE FACTOR VARIATIONS WITH COMPOSITION 92-
-vii -
-------
SECTION 1
INTRODUCTION
The U.S. Environmental Protection Agency initiated the work described in|
this document to explore alternate methods for determining volatile organic
compound (VOC) capture efficiency in industrial surface coating facilities. j
The information obtained from this program will be used to provide additionalj
technical support for EPA regulatory activity in the area of VOC emissions from
surface coating operations. The data obtained will also augment the existing!
VOC control technology data base. j j
! The remainder of Section 1 provides information pertaining to the program
background, objectives, and study appro'ach. Section 2 presents a summary of J
the results, conclusions, and recommendations arising from the study. Section
3 reviews the potential alternate capture efficiency measurement techniques j
considered. Sections 4 and 5 present the experimental testing of the |
liquid/gas-phase material balance metho:d in the laboratory and under field j
conditions. i
1.1 BACKGROUND
In recent years, the U.S. Environmental Protection Agency (EPA) through
the Office of Air Quality Planning and Standards (OAQPS) has developed gaseous
emission standards for new surface coating operations. The new source |
performance standards (NSPS) regulate emissions of VOCs from these operations;
The term "surface coating industry" is meant to include any operation which j
applies by spraying or roll coating or some similar method a coating which is |
dissolved or suspended in a solvent. The solvent may be organic or aqueous or
a combination of both. !
l i
j Most of the surface coating NSPSs !are written as emission limits based on
the coating solids applied. Compliance with these standards may be met by
using low solvent coatings within the limits of the standard. Compliance may |
also be met by installation of add-on controls and demonstration of the overal'l
' . . . l
VOC reduction required to meet the emission limits. Some of the regulations, :
specifically those for roll coating operations, offer an alternative using |
add-on controls and demonstration of overall VOC reduction efficiency (ORE) in
excess of a minimal established value. Operations utilizing add-on controls j
must conduct an initial performance test demonstrating the system's ability to
reduce the emissions by either the calculated or minimum value. Compliance is
later continually determined by monitoring certain normal operating parameters!
of the process and control system, which are indicative of the control I
performance. I
! ? !
I Performance demonstration testing .is normally conducted using material ;
balance methods around the regulated process and control system# In many ;
surface coating plants, carbon adsorption or recovery type control devices are;
utilized because of the economic advantages. In these cases, performance
d'emons'tration" test ing ~usua'l'ly~"i"nvo'l'ves'~*a""li*q'uid"-phase material ba lance " around
1
-------
the coating line and control device, i.e.. "comparison of"liquid VOC used with
that recovered, over a 30-day period.
Other plants utilize destruction type control devices and cannot conduct a
liquid-phase material balance for demonstration of performance. Any test
method for demonstrating ORE, other than a liquid-phase material balance,
requires a knowledge of the fraction of emissions actually captured by the
capture system, as well as the fractional destruction of the captured emissions
-by the control device. The fractional capture is called "capture efficiency"
and the fractional destruction is called "control device efficiency"* The
product of the two yields the overall reduction efficiency. ;
Capture efficiency has been attempted by several methods in the
development of the NSPSs. The only method that has produced satisfactory !
results is the gas-phase material balance method, utilizing an enclosure around
the regulated coating line to contain and direct fugitive emissions to a single
measurable stream. Where the regulated coating line is not isolated within a,
permanent enclosure, the EPA has indicated in some standards, such as that for
ihe pressure sensitive tape and label industry, that a temporary enclosure
should be erected and used to collect fugitive emissions for measurement.
This method may be costly and inconvenient. Therefore, this study has been
initiated to explore possible alternate; methods for determining capture
efficiency, which might be less costly and more convenient*
.2 TECHNICAL OBJECTIVE
The objective of this study was to develop an alternate technique for
measuring VOC capture * efficiency. The capture efficiency, as determined by !
this technique, and the control device efficiency, as determined by previously
developed techniques, may then be combined to determine the overall
effectiveness of a control system on emissions from a surface coating
operation.
1.3 PROGRAM APPROACH 1 !
i ;
j The study was designed and conducted in two phases. Phase I was primarily
a review of existing information/recent studies, and development and evaluation
of the feasibility of alternate assessment methods. Phase II was the detailed
testing phase of the project and involved testing the most feasible method, as;
determined from Phase I.
i 1
i
2
-------
SECTION" 2
1 SUMMARY OF RESULTS, CONCLUSIONS, AND RECOMMENDATIONS
i
\
i
1 In the study of alternatives for determining process capture efficiency,
alternate concepts were reviewed and evaluated for feasibility, and the most
acceptable alternative was tested* There were three major segments of the
study: ,
| • Review and evaluation of alternatives, 1
j • Laboratory testing, and \ |
[ • Field testing. j
I i
2.1 RESULTS AND CONCLUSIONS OF THE FEASIBILITY EVALUATION
I
i |.
• Alternate measurement concepts reviewed and considered included:
' i
; . I !
; • Material balance methods, I
• Tracer concepts, i j
• Modeling, and !
I • Other approximation techniques- I
i !
2.1.1 Results of the Feasibility Evaluation !
A review of available information .sources regarding alternate measurement
methods found surprisingly little supportive information in any area other than
material balances and tracers. Of the 'alternatives reviewed, several were '
I . I
considered feasible, but the liquid/gas.-phase material balance and the tracer1
gas concepts were the most acceptable approaches.
| A review of past tests involving Liquid and gas-phase measurements was !
inconclusive regarding the applicability of liquid/gas material balances to '
field determinations of capture efficiency. However, the majority of the tests
were not specifically designed around a liquid/gas material balance, involved;
short non-steady state test periods, or' had other shortcomings in the I
measurement of the mass flow variables.'. Several tests involving longer test !
periods and semi-steady state operating conditions did yield results which I
appeared to be representative of the capture efficiency. Therefore, the
potential applicability of this method seems limited to either steady-state or
relatively long term test conditions. -.Like the gas-phase methodology, the I
liquid/gas-phase material balance approach also can only be applied to j
operations which do not use zone incineration or direct firing of the drying I
oven. j
j Inert tracer gases, such as sulfur.; hexafluoride (SF^), have been used !
extensively in other air movement and monitoring studies. Some preliminary
studies indicate that the tracer gases might be used successfully to determine
hood capture efficiency and, in turn, the process capture efficiency. The
determination of hood capture efficiency appears very straightforward using
-------
tracer "ga'ses. However, determination" of'process capture ef ficiency' also
requires knowledge of the mass of emissions generated from the various process
segments, which is generally not available. A potential advantage of the
tracer gas method is the potential application to all operations, including
those employing zone incinerated or direct fired drying ovens.
2.1.2 Conclusions of the Feasibility Evaluation
! A comparison of the material balance and the tracer gas concepts concluded
that both methods might be feasible and potentially applicable, but that the
material balance approach is generally more acceptable. This conclusion was i
based on the premise that direct measurements are more readily accepted than I
indirect determinations, and that the material balance methods are further
developed than the tracer methods. Therefore, a decision was made to conduct
further evaluations and developmental testing of the liquid/gas-phase material
balance methodology. 1
1
, The experimental testing of the 1 iquid/gas-phase material balance method,
was divided into two phases: (1) laboratory testing and (2) field testing.
The laboratory testing was necessary because of the difficulties encountered in
assessing the test methodologies and results from previous field tests. It was
considered strategically important to begin testing under the simplest !
conditions and work up in complexity, and to be able to control and measure all
streams for verification of test results. The laboratory approach met both of
these criteria. ;
! ;
2.2 RESULTS AND CONCLUSIONS OF LABORATORY TESTING
j The 1iquid/gas-phase material bala,nce method of determining process VOC
capture efficiency was evaluated through a series of laboratory experiments.
The variability to be anticipated from the test methodology was investigated in
order to assess the actual feasibility and expected test performance under the
most ideal field conditions. ¦
I I :
1 (
| Individual methods were also evaluated in the two major sets of
experiments conducted, as shown below: ' j
i
Gas Flow Rate Experiments j
; • EPA Method 2 for the determination of gas flow rates, !
I Mass Balance (Evaporation) Experiments i
I ; 1
1 • EPA Method 25A for VOC determinations by total hydrocarbon analysis,!
' • EPA Draft Method 18 for VOC determinations by GC/FID, |
• NIOSH Method 127 VOC determinations by charcoal tube adsorption, and
i • EPA Method 24 for determination of volatile content of coating
j mixtures. * '
i «- 1
1
1
-------
2v2rl~ Gas Flow Rate" Experiment'Results ' - - .
The gas flow rate experiments were conducted according to a fractional
factorial test matrix# The objective of the study was to evaluate the main and
low order interaction effects associated with variables involving EPA Method 2
type flow rate determinations. The variables tested included pitot tube type,
the pressure measuring device, the sampling location, the relative flow rate,.
and the personnel conducting the test. ;
, Analysis of the gas flow rate data indicated excellent precision for all:
measurements with coefficients of variation not exceeding 4%. The greatest |
variability in the data was seen at the lowest flow rate, where minor
discrepancies have a more pronounced effect.
i i
! Analysis of variance (ANOVA) techniques were used to test the statistical
significance of first and second order |effects from each of the variables (e.g.
instrumentation used, location of measurement, operator). In only a few cases
were any of the interaction effects considered statistically significant. The
analysis indicated differences of marginal significance (<^=.10) in the data
obtained from different type pitot tube's at the lowest flow rate and from
differences in pressure measuring devices at the middle flow rate (oi=.05). The
combination of S-type pitot tube and Magnehelic® Gauge produced significantly
different data (a =.01) from other combinations at the highest flow rate. j
I Overall, no single variable or combination of variables appears to have !
any practical significance. However, i't is considered advisable to avoid j
combining the S-type pitot and Magnehelic®, particularly at higher flow rates.1
: | j
2.2.2 Mass Balance (Evaporation) Experiment Results
The mass balance experiments provided an opportunity to assess the ability
to close liquid/gas material balances using the available methodology, while
conducting comparative testing of three' different VOC analysis methods and
testing of a method for the volatile content of the coating. Varied solvent !
amounts, solvent types, and gas flow rates were introduced into the testing to
simulate some important field variables'. I
The test system was designed as a Jsimple flow-through evaporation chamber,
thus providing 100% capture and minimizing the number of measurement !
parameters. Therefore, a known-capture; efficiency was established for
comparison with the measured and calculated values, and sources of error in
measurement were reduced to the lowest "level.
I
j Method 25A closures were by far the most successful in this set of
experiments. Accuracy, bias, and precision were evaluated for the pure solvent
and the commercial mixture experiments .as separate groups. Accuracy seems more
than adequate for each group since the mean closure values were, res pect ive ly ,¦
99-9% (88.5-110%) and 102.2% (93.5-119). Bias for either group was not
statistically significant, since 100% was included within the 95% confidence
interval in both cases. Precision, or variability, estimates for the two
groups-were exce-l-lent>—s-i-nce- t-he coef f i-c ients- of-variat ion were 5.9- and -8 .9%-,-
5
-------
respectively. Test results for the commercial coatings test runs were less
accurate than pure solvent runs, probably due to the smaller change in liquid
mass and lower SLA concentrations.
2 The solvent amount was the only variable found to have any statistically
significant impact on the Method 25A data. As a practical matter, this
marginally significant effect would be difficult to explain since the mid-level
(1.0 liter) results seem to deviate from the low and high level results. Thus,
hone of the variables nor combination of variables seems to have any great
effect on the closure technique resultsj.
Method 18 closure data were evaluated for accuracy, bias, and precision i!n
the sane manner as the Method 25A data.; Based on mean closure values of 93.6%
(73-264%) and 106.8% (86.2-186%) for the pure solvents and commercial mixture-
\ • » . 1
tests, respectively, the method appearejd to be only marginally accurate. The!
wide 95% confidence intervals (83-1 to |104.2% and 71.1 to 142.6%) tended to j
discount bias as the primary problem. iThere was not sufficient evidence in the
Method 18 closure results to indicate significant main or interaction effects,1
at least to the i = 0.10 level. Howeve'r, the variability estimates as measured
by the coefficient of variation indicated that the particular methodology or |
instrumentation used was not adequate for measuring VOC capture efficiency j
during the laboratory testing phase. i
i . i
| Two possible explanations were postulated regarding the wide ranging
Method 18 data First, detailed review' of the THC concentrations for each run
showed variations with substantial up and down swings. The 7-minute interval ¦
data obtained by the Method 18 analyzer, very likely missed a number of the
peaks and/or valleys. Second, during s'ome of the runs, it was apparent that i
the sensitivity of the Method 18 analyzer changed during the run. At one |
point, the analyzer's detector jet was mechanically cleaned and the sensitivit'y
rose dramatically. The extent of this problem was not quantified.
: I
j Full statistical analysis was not performed on the charcoal tube data.
However, the data indicate that the method is not adequate for performance
testing. The most serious problem associated with this method seems to be its'
variability. As measured by the coefficient of variation, this was 62.2%. The
variability could not be explained based on the available data.
I '
2.2.3 Conclusions of Laboratory Experiments
The laboratory testing of the liquid/gas-phase material balance methods
concluded that reasonable closure results and capture efficiency determinations
can be obtained under controlled flow conditions. Gas flow rates determined by
EPA Method 2 are not significantly impacted by the instruments used nor the
operator, and they are suitable for use' in liquid/gas material balance type
capture efficiency determinations. EPA-. Method 25A was found to be the most
suitable method for measuring gaseous VOC concentrations of the three examined.
Based on these findings, it was recommended that the methods be further
evaluated in a rather non-complex field, environment.
-------
213 RESULTS AND CONCLUSIONS OF FIELD TESTING
VOC process capture efficiency determinations were
surface coating plant manufacturing magnetic tape and us
solvent system. Mass balance closure techniques and liq
methodologies were employed. Major efforts were focused
i
| • Coating Area Capture Efficiency Measurement,
; • Error Analysis in Capture Efficiency Determinations, and
i • Other Material Balances* \ ;
i 1 I
2.3.1 Results of Coating Area Capture Efficiency Measurements i
i i ;
? Measurement of VOC mass flow ratesj were made for the liquid (slurry)
applied and the captured gas from the coating area of the plant. Liquid
measurements included slurry feed tank >volume, slurry density, and slurry
volatile fraction. Gas phase measureme'nts of the capture stream included flow
rate, VOC concentration, temperature, and pressure. :
i i
Over a test period of 114 hours, consisting of 6 discrete batches, hourly
determinations of VOC capture efficiencies varied significantly using both EPA
Method 25A and Draft Method 18 data for, total gaseous hydrocarbon (THC). J
However, when averaging the data over approximately 24 hour periods, |
variability decreased. Over the entire' test period, the mean capture i
efficiency was 106.7% with a coefficient of variation (CV) of 7.4%, using
Method 25A data. Using Method 18 data,! the mean capture efficiency was 105.1%
with a CV of 1.3%. Evaluating the data| on a batch basis led to a mean capture
efficiency of 103.0% and a CV of 4.6%, !using Method 25A data. In each case the
calculated mean capture efficiency was higher than expected.
I 1
i In order to estimate the reliability of the capture efficiency results
obtained, a complete error analysis was. performed. Estimates for measurement!
errors were provided by an external audit or determined from repetitive
measurements. ;
I Average values and estimated errors were input to a Monte Carlo simulation
to obtain a series of 400 coating room closures that were statistically related
to the plant data set. This exercise indicated a relative bias of -3.9% in the
coating room closure data. Furthermore, the analysis showed that a single |
measurement has a 38% probability of being within +5% of its true value. ;
Finally, if 3 individual sets of closure measurements are made, the average j
value should be within +10*7% of the true value with a 95% confidence limit. !
For six measurements, the average valuerwould be within +7.6%. Using the data
collected at this site, it appears that, the limits can only be narrowed to i
approximately +4% of the true value. I
i ~ »
1 '
| An obvious discrepancy existed between the high results obtained in the 1
field test and the negative bias indicated by the error analysis. However, th'e
indicated bias in the error analysis comes strictly from an audit of the total;
hydrocarbon analyzers and does not include any opposing bias that might be
made at an industrial
ing a two-component
uid/gas conversion
on:
.7
-------
associated with the other measurements. Based on a review of • the--process
streams measured, the measurement methods, and quality assurance program, there
is no reason to believe that any source of VOC went unmeasured or that any
measurement bias existed that would cause the higher measurement values.
Therefore, the discrepancy is unexplainable.
2.3.2 Results of Other Material Balance Determinations
! Liquid inventory data were obtained for all storage and process areas.
The density and volatile fraction of each liquid were also determined.
| Gas-phase determinations were made of VOC content, flow rate, temperature
and pressure for the storage tank vent (collection stream, the main solvent
laden air (SLA) stream, and the carbon bed stack exhaust in addition to the
same measurements for the capture stream as described in the previous section.
VOC measurements were made by EPA Metho'ds 18 and 25A. 1
I
i
| The data were evaluated through various combinations to determine the j
acceptable methods of viewing capture efficiency and/or mass balance closures.'
One of the most exhaustive investigations involved a comparison of the main SLA
stream mass flow rate with the sum of the capture and vent stream mass flow
rates. These two comparative masses we're expected to be equal, since the vent:
and capture stream combine to form the main SLA stream. Results of this
i . i
comparison yielded 94.3% closure with a coefficient of variation of 4.3% for
Method 25A data and 24-hour averaging. For the Method 18 data, average closure
was 99»4% with a 9.1% coefficient of variation.
i
Further evaluation of this data on' a slurry batch basis produced closure
results of 92.9% with a coefficient of 'variation of 2.5%, using Method 25A
data. Method 18 data provided an average closure of 96.3% and a coefficient of
variation of 8.8%. The evaluation on a batch basis did not provide noticeable
improvement. This is probably related !to the continuous steady state nature of
the vent stream.
i .
j A potential source of error in the determination of the mass flow rates
and mass balance closures above may be associated with the Method 25A analyses
based on an erroneous assumption that all streams contained a 75/25 ratio of
the two solvents. The composition of the vent stream is known to have varied :
from this ratio to a higher ratio, causing measured values to be biased low.
This potentiaL source of error possibly, explains part of the discrepancy in the
^ethod 25A mass balances, but it does not appear significant enough to account
for the total differential.
1 ¦- !
I A plant-wide liquid inventory was conducted over the test period based on
measurements at approximately 8-hour intervals. Added to the measured liquid •
mass were the only known fugitive sources, the carbon adsorber stack outlet,
the solvent recovery water discharge, and the floor sweep exhausts within the .
plant. The purpose was to develop a total accountable VOC inventory. Running
a mass balance closure indicated that 95% of the VOC had been accounted for.
Other than product retention, an adequate explanation for the discrepancies
could-not-.be found.- - ~ — ......
8
-------
Several other comparisons of data were made. Many of these involved the
amounts of liquids transferred through metered pipes. In most cases, the
multiple combination comparisons indicated fairly wide error margins, in the
20% range.
2.3.3 Conclusions of Field Testing
It was concluded from the field test that estimates of the VOC capture
efficiency around a non-complex system can be obtained within approximately
+10% of the true value, based on an error analysis conducted on the field test
in this study. The accuracy of the determinations can be improved only to a :
limited degree by increasing the number, of tests. It is expected that |
variability of data will increase with :the complexity and non-steady state ]
conditions at most plants causing even wider confidence intervals.
1 i ;
The confidence interval is expected to be much narrower for the same site
using calculations based on the EPA proposed gas-phase material balance and an
exhausted measurement enclosure. This lis primarily due to the form of the j
gas-phase material balance equation. In the equation, the numerator and !
denominator are both composed primarily- of the same measurement value,
therefore, minimizing the impact of measurement bias or variability on the
results. !
] i
2.4 RECOMMENDATIONS ' 1
I
* i
, Presently, there is no apparent alternate method for measuring VOC capture
efficiency which would produce data equal or better in quality to that of thej
gas-phase enclosure method. However, the gas-phase methodology, like the j
iiquid/gas methodology, is not applicable to operations employing direct fired
ovens or zone incinerated ovens. In these cases an alternate approach must be
used unless permanent total enclosures are required. ¦
In this regard, criteria must be developed on which to base the
acceptability of an alternate method. iThis criteria should establish the j
required accuracy of measurements and confidence limits for calculated capture
efficiency values* ; !
1 ; '
i From this study, it appears that a; tracer gas method would be the most I
acceptable alternative. However, this methodology is not fully developed or
tested. Steps in the process of developing such a method should include the j
following: .. J
i • Identify or establish mass enfission factors or rates for typical |
| surface coating VOC sources. j
i l*'" !
j • Conduct a laboratory comparison between the tracer gas mass balance 1
; approach and the tracer gas relative measurement approach. i
i
i
9
-------
Conduct a laboratory ~eva luat ion-to determine the "most acceptab'Te*
tracer gas release technique (e.g. manifold, single point release,
multiple individual releases at different points on a source
surface).
Determine the. number of tests, release points, release rates, and
measurements required to obtain the desired accuracy.
Conduct a test involving direct flame contact with the SF^ to
evaluate the impact of direct fired ovens on test results.
10
-------
SECTION 3
REVIEW AND SELECTION OF ALTERNATE MEASUREMENT CONCEPTS
In the process of reviewing and selecting alternatives, four discrete
tasks were conducted:
• Identification and review of pertinent literature and other !
information. i
; t
• Review of past EPA sponsored material balance tests and test methods
• Formulation of conceptual techniques for determination of capture
efficiency. : ;
• Evaluation of the feasibility, of alternate techniques for the |
, determination of capture efficiency. i
J These tasks involved the search and review of literature in 9 computer
bases and off-the-shelf literature, numerous telephone contacts with
knowledgeable authorities, review of 31 field tests conducted on surface
coating facilities, and field visits arid reviews at 10 industrial surface
coating sites* Information obtained during these review and evaluation steps'
and the contractor's in-house technical; knowledge led to the formulation and
the feasibility evaluation of the conceptual techniques.
! The computer bases searched included the following:
j • Engineering Index
• Surface Coating Abstracts j
• NTIS '
j • APTIC J
• NIOSHTIC
• Pollution Abstracts |
• Conference Papers Index 1
• SSIE (Research in Progress)
!
1
Sources of information which were selected and reviewed from these bases and
from the shelf are presented in Appendix A. i
! ; ;
¦ The review of previously conducted field tests are summarized later in ;
this section. Individual reviews of al-1 tests are presented in Appendix B. .
i
I The methods conceptualized and reviewed included both direct and indirect
methods. Direct methods are material balance techniques, where the masses of 1
VOC in specific streams are measured and compared in order to calculate capture
efficiency. Indirect methods utilize techniques where the VOC masses are not
measured (e.g. tracers, modeling, or other approximation techniques).
i;
-------
¦ - The folVowing sections present "brief descriptions of "the'potential
alternate methods, evaluations of the methods, and conclusions regarding the
feasibility and applicability of each method to surface coating operations.
3.1 REVIEW OF MATERIAL BALANCES
Possible material balance methods ^applicable to capture efficiency
determination are gas-phase material balances and liquid/gas-phase material :
balances. All previous EPA sponsored work and tests involving capture !
efficiency determinations have utilized material balance techniques.
Figure 1 shows a typical surface coating operation and the VOC mass flows
in and around the coating line and control device. This figure and the denoted
VOC streams will be referred to throughout the following discussions. j
i i i
3.1.1 Gas-Phase Material Balance \
] The gas-phase material balance method using a permanent or temporary j
enclosure for measuring emissions is presently the only method recommended by'
EPA for demonstrating capture efficiency. From past efforts and evaluations,:
it is obvious that a gas-phase balance >can only be conducted using an exhausted
enclosure, as described by the EPA. Therefore, no alternate gas-phase testing
methodology is proposed. Referring to 'Figure 1, capture efficiency by the j
gas-phase method is calculated as shown in Equation 1 below:
C !
Capture Efficiency (%) = X 100 (1)
C + C'
3.1.2 Liquid/Cas-Phase Material Balances
In the liquid/gas-phase material balance approach, the mass of liquid VOC
applied at the coater is used to represent the mass of total emissions.
Therefore, the methodology involves a comparison of liquid VOC used at the
coaters and the gaseous VOC emission captured and directed to the control ;
device. Referring to Figure 1, capture efficiency by the liquid/gas material!
balance method would be calculated as shown in Equation 2 below:
I : :
I c •
I Capture Efficiency (%) = X 100 (2)
i B i
i
The liquid VOC (solvent) mass used-' at the coater can be determined by j
monitoring the mass or volume of coatings used during the test and by j
determining the VOC content of the coating material. The VOC mass flow in the
capture stream can be determined by measuring the gas flow in the stream and
the VOC content of the gas.
i
i This particular approach to determine capture efficiency is rather simplei
in concept since it requires the measurement of only two streams, liquid
solvent in and gas captured-. It also requires--none of the- time or expense
¦12
-------
FUGITIVE EXHAUST
TO ATMOSPHERE-C
CAPTURE SYSTEM
EXHAUST-C
HOODED PROCESS
EXHAUSTS-C"
OVEN
EXHAUST-C'
COATING SOLIDS
AND POSSIBLE
SOLVENT-A'
OVEN
BLENDING
COATED
PRODUCT
-E
SOLVENT APPLIED
AT THE COATER
-B
COATING
SOLVENT
VOC
CONTROL
DEVICE
VIRGIN
SOLVENT
RECOVERED SOLVENT
(WHERE APPLICABLE)
- F
W
FUGITIVE VAPORS
MAKEUP
AIR
CLEANED
EXHAUST
-D
Figure 1. Typical coating operation stream flow and emissions
-------
involved -wi th -erect ing-an-enclos ure, nor-does it involve the concerns that may
arise with regard to creating an altered test environment potentially
associated with a temporary enclosure test.
3.1.2.1 Evaluation of Methodology—
Theoretically, the product of the liquid/gas-phase capture efficiency and
the control device efficiency is equivalent to the liquid-phase material
balance equation for determination of overall reduction efficiency. Both
concepts similarly ignore any VOC retained in the product and assume it to be ]
part of the emitted VOC. i I
; ! I
| From an error analysis viewpoint, the capture efficiency determinations j
using the liquid/gas-phase equation are highly susceptible to variability. ]
Both variables in the equation will mos.t likely be large in magnitude, and the
calculations using this equation will be nearly equal in sensitivity to both, i
Since the values of the variables will he determined by different measurement;
methods on different streams, the impact of measurement inaccuracies may (
compound in the final results. Therefore, it can be concluded that accurate j
determination of capture efficiency by 'the liquid/gas-phase approach depends 1
critically on the precision and accuracy of the individual measurements.
In evaluating the test methodology, thirteen tests involving
1iquid/gas-phase measurements were reviewed, ten of which had adequate data to
attempt a capture efficiency determination. Table 1 presents a summary of the
material balance closure results obtained from these tests. Reviews of the
individual tests are presented in Appendix B. In most tests reviewed, the I
capture efficiency results did not compare well to expected efficiencies or
efficiencies determined by other proven methods (e.g. long term liquid-phase J
testing of overall efficiencies). However, several tests did appear to provide
accurate representations of the expected efficiencies. ;
j In many cases, there was not substantial detail to determine the reasons !
for the test failures. However, from the information in some of the tests the
following were derived as potential causes of variance: j
j • inadequate characterization o'f solvent and coating composition, |
: • difficulty in obtaining accurate measurements of liquid coating j
material used at the line tested, j
• use of intermittent gas stream test methods and monitoring at streams
| with non-ideal locations, and: J
i • short term tests involving non-steady state conditions. i
i . -¦ '
Each of these is discussed to some degree in the following paragraphs.
j Characterization of coating solvent content—Since the solvent content of
the liquid coating represents one portion of the determination of the VOC mass
used (total VOC emissions), the accuracy of this value is critical. Many of «
the plants tested used manufacturers data and plant formulation records for
determining the solvent (VOC) content of the coating rather than conducting
separate analyses. Past experience has. shown that manufacturers' data are not
14
-------
TABLE 1. LIQUID GAS-PHASE TESTING RESULTS
Plant
Industry
VOC
Type(s)
Control Method
Capture Control Overall3 Time
Efficiency Efficiency Efficiency Averaging
(X) (X) (X) Period
Method of Analysis Reference
Publication
rotogravure
Toluene Carbon adsorption
74.2C 124.5C 92.4C NA Analysis of gas
streams with semi-
continuous FID and
grab sample CC;
analysis of liquid
streams for solvent
content•
Publication
rotogravure
Toluene, Carbon adsorption
xylene
58.7C
144.3C
84. 7C
NA Analysis of gas
streams with semi-
continuous FID, CC/
FID, and Method 25;
analysis of liquid
streams for solvent
content •
11 Paper
Toluene,
ethyl
acetate,
isopro-
panol
Carbon adsorption
94
98
92
6 hours Analysis of gas
streams with Method
25; analysis of
liquid streams for
solvent content.
20 Vinyl
NA
Carbon adsorption
6o.5fc
NM
NM
50 hours Continuous FID
analysis of gas
streams, usage rate
and solvent content
of ink.
23 PSTL
NA Thermal incineration 40/159c><*
carbon adsorption
NA
NA
NA Bag samples with FID,
semi-cont i nuous FID,
and EPA Method 25 for
gas-phase analysis
24 Metal coil
NA
ThermaI inci nerat ion
NM
NM
99.9C NA GC/FID of incinerator
ef fluent; soIvent
usage rate.
25 Metal coil
Aromatic Thermal incineration NA
hydro-
carbons,
glycol
ethe re
NA
NA
NA
Analysis of gas
streams with
continuous FID and
EPA Method 25
26 Paper
NA
Thermal incineration 94-98
94-96
91-93 3 hours
Analysis of'gas
streams with Method
25; analysis of
liquid streams for
solvent content.
-------
TABLE 1. LTQUTD GAS-PHASE TESTING RESULTS
(continued)
Plant
Industry
VOC
Type(s)
Capture
Efficiency
Control Method (Z)
Control Overall®
F. f f i c iency Efficiency
(Z) (Z)
Averaging
Period
Method of Analysis
Reference
27 .
Paper
Toluene,
others
Carbon adsorption NM
79 NM
6 hours
Analysis of gas
streams with Method
25; analysis of
liquid streams for
solvent content.
3
28
Metal can
Cello-
solve
acetate,
butyl cel-
losolve,
xylene,
i6ophorone
others
Thermal incineration 221c
t
73 NA
NA
Analysis of gas
streams with Method
25; analysis of
liquid streams for
solvent content.
3
29
Metal can
Xylene,
diacetone
a lcohol,
MIBK, iso-
phorone,
butyl cel-
losolve
Thermal incineration 17AC
Catalytic incineration 90
26 NA
49 44
Analysis of gas
streams with Method
25; analysis of
liquid streams for
solvent content.
3
8
30
Beverage can
Cello-
solve
acetate,
others
None 87*8
NM NM
195 rains .
Analysis of gas
streams with Method
25; analyis of
liquid streams for
solvent content*
9
31
Beverage can
NA
Thermal incineration 79
NM NM
137 mins •
Analysis of gas
streams with Method
25; analysis of
liquid streams for
solvent content.
9
a Overall Efficiency - Capture Efficiency X Control Efficiency
^ data are reported to be inconclusive
c questionable data
^ FID analysis of gas stream
NA = data not available
NM = not measured
PSTL = pressure sensitive tapes and labels
FID = flame ionization detection
I GC = gas chromatography
| MIBK - methyl isobutyl ketone
-------
always~accurate and'are often generalized. On the otherhand,' plant formulation
records provide accurate coating data if the materials are measured with
reasonable precision. In either case, the solvent content at the time of
mixing does not necessarily accurately represent the solvent content evaporated
in the coating room, particularly if the coating is held for any period of time
before transfer.
i
Another typical source of error related to solvent content may be involved
when the coating is recirculated between the coating reservoir and the coating
pan. Use of manufacturer or plant formulation data combined with the amount of
coating used will not reflect the VOC Lost from any coating not applied but
continuously recirculated, the significance of this undetermined volume is
unknown.
i
j Addition of solvent to the coating reservoir for viscosity control is |
common and may be frequent at some plants. These added amounts also must be j
considered in the solvent content of the coating. Because of these adjustments
and the other potential sources of error above, it seems critical in most short
term tests to monitor the solvent content of the coating frequently. '
J Mass of liquid coating used—The amount of liquid coating lost or used |
represents the other half of the critical liquid VOC measurements. A number of
options exist for measuring the liquid use including: determination of the j
applied coating weight or thickness and; the web speed, depth gauging of coating
vessels and subsequent volume determinations, ard metering of the coating !
components Co the coating line. All of: the coating flow measurement techniques
are made difficult by the complex flow arrangements (e.g. recirculation and !
multiple vessels in the flow loop). ¦
i !
Depth gauging of coating vessels is by far the most popular and most j
applicable method in the tests reviewedj and plants visited. Depth gauging can
be applied to process configurations employing direct delivery of coatings to
the application point or those utilizing batch vessels, coating pans, and
recirculation at the coating line. Thel precision and accuracy of the volume !
determinations by depth gauging are uncertain and are definitely affected by }
the symmetry of the vessel, the depth to width ratio of the vessel, and the |
presence of objects occupying volume wi|thin the vessel (e.g. agitators, pumps,j
and hoses). No specific studies have been conducted to determine the effects i
of these items, nor do test reports contain sufficient detail regarding these
measurements to make an evaluation of their impact on the test measurements, j
In most plants visited, it was agreed that objects occupying space within the j
depth measurement zone are fairly insignificant (e.g. agitator blades) or are j
removable during the measurement period;. '
The applied coating thickness or weight may be determined in several j
different ways. Most of the coating thickness gauges produce relatively '
imprecise data, with the possible exception of the continuous beta gauges. 1
Applied coating weight determinations reportedly have been successful where the
coating weight is significant with respect to the weight of the substrate, and
the coating weight is determined by the difference in the weight of a web roll1
before- and after -coating-.- Using the applied coating weight -or-thickness and ¦
17
-------
the associated solvent content of the coating is only representative of the
total VOC emissions if the coating material is delivered directly to the
application point without recirculation. Otherwise, the amount of applied
coating will not represent the evaporated solvent lost from the coating vessel
or from recirculation at the process line. Visits at a variety of surface
coating plants indicate direct feed of the liquid coating without recirculation
is not typical.
Metering of coating material to the coating line for determining the
volume of coating material subjected to vaporization may have limited
application in some industries. One magnetic tape plant indicated that the
application of meters may be limited. The company claims that they are not
suited for handling highly viscous materials or abrasive components (e.g. iron
oxide in magnetic tape coating), A metal coil coating plant representative had
concern that the accuracy of measurements would be affected by changes in the
material density. However, in this study some publication rotogravure plants
were found to use metering of coating components to the line, since they use
large quantities of the same components. Metering, in these cases, is limited
to individual component lines feeding intermediate mixing or holding reservoirs
at the line. Reasonable accuracies have been experienced in metering the com-
ponents to the vessels, but depth gauging or another technique are ultimately
used to determine the volume or mass actually transmitted from the vessel to
the coater during the test period unless start and stop of the test are
scheduled at identical reference points (e.g. when the intermediate reservoir
is filled).
An untested option for measuring flow of liquid coating material to the
coating line and obtaining initial and final measurements is a mass determina-
tion approach. Rather than measure the volume of the coating vessel at the
start and finish of the test, the vessel could be weighed using the same basic
flow theory as that in the volumetric approach. Mass measurements at the start
and finish of the test could be made if the vessel is placed on a scale or
balance or mounted on load cells. Advantages of the mass measurement method
over the volume determinations are that the measurements are not affected by
the dimensions or symmetry of the vessels or by items occupying space in the
vessel.
Gas-phase flow rates and concentrations—The mass of gaseous VOC captured
is determined by the product of the gas flow rate measurements and the measure-
ments of the VOC content of the gas stream. In most tests reviewed, the VOC
concentration was determined by collection of full term integrated samples
(EPA Method 25) or by semi-continuous monitoring (EPA Method 25A), and the gas
flow rate was typically determined by one or two pitot traverses during the
test period.
In most capture systems the flow rate remains fixed while the concentra-
tion fluctuates due to process fluctuations. As long as the operating
conditions are nearly steady state, semi-continuous monitoring and continuous
sampling will account for any minor concentration peaks or valleys that may
occur. However, where non-steady state conditions exist, semi-continuous
monitors may not pick up rapid peaks and valleys and a variance may result in
the mass flow determinations.
18
-------
-0khe'r capture sys't'ems ," non'i tors "a'nd~~automa t ic controls hold " the
concentration of the gas stream steady by varying the gas flow rate. Since the
gas flow is typically not measured continuously, significant fluctuations could
take place unnoticed creating a major variance in mass flow determinations, but
they may be significant. Where testing involves capture streams with
potentially fluctuating flow rates, continuous monitoring of the flow rates
(velocity) will provide the proper adjustment factors for mass flow
determinations.
i From the tests reviewed, it was nojted that some did not follow EPA
reference methods outlining selection of gas flow measurement points in J
nonturbulent zones. The accuracy of pi'tot measurements may be severely \
affected by turbulent conditions. There was not sufficient information in each
test to assess the presence and effect :of this potential source of variance, j
but test results involving poor measurement locations may have been
significantly impacted. From plants visited during this study and past j
experience, it is realized that it is not unusual to encounter poor measurement
locations in existing systems. However, NSPS performance demonstration testing
will be conducted on systems which will have measurement locations meeting EPA
optimal criteria, as required by NSPSs,: and in those tests it should not |
present a significant source of error, j ]
! [ '
j Length of test—Most of the tests jreviewed were conducted over relatively
short test periods involving non-steadyi state conditions. When the sample
periods were relatively short (e.g. 3 h'ours), the tests with the greatest \
success were those which were conducted; during steady state operations. One I
test conducted over a relatively short period under steady state operations j
produced acceptable test results (3). I I
! . . i !
! Another test involving test periods of different lengths and non-steady 1
state conditions produced results indicating that closure data become less ;
variable when accumulated and averaged 'over longer test periods (4). In this]
test, it appears that the variability i'n closure results reached a relatively!
insignificant level at approximately nijne hours. It is expected that the j
required averaging period will differ for operations involving different !
operational parameters. ' j
j ¦ |
3.1.2.2 Assessment of Feasibility— i
; It was concluded from the review that the liquid/gas-phase material
balance approach is feasible and may prpduce capture efficiency results of
acceptable accuracy. However, previous, limited testing has not produced
conclusive results either negative or positive, and most tests were not [
designed as liquid/gas material balance's.
I . . ... . !
I Determinations of capture efficiency by this method are heavily affected i
by any variance of values within the eq-uation, which are in turn impacted by •
the measurement...of the individual tes t .parame ters . A number of sources
potentially causing inaccuracies in liq'uid and gas measurements have been .
outlined in the previous section. Each' of these potential sources is 1
considered correctable to some degree. . Short term testing under non-steady :
-------
s tat e~" cond'it ions, however, appears to compound these variances."" Both short
term testing under steady state conditions and long term testing under
non-steady state conditions appear to be potentially feasible in producing
acceptable results. The steady state conditions during short term tests tend
to reduce any drastic stream fluctuations which might not be detected or which
might create less accurate measurements. The longer test periods, where
non-steady state conditions are encountered, tend to provide a greater
averaging period for measurements thus reducing the impact of measurement
variability under fluctuating conditions.
; There is one additional limitation of the liquid/gas-phase approach, which
also affects all methods requiring the measurement of the captured gas stream
(e.g. gas-phase approach). It cannot be applied to operations employing direct
fired ovens (which destroy a portion of; the VOC), where the oven is in line
with the capture stream and is located ,upstream of the first accessible
measurement point on the capture stream^. In this case, a material balance
determination of capture efficiency usijng capture stream measurements is
technically infeasible because some of ithe VOC is destroyed prior to reaching'
the measurement location. j
j I ;
3.2 REVIEW OF TRACER CONCEPTS 1 i
: j
Three general classes of tracers were considered: (1) liquid tracers, (2)
gaseous tracers, and (3) aerosol tracers. They are described and evaluated in
the following paragraphs. 1
3.2.1 Liquid Tracers
i i i
j Liquid tracer concept would make use of a unique volatile liquid having
vaporization characteristics similar to the volatile components in the process
being evaluated. The tracer would possess unique qualities allowing it to be '
detected accurately at low levels and to be distinguished from emissions ;
emanating from adjacent operations not involved in the test. A known quantity
of a volatile liquid tracer material (e.g. fluorocarbons, tagged or labeled i
compounds, or radioisotopes) would be injected into the coating and allowed to
disperse. The gaseous tracer mass would then be measured in the capture stream
during the test period and compared to the mass of the injected liquid tracer1
material, following procedures similar to a liquid/gas-phase material balance.'
3.2.1.1 Evaluation of Methods-- I
| The liquid tracer approach was conceptualized during this study and, as
far as could be determined through a literature search and other direct
contacts, it has not been previously attempted or evaluated for quantitative
determinations. j
I
i
Because of the similarity of this"method to the liquid/gas material 1
balance, it would inherit the same difficulties and possible sources of
variance. The only obvious advantage is that this method would distinguish
between the emissions from the test source and those from adjacent non-test i
s ources.
20
-------
Stable- tagged or labeled compounds were the only materials identified-as
likely candidates to be used as tracers. The use of these compounds would
require custom compounding of liquids representing the properties of each
solvent used throughout all coating industries. Stable labeled compounds would
also require analysis by mass spectrometry. Another concern with liquid
tracers is that they may bond differently with the other coating components
causing the tracer to release differently than the normal solvents. \
3.2.1.2 Assessment of Feasibility— '
I . • .
» The feasibility of this approach technically is questionable since it is
unknown how the materials will react wi;th the other coating components. From •
discussions with industry representatives, they would only be agreeable to
injection of the foreign substance into! the process if they are provided
assurances that it will not affect the product quality. I
! . . . i . i
I Economically, the liquid tracer concept is considered infeasible because!
of the costs associated with custom compounding of tracers. Also, the cost of
mass spectrometric analyses associated with stable labeled compounds will :
probably be cost prohibitive. j j
I ' !
i Finally, this approach would be more costly and difficult than a typical i
1iquid/gas-phase material balance and would provide essentially the- same j
results. |
3.2.2 Gaseous Tracers
I The gaseous tracer concept takes a different approach from that of the
liquid tracers. Since the gas cannot by itself simulate the variable rate ofj
vaporization of the volatile liquid organic material used, a two step approach
is necessary: (1) determination of the1 efficiency with which emissions are •
captured from each major emission point and (2) determination of the VOC :
emission rate at each emission point or the relative contribution of each !
segment to the total emissions,
i !
] The first step requires that the tracer gas be released at the surface of
each unenclosed emission point and the amount collected in the capture stream I
be measured. This comparison or determination of the hood capture efficiency,
might be accomplished by either of two methods. The first is an absolute mass
balance determination, and the second is a relative determination,
j
j In the absolute mass determination, a known mass of the tracer gas would
be released at the emission point. The total mass of tracer passing through
the capture stream would be measured as: in a mass gas flow measurement. Then,j
the two masses would be compared to determine the capture efficiency. ,
j In the relative determination, the; tracer gas would be released at a
constant rate and the discharge nozzle placed inside the capture system ducting
to assure 100% capture. The capture stream direct reading monitor scale would
be marked to reflect this detection level as 100%. A linear scale would then'
be established between 0 and the 100% level. The tracer gas would then be *
released-at the same set cal-i-bration flow- rate-only at the surface of the
21
-------
emirssi'orr source this time. The capture stream monitor records the percentage
captured relative to the calibrated scale.
i The second step would be determination of the fractional emission
contribution or the actual emission rate for each emission segment. This could
either be based on generalized emission factors for emission source segments or
components for each industry category or on site-specific determinations of
emission factors or rates.
I The final capture efficiency determination for the total source (entire ,
coating line) would then be made by weighting the capture efficiencies for each
emission source segment (e.g. the product of the fractional contribution and j
the average capture efficiency for each segment) and summing the weighted !
capture efficiencies.
i . !
3.2.2.1 Evaluation of Methodology— j
| Gaseous tracers, such as sulfur hexafluoride and radioisotopes, have been
used rather extensively as tracer materials for air flow studies and to some
degree in capture efficiency studies (1.0-15). Sulfur hexafluoride (SF^)
seems to be the gas of choice in most s,tudies because of its low level of
detection by electron capture, its inert characteristics, its thermal ,
stability, and its low toxicity. SF^ has typically been used in outside j
dispersion studies and air exchange or 'dilution studies indoors. !
| i
I Only limited work has been conducted using SF^ for quantitative
determinations of capture efficiency (1'3, 14, 16). However, none of these
studies has involved determination of capture efficiency for an overall process
line comprised of multiple emission sources (e.g. open coating vessels, open
coating pans, coated web flash off area,, and coated web oven zone), each
possessing a different emission rate and a different fractional contribution of
the total emission. These studies havej found that the use of SF^ for
determination of capture efficiency produces reliable results, when releasing;
the tracer gas at a fixed rate from varjious points along the emission source. :
They have not, however, determined a statistical basis for the methodology '
(e.g. the number of release points required to accurately characterize the j
average capture efficiency of a source)] nor answered many other important \
questions required prior to the general1 application of this method. One minor
concern with this method is the amount :of time and effort that would be !
required to fully characterize all of the unenclosed emission segments with the
capture system operating under the same, conditions and influences (e.g. same
flow rate and same similar web speeds)/.
The release of SF^ at a constant rate and the measurement of relative 1
responses in the capture stream seems to have several advantages over the ;
method requiring an absolute mass comparison. It only requires measurement and
recording the relative responses detect'ed in the capture stream when the tracer
reaches steady state flow conditions, while the second method requires
determination of flow and concentration" of both the gas release stream and
capture stream with time and full integration of the measured results.
Therefore, the potential sources of error and the time per test are reduced.
The—de t ermi nat ion-of-rel-a t ive--res ponses—a-lso should negate any-effect of
-------
react"! vi'ty~""( chemical "or thermal") which might occur. It is likely that this
type of test might also be applicable to process lines involving direct fired
or zone incinerated ovens, where the captured gas stream cannot be measured
before the oven.
In order for this test method to be applied to capture efficiency as it
pertains to NSPSs for surface coating operations, emission factors (relative or
absolute) must be known for each source segment (e.g. coater and oven). 1
Industry representatives contacted were unaware of any existing data
characterizing the emissions from different source segments. NSPS background;
information documents for some coating industries (e.g. metal coil coating) do
contain some very general emission factors or ranges of factors for the major,
emission segments. ; ;
i !
1
I There are two possible approaches ;to defining emission factors for each !
source segment. The first is a general; approach of developing emission factors
for emission source segments within each industry category. The second is a \
direct approach of determining the emission factors for each site based on the
site specific conditions.
| In the general approach, the emission factors could simply be determined ¦
based on industry wide measurements of emissions contributed by the type of J
source. These general emission factors' would be determined on a one time basis
and would be applied to all plants within that industry category or within
individual subsets of an industry category. The obvious disadvantage of such
an approach is that it would provide a very broad approximation of actual
emissions where significant changes could occur from one line to another or
from one process to another due to many variables including solvent type, line
speed, and coating thickness.
The second approach, development of emission factors at the test site,
would provide emission factors based on the site specific operations and
configurations. Two techniques for determining site specific emission factors
have been considered: (1) direct measurement of emission levels or relative j
emission levels at different points on each source segment and (2) development
of emission factors per unit of exposed surface area for the specific coating
material being used and for the conditions under which it is being used (e.g. ,
temperature, velocity, and humidity). The applicability of the first technique
was evaluated during site visits at various surface coating plants during this
study and was found unacceptable. Attempts were made to determine relative
emissions from different points along the process line by conducting
measurements of VOC concentration and velocity using direct reading instruments
(e.g. TLV Sniffer® and velometers). In-'the cases evaluated, the measurements ,
of concentration and velocity were generally unstable and inaccurate, and the'
emission surfaces along the web were often difficult to access. The
feasibility of the second technique was" not fully evaluated during this phase
and was only conceptualized. This technique might involve the use of a small
evaporation test chamber and a sample of the liquid coating material being used
in the process. The test chamber could' be used either in the field or at a
laboratory. Emission data could be developed in terms of mass of VOC emitted
per unit-surface-area-of liq uid - coat ing--per-peri od--of time. Test -data of this
23
-------
type ' could -be"developed • using a- tes t -chamber- for each ¦ type- of ¦ coa ting or -— ¦
solvent mixture and plotted against varied conditions (e.g. temperature,
velocity, and pressure). This type of data would not be applicable to curing
surfaces (e.g. drying web); however, the drying web is located inside a totally
enclosed oven.
The overall advantages of the gaseous tracer measurement concept are that:
(1) it could provide a quantitative value of capture efficiency on any coating
line (including those with zone incinerated or direct fired ovens), (2) it j
would not require the installation of a temporary test enclosure, and (3) the.;
test could be conducted independent of the length or steady state 1
characteristics of the process being tested. The potential disadvantages are!
the unknown length of time that would be required to make the test (especially
if detailed emission factors must be determined) and the unproven methods being
considered.
i I
I !
3.2.2.2 Assessment of Feasibility— ;
| Based on the review conducted, the- use of tracer gases to determine |
capture efficiency might be a feasible approach. However, the potential I
applicability of this approach would be heavily dependent on the development of
acceptable emission factors or methods for determining emissions for each j
emission source. The use of SF^ tracer gas for measuring the capture |
efficiency of a single hood has been demonstrated in research by other groups |
for other purposes. It would certainly require some further development or
refinement before it could be applied to the testing of a process line for !
compliance determination purposes*
i
i This method is not envisioned as a potential replacement for the direct !
material balance techniques, where they are feasible. However, the tracer gas
method might provide an option to constructing an enclosure if the method could
tie developed. If so, it might provide a test method for determining capture
efficiency on process lines utilizing zone incinerated or direct fired ovens. |
3.2.3 Aerosol Tracers j
j The aerosol tracer concept follows! an approach identical to that of the |
gaseous tracers. The only difference in the approaches is the release of an |
aerosol rather than a gas. Therefore, no further description is provided here
of the method or procedures. !
1 ¦ :
3.2.3.1 Evaluation of Methodology—
! Aerosols used as tracers (e.g., uranine dye and ammonium chloride) have [
also received substantial attention in ;the past for air flow testing. Many of
the reported tests and uses have involved dispersion testing or other ambient I
air flow problems. Of the material reviewed, only one source discussed using i
aerosol tracers to determine capture efficiency on exhaust hoods (17).
I Although the aerosol tracer concept is similar to the gaseous tracer
concept and has many of the same advantages and disadvantages, it has a
distinct disadvantage associated with closed flow systems. Deposition of the •
aerosol can occur on surfaces- contacted (17).—With t-he - number-of turns- and - -
1U
-------
other-potential deposition surfaces that would be encountered prior to
measurement in capture systems, the measurements may be reduced in accuracy.
There is also the underlying concern that an aerosol will not behave in
flow identically to VOC gases or vapors. This is not believed to be a major
factor because of the disperse nature of the aerosols that would be used. ]
3.2.3*2 Assessment of Feasibility--
The use of aerosol tracers is considered potentially feasible. However, .
there is no apparent advantage and several disadvantages in comparison with the
use of tracer gases. .
i I
3*3 REVIEW OF MODELING CONCEPTS
|
i The modeling approach assumes the use of a code (model) that, when input
with the required conditional data, will define the capture efficiency of VOC;
emissions under the given emission, control, and site conditions through the j
use of a computer simulation. In orderj to represent capture efficiency as \
defined in this study and some NSPSs, codes are required to describe two
separate activities which interact. These two activities are: (1) the
formation of induced air flow patterns ;in and around the exhaust hoods and
process line and (2) the evolution and mobility of process VOC emissions. |
j Once the computer code or program (is selected, the model can be applied t!o
determine capture efficiency by conducting a site visit to collect the
conditional input data. |
3.3.1 Evaluation of Methodology ; i
i i !
! A review of available literature and contacts uncovered no existing models
specifically designed to meet the needs of this project. Most existing models
addressing the mobility of airborne masses are related to activity in the j
outdoor ambient atmosphere and not the jcomplex internal atmosphere of an
industrial plant. Additionally, there are no known models describing the j
movement of air in three dimensions around exhaust openings of varying
geometry, nor has the theory of industrial ventilation been developed to
address capture efficiency as defined in this study. Therefore, this approach
would require the development of new novels.
Model development would have to take into account every variable which
could possibly affect the capture system flow patterns, VOC evolution, and
mobility of the emitted vapor. The National Institute for Occupational Safety,
and Health (NIOSH) considered development of such models several years ago but
quickly abandoned the idea when the.required degree of effort was evaluated.
| If models were developed to depict the activities described, the combined
model must be verified or adjusted before it could be legitimately applied.
This process would require a significant amount of field testing to
statistically validate the model. The -field testing activities would have to ;
be conducted by valid and accepted methods, which at this time is limited to
the—gas-phase- material--ba-la-nce—methods-.- -- — - -
25
-------
" Because of the large number of affecting variables that would'change' from
operation to operation and plant to plant, the field task for gathering input,
data would still be significant in order to apply the model. The amount of
time required to collect input data tends to reduce the intended efficiency of
the model approach. The types of data and measurements that may be required I
from each plant site include: j
i
i
• types of solvent used and their characteristics, '
• surface area of product throughput as a function of time,
• product throughput speed, j i
1 • type of applicator (e.g. spray or roll), 1
• application rate, ; :
1 • surface area of applicator and reservoir exposed to the atmosphere,]
' • length of the process line (or product) through each segment, !
• surface temperature of the emission source at each segment of the ,
J process line, !
• velocity of any extraneous air currents (e.g. makeput air or dryer j
j air) impinging on the coated Iproduct and the particular surface of i
j the product impinged, j
, • location, configuration and dimensions of each hood (including all 1
i openings),
I • location, configuration, and dimensions of all air flow obstructions,
1 • volumetric flow rate through ieach exhaust hood, j
• portions of the process enclosed by any exhausted enclosure,
; • velocities and area of influence of any other external air currents.
! within the process area but not impinging on the coated product, and
1 • relative location of each emission source surface with respect to
j each exhaust point (three dimensional coordinates).
3.3.2 Assessment of Feasibility
Theoretically, it appears that a model could be developed to represent
capture efficiency. For modeling to be advantageous, it should require a field
testing and analysis effort to develop linput data, which would be no more
rigorous than that required for direct Itest methods. The effort that would be
required to develop and verify an accurate model, based on all of the i
potentially affecting variables, would ^probably be much greater than that 1
required to directly measure the capture efficiency of all new sources. There
seems to be no apparent advantage to thjis approach.
3.4 REVIEW OF APPROXIMATION OR ASSUMPTION CONCEPTS !
j Assumption concepts would involve the use of logical assumptions and j
non-quantitative methods to conservatively approximate the capture efficiency!
This type of approach may make use of visible chemical smoke. The smoke could
be used to visibly demonstrate capture and control of emission source segments
of the operation, which are exhausted by partial enclosures or external hoods'
to the control device. Source segments of the process would include coating
feed tanks, coating applicators, the web flashoff area, and the dryer.
In this approach, smoke would be released at the surface of the emission
source—segment or released—at -openings on-a partial enclosure around a segment.
26
-------
I'f "the "sinoke " ls"-vi sib'ly 'observed" to" be" drawn entirely "into the' "capture "arid"""
control system, then the segment may be considered totally controlled. In the
case where all emission source segments are visually demonstrated to be i
controlled, the capture efficiency may be assumed to be 100% or equivalent to ia
total enclosure. !
In cases where total control is demonstrated only on certain source i
segments and not the entire operation, an "all or nothing" approach might be '
used, whereby only totally controlled segments are given credit as being j
captured. To obtain a quantitative capture efficiency value for the entire i
operation, the relative significance of each totally controlled emission source
segment to the total operational emissions would have to be defined. The j
overall capture efficiency would be determined by summing the percentage J
contributions of each controlled source segment, as shown in the following
hypothetical example:
• Emission
! Source Segment
Emission
Contribution
Controlled
(Yes/No)
Total Emissions
Controlled
Coating Reservoir 2%
Coating Pan/Applicator 5%
Web Flash-Off 8%
Dryer/Oven 85%
No
No
Yes
Yes
0%
0%
8%
85%
I TOTAL
100%
93%
3.4.1 Evaluation of Methodology
j The use of this approach was conceptualized in this study and was not "
reported in the literature. However, the use of smoke tubes is a common ;
inexpensive practice to evaluate the performance of air movement and collection
systems, 1
| Smoke tubes can be used to effectively assess the total control of
emissions from a given point. Smoke tubes were used successfully to observe !
capture of emissions at numerous plantsj visited in a variety of surface coating
industries. There were concerns among ,only a few plant representatives
initially that the acid base smoke might affect the product quality if used 1
near the web. However, the smoke was used with their cooperation in the
vicinity of the product in all plants with no problems. The density of the
smoke does not necessarily represent the density of saturated solvent laden
air, but in spaces normally occupied by-.,workers the air movement is great ;
enough to control employee exposures and the solvent laden air is quickly
diluted and approaches the density of clean air at ambient conditions. ;
The application of the assumption concept appears fairly straightforward |
in the determination of total operation: control and could be easily implemented
to demonstrate total control, where applicable. Where only certain source
segments are totally controlled, there would be a requirement to develop
emission factors for each specific coating operation and source segment.
27
-------
3-4.2" Assessment of Feasibility
These conceptualized methods are viewed as feasible, particularly as a
screening technique for total control. Based on the conservative estimates of
capture efficiency that might be obtained at partially controlled operations, ;
the costs associated with the development of emission factors may not be
warranted. However, if emission factors are developed as an integral part of j
the tracer gas methodology, smoke tubes and the assumption concept may be used
effectively as a screening technique before deciding to test with the more
expensive tracer gas technique. I
i
3.5 SELECTION OF A PREFERRED METHOD j j
, i
i Several test concepts or methods were found feasible based on the reviews
in the previous sections. However, no single method was fully developed nor |
applicable to all possible test conditions and scenarios. Therefore, the
selection of a single preferred method to receive further testing was made
difficult. | j
[ The liquid/gas-phase material balance approach was selected as the \
preferred methodology based on the premise that direct measurements are more I
readily accepted for compliance demonstration than are indirect ones, and that
material balance methods are generally
further developed. Additionally, fromj
the review it was concluded that the liquid/gas-phase approach had not received
sufficient testing to conclusively evaluate the methods.
! It was concluded that further testing of the liquid/gas-phase methodology
was necessary to either validate or invalidate it as an acceptable alternative
to the gas-phase approach. It was further concluded that this testing should
be conducted in two steps: (1) laboratory testing and (2) field testing. 1
1 ! ;
I The laboratory studies (step 1) were considered necessary because of the;
complex array of measurement points, uncontrollable operational variables, and
difficulty in measuring VOC fugitive losses and product retention at even the \
simplest field setting. It was strategically important to begin testing under
the simplest conditions and work up in complexity, and to be able to control i
and measure all streams for verification of test results. The laboratory
allows both of these criteria to be met1. i
j
I
i
i
i
28
-------
SECTION "4f
LABORATORY TESTING
The objective of the laboratory experimental work was to assess the
reliability of the available measurement methods in determining a
liquid/gas-phase material balance under controlled laboratory conditions.
The conclusion of this work is an assessment of the alternate methods and
any limitations observed in the laboratory setting.
i •
i This section provides a description of the laboratory experimental
approach, the experimental work (apparatus, procedures, and methods), test
results, data quality, and conclusions-!
! I
4.1 EXPERIMENTAL APPROACH I
i . i
j From the review of the following subsection and the discussion of j
liquid/gas-phase material balances in Section 3 it should be obvious that j
field test conditions can be very complex- Even the simplest field test |
conditions are considered too complex of a setting to adequately assess the j
available measurement methods. Therefore, the laboratory tests were j
designed to allow a controlled evaluation of the available methods.
4.1.1 A Typical Liquid/Gas Material Balance
: i
' A liquid/gas material balance requires the measurement and comparison
of the liquid VOC mass used at the coat!ing operation and the gaseous VOC
mass captured from the operation and sent to the control device. The
capture efficiency is calculated as the1 percent ratio of captured VOC mass
to liquid VOC used (vaporized) as presented earlier in Equation 2.
i ;
i i
| The mass of liquid VOC's entering ithe system during a material balance
test period is determined by maintaining an inventory of the liquid usage
and determining the VOC content of the liquids. The liquid use is typically
determined by weighing aliquots, metering streams, or depth gauging tanks.
The liquid input includes the addition |of solvent for viscosity adjustment
or wash solvents used during the test period. Measurement of the liquid VOC
input in the field is often complicated! by recirculating coating streams,
multiple coaters on a single production- line, lack of measurable streams in
the coating area (e.g. the coating tank] which feeds the coater is located in
an area outside the test boundaries), and multiple coating lines (regulated
and unregulated) sharing the same coating tank.
| During the time that liquids are Heing processed, the flow rate and VOC
concentration of the captured gas stream(s) are measured to determine the
mass of captured VOC. The flow rate and concentration are monitored on a
continuous or semicontinuous basis because of fluctuations that typically
occur. Measurement of the captured VOC from the coating operation may be
complicated by any of the following:
29
-------
• poor flow measurement locations in a section where the flow
profile is nonuniform,
• addition of VOC to the capture stream from operations or sources
outside of the test area, or
! t
• widely fluctuating gas stream conditions (concentration and flow)
; from process starts and stops or control adjustments. 1
£.1.2 Experimental Design ' j
In order to effectively evaluate the performance of the analytical j
methodology, most of the variables experienced in the field were eliminated j
or strictly controlled during the laboratory testing. The test system was j
designed as a simple flow-through evaporation chamber, providing 100% \
capture and reducing the number of measurement parameters to the lowest !
possible state. Therefore, a known-capture efficiency was established for •
comparison with the measured and calculated values, and sources of error in j
measurement were reduced to the lowest .level. Only field test variables
that were easily simulated and controlled or that might directly affect the !
statistical evaluation of the methodology were allowed in the experimental j
design.
i ! !
i The experimental work was designed to evaluate specific test methods i
under a range of measurement conditions simulating some of the variable test !
conditions that might exist in the range of industries that could be tested.
The experiments were designed to involve numerous test runs varying these
conditions, while maintaining constant the measurement methods and the
system configuration.
i i ,
i For practical purposes and the elimination of an additional set of test .
data required during the mass balance experiments, a separate set of
experiments was designed to test the gas flow rate measurement methods.
Therefore, the experimental work and design is divided into two separate but !
related investigations: (1) gas flow rate experiments and (2) mass balance j
(evaporation) experiments. The designs; of these individual studies are '
further described in the following paragraphs.
4.1.2.1 Gas Flow Rate Experiments— ! !
| This study was designed specifically to assess the sensitivity of the
gas flow rate determinations to various- aspects of the velocity measurement
methods. Some work has been done towards investigating the variability of
the EPA method for gas flow rates which- indicates sufficient precision for !
capture efficiency determination (18, 1-9). This study should conclude in |
determining the best procedure and circumstance to be used for all velocity i
traverses in the later laboratory and field studies. !
Variations in pitot tube type, pressure measuring device type,
operator, flow rate and sampling location were included in the study to
determine their individual and combined effects on gas-phase flow rates. ;
-------
The following is a'listing of'the variables and the' specific levels
selected for testing.
Variable Levels to be Tested
2 (S-type and standard type)
2 (Magnehelic® and Inclined Manometer)
3 (20, 30, and 40 feet/second)
2 (Optimum and non-ideal)
3 (Field Test Engineer, Chemist, and
Technician)
The study was designed to utilize fractional factorial statistical
techniques to investigate the precision; of the flow rate mesurements. This
type of design is advantageous where there are several variables affecting a
measured parameter. For any test, the measured quantity is dependent on one
or more variables, each of which can exist in one or more states. Thus, if
X variables can have two possible states and Y factors can have three
states, an investigation of all possible combinations would require 2^ •
3y experimental runs. Through fractional factorial design techniques, it
is possible to conduct only a portion of the possible experiments without
sacrificing important information. The ability to estimate main effects and
low order interactions of interest is retained. However, higher order
interactions cannot be determined. Since high order interactions are
somewhat unusual and often difficult to interpret, this sacrifice seems
reasonable.
I
| A full investigation would have required 2^ * 3^ or 72 individual
traverses. Assuming that three factor and higher order interactions are
negligible, the number of traverses was; lowered to 36. Simultaneous
measurements proved to be impractical. : The test matrix is presented in
Section 4.2 (Table 3).
j i
4.1.2.2 Mass Balance (Evaporation) Experiments—
i This set of laboratory experiments: was designed to assess the
reliability of the best available liquid and gas-phase measurement methods,
and to evaluate their applicability in the determination of liquid/gas
material balances. !
! !
j The design of this investigation involved the testing and evaluation of
the measurement methods as a set and in. some cases on an individual basis.
The set of measurement methods for liquid and gas parameters are ultimately
evaluated on their ability to close thejoverall material balance. Where
several potentially acceptable methods existed for the measurement of a
given parameter, the tests were designed to conduct parallel measurements
under the same test conditions in orders to make a comparative evaluation of
the individual methods. The bias involving the total set of measurements is
strictly controlled by the design of rigorous quality control procedures for
each individual measurement method employed.
Pitot Tube
Pressure Measurement Device
Flow Rate
Traverse Location
Personnel
-------
The experimental-test matrix was designed to' evaluate the test methods
under a range of measurement conditions simulating some of the test
variables that might exist in the range of industries potentially tested.
Numerous test runs were designed into the matrix in which the measurement
conditions were varied, while maintaining the same system configuration and
measurement methods. The variables and their designed test levels are:
Variable Levels to be Tested
i
1 Composition of Organics 3 (pure solvent, binary mixture,
commercial coating)
j Mass Throughput of Organics 3 (0.5, 1.0, 2.0 liters)
| SLA flow rate 3 (20, 30, 40 ft/second)
A full investigation of all variables and their interactions would
require well over 100 individual evaporation tests. By making simplifying
assumptions and statistically evaluating the data, the number of tests was
reduced. For instance, it was assumed that data from an experiment using
only MEK would be equivalent in its contribution to the variability in a run
using only toluene. A full test matrix is given in Section 4.2 (Table 4).
Each of the test variables is discussed below.
I Composition of Organics—Surface coating operations cover a multitude
of process environments. The variety of solvents and solvent mixtures used
is very wide. However, two solvents, toluene and MEK, are in common usage
throughout the industry.
1
¦ These two VOCs were chosen to represent the solvents in coating
liquids. Toluene (BP = 231°F), with a lower explosive limit (LEL) of 14,000
ppm and a specific vapor volume of 9.2 ft^/L was chosen as a less volatile
compound than MEK (BP = 176°F), which has a specific vapor volume of 10.8
ft^/L and an LEL of 18,000 ppm. In addition to tests with pure solvent,
mixtures of pure solvents and a commercial coating mix were tested.
I The varied composition of organics in the liquid mixture or coating
material primarily presented a test of the measurement methods for gas-phase
VOC concentration.
\ Mass Throughput of Organics—In many plants, the surface coating
operation is somewhat of a batch process and process runs vary in length.
Changing coils or webs, product changes-, and periodic equipment failures are
some of the factors involved.
i
' To simulate variability in the length of runs, the amount of liquid
solvent placed in the evaporation chamber was varied from 0.5 to 2.0 liters.
This variable may have an impact on all' measurements because of the
averaging effect associated with time.
32
-------
It•is-anticipated that a VOC concentration versus time profile'wil1' ~
produce a rapidly rising front area, a relatively stable center area, and
finally, a slowly declining tail end. It is further anticipated that a plot
of incremental solvent weight loss versus time will produce a similar curve.
Under these circumstances, the data would provide knowledge of the relative
contribution of the non-steady state conditions (both ends) to the overall
uncertainty in the mass balance closure as the length of measuring time is
varied.
! For each experiment, the data from various time increments was to be
examined separately from the data analysis of the entire run. Ultimately,
the data may provide an estimate of the; minimum testing time required for
adequate characterization at actual plant sites under ideal conditions.
SLA Flow Rate—Throughout the industry, flow rates of the solvent laden
air (SLA) vary depending on blower size', duct dimensions, and flammability
characteristics of the solvent. The effects of SLA flow rate on mass
balance closure are examined by testing at three SLA flow rates. Nominal
values of 20, 30, and 40 ft/sec are selected. These represent typical flow
rates found in surface coating plants, j
I :
! In the laboratory studies the variation in flow rate will also affect
the rate of vaporization and the diluti'on of the volatilized solvent, thus
affecting the SLA concentration. I
i ;
! Liquid Organic Surface Area—For many of the plants involved in surface
coating, the VOC concentration varies oyer a wide range. These variations
arise from differences in substrate widths, substrate speeds, and coating
thickness. In plant operations, this phenomena is independent of the other
variables.
j
[ In the evaporation experiments, this variable was originally included
in the design. However, this proved to, be very impractical. With the
relatively high SLA flow rates, severe Iswirling and choppiness developed on
the surface of the VOCs. An airfoil was attached to the inlet chamber to
prevent the VOCs from actually being swept out of the evaporation chamber
and into the mixing chamber. Since this variable was not essential to the
success of the study, it was dropped from the experimental design.
i :
4.2 EXPERIMENTAL WORK :
j The experimental work accomplished* in this study involved designing and
building the test apparatus used in bot:h of the laboratory experiments and
conducting the gas flow rate and material balance experimental
investigations. This section presents a description of the test apparatus
and the experimental methods and procedures used in this work.
i
4.2.1 Test Apparatus
To satisfy the objectives of the experimental work, the test apparatus
was-designed- to -provide-a si-mplified-evaporation -chamber with the ability to
• 33'
-------
simulate-some field test conditions# The criteria for the design of this
apparatus were:
• the ability to control and to condition the air flowing through
the system,
• the ability to readily isolate the streams for mass balance
i determinations,
' • the ability to contain all of the volatilized solvent from the
I point of solvent evaporation through the measurement points,
! • accessibility of sampling points,
i
i • accessibility of system components, and
1 • provisions for adequate safety.
I ;
A schematic representation of the apparatus used is given in Figure 2. This
apparatus consisted of several segments described in the following
subsections.
I
I
4.2.1.1 Air Inlet Chamber— :
i The air inlet was a rectangular box consisting of asbestos sheeting and
measuring 12"W x 8"H x 24"L. A 12" x 1;2" sheet of 1/4" plexiglass was
installed at the inside end of this chamber to control the air flow and
minimize turbulence around the liquids contained in the evaporation chamber.
Controlling the temperature and relative humidity of the incoming air were
attempted but proved to be impractical due to the large air mass moving
through the system. Thus, all experiments were designed to be conducted at
ambient temperature and humidity. j
4.2.1.2 Evaporation Chamber—
The air inlet chamber was followed' by a detachable plexiglass chamber
measuring 12"W x 11"H x 12"L. For all evaporation tests, a heated aluminum
pan measuring 11 3/4"W x 2"H x 9 3/8"L was placed in the bottom of the
chamber to receive the simulated coating mixture or solvent charges. The
evaporation chamber rested on an analytical balance used to determine the
mass of material in the system at any given time. Attachment of the
evaporation chamber to the inlet and the mixing chambers was through
flexible sheeting to minimize interference with the weight determinations.
I Implicit in this design was the utilization of a total enclosure of the
VOCs in the capture system. With fugitive losses held at essentially none,
any discrepancies in the mass balance closures would be due to discrepancies
within the analytical system.
i
4.2.1.3 Mixing Chamber--
| After entering the gas-phase in the evaporation chamber, the SLA and
air were allowed to mix and their turbulence decreased in a mixing chamber
or-plenum.- This - chamber—consisted- of a—rectangu-lar- plexiglass -box -measuring
-34
-------
FLEXIBLE
EVAPORATION
CHAMBER
DUCTING
INLET
CHAMBER
MIXING
CHAMBER
12" DIAMETER
HOLE
AIR
INLET
ANGLED DEFLECTION PLATE
DETACHABLE
EVAPORATION CHAMBER
BALANCE
10' RIGID DUCTING = 10 DIAMETERS FOR 12" DUCT
ORTHOGONAL /
PITOT PORTS
(OPTIMUM LOCATION)
ORTHOGONAL
PITOT PORTS
(SECONDARY LOCATION)
THREE HOLE RAKE
VOC PROBE
VOC
SAMPLE
LINE
HEATED
DIAPHRAGM PUMP
AIR
FLOW
AIR EXHAUST
2000 SCFM
BLOWER
AUXILIARY AIR
INLET DAMPER
Figure 2. Test apparatus for capture efficiency laboratory study.
-------
12"W-x -9"H x¦36"Li The exit end of-this chamber was~a round hole through
which the main measuring duct extended.
4.2.1.4 Measuring Duct—
To ensure the applicability of the EPA reference method criteria for
gas flow rate measurements, the measuring duct was designed of a round,
rigid pipe, 12 inches in diameter and 120 inches long. To provide the
measurement points (optimum and non-ideal) for the velocity traverses, two
orthogonal holes were drilled at 24 inches from the exit end (optimum point)
and two more at the midpoint in the length (non-ideal).
| For collection of gas-phase samples to determine VOC concentrations a
piece of 1/4 inch stainless steel tubing with 1/16 inch holes centered at
2.0 inches, 6.0 inches, and 10.0 inches| was installed at a 45° angle and 12
inches upstream of the ideal velocity measurement ports. A slipstream
sample of SLA was pulled continuously from the duct with a heated (400°C)
diaphragm pump and analyzed for VOCs. (The sample transfer lines were not
heated since the entire system operated; at ambient temperatures.
4.2.1.5 SLA Exhaust Blower— .
All of the air through the entire system was drawn by a nominal 2000
SCFM blower attached to the end of the system. Air from the measuring duct
was directed to the blower inlet through approximately 20 feet of 12 inch
rigid ducting. A series of damper holes are cut into this ductwork to
regulate the air flow rate through the ,test apparatus.
I
: i
4.2.2 Test Procedures and Methods
,
I
| The sampling and analytical method's used in this project are summarized
in Table 2. Each is described in more detail in the sections following,
which present the procedures and methods used in the individual
investigations. \
I i
I i
4.2.2.1 Gas Flow Rate Experiments— |
| In this investigation, thirty-six sets of gas flow mesurements were
performed while varying the flow rate, the measurement location, the pitot I
tube type, the differential pressure measuring device, and the personnel. !
The test matrix presenting particular variables involved with each run is
shown in Table 3.
I i !
| EPA Method 2 was used as the standard method for conducting pitot j
traverses to measure differential pressure. This data when combined with j
stream pressure and composition, flow cross-sectional area, and static \
pressure yields a volumetric flow rate determination. Method 2 calls for 6 |
point traverses (per diameter) at the optimum locations and 8 point (
traverses at the non-ideal ports. The;method does allow for measurements at j
additional traverse points if desired by the analyst.
All velocity traverses in this experimental work, consisted of 9 point
measurements. Eight were at locations .specified by Method 2• The odd point
was—at- the center of the-ductData collected at-this center point were not—¦
36
-------
TABLE 2. SUMMARY OF VOC CAPTURE EFFICIENCY SAMPLING/ANALYTICAL METHODS
Measurement Parameter
Measurement Method(s)
Gas Velocity/Volumetric Flow Rate'
(1) EPA Methods 1 and 2 using an S-type pitot
(2) Traverses using a standard pitot
(3) Non-standard method using a continuous, center
point pitot ;
Gas-Phase VOC Concentration
(1) EPA Method 25A (Byron 401 THC Analyzer)
(2) Draft EPA Method 18 (speciation using GC/FID)
( 3) NIOSH" Me'tHcTd T27~Tcharcoal'"tube-absorp tion,
CS2 extraction, GC/FID analysis)
Liquid-Phase Mass
Direct Gravimetric Measurement
Liquid-Phase Volatile Content
EPA Method 24
aTemperature measured using type K thermocouples; since the sample gas consists primarily of air, a
molecular weight of 29.0 will be assumed for flow rate calculations.
-------
TABLE 3. TEST MATRIX FOR FLOW RATE INTERACTION EFFECTS
Pi tot • Pressure - • — - ••• - Nominal
Test Tube Measurement Sampling Flow Rate
No, Type Device Location ( ft/sec) Personnel
1
S
Mag
NI
20
Chem
Std
IM
2
S
Mag
OPT
30
Tech
Std
IM
3
Std
Mag
01
•T
30
Chem
S
IM
4
S
Mag
NI
40
Chem
Std
IM
5
Std
IM
OPT
40
Tech
S
Mag
6
Std
Mag
OPT
20
Tech
S
IM
7
S
IM 1
NI
20
Engr
Std
Mag
8
S
Mag
i
NI
40
Engr
Std
IM
9
Std
Mag
OPT
40
Engr
S
IM
10
Std
Mag
OPT
20
Chem
S
IM
11
S
IM
NI
40
Tech
Std
Mag
12
S
Mag
NI
30
Engr
Std
IM
13
S
Mag
NI
20
Tech
Std
IM
14
Std
IM
OPT
20
Engr
S
Mag
15
Std
Mag
OPT
40
Chem
S
IK
16
s
IM
NI
30
Chem
Std
Mag
17
S
Mag
NI
30
Chem
Std
IM
18
Std
Mag
OPT
30
Engr
S
IM
Std =
Standard p
i tot.
S = S-type p
i tot
Mag =
Magnehelie
gauge.
IM = inclined
manome te r
OPT =
Optimum location
(2 diameters
ups
tream/8 diameters
downs trean).
NI " =
Non-ideal
location (5 diamete
rs upstream and downst
ream)•
Chem :
5 Chemist.
Engr =
Engineer.
Tech
= Technician.
38
-------
used' in-calculating average" differentral~pressure,~~but 'were collected to -' ~
provide data for a single point reference.
Other miscellaneous measurements were made during the Method 2 type
traverses. Temperature of the gas was measured with a K-type thermocouple
and a calibrated digital read-out. Atmospheric pressure was determined by
an aneroid barometer. The relative humidity was not measured, but the mole
fraction of water in the gas stream was assumed to be zero. The dry
molecular weight of the gas stream was assumed to be 29.0 g/g-mole as
described in Section 3.6 of Method 2. j
1 i
i The tests were not necessarily conducted in the order shown in Table 3.
J i
Test runs were conducted in groups base'd on set flow rates in order to
maintain a common condition for evaluation of the other variables.
i '
4.2.2.2 Mass Balance (Evaporation) Experiments—
i In this investigation, a series of; twenty-four experimental runs was
conducted using the best available liquid and gas measurement methods. The
composition of organics, the mass throughput of organics, and the SLA flow
rate were varied from one run to the next according to the experimental
design and the resultant test matrix presented in Table 4.
i :
i In these experiments, a known quantity of liquid solvent or coating was
placed in a pan located in the evaporation chamber, and the pan was heated
to promote evaporation. Air was pulled through the system at discrete flow
rates to produce an SLA flow through the measuring duct. While the known
quantity of liquid solvent was being evaporated, the resulting VOC
concentration and the flow rate of the SLA stream were monitored. Each of
these measurements, including loss of liquid, was recorded at a
corresponding frequency for incremental determination of mass balance
closure. Each experimental run was terminated when the VOC concentration in
the SLA stream fell below the background level.
I i
j Details of the procedures and methodology used in the measurement of
each parameter are given below.
i |
Gas-phase VOC concentration—Usingi the three-hole rake type probe,
samples were collected from the SLA stream and analyzed for VOC
concentration concurrently by different! methods. These methods included EPA
Method 25A, EPA Draft Method 18, and NI.0SH Method 127. The first method
uses flame ionization detection (FID) to provide total hydrocarbon (THC)
concentrations. Similarly, the second method also uses an FID for
quantitation. However, organic species': in the sample are
chromatographically separated prior to detection. The third method uses
charcoal tube adsorption of VOCs followed by solvent extraction. The extract
is analyzed by GC/FID. v
! EPA Method 25A and Draft Method 18 were utilized in each run of the
experimental set, while NIOSH Method 127 samples were collected only on
limited runs. The Method 25A measurements were made at one minute
intervals The Draft _Method~~18 measurements were conducted at seven minute
intervals.
39
-------
TABLE 4. TEST MATRIX FOR MASS BALANCE EXPERIMENTS
Organic ] Organic Nominal SLA
Composition ! Throughput Flow Rate
(Liters) (ft/sec)
1
KEK
;
2
20
2
To 1uene
,
0-5
30
3
Toluene
1
1
20
4
Mixture^
i
1
30
5
MEK
i
2
40
6
MEK
i
1
30
7
Mixture*
i
0.5
30
8
Toluene
'
'
2
30
9
Mixture*
0.5
20
10
Mixture*
i
1
40
11
Mixture*
j
1
20
12
MEK
1
0.5
40
13
Mixture*
i
2
30
1
14
Toluene
!
1
40
1
15
MEK
;
0.5
20
1
16
Mixture*
0.5
40
17
Mixture*
2
40
18
Mixture*-
2
20
19
MEK
1
30
26A
Comm 1^
1
30
22
Comm 2^
1
40
23
Comm 2^
1
20
2 6B
Comm 2^
1
30
,
31
Comtr. 3^
1
40
1
Mixture
of 50% each by volume of MEK
and
toluene.
Individual weights
were used in the closure calculations
•
2
Commerc:
al mixture of rubber in MEK,
spec
ified by
the
manufacturer to
contain
31% solids by weight.
3
Mixture
described in 3 above diluted
approximately
10
:1 with MEK.
4
Mixture
described in 3 above diluted
appr
oximately
2:
1 with MEK.
40
-------
Method 25A specifies' that" THC determinations' "be 'performe~d~usin'g"a
continuous monitor equipped with an FID. Although an FID was used on this
project, monitoring was performed on a semi-continuous basis using a Byron
Model 401 THC analyzer connected to the VOC sampling manifold. The Byron
401 can be operated in one of two modes: THC or full scan. For this
project, measurements were made only in the THC mode.
I
In the THC mode, sample is introduced continuously into a sample loop
from the VOC manifold. At sample injection, the loop contents are flushed
into a restrictor column (empty 1/16 inch tubing) and eventually flow into
the detector. A complete cycle is completed in one minute* During each '
cycle, a built-in integrator is first cleared and then allowed to accumulate i
electrical charge corresponding to the detector response. At the completion ;
of each analysis, this accumulated charge representing the integrated area
is available for data recordation.
t
i
j All data from the Byron instrument, was automatically captured, stored
on disk, and reduced using an Apple® II1 microcomputer. Software to j
accomplish this was tailored for this project. In addition, the computer
recorded the date and time of analysis, instrument, analysis type, sampling
location, and run number. As a backup to the microcomputer data acquisition
system, integrated areas for the instrument were recorded using a strip
chart recorder to provide hardcopy records of the analyses.
i i
j EPA Draft Method 18 differs from Method 25A in that sample acquisition
may be performed either by direct interface or by grab sampling. In
addition, the organic components of the' stream are chromatographically
separated prior to detection by FID* Sample gas for Method 18 was delivered
via the multiport manifold, and analyzed concurrently with Method 25A
analyses.
!
| The instrument and analytical conditions used were:
: I
j • Instrument AID Model 511
! • Column 18 in. x 1/8 o.d. s.s., Porapak Super Q,
I 80/100 mesh
| • Carrier Gas Nitrogen, approximately 40 mL/min
¦ • Column Temperature Isot'hermal at 170°C
j • Injection 5 mLl fixed volume loop
j • Integrator Hewlett Packard Model 3390A
i ' :
Gas from the VOC Manifold was flushed continuously through the sample [
loop. The injection valve was fitted with a Carle Model 4200 actuator '
controlled by an adjustable Eagle Signal timer. Cycle time was set at 7
minutes for all analyses.
i
! The integrated areas from the integrator hardcopy were input manually
to _the Apple® II data acquisition system. __Thi_s._step. was ..very_ helpful .in
reporting results on a consistent basis. Additionally, it facilitated
simple statistical analysis of the data using computer techniques.
41
-------
N*I-0SH-Method"127 was—followed in analyzing charcoal tube samples from -
selected experimental runs. However, sample collection deviated
considerably from the methodology specified in Method 127.
i Rather than use a personal pump calibrated with a typical tube, as
specified in the method, each charcoal tube was attached directly to a line
from the VOC manifold. Since this line was already pressurized from the
heated diaphragm pump, a separate pump for the tube was not needed.
: The exhaust end of the tube was directed into a Byron Model 90 mass
flow meter/totalizer. This provided not only a continuous indication of
flow rate, but also measured total air ;flow through the tube. The tube
sample flow was adjusted for each run in an effort to provide maximum
loading without breakthrough. ,
| In the analysis step, the front section of the tube was desorbed with
CS2 for 1/2 hour. A 5 uL sample was th'en injected into a GC equipped with
an FID. Instrumental and analytical conditions used were:
1 :
! 1
m Instrument Vatican Series 1400
• Column 20 ft. x 1/8 o.d. s.s., 10% SP 1000 on
! Supelcoport, 80/100 mesh
; i
• Carrier Gas Nitrogen, approximately 25 mL/min
• Detector Temperature 230°C
• Column Temperature 60°C for 3 min. then 10°C per min. to
150°.C, end at 20 min.
• Injection 5 uL' by syringe into port at 210°C
• Integrator Hewlett-Packard Model 3390A
I
Since none of the front sections indicated extremely high loading levels,
hone of the back sections were analyzed.
1
! Volumetric flow rate—SLA flow rates were measured at approximately
midway through each run by EPA Method 2i without modification. An S-type
pitot and inclined manometer were used !to measure differential pressures.
Details of these measurements and the other necessary data collection steps
were described in Section 4.2.1. A complete traverse was performed during
each evaporation run.
I Additionally, the SLA flow rate was continuously monitored during each
run with a standard pitot at the center of the non-ideal port. Differential
pressure at this location was measured by a digital manometer and recorded
on a strip chart recorder. The purpose: for this step was to monitor any
change in flow rate.
Mass of Liquid VOC—Determinations of VOC capture efficiencies require
that the total mass of liquid VOC entering or existing in the system be
known.—For- field operat-ions~-this—i-s- anticipated to be a -formidable task.
For the experiments, this task was relatively straightforward and simple.
-------
Mass-measurements - were made-on a calibrated laboratory balance capable
of measuring to the nearest 0.1 g. The balance was permanently placed under
the evaporation chamber to monitor the weight of the chamber and any liquid
that it contains. During each evaporation run, an initial weight of liquid
was recorded. Periodically, at 5 or 10 minute intervals, during the
evaporation phase, instantaneous weights were recorded. The mass of solvent
evaporated was determined by difference.
i Verification of the mass of each liquid charge to the evaporation
chamber was performed by weighing the solvent storage bottle before and
after applying its contents.
j
In the tests involving pure solvents, i.e., no solids, virtually all of
the solvent evaporates. In the tests involving a commercial coating
mixture, the dissolved or suspended solids remained in the pan. For the
laboratory tests, the solvent VOC weight loss was all that was required to
perform a mass balance calculation. However, in the field where the solids
completely disappear onto the web, the volatile content of the coating
mixture will have to be determined, probably by EPA Method 24.
i ;
A brief examination of EPA Method 24 (ASTM Method D2369-81) was made on
the commercial coating mixture used in the simulation tests. EPA Method 24
specifies that volatile content of coating mixtures be determined by running
duplicate analyses using ASTM Method D2;369-81 until the two values are
within 1.5% of each other. The ASTM procedure is basically a weight loss on
heating test. In the method, a discussion of precision indicates that a
60-minute heating period should provide! a relative deviation of less than
1.5% at the 95% confidence level. A forced draft oven was used to minimize
the possibility of a fire. j
To perform Method 24 analysis, an accurately weighed sample (0.3 +0.1 g
to the nearest 1 mg) was mixed with 3 mL of solvent for dispersion. This
mixture was then heated for 60 minutes |at 110 +5°C in the oven. The sample
was then removed, cooled in a desiccator and reweighed. The weight loss was
assumed to be the volatile content of t;he coating mix.
i ;
4.3 EXPERIMENTAL RESULTS AND DISCUSSION
i
The results of the laboratory tests are presented in this section for
the separate series of experiments. The calculation methods and raw test
measurement data are presented in Appendix C.
4.3.1 Gas Flow Rate Test Results ,*7
| A fractional factorial designed series of experiments was conducted to
evaluate the main and low order interaction effects surrounding EPA Method 2
type flow rate determinations. Thirty-six individual pitot traverses were
performed to study the variability of flow rate measurements with pitot tube
type, differential pressure measuring device, flow rate, personnel, and
sampling location. Table 5 shows the variables studied, the level of each
vari-able7 -and- the- resul-t-s-of- the testing-.
43
-------
TABLE 5- TEST MATRIX AND RESULTS FOR FLOW RATE INTERACTION EFFECTS
Test
No-
Pitot
Tube
Type
Pressure
Measurement
Device
Sampling
Location
Nominal
Flow Rate
(ft/sec)
Personnel
Velocity
Measurement
(ft/sec)
1
S
Std
Mag
IM
NI
20
Chem
18.2
18.3
2
S
Sta
Mag
IM
OPT
30
Tech
31.4
31.0
3
Std
S
Mag
IM
OPT
30
Chem
32.0
30.5
~
S
Std
Mag
IM
NI
40
Chem
37.1 !
38.8 i
5
Std
S
IM
Mag
OPT
40
Tech
40.1 ,
37.0 .
6
Std
S
Mag
IM
OPT
20
Tech
18.9 ;
17.3 1
7
S
Std
IM
Mag
NI
20
Engr
18.8 |
i9.6 ;
i8
i
S
Std
Mag
IM
NI
40
Engr
38.1
39.3
| 9
Std
S
Mag
IM
OPT
40
Engr
40.3
39.2
10
Std
S
Mag
IM
OPT
20
Chen;
19.2 '
18.4
11
S
Std
IM
Mag
NI
40
Tech
39.2
39.6
12
S
Std
Mag
IM
NI
30
Engr
32.0
31.5
13
S
Std
Mag
IM
NI
20
Tech
18.6
18.3
14
Std
S
IM
Mag
OPT
20
Engr
20.0
18.5
15
Std
S
Mag
IM
OPT
40
Chem
39.7
40.9
16
S
Std
IM
Mag
NI
30
Chem
30.8
31.4
17
S
Std
Mag
IM
NI
30
Chem
31.6
30.7
18
Std
S
Mag
IM
OPT
30
Engr
31.5
31.3
Std = Standard pitot. S = S-type pitot
Mag = Magnehelic gauge. IM = inclined manometer
OPT = Optimum location (2 diameters upstream/8 diameters downstream)•
NT- •-= Non-ideal -locat-ion-(5 diameters upstream and downstream)-.
Chem = Chemist# Engr = Engineer. Tech = Technician.
44
-------
In-the gas flow rate experiments, all test runs calling-for -a - • •
particular fLow rate were conducted in sequence at the sane setting with no
change in the duct work or blower configuration. The velocity was set in
the range of the specified level (e.g. 20, 30, or 40 feet per second) to
produce data for a widely varied range of flow rates. Thus, the absolute
values obtained for velocity are not as important as the relationship
between grouped data for each velocity range setting.
| Analysis of variance (ANOVA) techniques were used to test the
statistical significance of first and second order effects from each of the
variables. ANOVA is a statistical technique that separates the variations
found within a data set into exclusive components and then tests various
hypotheses about those components. A comprehensive discussion of ANOVA has
been written by Cochran and Cox. (20) Results of the ANOVA study are
presented in Table 6.
i
\ For the lowest flow rate, nominally 20 ft/sec, the type of pitot tube
was found to affect the results. This was partly anticipated since the flow
rate was near the lower limit of applicability for an S-type pitot.
However, the significance level of a = 0.10 indicates that the effect is
only marginal, i.e., we are wrong in concluding a significant effect in 10%
of the cases. Additionally, the relative difference between the two mean
values is 4.0%. j
i
For the middle flow rate tests, the type of pressure gauge was found to
have a significant effect at the 0.05 level. It should be noted that the
relative difference between the two mean values is only 2.2%.
For the highest flow rate, none ofj the individual variables was found
to have a significant effect on flow rate measurements. However, through
multiple comparison tests, the combination of S-type pitot and Magnehelic®
gauge was found to provide different results at a significance level of
0.01. The relative difference between jthe mean for this combination and all
other combinations is 8.1%. i
i i
| For all data sets, the precision looks excellent. Coefficients of
variation do not exceed 4%. The greatest variability is seen at the lowest
flow rate where minor discrepancies in interpreting pressure gauge readings
have a more pronounced effect. For instance, reading a value to the nearest
0.01 inch implies a potential error of 0.005 inch. If the correct value is
0.10 inch, the variability is +5%. However, if the correct value is 1.00
inch, then the variability is only +0.5%.
In the final analysis, it was concluded that no single variable or
combination of variables provided distinctly better flow rate data. As a
practical matter, this means that any combination of pitot tube, measuring
device, personnel, and sampling location provides adequately precise flow
rate measurements over the range studied.
45
-------
TABLE 6.' -ANALYSIS -GF- GAS-PHASE-FLOW-RATE-TEST- RESULTS
Nominal
Flow Rate
(ft/sec)
Data
Mean—ft/sec
(95% C.I.)
Number
of
Observations
Standard
Deviation
(ft/sec)
Coefficient Probability
of Variation Level for
(%) Significance*
20
S-Type Pitot
18.30 (17.74,
18.86)
6
0.53
2.89
Standard Pitot
19.05 (18.33,
19.77)
6
0.69
3.62 0.10
Overal1
18.68 (18.24,
19.12)
12
0.70
3.77
30
Inclined Manometer
30.97 (30.57,
31.36)
6
0.38
1.22
Magnehelic® Gauge
31.65 (31.36,
31.94)
6
0.28
0.89 0.05
Overall
31.31 (31.01,
31.61)
12 - •
... - 0.48- -
1.53
40
S-Type Pitot/
Magnehelic® Gauge
37.40 (35.89,
38.91)
3
0.61
1.63
0.01
Otherwise
39.68 (39.18,
40.18)
9
0.66
1.65
Overall
39.11 (39.43,
39.79)
12
1.07
2.74
" If this
1 only 5%
probability is small, i.e., 0.05,
of the time.
then
you are wrong
in concluding
a significant effect
-------
4«3t2'~'Mass~Balance (Evaporation) Test Results - - -
For this study, variations in solvent type, solvent amount, and SLA
flow rate were designed to represent an adequate compromise between expected
field conditions and the limitations of bench-scale testing. In order to
reduce the number of tests and facilitate the analysis of the data, it was
assumed that the data from one pure solvent is equivalent to the data from
another pure solvent. This assumption was evaluated from the data collected
and found to be correct. (
A series of 19 experiments was conducted wherein known quantities of I
pure solvent were evaporated while the resulting VOC concentration and the j
flow rate of the SLA were monitored. These were conducted using MEK, j
toluene, and a 50/50 by volume mixture of the two. j
An additional five experiments were conducted using a commercial
coating mixture containing rubber dissolved in MEK. The initial test using i
this mixture as received proved unsatisfactory due to a crusting effect. To j
alleviate this problem, the other runs were made on a portion of the mixture I
diluted with MEK. !
i I
I I
| Liquid/gas closure values were independently obtained for each VOC :
measurement type (Method 25A, Method 18;, and NIOSH Method 127). These are i
presented for each run in Table 7. The. data were analyzed using techniques !
similar to those used for the flow rate study. To test the statistical
significance of main and interaction effects, ANOVA methods were used. The ;
results of this are given in Table 8. The results of each method are
discussed in the following subsections. !
, i
4.3.2.1 Method 25A Closure Results— j
I 1
i The Byron 401 analyzer was calibrated daily with certified gas !
standards containing MEK and toluene. Response factors were obtained for |
each compound and used to calculate concentrations in the SLA stream. For
those experiments where the 50/50 mixture was used, an average of the two RF i
values was used to calculate concentrations. i
! A sample calculation showing conversion of the VOC concentration to )
mass of solvent is given in Appendix C. This calculation takes into account \
the SLA flow rate and utilizes temperature and pressure corrections. The !
total mass thus obtained was then compared to the weight loss measured with i
the balance under the evaporation chamb.er. j
I
| Mass balance closures were calculated for the individual time j
increments and for the total time elapsed interval, as shown in the sample '
calculations in Appendix C. Additionally, total mass balance closures were i
calculated from comparison of the initial solvent weight with the total sum !
of the measured VOC weights. Table 7 summarizes the conditions and both THC
closure values for all of the runs. Also included in the table is similar
information from the speciated GC-FID analyzer and charcoal tube analyses.
< ~
-------
TABLE 7. LIQUID/GAS MASS BALANCE CLOSURES
Run Elapsed
No. Time f
(mi n)
So lve nt
Type
Solve nt
Amount
(liters)
Measured
SLA
Flow Rate'
(SCFM)
THC• Closures
Method 18 Closures
NIOSH 127 Closures
Aggregate -
(Z)
Total^
(%)
MEK
U)
To luene
(2)
Total 5
U)
MEK
(Z)
To luene
(X)
Total*
(X)
1
100
MEK
2
832
100
100
73.0
—
73.0
397
—
39 7
2
56
Toluene
0.
5
1400
101
101
—
98.0
98.0
—
210
210
3
140
Toluene
1
843
98.0
96.1
—
88.5
88.5
—
2356
235
4
61
Mixture ^
1
1417
107
108
84.7
106
95.3
90.1
216
153
5
86
MEK
2
1825
88.5
88.5
103
—
103
117
—
117
6
51
MEK
1
1471
109
107
88.5
—
88.5
NM
—
NM8
7
53
Mixture ^
0.
.5
1395
93.5
100
103
156
129
NM
NM
NM
8
150
Toluene
2
1406
98.0
99.0
—
119
119
—
NM
NM
9
70
Mixture ^
0.
.5
844
90.1
89.3
75.8
90.1
82.9
69.0
155
112
10
70
Mixt ure ^
1
1753
110
110
174
546
360
NM
NM
NM
11
105
Mixture ^
1
918
102
101
78.7
97.1
87.9
123
204
163
12
35
MEK
0.
.5
1718
98.0
100
76.9
—
76.9
45.2
45.2
13
81
Mixture '
2
1461
102
100
84.7
95.2
89.9
131
350 5
240
14
77
Toluene
1
1700
96.1
95.2
—
87.0
87.0
—
111
111
15
56
MEK
0.
.5
844
93.9
103
92.6
—
92.6
NM
—
NM
16
49
Mixture ^
0.
.5
1787
99.4
100
96.1
95.2
95.6
50
149
99.5
17
60
Mixture ^
2
1771
98.0
102
90.9
61.3
76.1
122
290
206
18
175
Mixture ^
2
859
104
95.2
76.3
89.3
82.8
403
369
386
19
50
MEK
1
1834
100
107
77.5
—
77.5
NM
—
NM
26A
156
Comm 1 ^
1
1411
119
ND 10
90.1
90.1
NM
—
NM
22
190
Comm 2 ^ ^
1
1769
93.5
ND
117
~
117
NM
~
NM
23
121
Comm 2 ^ ^
1
922
99.0
ND
86.2
—
86.2
118
—
118
26B
120
Comm 2 J ^
1
1427
104
ND
98.0
—
98.0
NM
—
NM
31
241
Comm 3
1
1730
99.0
ND
186
—
186
NM
—
NM
1. Elapsed time for each analysis varied due to cycle time and sensitivity differences.
2. Flow rate measured by EPA Method 2 during each run. (SCFM 47.124 X feet per second.)
3. Calculated from ( MW x 100/ CW) where MW = sum of the balance weight losses and CW = sum of the weight losses from the
VOC concentration and the SLA flow rate. For the commercial mixtures, MW is the difference between the initial and final
balance weights.
4. Calculated from (TW x 100/ CW) where TW is the initial total weight of solvent placed in the system.
5. These data are the arithmetic average of the MEK and toluene closures.
6. Average value of duplicate determinations.
7. Mixture of 502 each by volume of MEK and toluene. Individual weights were used in the closure calculations.
8. NM » not measured.
9. Commercial mixture of rubber in MEK, specified by the manufacturer to contain 312 solids by weight.
10. ND s not determined since all of the material was not evaporated.
11. Mixture described in 9 above diluted approximately 10:1 with MEK.
12. Mixture described in 9 above diluted approximately 2:1 with MEK.
-------
TABLE 8. ANALYSIS OF MASS BALANCE TEST RESULTS
Analysis
Data
Mean of
95% Confidence
Number of
Standard
Coefficient
Type
Type
Closures
Interval
Observations
Deviation
of Variation
(%)
(%-%)
(%)
(%)
EPA Method
Pure solvent
99.9
97.1 - 102.7
19
5.9
5.9
25A/Byron
(overal. 1)
401 THC
Ana lysis
Commercial
102.2
89.6 - 114.9
5
9.1
8.9
mix (overall)
EPA Method
Pure solvent
105.4
75.0 - 135.9
19
63.9
60.0
18/GC-FID
(overal1)
with
Speciation
Commercial
115.5
64.4 - 166.7
5
41.2
35.7
mix (overall)
NIOSH Method
Pure solvent
190.4
126.1 - 254.6
13
106.2
55.8
127/Charcoal
(overal 1)
Tubes
i
-------
From Table 8, the "only variable found to have"a' "statistically
significant effect on THC closures was the solvent amount, with a
probability level of 0•10. As a practical matter, this marginally
significant effect would be difficult to explain since the mid-level (1.0
iiter) results seem to deviate from the low and high level results. Thus,
none of the variables nor combination of variables seem to have any great
effect on the closure technique results.
i Accuracy, bias, and precision were evaluated for the pure solvent and
the commercial mixture experiments as separate groups. Accuracy seems more
than adequate for each group since the 'mean closure values are 99.9 and
102.2%. Bias is not indicated for either group since 100% is included
within the 95% confidence interval in both cases. Precision or variability
estimates for the two groups are excellent since the coefficients of
variation are 5.9 and 8.9%. I
i
Finally, the statistical evaluation indicates that the mass balance
method using EPA Method 25A for VOC concentrations is adequate for VOC
capture efficiency determinations at surface coating operations. Further
discussion concerning accuracy and testing requirements will be given in
following sections. ;
! i
A.3.2.2 Method 18 Closure Results—
j The Method 18 closure values are shown in Table 7. Separately measured
solvent masses for the SLA stream were calculated in a manner analogous to
that described for THC measurements. A sample calculation appears in
Appendix C. Since the evaporated weight loss data were obtained at five-
(in some cases ten-) minute intervals, and the SLA weight measurements were
obtained every seven minutes, direct comparison of the masses could be made
only at 35- or 75-minute intervals. As! indicated by the elapsed time column
of Table 7, most of the runs would include only one or two such comparisons.
Thus, only total weights were used in the Method 18 closure calculations.
The overall closures for mixture runs are the average between the separate
MEK and toluene values. \
j From Table 8, there is not sufficient evidence to indicate significant
main or interaction effect in the Metho'd 18 closure results, at least to the
0.10 level. This may be due, in part, to the rather large variability
associated with the method itself.
I As with the Method 25A data, accuracy, bias, and precision were
evaluated for the pure solvent and the commercial mixture experiments as
separate groups. With mean closure values of 105.4% and 115.5% for the pure
solvent and the commercial mixture tests, respectively, the method appears
to be marginally accurate. The wide 95% confidence intervals (75.0 to
135.9% and 64.4 to 166.7%) tend to eliminate bias as a potential problem.
However, the variability estimates as measured by the coefficient of
variation indicate that the particular methodology used is not adequate for
measuring capture efficiencies. This problem was verified by the QC
activities which are further discussed -in this section. While no actual
cause-of— this- inadequacy-was-determined—two- major-possibili-ties were ¦
postulated.
"50
-------
The' first"possibility involves variability in stream"VOC"concentrations
and the measurement frequency used. A detailed review of the THC
concentrations measured using Method 25A for each run shows variations with
substantial up and down swings. These data, taken at 1-minute intervals,
seem to have adequately followed these variations. The 7-minute interval
data obtained by the Method 18 analyzer very likely missed a number of the
peaks and/or valleys. This factor may have played a major role because of
the relatively short test run periods.
j
j During some of the runs, it was apparent that the sensitivity of the
Method 18 analyzer changed during the run. This effect was particularly
evident during Run Number 10 where the KKK concentration for the third
sample had dropped to 26 ppm. At that point, the analyzer's detector jet
was mechanically cleaned of a white powder which had collected on it. As
evidenced by the data, sensitivity rose; dramatically. The overall results
show an excessive lack of recovery of the evaporated solvents. The extent
of this problem was not quantified for any of the runs.
1 The GC/FID analyzer was operated at maximum sensitivity during each of
the runs. Additionally, a rather large; (5 mL) sample loop was used to
obtain a relatively low detection limit1. It is very possible that a more
sophisticated or sensitive GC/FID instrument could provide better results.
! i
i This sensitivity problem could possibly have been quantified by
following the dictates of the draft method, i.e., analyzing a standard after
every three samples. However, this would certainly have exaggerated the
effects of the first problem associated, with the frequency of sampling.
i
4.3.2.3 NIOSH Method 127 (Charcoal Tube) Results—
Closure values for the charcoal tube samples taken during some test
runs are given in Table 7. Individual VOC masses were calculated from the
desorbed solvent mass and the ratio of sampled SLA to the total SLA flow. A
sample calculation is given in Appendix: C.
• i
i
' Full statistical analysis was not performed on the charcoal tube data.
However, the data shown in Table 7 indicate that the method is not
sufficiently accurate for performance testing. The most serious problem
associated with this method seems to be; its variability. As measured by the
coefficient of variation, this was 55.8%. Some of the more obvious
potential sources of this variability are discussed below.
Errors in charcoal tube flow rates do not seem likely since they would
cause a bias in the results. The high recoveries indicated by the toluene
results could support this type of error. However, the mixed high and low
recoveries for MEK seem to eliminate it.
! Duplicate analyses as well as other quality control efforts indicate
that the problem does not lie with the laboratory analysis. A summary of
the QC data is presented in Section 4.4".
51
-------
"Severe problems in the adsorption/desorption stage are indicated by the
results obtained for Runs 3 and 13. The closure value of 235% for Run 3 is
the result of averaging values of 358% and 112%. For Run 13, the MEK value (
of 131% came from averaging values of 240% and 22.3%, while the 350% closure 1
for toluene was derived from 571% and 129%. In both runs, the tube samples
were taken side-by-side from the same VOC sample line.
4.3.3 Volatile Content of Coating Mixture (EPA Method 24) Results
I ;
i A brief examination of EPA Method 24 for determination of volatile
content was made on three of the commercial coating mixtures. The samples I
were: ;
! i
i Comm 1 - mixture of rubber in MEK which contained 31% by weight of
| solids, ;
j Comm 2 - an approximately 10 to 1 dilution of Comm 1 with MEK, and
| Comm 3 - an approximately 2 to 1 dilution of Comm 1 with MEK. |
1 Attempts to obtain the specified 0.3 mL of sample for Comm 1 failed. |
The material was simply too viscous to pull up into a syringe. j
I i
! Obtaining a sample of Comm 3 was possible, but with great difficulty.
As with Comm 1, the viscosity of the material was the major source of
mechanical difficulty. j
! Obtaining samples of Comm 2 was relatively easy. However, after
dispensing the 0.3 mL sample into the weighing pan, it was obvious that the
solids content of duplicate samples was: different. i
! I !
j The results of analyzing Comm 2 arid Comm 3 are given in Table 9. As I
previously described, the method calls for running samples until duplicate :
results are within 1.5% of each other. ' j
\ i
i
4.3.4 Evaluation of the Effect of the Response Factors |
! ;
; In Section 4.3.2.1, the accuracy of the THC values was concluded to be ]
sufficient for capture efficiency determinations. This conclusion included
evaluation of the closure data from the. mixture runs. These values were j
based on an instrument response factor .(RF) derived from averaging the I
individual responses for MEK and toluene. The language of Method 25A allows j
for deriving instrument RFs from other 'sources such as propane standards. ;
j An investigation of the effects of- such calibrations was conducted. I
The THC data from each of the mixture runs were recalculated on the basis of :
the MEK response and on the basis of the toluene response# The results of
these calculations are given in Table 10.
The bias imparted to the results is relatively obvious in both cases. '
Basing the VOC concentrations of MEK leads--to--l_2,2%— closure while toluene
52
-------
TABLE 9. VOLATILE CONTENT OF COATING MIXTURES
Sample
Run No.
Average
Volatile
Content
(%)
Coefficient
of
Variation^
(%)
Meets
Method
Requirements
(Y or N)
Comm 2
Comm 2
Comm 2
Comm 2
26
23
23
23
2.19
2.43
2.57
2.65
2.48
2.41
2.64
2.58
2.34
2.42
2.61
2.62
8.8
0.6
1.9
1.9
N
Y
N
N
Comm 3
Comm 3
Comm 3
31
31
31
19.8 13.3 16.6
13.0 11.8 12.4
14.4 14.5 14.5
27.8
6.8
0.5
N
N
Y
1 Coefficient of variation = standard deviation as a percentage of the mean.
-------
TABLE 10. COMPARISON OF RF FACTORS ON CLOSURES FOR MEK/TOLUENE MIXTURE EXPERIMENTS
Method 25A - THC
Method 18 - Speciated
Run
No.
Closure
from
MEK. RF
(*)
Closure
from
Toluene RF
U)
Closure
from
Combined RF
(%)
Closure
from
MEK
U)
Closure
from
Toluene
(%)
CIosure
from
Average
(%)
4
113
71 .2
92.8
118
94.0
106
7
122
76.0
100
97.1
63.9
80.5
9
136
85.7
112
132
111
122
10
111
69.2
91.1
57.4
18.3
37.9
11 ¦ ;
121
- 7-5-.-8
99.2 -
127
103
- 115
13
121
76.5
99.8
118
105
112
16
123
76.7
100
104
105
105
17
120
74.3
97.7
110
163
137
18
128
80.4
105
131
112
122
Mean
122
76.2
99.7
110.5
97.2
104
CV (%)
6.1
6.3
6.2
21.0
40.2
28.1
-------
gives 76;2%- closure. Each of these is far-outside the allowable 10%
variance.
1 By mathematically averaging the RFs before computing the
concentrations, the more acceptable mean closure value of 99.7% is obtained,
it is not clear from this limited data whether mathematical combining of RFs
is appropriate for other ratios of solvent mixtures. The only clear
conclusion is that calibration of the analyzer must be performed using the
same solvents that will be later analyzed.
j
; I
Calibration of the Method 18 analyzer was effected by determining
individual RFs. Additionally, in calculating closures, each RF was applied
to the appropriate peak to calculate it's concentration. For completeness,
the Method 18 data for the mixture experiments is shown in Table 10. All of
the conclusions given in Section 4.3.2.jl remain valid for the individual
compounds. In fact, the variability experienced with toluene separately is
even higher than for any of the combined data.
i
4.4 QUALITY ASSURANCE AND QUALITY CONTROL
i i
; The assessment of overall performance of the VOC capture efficiency
bench-scale testing was based upon performance and systems QA audits and
upon internal QC data generated throughout the testing program. This
section summarizes the results of the QA/QC program, which are reported in
detail in a separate report titled "Measurement of Process Capture
Efficiency Quality Assurance for Bench-Scale Testing".
i
I Since the primary emphasis of the liquid/gas-phase mass balance tests
was upon evaluation of Method 18 and Method 25A VOC measurements, this
analysis emphasizes assessment of error, (i.e., uncertainty) associated with
Method 18 and Method 25A measurements. ; This assessment indicates that the
measurement data collected during the bench-scale tests are sufficiently
reliable to be used for evaluation of the feasibility of the
liquid/gas-phase mass balance approach.1
j Multipoint performance audit data for the two VOC measurement methods
(i.e., Method 18 and Method 25A) generally indicated that, at the time of
the audit, there was no statistically significant, overall difference
between the two methods, with respect to bias (i.e., systematic error). The
multipoint audits indicated that, for both methods, bias for measurements of
methyl ethyl ketone (MEK) and toluene concentrations was not significant.
Observed bias for multipoint audit data*using propane gas mixtures was,
however, statistically significant for -both Method 18 and Method 25A
analyses, at -7.2% and -8.8%, respectively.
While the performance audit data indicate comparable results for both
methods on the day these analyses were performed, control sample results
indicate that, over the long-term (i.e.', over the duration of bench-scale
testing), there was a statistically significant difference in measurements
made by the two methods. These data, based on daily analyses of a quality
control—gas mixture- containing- MEK--and—'toluene,- indicate that Method 18
55
-------
results" averaged about 5% lower than corresponding results for Me'tHo'd 25A'
analyses. This same trend is evident in the mass balance closure data for
the evaporation tests using MEK and toluene. Differences between closure
results for the two methods were also statistically significant, even when
tested at the 99% confidence level. The average difference for closure data
was approximately 12%.
i
In addition to providing a measure of reLative bias between the two
measurement methods, the control sample data also provide estimates of
precision (i.e., measurement variability, or random error) for the two
methods. For both methods, day-to-day ^variability (repeatability) was
greater (i.e., the precision was poorer) than within-day variability
(replicability). For both Method 18 and Method 25A, within-day variability
was approximately 0.8%, expressed in terms of the coefficient of variation
(CV), or relative standard deviation. .Repeatability for Method 18 was not
as good as that for Method 25A, with day-to-day CVs of 3.2% and 1.5%,
respectively.
I Control sample analyses represented an internal QC procedure designed
to provide an estimate of sampling/analytical precision, as well as
controlling this aspect of data quality, within acceptable limits. In
addition to control sample analyses, several other QC procedures were used
to control measurement data quality. These included checks of and
acceptance criteria for instrument response linearity, response factor
variability, blank, analyses, and instrument drift. Specified acceptance
criteria were met in all but a few cases, and these QC checks generally fell
within or near the specified limits.
-------
SECTION "5"
FIELD TESTING
The laboratory testing led to the conclusion that the liquid/gas-phase
material balance methodology tested was sufficient to determine VOC capture
efficiencies under controlled conditions. Recommendations included a plant or
field test of the liquid/gas-phase material balance methodology under
conditions which are more complex than the laboratory tests, yet simple enough
to evaluate the methods in a stepwise manner. I
: I
Selection criteria for a field test site included the following: I
; • a single coating line within a relatively confined area, i
1 • long steady state production runs absent of frequent long breaks or
process changes, i ;
• a coating formulation containing only a few solvent components, 1
\ • good measurement locations on all liquid and gas streams, ;
t • an indirect fired drying oven, and \
• a solvent recovery type control device. ;
; i
1 A number of plants were reviewed and considered for testing, including the
plants visited in the Phase I study (feasibility evaluation). One plant
reviewed, Plant A, met the majority of the criteria. It was finally selected
as the test site after a pre-test visit, and evaluation at the plant. I
5.1 TEST SITE DESCRIPTION
j Plant A operates a single magnetic tape coating line. The manufacturing
process involves the application of a magnetic iron oxide coating to a
polyester film substrate. After application of the coating, the polyester film
is sliced into tape and loaded onto tape cassettes. The tape coating process
typically operates 24 hours per day, five days per week. j
] For the purposes of the current pr.ogram, the tape coating operation can be
categorized into the following process units: (1) solvent storage, (2) coating
preparation, (3) coating application, (4) drying, and (5) solvent recovery. A
flow diagram of the entire coating operation is shown in Figure 3, and the
separate process units comprising the manufacturing operation are described in
the following subsections.
5.1.1 Solvent Storage
! !
The solvent storage process area supplies and receives the two solvents 1
used in the coating operation, solvent A and B. The solvent storage area
consists of the following equipment:
I • a 10,000-gallon solvent storage tank (solvent A),
• a 10,000-gallon solvent storage tank (solvent B),
• a 10,~000-gal~ron~premix" so'lvent~~storage~"tank,
57
-------
• "a 10, OOO-ga l'lon distillate solvent storage tank,""" arid ~
• associated distribution equipment (transfer pumps, meters, and
transfer piping and ductwork).
!
These equipment items are shown in Figure 3 and described below.
i i
The solvent storage tanks are used to receive periodic shipments of virgin
solvents. These tanks are also used to receive purified solvent fractions froim
the solvent recovery area. Solvents from these two tanks are metered as
necessary into the premix tank to formulate a 75/25 weight percent solution of
solvent B/solvent A. The solvents are metered and pumped as needed to various
parts of the plant. The distillate tank is used as a temporary holding tank |
for unseparated solvent mixtures from the solvent recovery area. j
i
1 All four solvent storage tanks arej vented to the solvent-laden air (SLA)
system. In this manner, solvent vapors, generated by diurnal temperature
changes and tank volume displacements are routed to the solvent recovery unit,j
resulting in a higher total plant solvent recovery efficiency. j
i \
5.1.2 Coating Preparation ! j
| The coating preparation process area consists of the following equipment,]
as shown in Figure 3; j '
i • a 600-gallon mixer, .
two 1000-gallon mix tanks (T101 and T102), j
one 600-gallon mix tank (T103),
mills, 1
two 800-gallon milled slurry tanks (T151 and T152), j
three 1000-gallon staging tanks (T201, T202, and T203), and
associated distribution equipment (transfer pumps, meters, and
| transfer piping and ductwork..
i ;
The purpose of the coating preparation process unit is the formulation of a ,
coating slurry of the required composition, density, viscosity, and throughput.
The processing steps of the coating preparation phase are described below.
| As shown in the figure, the coating preparation process begins with the
mixture of resins, magnetic iron oxide powder, and other solids with a 75/25
weight percent solvent solution of solvent B/solvent A. These components are
blended to achieve a final mixture containing 40 weight percent solids. The
mixing operation is performed on a batch basis using the mixer and mixing tanks
T101, T102, and T103. .j- I
i
! Following the mixing operation, the 40 weight percent solids solution is
milled. Final solvent adjustments are .made in milled slurry tanks T151 and
T152.
i After the slurry has been milled and adjusted to the desired
specifications, the coating slurry batch is pumped to staging tank T201. Tank
T201—is-a--supply/'holding- vessel—that--drscharges""to~tank T20 2~ one of the
58
-------
(.XI1AUM
AIM
SOl Vf N t f [ZT)~ co-i
r
Figure
3. Detailed plant-wide flow diagram.
-------
primary" feed "tanks for the" coating "station'*' The "slurry is pumped in batch'
quantities from T201 to T202 between one and two times per each 24-hour period.
Occasionally, solvent (solvent B) additions are made to the feed tank to adjust
the viscosity, because of solvent losses during the holding/feed period. Tank.
T203 served as a spare feed tank during the test period.
As with the solvent storage process area, all tanks in the coating
preparation process area are vented to the SLA system. Also, a covered wash
sink located in the milling area is vented to the SLA system. The sink j
contains a solvent bath for intermittent manual cleaning of tools and parts. :
' i
Two floor sweep exhausts are located at the floor levels behind the tanks
in each of the mixing and milling rooms. The floor sweeps are exhausted to the
roof and discharged to the outside ambient atmosphere. j
i ! i
5.1.3 Coating Application :
I i i
| The coating application process unit consists of the following equipment:
i : I
j • a single coating line, | j
• supplemental coating tanks, and I
| • associated distribution equipment (transfer pumps and transfer piping
i and ductwork). i ;
' i
These equipment items are shown in Figure 3 and are described below. ;
' From primary feed tank T202 (in the milling room), the coating slurry is |
pumped to the coating station which is -located in the adjacent coating room, i
The coating station is maintained within a small vented permanent enclosure j
extending to the entrance of the drying oven. The enclosure exhaust is vented
to the overhead drying oven. Within the coating station, supplemental coating
containing VOC is added to the slurry, and the slurry is applied to the '
synthetic web material. The web is coaled continuously, except for a very !
brief period during the splicing of the1 web rolls. During this period, no
slurry is applied, but instead is recycled through a loop back to the feed tank
(T202). During normal operation, approximately 28 gallons of slurry are used:
per hour. j j
1
! The coating room has one exhaust in the general room area (outside the
coating station enclosure) leading to atmosphere; however, it was not
operational during the test. Use of chemical smoke tubes indicated that the 1
coating station enclosure did not capture all of the emissions from the j
applicator and that some escaped into the general room area. During the test j
period, the enclosure doors were left open because of concentration buildups. !
Plans were being made at that time to increase the capture and exhaust in the \
enclosure. Smoke tests also indicated Cthat the milling room was negative in !
pressure relative to the coating room, causing some minor air flow into that
area.
60
-------
5.1.4 Drying ...
The drying process unit consists of the following equipment:
• a steam-heated drying oven, ;
• a supply air preheater, and
• ductwork for transport of the various air streams.
These equipment items are shown in Figure 3 and described below.
} j
After coating, the wet web is dried, i.e., the solvent is evaporated, in]a
steam-heated drying oven. The drying medium is a mixture of preheated ambient
air and air from the coating station enclosure. The solvent laden oven exhaust
air, referred to as the capture stream,, is cooled by the supply air preheater !
and a chilled water coil. It then joins the tank vent stream to form the main
or total SLA stream. |
! I
5.1.5 SLA System and Solvent Recovery ; i
i i
I The solvent recovery process unit consists of the following major ]
equipment items: I
! , ;
• a fluidized carbon adsorber bed (CAB) and related equipment,
• a recovered liquid decanter, I
• fractionating columns, and j
• associated piping and ductwork. 1
These equipment items are shown in Figure 3 and described below.
i i
The feed for the solvent recovery system is the main SLA stream which i
collects the gaseous emissions from thej plant through the SLA system. This
system is comprised of two sub-systems:; (1) the vent system from the various i
tanks and wash sink previously described and (2) the capture system from the ¦
coater and drying oven. The vent stream and capture stream join to form the !
main SLA stream on the roof immediately after the chiller on the capture !
stream.
I , !
j The solvent recovery system features a fluidized bed carbon adsorber sized
to process a total SLA flow rate of 11,000 scfm. The system uses beaded
activated carbon in a continuous adsorption/regeneration process to remove i
solvent from the SLA at efficiencies of,nominally 95 percent and higher |
depending on carbon age and other factors. After adsorption of solvent, spent
carbon is regenerated with high-temperature nitrogen. A water-cooled condenser
is then used to separate the nitrogen from the desorbed solvent. The nitrogen
is recycled; the condensed solvent phase is continuously dewatered in a |
decanter and sent to the distillate holjding tank. On a batch basis, condensed
solvent is sent through the distillation unit separating it into the fractional
components. The individual components (solvent A and solvent B) are metered
and returned to the individual solvent storage tanks for reuse in the coating
process.
61
-------
5v2~"TESTING•APPROACH
The technical approach to the field testing can be logically divided into
three areas. These are: ,
• Measurement of capture efficiency of the coating line (coating and
drying only), using a liquid/gas-phase material balance.
• Measurement of other streams and calculation of material balances, ;
including the overall plant recovery efficiency.
I ;
; ~ Analyses of errors in material balance measurements and calculations.
j I
5.2.1 Capture Efficiency Measurement !
! i 1
• As previously described, capture efficiency by the liquid/gas-phase !
approach is determined by comparing the; VOC mass of the captured gas stream |
with the VOC mass of liquid used or evaporated. ;
i Figure 4 provides a graphic summary of the liquid and gaseous VOC streams
flowing into and out of the coating area at Plant A. i
i t
' Determination of the gaseous VOC m'ass flow rate (GPMF) involves '
determination of the following: I ;
i j i
• gas-phase VOC concentration in the capture stream, j
• the volumetric flow rate of the capture stream, and
• the molecular weight of the VOC material. j
i
At plant A, the captured VOC is directed into a single capture stream. ;
Therefore, only a single measurement site is required. j
| 1
] Determination of the liquid VOC usage rate (LPMF) involves determination ;
of the following: |
* i
•; • the coating slurry volumetric consumption rate, I
i • the slurry density, and j
i • the VOC weight fraction of the coating slurry- j
At Plant A, it is not feasible to make the above determinations on a single j
coating slurry stream at the coater, because of the applicator configuration. |
Therefore, the determinations for each of the above must be made from the i
; ' j
liquids at the slurry feed tank (102) and the supplemental coating tanks. The
liquid mass flow rates are then added to yield the total liquid mass input. I
! I
5*2.2 Other Material Balances '
i ¦ - i _.
' Other stream measurements and material balance determinations seem j
advantageous during the field testing for verification of the coating room
capture determinations and for determination of plant-wide emission reductions.
They include: . .
62
-------
VOC CAPTURE
STREAM TO
SOLVENT
RECOVERY
SUPPLEMENTAL
COATING
SLURRY FEED
FEED
ROLL
TAKE UP
ROLL
SLURRY RECYCLE
i
Figure |4. Coating area flow diagram .
-------
' a gas-phase balance of SLA system streams,
• a liquid/gas-phase balance of recoverable VOC,
• the plant-wide capture efficiency, and
• the plant-wide recovery efficiency.
Figure 5 provides a simplified plant-wide flow diagram which may be useful
in visualizing the following discussions.
5.2.2.1 Gas-Phase Balance of SLA Streams—
¦ The purpose of this exercise is two-fold: (1) to provide a comparison !
between the sum of the measurements of the two individual SLA sub-system ,
streams (capture + vent) and the main (total) SLA stream, and (2) to determine
the fractional contribution of the two SLA subsystems to the total. Figure 5;
shows the VOC mass flows of the two SLA' subsystems into the main SLA stream.
The VOC mass flow rate for each of the |three streams is determined the same as
that for the capture stream described In Section 5.2.1.
; [ !
• In the first case above, the sum of the measured vent and capture stream]
VOC mass flow rates should equal the measured VOC mass flow rate for the main!
SLA stream during the same period. Thi's provides a relative verification of
the VOC mass flow rate determinations for the two individual substreams. ;
Referring to Figure 5, the comparison is made following Equation 3 below:
GPMFven£ + GPMF,capture
1 Closure (%) = ; X 100 (3)
GPMF^ainl
where, GPMF = mass flow rate of VOC in the various gas streams, (lb/hr).
In the second case, a comparison of the two individual SLA subsystem mass
flow rates to the sum of the two determines the fractional amount of VOC
collected from each system. Therefore,: a comparison can be made between the
VOC collected from the coating operatio,n and that collected from the other
process units. !
i
l
5.2.2.2 Liquid/Gas-Phase Balance of Recoverable VOC—
The purpose of this exercise is to] verify the gas stream VOC mass flow
measurements by comparison with liquid VOC mass flow measurements. The VOC
mass entering the control device should equal the VOC mass exiting. Referring
to Figure 5, Equation A below is used to calculate the liquid/gas closure:
! ^^Recovered Solvent = GPMFyent ^^Capture ~ ^^Outlet Stack (&)
i i
I where, LPMF = mass flow rate of liquid VOC in the form of recovered \
solvent (lbs/hr). 1
GPMF = mass flow rate of VOC in each gaseous stream (lbs/hr).
All gaseous VOC mass flow rates are determined as described for the captured
VOC stream in Section 5.2.1.
64
-------
OUTLET STACK
FUGITIVE VOC
BENDIX THC
ANALYZER
BENDIX THC
ANALYZER
BYRON THC
ANALYZER
A MAIN SLA
STREAM
BYRON THC
ANALYZER
VENT
STREAM
CAPTURE
STREAM
DISTILLATION
SYSTEM
RECOVERED SOLVENT
SOLVENT
RECOVERY
MILLS
MIX TANKS
STORAGE TANKS
WASH SINK
COATING
PROCESS OVEN
AND VOC
CAPTURE
SYSTEM
Figure 5. Simplified plant-wide VOC flow diagram..
-------
•Since the recovered liquid solvent stream is not metered and- is not -
measurable at a single point on a continuous basis, the simplest determination
is the combination of mass determinations for resultant streams. The total
recovered VOC mass is equal to the sum of the metered pure solvent masses
recovered and the net increase in the distillate tank VOC inventory from the
start to the end of the period.
5.2.2.3 Plant-Wide Capture Efficiency—
The purpose of this exercise is to determine the percentage of the total :
VOC emissions from all VOC handling areas that are captured and delivered to !
the control device. This determination; is similar to the liquid/gas capture
efficiency determination for the coating area, described in Section 5.2.1, and
requires measurement of the same parameters. However, it includes all of the j
process units in the tape coating plant and, therefore, requires a slightly !
different measurement approach. The plant-wide capture efficiency is ;
calculated according to Equation 5 below: j
i |
i GPMp , |
; CEp = X 100 1 (5) !
; lpmp |
I
\ where, CEp = Plant-wide Capture Efficiency, (%). j
i GPMp = Mass of gaseous VOC collected by the SLA system from the j
j entire plant during a- test period, (lbs).
: LPMp = Mass of liquid VOC vaporized throughout the tape plant
j during a test period,j (lbs).
; In the plant-wide capture efficiency determination, the liquid mass of VOC
vaporized, i.e. VOC emissions, is not as easily determined as in the coating
room area. This is complicated by the 'recycling of recovered solvent to the
solvent storage and feed tanks, exhausting VOC mass from the feed tanks through
the tank vent system, and the large fluctuations of liquid mass held in tanks j
within the plant. There are two methods for estimating the total VOC mass <
vaporized, however, both involve complex indirect determinations. The first (
solely involves liquid-phase measurements of the solvent flowing into the plant
adjusted by net change in the inventory' level within the plant during the test
period. The second involves a determination of the total VOC emissions as the
sum of captured and uncaptured emissions. The uncaptured emissions are
determined indirectly by subtracting the mass exhausted from the control device
from the total plant loss value (difference between starting and ending liquid
inventories). _ j
* I
The gaseous VOC mass for the plant-wide capture efficiency determination i
is the total SLA stream or the sum of the VOC mass flows for the vent and |
capture streams.
i ~ •
5.2.2.4 Plant-Wide Recovery Efficiency— ¦
! The purpose of this exercise is to determine the percentage of the total
vaporized VOC that is recovered. Conversely, the percentage of the total
solvent used and lost to fugitive emiss.ions can be determined.
66
-------
The" plant-wide recovery efficiency is determined by' coraparing 'the measured
mass of recovered solvent with the total measured mass of VOC vaporized, as
shown in Equation 6 below:
' i
T PM
Lrurecov. ,
REp X 100 (6) ;
lpmp !
where, REp = Plant-wide recovery efficiency, (%). !
LPMrecov. = Mass liquid VOC recovered during a test period, (lbs). j
LPMp = Mass of liquid VOC vaporized during a test period, (lbs). |
! i
The mass of VOC vaporized is determined as described in Section 5.2.2.3. Thej
total mass of recovered solvent is determined as described in Section 5.2.2.2;
i
i •
S.2«3 Error Analysis
J In order to determine the reliability of the material balance data 1
obtained in this test, it is necessary to determine the precision and accuracy
of the measurements and to conduct an error propagation analysis of the
theoretical calculations.
i i
I |
| The precision may be estimated based on observed variability in results s
from multiple determinations. The reliability of such an estimate increases '
with the number of measurements.
Once the precision of each parameter is estimated, statistical techniques
can be used to estimate the overall uncertainty in the capture efficiency
determination. Standard error propagation techniques involving Monte Carlo \
simulations yield estimates of confidence intervals and the minimum time I
necessary to adequately characterize the capture efficiency. !
' ' I
5.3 TEST METHODS AND PROCEDURES |
1 ! 1
1 This section describes the sampling and analytical methods and procedures
used in measuring the individual parameters described in the previous section.-
The methods are summarized in Table 11. j
5.3.1 Liquid-Phase Measurements
Liquid-phase measurements included, liquid volume, density, and solvent j
content at set frequencies throughout the plant. These determinations were i
made on each accessible VOC containing liquid stream and tank throughout the 1
tape coating plant. j
5.3.1.1 Liquid Volume— ~ I
The liquid volume used during any time period was determined by the
difference between inventory levels at the beginning and end of each period ;
plus any volumetric adjustment transfer made on the tank. Liquid volumes were
determined for the main coating slurry tank (102) at least hourly and at the 1
beginning-and • end- of tank-refil-1- period. -All-other streams and tanks were
measured approximately every eight hours.
67
-------
TABLE 11. SUMMARY OF SAMPLING/ANALYTICAL METHODS
Measurement Parameter
Measurement Method(s)
! Liquid-Phase Mass
Gravimetric Measurement of Density and Volumetric
Measurements by Tank. Depth or Meter Reading
Liquid-Phase Volatile Content
Gas Velocity/Volumetric Flow Rate3
EPA Method 24
EPA Methods 1 and 2 using an S-type pitot
ON
oo
Gas-Phase VOC Concentration
(1) EPA Method 25A (Byron 401 and Bendix Continuous THC ¦
Analyzer) |
(2) Draft EPA Method 18 (speciation using GC/FID) '
' aTemperature was measured using type K thermocouples; since the sample gas consisted primarily of
air, a molecular weight of 29.0 was assumed for flow rate calculations.
-------
• Liquid"inventory volumes in -tanks were determined- by making depth - •
measurements and applying depth/volume conversion factors for each tank to the
measurements# Depth measurements were performed using attached sight glasses
and depth scales on tanks containing transparent liquids, and by hand held
meter sticks on tanks containing coating slurry. Depth gauge readings were
taken to the nearest 1/8 inch.
i Volumetric transfers of pure solvents to and from tanks were determined by
reading totalizer meters on the streams. Readings were to the nearest gallon.
5.3.1.2 Density— ! i
I Density of each liquid stream was .determined at approximately eight-hour:
intervals. Individual 25 milliliter liquid samples were collected from each :
tank and weighed on a Mettler laboratory balance to the nearest 0.01 gram. The
density was determined by calculating the mass to volume ratio.
i !
5.3.1.3 Volatile Content—
! The volatile or solvent content determinations were conducted at the same
frequency as the liquid densities. The volatile content of liquid in each tank
was determined by collecting liquid samples and conducting EPA Method 24 1
analyses. In this method, an accurately weighed slurry sample is mixed with a
dispersing solvent and heated for 60 minutes at 110 +_ 5°C. The weight lost I
during heating is assumed to be the volatile content of the slurry. For most j
analyses, a larger than specified sample size was used,
i !
5.3.2 Gas-Phase Measurements
i
: i
i Gas-phase measurements in this tes,t included volumetric flow rates and VOC
concentrations in each of the major branches of the SLA system. These are j
described in the following paragraphs.
i : i
5.3.2.1 Gas-Phase Volumetric Flow Rates— I
j Flow rates were measured for the VOC capture stream, the tank vent stream,
and the main SLA stream (CAB inlet). In general, each stream was measured ]
three times per day using EPA Method 2*
! ¦ '
I Other miscellaneous measurements Wjere made during the Method 2 traverses.)
Temperature of the gas was measured with a K-type thermocouple and calibrated j
digital read-out. Atmospheric pressure, was obtained from the U.S. weather '
service. The relative humidity was measured during each traverse. The dry j
molecular weight of the gas stream was assumed to be 29.0 g/g-mole as described
in Section 3.6 of Method 2. ~ !
Additionally, the gas flow rate in the SLA stream was monitored hourly j
using the plant in-line monitor. This 'provided observation for any major
change in the flow during the test period. j
!
5.3.2.2 Gas-Phase VOC Concentration—-
; Using a three-hole rake type probe, samples were collected from each SLA ¦
stream (vent, capture, main SLA and CAB outlet) and analyzed for VOC
concentration. Analyt-ical--methods--included EPA- Method- 25A- and EPA Draft Method
-------
18-.—-The first method-uses flame Ionization detection- (FID) to-provide total-
hydrocarbon (THC) concentrations. Similarly, the second method also uses an
FID for quantitation. However, organic species in the sample are
chromatographically separated prior to detection.
The VOC concentration of each stream was measured continuously throughout:
the test period, except during calibration periods and one short period when :
the bottled air supply was depleted. Each measurement point is indicated on
Figure 5.
i ; !
EPA Method 25A—Method 25A specifies that THC determinations be performed
using a continuous monitor equipped with an FID. This requirement was I
fulfilled in monitoring the carbon bed anlet (main SLA stream) and outlet (SLA
exhaust) streams with Bendix Total Hydrocarbon Analyzers. The Bendix analyzers
were supplied by the plant but calibrated according to the requirements
specified in the QA project plan. Each of these analyzers was allowed to I
continuously analyze the appropriate stream during the entire test period. j
i : 1
j Simultaneously, individual Byron model 401 analyzers were monitoring the|
THC content of the coating capture stream and the tank vent stream. In the THC
mode, sample is introduced continuously into a sample loop. At sample I
injection, the loop contents are flushed into a restrictor column (empty 1/16 i
inch tubing) and eventually flow into the FID. A complete cycle is completed \
in one minute. At the completion of each analytical cycle, a built-in •
integrator provides the peak area data for recordation,
I i
I All data from the Byron instruments were automatically captured, stored on
disk, and reduced using a microcomputer. Software to accomplish this was J
tailored for this project. In addition, the computer recorded date and time of
analysis, instrument, analysis type, sampling location, and run number. As a ;
backup to the microcomputer data acquisition system, integrated areas for each
instrument were recorded on strip chart's to provide hardcopy records of the j
analyses,
!
EPA Draft Method 18—Draft Method 18 differs from Method 25A in that i
sample acquisition may be performed either by direct interface or by grab
sampling. In addition, the organic components of the stream are 1
chromatographically separated prior to detection by FID.
The instrument and analytical conditions used were: j
c i
• Instrument Shimadzu Model Mini-2
• Column 6 ft. x,. 1/8 o.d. glass with SP2100 at 20% plus
Carbowax 1500 at 0.1% on Supelcoport 100/120 I
• Carrier Gas Nitrogen, approximately 40 mL/min !
• Column Temperature Isothermal, approximately 100°C J
• Injection 1 mL fixed volume loop
• Integrator Hewlett-Packard Model 3390A
!
Gas from the sample was flushed continuously through the sample loop. The
injeet-ion—•valve - was fi tted - with- a -Car-le -Model 4200 actuator- control-led by- an -
-------
adjustable Eagle Signal timer. The cycle tine was set at 4 minutes "for" all"
analyses.
The integrated areas from the integrator hardcopy were input manually to'
the microcomputer data acquisition system. This step was very helpful in
reporting results on a consistent basis. Additionally, it facilitated simple
statistical analysis of the data using computer techniques.
5.4 RESULTS AND DISCUSSION
; i i
On-site testing consisted of 114 hours of analyses beginning at 0200 on 13
February 84 and ending at 2000 on 17 February 84. During this period, six
discrete batches of slurry were prepared and applied to the web material. !
Detailed results of all analytical tests are presented in Appendix D. Sample
calculations for determination of the liquid and gas masses are also given in.
Appendix D. ! j
This section contains summaries of: the results and discussions of those j
results. Testing results are presented1 in two categories: (1) coating area !
capture efficiency results and (2) other material balance results. Potential j
errors and their combined effects on the results and calculations are described
in a third section. i j
I i
; I ,
The overall findings of the tests were as follows:
Coating Room Capture Efficiency (Method 25A) - 106-7%/l03.0% j
Coating Room Capture Efficiency (Method 18) - 105.1% !
! Gas-Phase Balance of SLA Streams (Method 25A) - 94.3%/92.9% j
j Gas-Phase Balance of SLA Streams (Method 18) - 99-4%/96.3% !
Liquid/Gas Balance of Recoverable VOC - 81% j
Measured Plant-Wide Capture Efficiency - 120%/94%
Measured Plant-Wide Recovery Efficiency - 86%/67% '¦
! i
5.4.1 Coating Area Capture Efficiency ' 1
i ' i
| As described m Section 5.2.1, the. VOC capture efficiency in the coating;
area is determined as the ratio of gaseous VOC mass captured versus the liquid
VOC mass used. A summary of the 1 iquid'/gas-phase mass flow rates and capture 1
efficiencies from the tests is given in' Table 12.
j Initially, it was thought that the.: results obtained from the test data [
could be compared with an assumed capture efficiency value of 100 percent,
since the coating room was totally enclosed and only vented to the capture i
stream. However, simple smoke tests during the test period indicated that this
assumption was incorrect, and that in fact the potential existed for some minor
uncaptured fugitive emissions. Therefore, the capture efficiency had to be
considered somewhat less than 100 percent and the capture data evaluated by '
statistical methods.
The capture efficiencies presented- in Table 12 demonstrate a substantial
amount" of" variabi'l"ity~~on~ an—hourly ba's is".' This variability" i's more easily seen
-71
-------
TABLE' 12T LIQUID~AND~'C-AS 'MASS "FLOW "RATES'" AND "CAPTURE
EFFICIENCIES FOR THE COATING ROOM
Capture Stream VOC
VOC Mass Flow Rate . Capture Efficiency '
Tine * Applied Method 25A Method 18 Method 25A Method 18
Date Interval (Lb/Hr) (Lb/Hr) (Lb/Hr) (%) (%)
0200-0259
141
164
136
116
96.5
0300-0359
118
146
134
123
113
0400-0459
177
146
NA
82.5
NA
0500-0559
88.8
146
NA
164
NA
0600-0659
155
147
NA
94.8
NA
0700-0759
133
144
138
108
103
0800-0859
162
146
138
90.1
85.2
0900-0959
69.4
145
137
208
197
1000-1059
118
145
137
122
116
1100-1159
139
143
136
102
97.8 ,
1200-1259
130
146
137
112
105
1300-1359
95.6
146
137
152
143
1400-1459
132
146
135
110
102 i
1500-1559
160
145
134
90.6
83.8 I
1600-1659
57
145
NA
254
NA
1700-1759
133
126
114
94.7
85.7 i
1800-1859
126
144
135
114
107
1900-1959
142
143
135
100
95.1 '
2000-2059
137
143
136
104
99.3
2100-2159
119
119
123
100
103 i
2200-2259
130
129
92.9
99.2
71.5!
2300-2359
139
143
NA
102
NA j
0000-0059
160
KA
NA
NA
NA !
0100-0159
114
NA
NA
NA
NA 1
0200-0259
114
NA
NA
NA
NA !
0300-0359
137
NA
NA
NA
NA |
0400-0459
137
NA
NA
NA
NA ;
0500-0559
160
NA
NA
NA
NA
0600-0659
160
NA
NA
NA
NA 1
0700-0759
23.3
135.
- NA
579
NA ,
0800-0859
134
137 '
NA
102
NA
0900-0959
121
104 i
NA
86
NA j
1000-1059
110
138 .
NA
125
NA
1100-1159
143
141 '
NA
98.6
NA '
1200-1259
111
147
NA
132
na ;
1300-1359
129
144 -
NA
111
NA 1
1400-1459
136
145
NA
106
KA :
1500-1559
92.2
104 '
NA
112
na ;
1600-1659
137
150
NA
109
NA
1700-1759 ¦
132 "
¦ ---14-4-
" -- NA "
- T09 ' "
- "NA"--
(continued
72
-------
TABLE' 12 ('continued)
Capture Stream VOC
Date
VOC
Mass Flow
Rate
Capture Ef
ficiency
Time
Applied
Method 25A
Method 18
Method 25A
Method 18
Interval
(Lb/Hr)
(Lb/Hr)
(Lb/Hr)
(%)
(Z)
1800-1859
146
137 ,
NA
93.8
NA
1900-1959
123
134
NA
108
NA
2000-2059
139
138
NA
99.3
NA
2100-2159
140
134
NA
95.7
NA
2200-2259
140
139
NA
99.3
NA
2300-2359
98.3
113 .
NA
115
KA
0000-0059
151
NA
NA
KA
KA
0100-0159
93.2
NA
NA
NA
NA
0200-0259
163
NA
NA
NA
NA
0300-0359
93.2
NA j
NA
NA
NA
0400-0459
94.2
NA :
NA
NA
NA
0500-0559
141
116
NA
82.3
NA
0600-0659
118
132
NA
111
NA
0700-0759
140
132
NA
94.3
NA
0800-0859
104
85.8
NA
82.5
NA
0900-0959
151
135
NA
89.4
NA
1000-1059
81.1
109
NA
134
NA
1100-1159
78
58.5
NA
75
NA
1200-1259
132
68.7
NA
52
NA
1300-1359
113
144 ;
NA
127
NA !
1400-1459
107
95.6
NA
89.3
NA
1500-1559
171
143 '
NA
83.6
na
1600-1659
114
121 :
NA
106
NA
1700-1759
131
192 j
KA
146
NA
1800-1859
147
155
NA
105
NA
1900-1959
132
146 |
NA
110
NA
2000-2059
148
146 j
NA
98.6
NA
2100-2159
125
146 '
KA
116
NA
2200-2259
117
146 :
KA
124
NA
2300-2359
91 .2
84.9
KA
93.1
KA i
0000-0059
184
KA "
NA
KA
NA j
0100-0159
93.2
NA
NA
KA
ka :
0200-0259
140
KA ¦
KA
NA
ka :
0300-0359
140
NA
NA
NA
NA 1
0400-0459
150
NA
KA
NA
NA ;
0500-0559
100
NA
NA
NA
KA
0600-0659
166
122 "
NA
73.5
NA
0700-0759
87.1
122 '
NA
140
KA
0800-0859
91.2
125
NA
137
KA
02/15
02/16
(continued)
73
-------
TABLE 12 (continued)
Date
02/17
Capture Str
earn VOC
VOC
Mass Flow
Rate
Capture Eff
iciency
T ime
Applied
Method 25A
Method 18
Method 25A
Method 18
Interval
(Lb/Hr)
(Lb/Hr)
(Lb/Hr)
<%)
(%) '
0900-0959
133
126
NA
94.7
NA
1000-1059
147
126
NA
85.7
KA
1100-1159
112
128
NA
114
NA
1200-1259
135
137
KA
101
NA
1300-1359
127
136
KA
107
NA
1400-1459
76
62.9
NA
82.8
KA
1500-1559
121
136
NA
112
KA
1600-1659
143
135
NA
94.4
NA
1700-1759
112
136
NA
121
KA
1800-1859
146
136
NA
93.2
NA
1900-1959
104
135
NA
129
NA
2000-2059
99.3
133
NA
133
KA
2100-2159
137
132
NA
96.4
NA
2200-2259
139
135
NA
97.1
KA
2300-2359
160
135
NA
84.4
NA
0000-0059
121
NA ,
NA
NA
NA
0100-0159
137
NA
NA
NA
NA
0200-0259
102
NA
NA
NA
NA
0300-0359
126
NA :
NA
NA
NA
0400-04 59
137
NA
NA
NA
NA
0500-0559
137
NA
NA
NA
NA
0600-0659
126
NA
KA
NA
NA ;
0700-0759
131
NA |
132
NA
100 ;
0800-0859
140
133 !
131
95
93.6 '
0900-0959
147
138 ;
130
93.9
88.4 ;
1000-1059
104
140 ;
131
134
126
1100-1159
183
137 |
128
74.9
69.9 '
1200-1259
108
140 ;
128
129
1 18
1300-1359
111
136 '
126
122
113
1400-1459
149
55.3
103
37.1
69.1 .
1500-1559
141
NA
KA
NA
NA
1600-1659
168
146 '
NA
86.9
na :
1700-1759
96.3
97.4
129
101
134 ;
1800-1859
111
114
111
102
100 ;
1900-1959
83.1
108
111
130
133 i
NA - Not Analyzed
-------
in the average capture efficiencies over an entire day as shown in Table 13.
The coefficient of variation for individual method 25A data varies from 10.9 to
34.3% with an average of 23%. However, after being averaged or smoothed over a
days time, the variation is reduced to 7.4%. For method 18 data, the reduction
in variability is from 24.3% to 1.3%.
i Another logical method of smoothing the data is to evaluate capture
efficiencies during each discrete application batch. The results of such
analyses are shown in Table 14. The amount of method 18 data was insufficient
to include it in the table. For several batches, the start up and/or ending
data were not included, because of non-steady state conditions. j
i * i
These evaluations can be viewed as', steady state determinations. The batch
average capture efficiency of 103% appears to be somewhat better than the j
106.7% when averaging by day. Additionally, the batch coefficient of variation
of 4.6% is smaller than the 7.4% value 'from the daily average.
t
The fact that both mean values are; greater than 100% is puzzling. The i
maximum capture efficiency possible is ;100%, and the efficiency of this system
is obviously expected to be somewhat less than that value. With the higher <
than expected values, one has to consider the entrance of an unmeasured source
into the system or measurement bias as possible causes. A thorough review of I
the process revealed no unmeasured source of VOC. Based on the literature j
regarding the measurement methods used and on the detailed QA/QC program !
followed, there is no reason to believei that any of the measurement methods is
biased high. Therefore, no reasonable explanation can be provided for the
apparently high values.
' : I
! Overall, the test data indicate that the liquid/gas conversion methodology
can be used to determine reasonably reliable estimates of VOC capture i
efficiency around the coating room, as long as the limitations of this approach
are taken into consideration. The limitations that exist include the need for
long term averaging periods and fairly steady state operations. Only data
collected during steady state conditions should be used in efficiency ,
calculations. Also, the reliability of the estimate improves as the number of
individual measurements increases.
I
The uncertainty in the capture efficiency determinations is analyzed in
detail through an error analysis later in this section.
5.4.2 Other Material Balance Results
As described in Section 5.2.2, other material balances were included as •
part of the coating room capture efficiency test to verify measurements made in
that test and to determine other efficiencies associated with the plant VOC
control program. The results of these '¦material balances are presented in the !
following paragraphs.
5.4.2.1 Gas-Phase Balance of SLA Streams—
From the discussion in Section 5.2.2.1, it should be clear that the sum of
the -vent stream and capture-s tream-VOC masses should-equal- the- main-SLA stream
75
-------
TABLE 13. CAPTURE EFFICIENCY AVERAGES BY DAY
Average Capture Efficiency Coefficient of Variation (CV) for Hourly Data
Method 25A Method 18 Method 25A Method 18
Date (%) (%) (%) (%)
02/13
120.1
106.1
34.3
26.5
02/14
106.4
10.9
02/15
101.0
22.4
02/16
105.3
19.0
02/17
100.5
104.1
28.5
22.0
Average
' J 106.7
105.1
23.0
24.3
CV for 24
Hour Data
7.4%
1.3%
-------
TABLE 14. CAPTURE EFFICIENCY AVERAGES BY SLURRY BATCH
Test
Period
Liquid-Phase Mass
Gas-Phase Mass
Capture Efficiency
Start
Stop
' Method 25A
(lbs)
Method 25A
(%)
(Date)
(Time)
(Date)
(Time)
(lbs)
02/13
0200
02/13
0900
1037
1039
100.2
02/13
0900
02/14
0700
2860
3092
O
00
«
02/14
0800
02/15
0300
2430
2552
1
105.0 I
02/15
0400
02/16
0400
3071
2977
96.9
i
02/16
0500
02/16
2000
1751
1889
i
107.9 |
.02/16
2100
- , 02/17
1600
2543
2534
99.6
Average
103.0% :
i
1
CV for
Batch Data
4.6%
:¦
-------
mass~ A" gas-phase mass balance is~conducted by calculating the ratio of the
sum of the two component stream masses to the main stream mass. The gas-phase
mass flow rates and the associated material balances for the entire test period
are presented in Table 15 on an hourly basis. Evaluations of the gas-phase ;
closure results after smoothing are shown in Tables 16 and 17. j
i
Table 16 shows the closure values and coefficients of variation when the:
hourly data are averaged over day long periods. The average closure by Method
25A was 94.3% and 99.4% by Method 18. The average coefficients of variation !
for the daily closures are 4.3% and 9.1-%, respectively. This is considerably!
better than the 12.7% and 13.9% average coefficients of variation for the i
Method 25A and Method 18 hourly closure results. ;
! Table 17 shows the closure values ;and coefficients of variation when
hourly data are averaged over individual slurry batch use periods. The average
closures by Method 25A and Method 18 were 92.9% and 96.3%, respectively, with
average coefficients of variation at 2.5% and 8.8%.
j i i
j In evaluating the discrepancy in the Method 25A closures, one potential j
source of error surfaced. It was determined from the Method 18 analyses that j
the vent stream contained a solvent B/solvent A ratio greater than the 75/25 |
ratio assumed for calculation purposes. The exact ratio could not be specified
over the test period because only limited data were collected. An assessment!
of this factor indicates that the calculated VOC mass flow rates for the vent :
stream would be slightly lower than the actual values. The basis of this i
evaluation is presented in Section 5.4.:3. However, the significance of this
potential source of error is minimized since the vent stream mass flow rate is
relatively small in comparison to the capture stream. Additionally, the same j
potential source of error likely affected the main SLA mass flow determinations
to a lesser degree, thus tending to negate the net effect on the balance. j
i
The closure values from both daily and batch data tend to confirm that the
individual gas stream measurements are in agreement.
i i
I l
• From the gas stream data for the entire test period, the fractional i
contributions for the vent system and the capture system to the total SLA load
were determined to be approximately 20%. and 80%, respectively.
5.4.2.2 Liquid/Gas Material Balance of Recoverable VOC— !
j From the discussion in Section 5.2.2.2, it can be seen that the liquid VOC
mass recovered (recovered solvent) should equal the net recoverable VOC mass in
the gaseous form (sum of gas-phase VOC mass input to the carbon bed minus the •
exhaust). The purpose of this comparison was to verify that the liquid and j
gas-phase mass measurements were equal.. !
!
I The total recovered liquid VOC during the test was determined to be 13,450
pounds. When compared to 16,586 pounds., as determined by the gas-phase
measurements over the same period, the "net difference is approximately 21%.
' The difference between the two calculated masses was greater than 1
orrgrna-l-ly expected.—-However,- it should- be—noted--that the-calculated gaseous
78
-------
TABLE L5• GAS-PHASE MASS FLOW RATES AND MATERIAL BALANCES FOR THE SLA SYSTEM
1
Method 25A
Mass Flowrates
Method 18
Mass Flowrates
Gas-Phase
Closur
Capture
Ve nt
Main SLA
Capture
Vent
Main SLA
EPA 25A
EPA 18
Date
Interval
(Lb/Hr)
(Lb/Hr)
(Lb/Hr)
(Lb/Hr)
(Lb/Hr)
(Lb/Hr)
(»
<*)*
02/13:, 0200-0259
164
32.4
NA
136
NA
NA
NA
NA
; 0300-0359
146
27.7
181
134
29
NA
96
90.1
i 0400-0459
146
28.6
183
NA
28.6
NA
95.4
95.4
0500-0559
146
29.6
176
NA
NA
162
99.8
108
0600-0659
147
29.7
186
NA
NA
NA
95
NA
0700-0759
144
30
188
138
NA
171
92.6
98.2
0800-0859
146
30.1
188
138
NA
NA
93.7
89.4
i 0900-0959
145
31.7
188
137
NA
NA
94
89.7
1 1000-1059
145
30.1
186
137
NA
NA
94. 1
89.8
i 1100-1159
143
29
186
136
NA
NA
92.5
88.7
1200-1259
146
30
183
137
NA
NA
96.2
91.3
1300-1359
146
29.6
183
137
NA
NA
96
91
1400-1459
146
29.6
185
135
NA
NA
94.9
89
1500-1559
145
29.3
182
134
NA
NA
95.8
89.7
1600-1659
145
31.9
187
NA
NA
NA
94.6
NA
1700-1759
126
29.3
160
114
NA
NA
97.1
89.6
1800-1859
144
36.1
191
135
NA
NA
94.3
89.6
1900-1959
143
33.3
191
135
NA
NA
92.3
88.1
2000-2059
143
33.8
191
136
NA
NA
92.6
88.9
2100-2159
119
33.1
135
123
NA
NA
112
115
1 2200-2259
129
32.3
191
92.9
NA
NA
84.5
65.5
1 2300-2359
1
143
30.8
191
NA
NA
NA
91
NA
02/14
0000-0059
NA
NA
191
NA
NA
NA
NA
NA
0100-0159
NA
NA
188
NA
NA
NA
NA
NA
! 0200-0259
NA
NA
186
NA
NA
NA
NA
NA
; 0300-0359
NA
NA
188
NA
NA
NA
NA
NA
0400-4059
NA
NA
194
NA
NA
NA
NA
NA
: 0500-0559
NA
NA
209
NA
NA
NA
NA
NA
' 0600-0659
NA
NA
194
NA
NA
NA
NA
NA
' 0700-0759
135
41.3
197
NA
NA
NA
89.5
NA
; 0800-0859
137
43.8
192
NA
NA
NA
94.2
NA
0900-0959
104
40
162
NA
39.8
NA
88.9
88.8
1000-1059
138
40.4
196
NA
42.8
NA
91
92.2
; 1100-1159
141
34.8
180
NA
35.6
NA
97.7 '
98.1
i 1200-1259
147
31.6
180
NA
31
NA
99.2
98.9
! 1300-1359
144
30.6
183
NA
30.1
NA
95.4
95.1
1400-1459
145
30.6
179
NA
29.9
NA
98.1
97.7
1500-1559
104
19.8
122
NA
29.8
NA
101
109
1600-1659
150
29.1
196 ,
NA
30.3
NA
91.4 .
92
; 1700-1759
144
29.3
176 ''
NA
28.5
NA
98.5
98
! 1800-1859
137
29.2
194
NA
29.3
NA
85.7 :
85.7
i 1900-1959
134
28.3
189 .
NA
28.5
NA
85.9
86
.(continued)
-7-9-
-------
TABLE 15 (continued)
Date
Interval
Method 25A
Mass Flowrates
Method
18 Mass Flowrates
Gas-Phase
CI osur<
Capture
(Lb/Hr)
Vent
(Lb/Hr)
Main SLA
(Lb/Hr)
Capture
(Lb/Hr)
Vent
(Lb/Hr)
Main SLA
(Lb/Hr)
EPA 25A
(%)
EPA 18
(20*
2000-2059
138
28.4
189
NA
28.8
NA
88
88.3
2100-2159
134
29.1
191
NA
30.4
NA
85.4
86.1
2200-2259
139
31.2
194
NA
32.3
NA
87.7
88.3
2300-2359
113
32.4
162
NA
34.4
NA
89.8
91
02/15
0000-0059
NA
NA
194
NA
NA
NA
NA
NA
0100-0159
NA
NA
128
NA
NA
NA
NA
NA
0200-0259
NA
NA
194
NA
NA
NA
NA
NA
0300-3059
NA
NA
194
NA
NA
NA
NA
NA
0400-0459
NA
RA
118
NA
NA
NA
NA
NA
0500-0559
116
25.9
186
NA
NA
NA
76.3
NA
0600-0659
132
25.8
186
NA
NA
161
84.8
98
0700-0759
132
33.1
194
NA
NA
161
85.1
102
0800-0859
85.8
34.6
163
NA
NA
94.9
73.9
126
0900-0959
135
38.7
176
NA
NA
NA
98.7
NA
1000-1059
109
35.9
115
NA
NA
NA
126
NA
1100-1159
58.5
30
139
NA
NA
NA
63.7
NA
1200-1259
68.7
18.8
174
NA
NA
NA
50.3
NA
1300-1359
144
35.6
172
NA
NA
NA
104
NA
1400-1459
95. 6
23
172
NA
NA
NA
69
NA
1500-1559
143
34.9
166
NA
NA
153
107
116
1600-1659
121
39.1
170
NA
NA
148
94.2
108
1 1700-1759
192
51.7
170
NA
NA
230
143
106
! 1800-1859
155
40.1
180
NA
NA
176
108
110
j 1900-1959
146
39.2
180
NA
NA
166
102
111
; 2000-2059
146
38.5
180
NA
NA
163
102
113
j 2100-2159
146
38.6
180
NA
NA
160
102
115
¦= 2200-2259
146
42.7
182
NA
NA
167
103
113
: 2300-2359
84.9
44.7
131
NA
NA
114
98.9
113
02/16
0000-0059
NA
NA
184
NA
NA
170
NA
NA
. 0100-0159
NA
NA
142
NA
NA
NA
NA
NA
: 0200-0259
NA
NA
184
NA
NA
NA
NA
NA
0300-0359
NA
NA
182
NA
NA
NA
NA
NA
; 0400-0459
NA
NA
184
NA
NA
NA
NA
NA
j 0500-0559
NA
NA
187 "
NA
NA
NA
NA
NA
] 0600-0659
122
33.6
187 -
NA
NA
NA
83.2
NA
' 0700-0759
122
34.7
191
NA
NA
NA
82
NA
0800-0859
125
36.7
191
NA
NA
NA
84.7
NA
; 0900-0959
126
37.5
189 ¦:
NA
NA
NA
86.5
NA
j 1000-1059
126
34.3
186
NA
NA
NA
86.2
NA
1100-1159
128
31.8
184
NA
NA
NA
86.8
NA
1200-1259
137
30.4
181 '
NA
NA
NA
92.5
NA
(continued)
-------
TABLE 15 (continued)
Date
Interval
Method 25A
Mass Flowrates
Method 18
Mass
Flowrates
Gas-Phase
Closure
Capture
(Lb/Hr)
Vent
(Lb/Hr)
Main SLA
(Lb/Hr)
Capture
(Lb/Hr)
Vent
(Lb/Hr
Main SLA
) (Lb/Hr)
EPA 25A
(%)
EPA 18
(%)*
1
1300-1359
136
30.4
177
NA
NA
NA
94
NA
i 1400-1459
62.9
31
87.4
NA
NA
NA
107
I NA
1 1500-1559
136
30.3
177
NA
NA
NA
94
NA
! 1600-1659
135
31.2
180
NA
NA
NA
92.3
NA
1700-1759
136
30.8
180
NA
NA
NA
92.7
NA
1800-1859
136
31.8
187
NA
NA
NA
89.7
NA
1900-1959
135
31.2
187
NA
NA
NA
88.9
NA
2000-2059
133
33
192
NA
NA
NA
86.5
NA
2100-2159
132
32
189
NA
NA
NA
86.8
NA
! 2200-2259
135
31.8
187
NA
NA
NA
89.2
NA
1 2300-2359
l
135
32.6
189
NA
NA
NA
88.7
NA
02/17
0000-0059
NA
NA
189
NA
NA
NA
NA
NA
! 0100-0159
NA
NA
187
NA
NA
NA
NA
NA
i 0200-0259
NA
NA
187
NA
NA
NA
NA
NA
j 0300-0359
NA
NA
65.5
NA
NA
NA
NA
NA
j 0400-0459
NA
NA
187
NA
NA
NA
NA
NA
: 0500-0559
NA
NA
189
NA
NA
NA
NA
NA
j 0600-0659
NA
32.5
189
NA
NA
NA
NA
NA
! 0700-0759
NA
33.7
189
132
NA
NA
NA
87.7
! 0800-0859
133
34.2
189
131
NA
NA
88.5
87.4
j 0900-0959
138
34.8
182
130
NA
NA
94.9
90.5
1000-1059
140
35.2
173
131
NA
NA
101
96.1
1100-1159
137
32.8
175
128
NA
NA
97
91.9
; 1200-1259
140
31.2
173
128
NA
NA
99
92
; 1300-1359
136
30.8
173
126
NA
NA
96.4
90.6
1400-1459
55.3
34.7
143
103
NA
NA
62.9
96.3
; 1500-1559
NA
NA
176
NA
NA
NA
NA
NA
. 1600-1659
146
40.3
164
NA
NA
NA
113
NA
; 1700-1759
97.4
32.7
85.8
129
NA
NA
151
188
' 1800-1859
114
32.7
35.7
111
NA
NA
410
402
1 1900-1959
108
32.2
28.6
111
NA
NA
490
500
i
i
* Closure determined by combining the avail-able Method 18 data with the other necessary
Method 25A data.
-81-
-------
TABLE 16. SLA GAS-PHASE MATERIAL BALANCES USING DAILY AVERAGING PERIODS
Date
Average Gas-Phase Closure
Method 25A Method 18
(I) (%)
Coefficient of Variation (CV) for Hourly Data
Method 25A Method 18
(%) (%)
02/13
02/14
02/15
02/16
02/17
95.0
92.2
94.3
89.4
100.4
91.5
93.0
110.9
102.3
5.1
5.7
23.1
6.2
23.2
10.7
7.0
6.4
31 .6
Average
94.3
99.4
12.7
13.9
CV for Daily Data
4.3%
9.1%
-------
TABLE 17. SLA GAS-PHASE MATERIAL BALANCES USING SLURRY BATCH PERIODS
Test
Period
Average Gas-
¦Phase Closure
Coefficient of Variation
for Hourly Data
Start
Stop
Method 25A
Method 18
Method 25A
Method 18
(Date)
(Time)
(Date)
(Time)
(%)
(%)
(%)
(%) ;
02/13
0200
02/13
0900
95.4
96.2
2.6
7.8
02/13
0900
02/14
0700
94.8
89.7
5.9
11.3 1
02/14
0800
02/15
0300
92.4
93.0
5.8
7.0 i
02/15
0400
02/16
0400
94.3
110.9
23.1
6.4 i
02/16
0500
02/16
2000
90.0
7.0
.02/16
2100
" > .
02/1 7
_160.0 .
.90.4.,
. . 91..6
.12.0.
3.6. |
Average
92.9
96.3
9.4
7.2
CV for Batch Data
2.5%
8.8%
-------
mass value was based on measurements of two separate streams, and the recovered
liquid mass was calculated from a combination of two separate meter readings
and one tank measurement. The gas-phase measurements were discussed in the
previous section.
Brief discussions are presented below regarding the reliability of the
liquid measurements. Comparisons were made between the recovered solvent
metered values and those determined by tank measurements. i
¦ I
The plant solvent A feed meter indicated that a total of 176 gallons or
about 1271 pounds (solvent A density = 7.22 lb./gal.) was removed from the
virgin tank during the test period. According to the depth gauge readings only
715 lb. were used. The difference of 556 pounds must be attributed to
recovered solvent A. This value, when compared to the 91 gallons or 657 pounds
registered by the dedicated solvent A recovery meter, indicates a discrepancy
of approximately 15%.
i ;
j Data for solvent B recovery corresponding to the analysis above show 669 .
gallons or 4951 pounds of solvent (solvent B density = 7.AO lb./gal.) removed
through the main tank meter. Depth gauging indicates a gain of 3665 pounds
during the test for a total of 8616 pounds. The recovered solvent B meter
showed recovery of 1182 gallons or 8747 pounds. The difference of 1.5% is i
significantly less than that for solvent A.
5.4.2.3 Plant-Wide Capture Efficiency-,-
The concepts for determining plant|-wide capture efficiency were discussed
in Section 5.2.2.3. The approach followed in this determination was a
comparison of VOC mass captured with VOC mass vaporized. Determination of the
captured VOC was straightforward. However, determination of mass vaporized was
extremely complex and indirect, because1 of the complicated recycling of
recovered solvent and venting of the majin storage tanks to the SLA system.
The total VOC captured and sent to; the control device can be estimated by
either the sum of the vent and capture ^streams or by the carbon adsorber inlet
stream. Over the testing period, these values were 17,735 pounds and 18,825 !
pounds, respectively. For the remainder of this section, the captured mass is
assumed to be that of the CAB inlet (18,825 pounds). The CAB inlet value was;
selected because it represents a single; continuous measurement point and the '
other value involves combining measurements and potential sources of error.
' An estimate of the total solvent evaporated during the test period can bei
obtained from the total amount flowing "into the plant adjusted by the net
change inside the plant. From the metex readings reported in Appendix D, 4900
pounds of solvent B, 1300 pounds of solvent A, and 5600 pounds of premix were 1
fed into the plant during the test period. From Table 18, liquid inventory i
data, a net decrease of 3815 pounds is 'indicated for the inside inventory. The
sum of the values is 15,619 pounds and .represents a liquid measure of the
solvent evaporated inside the plant. It should be noted that this value does
not include the VOC vaporized in the outside tanks and distillation area which
are captured by the vent system and sent to the recovery unit.
84
-------
TABLE 18. LIQUID MASS INVENTORY DATA*
2/13 2/13 2/13 2/14 2/14 2/14 2/15 2/15 2/15 2/16 2/16 2/16 2/17 2/17 2/17 Inventory Change
Tank 8:30 16:00 22:15 8:45 16:00 22:15 8:45 16:00 22:15 8:45 15:00 22:15 8:45 15:30 20:00 (Start to Finish)
Mix Room
T101
T102
T103
Myera
4430
4178
247 5
0
4404
4164
2212
0
4397
4178
787
0
1908
1856
2385
0
1863
6158
2361
0
1921
6144
2344
0
1942
6122
2871
0
1870
6122
0
0
0
5932
0
3326
4773
4113
2568
0
4615
4149
2399
0
4602
4149
2370
0
4704
4295
754
0
2637
4164
0
0
2562
4178
0
0
-1868
0
-2475
0
Mill Koouj
T151
T152
T201
T202
Was h
Sink
4563
1846
6125
3523
574
4563
300
5755
5007
581
4550
1808
5689
4214
552
4507
5239
2379
5906
559
1897
5170
4908
4921
552
1371
5176
5588
4129
552
0
4919
4743
5211
574
1843
4926
4743
4357
581
4149
4919
4773
3587
581
4315
3567
3071
5125
596
4315
1237
5365
4356
581
4315
669
4258
5535
588
4471
1476
5037
4054
596
4445
4374
5001
3285
581
3083
4473
4062
5004
537
-1480
2627
-2063
1481
-37
Inside Subtotal
-3815
Bulk Storage
Solvent
A 37916
Solvent
B 41167
l'remix 27409
Dist il-
late 43066
35162
42041
27 284
43794
38393
41789
27284
43794
38155
42285
27048
43911
38393
42640
22798
45132
38093
43122
22445
44890
37916
43492
22209
45728
38393
45313
22445
45868
38039
42404
22209
45368
36839
42404
22092
45250
37078
43729
22209
45853
370 78
43729
21974
45368
37078
43729
21621
46802
374 39
44951
21739
47044
37201
44832
21739
47280
-715
3665
-5670
4214
Outside Subtotal
1494
Tota 1
Inve n-'
tory
177272
175267
177435
176138
176793
175721
175727
176461
175281
1 7471 3
175886
174635
17461 7
175660
174951
-2321
*A11 values are in pounds.
-------
The total mass of-VOC vaporized can also be viewed as the total mass of
vapors formed. This mass can then be determined by summing the captured and
uncaptured VOC masses. The captured mass is directly determined, while the
uncaptured mass can only be indirectly determined. The uncaptured mass is
equal to the 2321 pounds of total VOC lost from the plant (difference between
beginning and ending liquid inventories) minus 1149 pounds of VOC mass
exhausted from the control device. The mass of VOC captured and sent to !
control device was 18,825 pounds. The uncaptured VOC mass was 1172 pounds
(2321 pounds - 1149 pounds). The total mass of vaporized VOC was therefore
19,997 pounds.
| The overall capture efficiency determination using the first concept forj
approximating the mass of vaporized VOC is 120%. This compares to j
approximately 94% as determined using the second concept. Obviously, there
cannot be more VOC captured than vaporized, thus demonstrating a significant «
amount of error in the first calculation. An obvious source of error is the
exclusion of the tank and distillation vapors from the total emissions, and
their inclusion in the total captured mass. Although the second calculation ;
yields a value in the expected range, it is based on data which were obtained
as indirectly as that for the first calculation, and must therefore be !
considered as equally prone to error.
! |
5.4.2.4 Plant-Wide Recovery Efficiency— '
The plant-wide recovery efficiency, is the percentage of the total VOC mass
vaporized which is recovered from the solvent recovery unit. The liquid
recovered (13,450 pounds) can be compared with the VOC mass evaporated (15,619
pounds) or the mass of VOC vapors formed (19,997 pounds), as discussed in the:
previous section. The recovery efficiency values calculated by both methods
are 86% and 67%. As in the previous section, these values were obtained using
indirect methods and may be somewhat error prone.
In evaluating plant-wide recovery determinations, an attempt was made to,
close a complete material balance around the plant. From the VOC mass
evaporated and recovery data, it can be' seen that only 2169 pounds of the
15,619 pounds are unaccounted for. The' most obvious source of fugitive losses
is the carbon adsorber. Measurements of the carbon adsorber exhaust indicate •
that 1149 pounds of VOC were lost during the period. Measurements and
calculations of VOC mass lost with the water discharge from the solvent
recovery unit estimate a maximum loss of 2 pounds during the test period. The
only obvious paths of VOC loss within the plant were floor sweep exhausts in ¦
the mixing and milling areas and product retention. Measurements of VOC
concentrations and determinations of floor sweep flow rates yielded an
approximation of fugitive losses totalling 192 pounds. No value was determined
for the product retention. Totalling all sources accounts for 14,793 pounds or
95% of the solvent throughput. 1 i
! i
5.4.3 Error Analysis
! Monte Carlo simulations were used to derive estimates of overall
uncertainty in the liquid/gas capture efficiency results for the coating room.
Using-the-Monte-Carlo—simu-l-ation- technique; 400 closure estimates were - - -
86
-------
generated from'a "typical" set "of" measurement "data. With this technique, each
measurement parameter (e.g., gas flow rate for the capture stream, VOC
concentration of the capture stream, liquid density, etc.) was perturbed
randomly to simulate the actual variation that would occur in repeated
measurements under a single set of process conditions. A Guassian distribution
was assumed in simulating this variation.
Error estimates used in the simulation included both random and systematic
error components corresponding to error resulting from imprecision (i.e., i
random variability in replicate measurements) and to error resulting from i
measurement bias. For instance, replicate measurements of liquid density were
found to have a precision of 1.3%, expressed in terms of the coefficient of \
variation. Performance audits of the density measurements indicated an average
bias of -0.04% with a standard error of. 1.3%. For the coating line liquid/gas
closure simulation, a "typical" liquid 'density of 1.25 g/ml was used. This ¦
input value was varied in each of 400 closure calculations, according to the 1
input error estimates. In other words,! the error statistics were used to
generate the simulated density measurement values which were in turn used to :
calculate liquid/gas closure. The resulting set of closure results is thus aj
simulation of results which would theoretically be obtained if 400 closure |
measurements could be made under a single set of process conditions. The
statistical distribution of this simulated data set were then used to derive
estimates of uncertainty for the actual closure data as discussed below. Input
values and error estimates for measurement parameters used in the Monte Carlo !
simulation of coating line liquid/gas closure are shown in Table 19.
i '
i The output of the Monte Carlo simulation is a set of closure values, which
is assumed to be statistically representative of the population of all possible
measurements which could be made under !the given set of process conditions. !
Assuming that a similar relative distribution would be expected under any other
set of process conditions, these result's may be used to estimate relative i
uncertainty in the actual closure estimates. For the coating line liquid/gas .
closure, the Monte Carlo simulation indicated a relative bias of -3.9% in
measured closure. In other words, if repeated measurements were made of a !
process where the actual closure was 100%, the mean measured value would j
approach 96.1%. The simulation further indicated that approximately 95% of the
individual measurements in such a case would be between 80% and 120%, since
more than 95% of the 400 simulated values fell within +20% of the "true" value.
Approximately 74% fell within _+10% of the true value and slightly more than 38%
fell within +5%. j
j The error distribution described above may be used to define the probable
uncertainty associated with any single closure measurement. For a single j
measurement for example, there is a probability of 38% that it is within +5% of
the true value, and a 95% probability that it is within +20% of the true value.
In practice, however, closure estimates]; would rarely be based on a single
determination. Rather, repeated measurements would likely be made and the
average of these reported as the closure estimate. Error in such an estimate
is substantially reduced since some of "the random measurement variability tends
to be averaged out. Overall error in the closure estimate becomes smaller as
the—number' of individual—measurements used to~*compute the estimates" becomes
87
-------
TABLE 19. INPUT VALUES FOR MONTE CARLO SIMULATION OF COATING LINE
LIQUID/GAS CLOSURE MEASUREMENTS
Measurement
Parameter
Input
Valuea
Coefficient
of Variation*5
Relative Biasc
Mean
Standard Error
Capture Stream VOC
Concentration
j
i Capture Stream Flow
i Ra te
Capture Stream
; Molecular Weight
Liquid Density
| Coating Mixture
! Volatile Fraction
i
I Slurry Tank Hourly Volume
j Change from T to T2
3150 ppm
3850 ft^/min.
7 7 g/g-mole
1.25 g/ml
0.60
18 gal
4%
4%
0%
1.3%
0.3%
0%
-3.1%
0%
0%
-0.04%
0.8%
0%
4.2%
3.1%
0%
1.3%
0.7%
5%
a Selected as representing a typical measured value.
k Standard deviation for replicate measuremnts, expressed as a percentage of the input value.
c Systematic error expressed as a percentage of the input value.
-------
larger-.—Based-on-the Monte-Carlo results for coating-line™liquid/gas ¦ closure,
the mean of three measurements would, at the 95% confidence level, be expected
to be within +10.7% of the true value, while the mean of six measurements would
be expected to be within +7.6%. A plot of this relationship for different
numbers of measurements is shown in Figure 6. These estimates are based on the
relationship:
: c_ = -j/ /n~ i
1 x
where, ;
! standard deviation of means, I
cx = standard deviation of the population from which the samples are j
j drawn, and j
I n = number of samples used to compute the mean.
I f !
A Gaussian distribution is assumed in the above expression. |
i
5.5 QUALITY ASSURANCE AND QUALITY CONTROL
l ;
j This data quality assessment was based upon performance and systems QA
audits and upon internal QC data generated throughout the field testing
program. This section summarizes the results of the QA/QC program. Detailed
QA/QC data and discussions are provided' in a separate document titled
'('Measurement of Process Capture Efficiency, Quality Assurance Report for Field
Testing".
! The primary emphasis of this section is upon the gas phase VOC
measurements using Methods 18 and 25A, although other measurements necessary j
for determining capture efficiency (e.g., density, volatile fraction, etc.) are
also addressed. Overall, the QA/QC data indicate that the measurement data
collected during the plant tests are sufficiently reliable to be used for
evaluation of the feasibility of the liquid/gas-phase mass balance approach.
I !
5.5.1 Method 25A
t Performance audit data indicated that Method 25 VOC measurements were
within the +5% objectives for bias for all four instruments/sampling locations.
Although the solvent B/solvent A ratio in the calibration gas mixture was
different than that for the sample gas at some locations, there seemed to be no
appreciable error in the results. ThiSj item is further discussed in Section
5.5.4. Control sample data for the Method 25A analyses indicated within-day
precision ranging from 0.9% to 2.7%, expressed in terms of the pooled
coefficient of variation. Between-day variability ranged from 1.2% to 4.0% for
the four instruments. ;
: i
I i
5.5.2 Method 18 t ¦
j
I
; Method 18 performance audit results indicated that these analyses met the
+10% bias objectives specified in the QA Project Plan. Control sample data
indicated within-day precision of 1.1% for solvent B and 0.7% for solvent A.
Day-to-day• variability-of -the- control-- sample—results was -4v8%- and -4%0% for - - ¦
solvent B and A, respectively.
89
-------
1 3 5 7 9 11 13 15 17 19 21 23
Number of Measurements
Figure 6. Range of error for means of multiple
measurements of coating line liquid/gas closure.
-------
5• 5•3—Other Measurements
Performance audits of the density and volatile fraction determinations
indicated that the relative error of both of these analyses was within +2%.
Audits of the balances indicated a maximum error of 0-04 g for a 100 g load.
5.5.4 Byron 401 Response Factors
] All of the individual THC data given in Appendix D and summarized in ;
Section 5 of this report were based on the response factors obtained from
multipoint calibrations using certified mixtures of solvent A and solvent B.
! j
j Since flame ionization detectors used in VOC analyzers have differing I
detection efficiencies for different compounds, the response factor for the i
mixture changes with the ratio of the components. Therefore, if the instrument
response factor is based on responses using a calibration stream with a ,
different ratio than that in the measurement stream some error may exist in !
values determined. j
j In the field test, several streams such as the vent stream and carbon bed
outlet are known to have component ratios different from those in the
calibration. In particular, the CAB stack outlet is almost pure solvent B.
This creates a question of what the impact might be on VOC data collected from
streams with ratios different than that, during calibration and on the resultant
mass flow calculations. i
; I
; To assess the potential impact, a series of known mixtures of solvent A
and B were analyzed on a Byron 401 and |the response factors calculated. The
results of these are given in Table 20 and are shown graphically in Figure 7. ;
These data yields a correlation coefficient of 0.9917. The slope of the linear
least squares line is 0.002524 RF units, per % solvent B. This model predicts :
an RF factor of 0.8760 for pure solvent A and 0.6237 for pure solvent B. i
j The plot of the experimental data for solvents encountered in this test !
yields a relatively shallow slope. This indicates that the difference in !
response factors for any combination of the two solvents should not be large,
unless the difference between the expected ratio and actual ratio is large.
For example, if the expected ratio of the measurement stream and the ratio used
in calibration is 75/25 and the actual ratio is 80/20, the respective response
factors should be 0.6867 and 0.6741. The difference in the response factors is
less than 2 percent. In addition to the response factor, the carbon number and
molecular weight for the mixture also change with the ratio of the components !
and play integral roles in the calculation of the mass flow rate. For the case
above, inclusion of the carbon number and molecular weight for the calibration
mixture also cause the calculated mass flow rate to be lower than the true j
value. For typical VOC mass flow rates/in the capture stream (e.g. 130 :
lb./hr.) and the difference between the.calibration and the actual ratio
described above, the calculated value would be about 3.5 percent lower than the
actual mass flow rate.
91
-------
TABLE 20. BYRON 401 RESPONSE FACTOR VARIATIONS WITH COMPOSITION
Run No.
Solvent B
(ppmc)
Solvent A
(ppmc)
THC
(ppmc)
Solvent
(Z)
»u)
Solvent A
(%)
(1)
RF
vo
K3
1
2
3
4
5
6
7
8
9
10
11
0
4160
1335
441
208
, 998
1726
2721
3370
703
2184
3448
0
2341
3082
3275
2620
2017
1193
655
2865
1638
3448
4160
3676
3 523
3483
.3618
3743
3913
4025
3568
3822
0
100
36.3
12.5
6.0
- 27.6
46.1
69.5
83.7
19.7
57.1
100
0
63.7
87.5
94.0
72.4
53.9
30.5
16.3
80.3
42.9
0.8990
0.6367 I
0.7888 :
0.8373 :
0.8469 :
0.8014 ;
0.7614 i
0.6899 1
0.6584
0.8268
I
l
0.7327
(D % by volume of the total VOC,
-------
SOLVENT A (%)
09000-
90
_J
80
I
70
_L_
60
I
50
—I
40
30
20
I
10
0.8000-
RF
0 7000
06000
I-
30
I-
40
"T"
50
-T"
60
-T"
70
T"
80
I
10
20
90
I
100
SOLVENT B (%)
Figure 7. Response factor variations with compositions for a Byron 401.
-------
For other solvent mixtures at other test sites, the degree of potential
error due to assuming the wrong mixture ratio may be greater or less than that
for the two solvents in this test. The magnitude of the potential error will
be dependent on the degree of difference between the response efficiencies,
carbon numbers, and molecular weights for the various organic components.
i
i
!
i
i
i
i
-------
" ¦••• REFERENCES - ""
1. Publication Rotogravure Printing - Background Information for Proposed
Standards, EPA-450/3-80-031a, OAQPS, U.S. Environmental Protection
Agency, Appendix C, May, 1980.
i i
2. Graphic Arts Emission Test Report, EMB Report 79-GRA-3, EPA Contract No.
' 68-02-2818, Monsanto Research Corporation, April, 1979. i
i 1
3. Determination of Capture and Destruction Efficiencies of Selected
| Volatile Organic Compound Control Devices in the State of Illinois, EPA
i 905/2-80-005, U.S. Environmental Protection Agency, April, 1981.
, 1
4. Flexible Vinyl Coating and Printing Operations - Background Information
for Proposed Standards, Radian Corporation for EPA/OAQPS, Draft. ¦
5. Pressure Sensitive Tapes and Label's, Emission Test Report, EPA Contract j
No. 68-02-2818, Monsanto Research .Corporation, June, 1979.
I
6. Metal Coal Surface Coating - Background Information for Proposed 1
] Standards, EPA-450/3-80-0359, OAQPS, U.S. Environmental Protection
; Agency, October, 1980. ; 1
7. Emission Test of a Coal Coating Plant in St. Louis, Missouri, EPA ¦
Contract No. 68-02-2814, Midwest Research Institute, October, 1980. j
8. Industrial Surface Coating Cans, Emission Test Report, EMB Report
79-ISC-7, EPA Contract No. 68-02-2&20, TRC Environmental Consultants,
! August, 1979. ; j
I \
9. Industrial Surface Coating (Can Coating), Emission Test Report, EMB
Report 79-ISC-8, EPA Contract No. 68-02-2820, TRC Environmental j
Consultants, December, 1979. ! !
10. Deuble, J. L., Atmospheric Tracer Gas Technology, Techniques and
! Applications. Technical Bulletin |79-5, Systems, Science and Software, ,
I La Jolla, California. ; j
11. Gilath, C., Ventilation and Air Pollution Studies Using Radioactive ]
Tracers - A Critical Review. International Journal of Applied Radiation !
and Isotopes, Vol. 28, pp. 847-854, 1977. I
i
i - i
12. Lamb, B. K., Vitols, V., and Skogvold, 0., Atmospheric Tracer Techniques j
j and Gas Transport in the Primary Aluminum Industry. Journal of the Air |
j Pollution Control Association, Vol:. 30, No. 5, May, 1980. I
i
13. Hampl, V., Industrial Local Exhaust Efficiency Evaluation by a Tracer
| Gas Technique. Presented at the American Industrial Hygiene Conference,
| Cincinnati, Ohio, 1982.
95
-------
14——-Hamplj Vladimir. Evaluation of Industrial Local Exhaust Hood-Efficiency
by a Tracer Gas Technique. American Industrial Hygiene Association
Journal, 45(7):485-490, 1984.
15. Drivas, P. J., Simmonds, P. G., and Shair, F. H. , Experimental
Characterization of Ventilation Systems in Buildings. Environmental
Science and Technology, Vol. 6, No. 7, July, 1972.
16. Ellenbecker, M. J., Gempel, R. F., and Burgess, W. A., Capture
Efficiency of Local Exhaust Ventilation Systems, American Industrial
Hygiene Association Journal, Vol. 44, No. 10, October, L983.
i
17. Caplan, K. J., Energy Conservation and Comfort Performance of Auxiliary
! Air Laboratory Fume Hoods. Draft,' 1980.
j ;
18. J. C. Williams, III and F. R. DeJa'rnette, A Study of the Accuracy of the
| Type-S Pitot Tubes. U.S. Environmental Protection Agency, Research
| Triangle Park, NC, EPA-600/4-77-03,0. June 1977.
19. M. R. Midgett, The EPA Program for, the Standardization of Stationary
Source Emission Test Methodology - A Review. U.S. Environmental
j Protection Agency, Research Triangle Park, NC, EPA-600/4-76-044. August
| 1976'
20. Cochran, W. F. and Cox, G. M. Experimental Design, Second Edition, John
j Wiley and Sons, Inc., 1957.
96
-------
APPENDICES
APPENDIX A. Information Sources Reviewed 98
APPENDIX B. Previous Liquid/Gas-Phase Tests Reviewed 102
APPENDIX C. Laboratory Test Data and Calculations *
APPENDIX D. Field Test Data and Calculations *
*Appendices C and D may be obtained by writing to the U.S. EPA, Hazardous Waste
Engineering Research Laboratory, 26 West St. Clair St., Cincinnati, OH 45268
97
-------
APPENDIX A
INFORMATION SOURCES REVIEWED
98
-------
. -APPENDIX A -
OTHER REFERENCES
Bender, M. and W. D. Baines. Operation of an Open Canopy Fume Hood in a
Crossflow. Journal of Fluids Engineering, 242-243, June 1975.
Bender, Manfred. Fume Hoods - Open Canopy Type. Their Ability to Capture i
Pollutants in Various Environments. Air Pollution Control Association, 2-19,!
June, 1977.
i
Caplan, Knowlton J. Energy Conservation and Comfort Performance of Auxiliary
Air Laboratory Fume Hoods. Industrial Health Engineering Associates, Inc. ;
December 1980. ; i
I ' !
Cheremisinoff, R. N., P. E. and N. P. Cheremisinoff. Calculating Air Flow \
Requirements for Fume Exhaust Hoods... total and partial enclosure types. j
Plant Engineering, 111-114, February 1976. |
i ;
Dallavalle, J. M. The Importance of Velocity Characteristics in the Design
of Local Exhaust Hoods. Journal of Industrial Hygiene, 15(1), 18-26, January
i 93 3.
! ; i
Deuble, J. L. Atmospheric Tracer Gas Technology, Techniques and
Applications. Technical Bulletin 79-5.: Systems, Science and Software, La
jolla, California.
i
Drivas, Peter J., Peter G. Simmonds, and Fredrick H. Shair. Experimental
Characterization of Ventilation Systems in Buildings. Division of Chemistry
and Chemical Engineering and Jet Propulsion Laboratory, California Institute !
of Technology, Pasadena, California. Environmental Science and Technology,
Volume 6, Number 7, July 1972. !
I
Ellenbecker, Sc.D., Robert F. Gempel and William A. Burgess. Capture
Efficiency of Local Exhaust Ventilation Systems. American Industrial Hygiene ,
Association, J. 44(10):752-755, 1983. j
I ' !
i
Fitzsimmons, Charles K. and Kummler, Ralph H. A Chemiluminescent Detection
System for Monitoring Volatile Organic Compounds. TS-AMD-82080,
Environmental Monitoring Systems Laboratory Office of Research and
Development, U.S. Environmental Protection Agency, Las Vegas, Nevada, 1982. ;
i i
1 i
Flanigan, L. J., Talbert, S. G., Semones, D. E., and Kim, B. C. , Development [
of Design Criteria for Exhaust Systems for Exhaust Systems for Open Surface ]
Tanks. HSM 099-71-61, U.S. Department -of Health, Education, and Welfare,
Cincinnati, Ohio, 1974. 1-66 pp.
Fraser, David A., Sc.D. An Innocuous Tracer Technique for Testing the
Performance of Ventilation Systems. American Industrial Hygiene Association
journal., _26.(5.): 490 , -19.66.„
99
-------
CiTat'K7 Chaim. Ventilation arid Air Pollution Studies Using R'a'dioa'ctive
Tracers. A Critical Review. International Jorunal of Applied Radiation and
Isotopes, 28, 247-854, 1977.
Goodfellow, H. D. and M. Bender. Design Considerations for Fume Hoods for
Process Plants. American Industrial Hygiene Association Journal
(41) :473-483, July, 1980.
Hagopian, John H., and Bastress, E. Karl. Recommended Industrial Ventilation
Guidelines. CDC-99-74-33, U.S. Department of Health, Education, and Welfare 1
Public Health Service, Center for Disease Control, National Institute for ,
Occupational Safety and Health, Cincinnati, Ohio, 1976. 1-329 pp. i
i •:
Hampl, Vladimir. Evaluation of Industrial Local Exhaust Hood Efficiency by a:
Tracer Gas Technique. American Industrial Hygiene Association Journal,
45(7):485-490, 1984. !
: 1
Hampl, Vladimir, Ph.D. Industrial Local Exhaust Efficiency Evaluation by a ;
Tracer Gas Technique. Presented at the: American Industrial Hygiene
Conference in Cincinnati, Ohio, 1982.
i
Herget, William F. and James D. Brasher. Remote Fourier Transform Infrared
Air Pollution Studies. Optical Engineering, 19(4):508-514, 1980.
Herget, William F. and James D. Brasherj. Remote Measurement of Gaseous
Pollutant Concentrations Using a Mobil Fourier Transform Interferometer
System. Applied Optics, 18(20): 3404-34'20, October 1979.
| i
Kenson, Robert E. and John A. Ripp. Task Report on Hood Capture Efficiency
Preliminary Investigation. EPA-68-02-2133, Process Measurments Branch,
IERL-RTP, EPA, 1978. i
I j ;
Lamb, Brian K. Atmospheric Tracer Techniques and Gas Transport in the j
Primary Aluminum Industry. Air Pollution Research Section, Washington State :
University.
1 1 1
National Institute for Occupational Safety and Health. Industrial
yentilation (588). U.S. Department of Health, Education, and Welfare, Public!
Health Service, Center for Disease Control, Cincinnati, Ohio. February 1980. j
i '
Radian Corporation. Determination of Capture and Destruction Efficiencies of j
Selected Volatile Organic Compound Control Devices in the State of Illinois. !
EPA-905-2-80-005, U.S. Environmental Protection Agency, Region V, Air and
Hazardous Materials Division, Air Programs Branch, Chicago, Illinois, 1981.
1—78 pp.
! < !
Sandberg, Mats. What is Ventilation Efficiency. Building and Environment,
16(2):123-134, 1981.
100
-------
She'e't" Meta'l and'Air Conditioning ContractorsNati'o'n'al Association," Inc.
Industrial Ventilation, A Manual of Recommended Practice, 16th Edition.
American Conference of Governmental Industrial Hygienists.
TRC Environmental Consultants, Inc. Protocol for the Measurement of
Inhalable Particulate Fugitive Emissions from Stationary Industrial Sources.
EPA-68-02-3115, Process Measurements Branch Industrial Environmental
Protection Agency, RTP, North Carolina, 1980.
I
Wilson, Ayjay. Methods for Attaining V,0C Compliance. Pollution Engineering,
April 1983. :
i
i
i
i l
i
i
1Q1
-------
( APPEND IX'"B
| i t
! PREVIOUS LIQUID/GAS-PHASE TESTS REVIEWED
I : !
i ... ; . ... 1
| Data from previous liquid/gas-phase testing are summarized in this '
,appendix. Using the liquid/gas material balance technique, the amount of
[liquid solvent used at the coater is compared to the measured amount of VOC
'the capture stream and the measured solvent removal of the control device. The
jliquid/gas-phase testing approach incorporates test methods used in both the
liquid- and the gas-phase approaches. As discussed below, comparisons of
liquid- and gas-phase measurements to complete a solvent mass balance around
the system have met with varied degrees of success.
B.l PLANT 7
Plant 7 is a publication rotogravure facility using toluene as the solve;nt
in the printing process. Two of the processes are equipped with unique VOC \
capture systems consisting of cabin-type structures enclosing the upper third|
of the presses. Air is drawn from thej pressroom and up through the enclosure's.
Solvent vapors are captured by the induced air flow and are ducted to a 49,000
scfm carbon adsorption system for recovery. j
! |
The capture/control system at Plant 7 was performance tested in December,',
'1978 (B-l, B-2). The following measurements were made: 1) Solvent input to ;
|the system was determined by monitoring plant ink and solvent flow meters, by|
.'analyzing grab samples of ink for solvent content, and by estimating solvent j
jcontribution of the system make-up air stream; 2) Captured solvent was j
'determined by monitoring solvent concentrations (and air flow rates) in the j
jcarbon adsorber inlet duct on a semi-continuous basis using GC and FID; 3)
.Recovered solvent quantity was determined from plant flow instrumentation; 4):
iCarbon adsorber stack solvent losses were determined on a semi-continuous basis
jusing GC and FID; and 5) Solvent losses in the wastewater were calculated by |
^determining wastewater flow rate and solvent concentration. Many of the j
'measurements were made on a grab (non-integrated) basis, but details of testing
'period times are not reported. i j
i
l ' '
: The average overall efficiency as•determined by comparison of recovered !
solvent and inlet solvent rates (liquid-phase data) is reported as 92.4 percent
(range of 88-97%). The liquid-phase measurement methods are straightforward,'
reasonably integrated, and capable of yielding fairly accurate apparent overall
efficiency data. In contrast, the overall recovery efficiency by the
1iquid/gas-phase method ranged from 62 to 87 percent (capture efficiency
63-89%, and control efficiency 98-99%). The data representing the capture
efficiency by liquid/gas-phase comparison are obviously inaccurate. The
liquid/gas comparisons around the coater and capture system and around the
recovery device both indicate inaccuracies. These data suggest that up to 30'
percent of the solvent entering the press was lost but that the carbon adsorber
simultaneously recovered over 20 percent more solvent than the amount that
entered it. Considering this contradiction, a material balance could not be
closed—around-the- ent-ire -sys-tem-based- on—gas—and- -liquid-phase measurements .
-1-02-
-------
J* SiTpp 1 ernerita"!~measurenie"rits ,_ mad e~~i~rf~"Ja-n'uaT'y] 1'980"™indicafed "t"fra~t~~~amb"~ie~ri £".
gas-phase solvent contributions to the'press may have been greater than !
realized during the 1978 testing. These greater inlet gas-phase solvent j
concentrations, resulting from infiltration from other areas in the plant, may
have inflated the overall efficiency results by about 3 percent.
I I
Errors in capture, control, and overall efficiency may be due to several
jfactors, including variable air flow rates and solvent concentrations, the |
.failure to identify and characterize all VOC flows into and out of the system',
'and errors introduced by the gas- and liquid-phase measurement and analytical
methods, including noncontinuous analysis.
[ i
¦B.2 PLANT 8
I .1 1
I Plant 8 is also a publication rotogravure facility using carbon adsorption
jfor solvent recovery. The 75,000 scfmjsystem recovers a solvent formulation of
ilactol spirits, toluene, and xylene from the VOC stream captured from three [
.printing presses.
The capture/control system at Plant 8 was performance tested in April, j
1974 (B-l). Flow and VOC measurement/analysis methods used were similar to I
those used at Plant 7, except that EPA-Method 25 and 25A testing were performed
jin addition to GC/FID analysis. The VOC sampling was conducted in a |
isenii-continuous manner. ;
' ; !
| Based on comparisons of liquid solvent recovered and liquid solvent j
jcharged to the presses, overall efficiency for the test period averaged 84.7 '
'percent. As with the Plant 7 data, however, the Plant 8 capture efficiencies!
based on liquid/gas comparisons are obyiously in error. Based on the
;comparison of inlet liquid solvent flows and gas-phase solvent flow to the
adsorber, the average capture efficiency was determined to be 59, 58, and 122•
percent, respectively, for the GC/FID,,Method 25A, and Method 25 analyses.
Control efficiency, as determined by the adsorber inlet gas-phase VOC mass rate
and the recovered liquid solvent rate,;was measured to exceed 100 percent.
Control efficiency, based on gas-phase«analysis of inlet and outlet adsorber |
streams, was determined to be 94, 94, and 97 percent, respectively, for the
GC/FID, Method 25A, and Method 25 analyses. Considering these contradictions',
a material balance could not be closedjaround the entire system. j
Errors introduced into the gas/1iquid-phase comparisons are apparently due
;to largely unidentified inconsistencies, in the sampling/analysis methods. 1
•Among the inconsistencies may be variable air flows and VOC concentrations and
!the failure to identify and properly characterize the solvent and solvent flows
'into or out of the system. The Method,25 results were generally very
dissimilar to the GC/FID and Method 25A results.
B .3 PLANT 10 •
* Plant 10 operates two paper coating lines which are ducted to a single
10,000 scfm carbon adsorption system. Solvents used in the coating process
inc lude toluenee thy!"acetate; ~and' isopr opanol"; ~ -Plant~l-0—opera tes" with" the *
¦ 103
-------
|C6ater"r"oo"in-a't—a-sTi"gh't^n'eg"a"t i"ve—pre's'sure~wi"th~"res'pe"c't—to-the-ou'ts"i"d'e—amb'rent.
pressure. The overall effect is for all of the drying oven makeup air to come
from the coater room and be pulled through the oven. !
| i
The plant routinely maintains records of solvent usage and solvent |
recovery, using measuremnt methods of unknown type and accuracy. During a I
jl2-month period ending in November, 1980, overall VOC reduction efficiency
averaged about 92 percent. The carbonjin the adsorber beds had been in service
'about 5-6 years during this monitoring«period. j
| !
Plant 10 was performance tested in October, 1980 (B-3). The liquid
solvent rate to the coating lines was monitored in addition to carbon
adsorption system exhaust VOC rate (Method 25) for a six-hour period.
j
The overall efficiency was assumed to be 92 percent, based on the
liquid-phase data. Multiplying the overall efficiency level by the measured
iinlet liquid solvent mass yields the mass of recovered solvent. Totaling the
[recovered mass and the VOC mass measured in the carbon adsorber exhaust j
provides an approximation of the total VOC mass in the capture stream leadingj
Ito the carbon adsorber. Utilizing the!data from the inlet liquid solvent, the
'approximated gaseous VOC in the capture stream, and the gaseous VOC in the j
carbon adsorber exhaust, capture and control efficiencies of 94 and 98 percent,
respectively, were determined. 1
No data were available for comparing results by different methods. !
t
B.4 PLANT 20 j
I Capture efficiency for the print line at Plant 20, a flexible vinyl wall
Jcovering manufacturer was determined by gas- and liquid-phase measurements in
^September - October, 1980. The determination consisted of measuring the mass
iof solvent applied at the print line, the mass of gaseous VOC emissions ducted
jto the carbon adsorption system, the mass of gaseous VOC fugitive emissions in
print room exhaust streams, and the amount of VOC retained in the coated
^product. VOC measurements were made oh a continuous basis using EPA Method 25A
at print room exhaust areas and the carbon adsorber inlet duct. Ink usage was
determined by measuring changes in the,level of the ink tanks supplying the
print line* Ink and vinyl product samples were analyzed for solvent content.i
Calculated capture efficiency based on1 liquid/gas-phase data averaged 60.5 |
percent for testing conducted over 50 hours in nine days. The range of |
calculated capture efficiencies was from 47 to 70 percent, excluding one
labnormal measurement. The efficiency results are reported to be inconclusive
due to failure to close the solvent material balance, as discussed below. ,
A room air supply fan caused excessive turbulence around the print
stations, affecting and lowering capture efficiency. Also the net flow of
solvent to the printing process was difficult to accurately characterize. i
Finally, the mass flow of solvent through open doorways was high, due to the
.room air supply fan, and difficult to accurately quantify. Measurement of
flows through doorways and areas other, than ducting, through the use of
hand -hel-d- anemometers*;—may -be -prone--to—greater-inaccuracies-,- and EPA reference
methods do not apply.
-------
-------
; Source testing was performed at Plant 25, a metal coil coating plant, inl
August - September, 1979. Most of the metal coated by Plant 25 is used in the
building industry for siding, roofing,
gutters, awnings, and other items
Types of coatings applied include epox^ resins and primers, polyesters, j
acrylics, fluorocarbons, siliconized acrylics and polyesters, and polyvinyl j
'chlorides. Solvents used in the coating process are aromatic hydrocarbons anii
glycol ethers. The coating line tested (one of two) is equipped with thermal
incinerators for control of solvent VOG emissions.
The following VOC analyses were performed:
+ total hydrocarbon (THC) sampling at the incinerator outlets,
• EPA Method 25 sampling at the incinerator inlets and outlets,
• EPA Method 25 and THC sampling at a quench zone vent, and
• EPA Method 24 analysis of coatings for solvent content.
Data representing solvent usage rate (EPA Method 24) and incinerator inlet VOC
rate (EPA Method 25) indicate that process capture efficiency averaged {
approximately 71 percent for the test period. Corresponding control
Efficiency, based on incinerator inlet|and outlet VOC rates (Method 25), |
averaged 79 percent, for an overall VOG reduction efficiency of about 56 '
percent. However, serious inconsistencies are reported for the EPA Method 25l
and THC analyses, particularly at low VOC concentration levels. The resulting
efficiency values, based on EPA Method!25 analyses, are therefore subject to \
'error (material balance data based on THC analyses are not available). Also j
contributing to possible measurement error are the relatively short test ;
periods; the above efficiency values are based on the results of 13 nominally|
60-minute test periods conducted over six workdays.
B.8 PLANT 26 I
i !
\ The VOC system at Plant 26, a paper coating facility, was performance j
tested for a several-hour period in October, 1980 (B-3)« The tested facility
'consisted of a single coating line controlled by an 18,000 scfm thermal
incinerator operated at 1250°F. Gas-phase measurements at the incinerator
jinlet and outlet were made using EPA Method 25. Liquid-phase measurements of
solvent charged to the coating applicator were made by monitoring the inlet
solvent (carbon) mass rate. Capture efficiency was measured to range from 94
to 98 percent; incinerator control efficiency averaged 95-96 percent.
Therefore, the overall efficiency averaged 91 to 93 percent.
fe.9 PLANT 27
j
1
VOC testing was performed at Plant 27, another paper coating facility, in
•October, 1980 (B-3). The plant operates two coating lines which are controlled
by a single 15,000 scfm carbon adsorption system. The sequencing of the 3-bed
carbon adsorption system, which recovers toluene solvent, is controlled by
exhaust stack hydrocarbon analyzers. The only streams measured were the liquid
solvent "charged to"the"process-and the "gaseous"VOC~"emi"Ssions~-rn~the" carbon
106.
-------
-adsorption—exhaust—air—("Method—25")";—Control*—ef f i-c i-ency-was—es t ima ted -as—79—,
'percent, based on these two flow determinations. This value was calculated j
assuming a capture efficiency of 100 percent, which appeared reasonable due to
the efficient design of the VOC capture system. There was no obvious |
explanation for the low carbon adsorber control efficiency, but the observation
was consistent with above-average solvent purchases by the plant. \
i
>5 • 10 PLANT 28
Plant 28 is a can coating facility which applies a single enamel coat to
sheets of tin plate that are used to make three piece cans. After the coating
process, the sheets are oven-dried. The oven exhaust is ducted to a 8,000 sc'fm
thermal incinerator operating at 2200°F. The sheets are then cooled by ambient
air, which is exhausted to the atmosphere.
The plant was tested in October, 1980, using Method 25 for gas-phase VOC
analysis and measurement of inlet liquid solvent mass flow to the coating
process (B-3). Gas-phase sampling wasjperformed at the incinerator inlet,
incinerator outlet, and cooling zone discharge ducts.
When the amount of solvent ducted to the incinerator inlet plus that in
Ithe cooling zone gas streams was compared with the amount of solvent applied at
'the coating applicator, the VOC capture efficiency was calculated as 221 j
.'percent. The inconsistency was initially thought to be due to plant data on j
the applied coating. However, all of these numbers were verified by Plant 28|
as correct. No explanation for the poor material balance was determined. From
•the gas-phase analyses, the incinerator VOC control efficiency was calculated^
as 73 percent. |
I |
B.11 PLANT 29 ] >
; j
Plant 29, another beverage can coating facility, was performance tested in
October, 1980 (B-3). The plant has two coating lines with separate capture and
.control systems. 1 j
The first coating line had a drying oven, controlled by a 4,000 scfm j
thermal incinerator operating at 550°F, and a cooling zone, which was vented j
directly to the atmosphere. Gas-phase VOC analysis using EPA Method 25 and :
'flow measurements were made at the incinerator inlet, incinerator outlet, and'
jcooling zone exhaust. Solvent feed rates to the coating line were also
monitored.
I
| Capture efficiency was calculated, to be greater than 100 percent and
^incinerator control efficiency was determined to be only 26 percent. The
;ineonsistencies inherent in the capture efficiency value are attributed to
questionable gas flow rate data and to:frequent process startups and shutdowns.
The low incinerator control efficiency may be partially attributed to
•inaccurate gas-phase flow measurements,' but the low incinerator temperature may
.also contribute to relatively poor VOC destruction.
.107
-------
i Tbe~sec*o nd~~coa ting—li-ne—cons*is-ted~~of~~a—dry irig—oven~contro"l~led—by-a~7T000
scfm catalytic incinerator operating at 550°F, and a cooling zone, which was :
jvented to the atmosphere. Solvent mass rates were determined using methods
^similar to those used for testing of the first coating line. ¦
! The material balance for the second line demonstrated a 74 percent VOC I
jcapture efficiency and the potential for 90 percent capture efficiency if thel
jcooling zone air was directed to the incinerator. The VOC control efficiency-
jof the catalytic incinerator was determined to be a relatively low 49 percent',
Iwhich may be due to the low-temperature operation. Overall efficiency is '
jcalculated as 36 percent from the capture and control efficiencies with the
Ipotential for 44 percent. j
|B . 1 2 PLANT 30
| A base coating operation at Plant 30, a three-piece can manufacturing
plant, was tested for VOC material balance quantification in October, 1979
(B—8)• The base coating operation consists of the application and drying of a
base coat on metal sheets. The coatedjsheets are dried at 400°F in the heating
zones of an oven. The sheets then enter cooling zones within the oven prior to
further processing. Oven exhaust air is vented to the atmosphere. !
i ;
Six nominally 30-minute material balance tests were performed. For each:
test, solvent usage, dry coating application rate, and VOC emission rate were!
quantified. Solvent usage was determined based on coating rate and the j
manufacturer's specification for solvent content. Dry coating application rate
was determined by before-coating and after-coating weighings of the metal |
sheets. VOC emissions from the heating and cooling zone exhausts were «
quantified using EPA Method 2 determinations of flow rate and FID and EPA !
Method 25 analyses for hydrocarbon concentration. Based on liquid solvent
usage rate and the method 25-determined oven emission rate, capture efficiency
for the system was estimated to range from 69 to 75 percent for five of the six
tests (one of the tests indicated capture efficiency in excess of 100 percent).
FID and Method 25 VOC emission data agreed within an average of 7 percent for;
the six tests.
The capture efficiency determinations at Plant 30 were based on a
comparison of the solvent application rate and the solvent emission rate
(captured emissions). The application rate was based on metered coating
application and manufacturers figures for the solvent/coating ratio, and the
solvent emission rate was based on EPA'Method 25 and gas flow measurements. .To
develop a comparative basis between the solvent applied (lb. solvent) and
measured solvent emitted (lb. carbon) both were converted to propane
equivalents. This conversion was made since the solvent content was unknown
and it was assumed that the carbon/hydrogen ratio of the solvent was likely
similar to propane. The comparison based on this assumption will be somewhat;
in error if the solvents are not similar to propane (e.g. oxygenated
compounds).
.108.
-------
B—1-3—PL-ANT-3-1 : ——
; (
An interior spray coating operation at Plant 31, a two-piece aluminum can
manufacturing plant, was tested in October, 1979, in a manner similar to that;
[used for Plant 30 (B—9)~ After base and top coats are applied to the exterior
'surfaces of the aluminum, the cans receive 80 weight percent solvent, 20 weight
percent solids interior spray coats. The cans are then cured in a gas-fired
Joven. Emissions from the base coat oven, the top coat printer and oven, and
the interior spray coater and oven arejducted to a thermal incinerator
'operating at 1150°F. The referenced source test focused on VOC emissions from
jthe interior spray coating operation, which includes a can transfer conveyor j
'and the gas-fired oven
Test results for four of the six nominally 30-minute tests indicated
capture efficiencies ranging from 58 to 97 percent (based on solvent usage rate
'and EPA Method 25-determined VOC emission rate). For the remaining two tests',
jcapture efficiency measured in excess of 100 percent, for unidentified reasons.
'EPA method 25 and FID results did not correlate closely on the average. The •
[inconsistencies were attributed to organic condensation in the FID integrated'
bag samples, resulting in artificially!lowered gas-phase concentrations.
! 1 I
I Inaccuracies in the tests at Plant 31 may be somewhat attributable to Che
ipotential error described in the last paragraph on Plant 30. The same j
'assumptions were followed not knowing the solvent content. 1
i
109.
-------
APPENDIX B REFERENCES"
B-l Publication Rotogravure Printing - Background Information for Proposed
Standards, EPA-450/3-80-031A, OAQPS, U.S. Environmental Protection Agency,
Appendix C, May, 1980.
B-2 Graphic Arts Emission Test Report, Meredith/Burda, Inc., Lynchburgh, |
Virginia, Test 2, EMB Report No. 79-GRA-1A. OAQPS, U.S. Environmental
Protection Agency, February, 1980.! j
l !
B-3 Determination of Capture and Destruction Efficiencies of Selected Volatile
Organic Compound Control Devices in the State of Illinois, EPA i
905/2-80-005, U.S. Environmental Protection Agency, April, 1980. !
i :
! . . I
B-4 Background Information Document for the Pressure Sensitive Tape and Label'
Surface Coating Industry, EPA-450/3-80-003A, OAQPS, U.S. Environmental !
Protection Agency, Appendix C, May;, 1980. <
B-5 Draft Pressure Sensitive Tapes and' Labels Emission Test Report, OAQPS, ]
U.S. Environmental Protection Agency, June, 1979. j
B-6 Metal Coil Surface Coating - Background Information for Proposed
Standards, EPA-450/3-80-0359, OAQPS, U.S. Environmental Protection Agency,
October, 1980.
i
B-7 Industrial Surface Coating (Coil) Emission Test Report, Precoat Metals j
Incorporated, St. Louis, Missouri, | EMB Report 79-ISC-10, OAQPS, U.S.
Environmental Protection Agency, October, 1980. !
j ¦ j
B-8 Industrial Surface Coating Cans Emission Test Report, American Can, Forest
! Park, Georgia, EMB Report 79-ISC-7, OAQPS, U.S. Environmental Protection1
Agency, August, 1979.
B-9 Industrial Surface Coating (Can Coating) Emission Test Report, Metal
j Container Corporation, Jacksonville, Florida, EMB Report 79-ISC-8, OAQPS,;
i U.S. Environmental Protection Agency, December, 1979. i
110
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA/600/2-35/011
3. RECIPIENT S ACCESS.!QN-^O/vv?
5 17 >2 4 J fm
A. TITLE AND SUBTITLE
MEASUREMENT OF VOLATILE ORGANIC COMPOUND
CAPTURE EFFICIENCY
5 REPORT OATE
February 1985
6 PERFORMING ORGANIZATION CODE
7. AUTNOR(S)
D. B. Hunt and J. L. Randall
8. PERFORMING ORGANIZATION REPORT NO
9 PERFORMING ORGANIZATION NAME AND ADDRESS
Radian Corporation
8501 Mo-Pac Blvd.
Austin, TX 78766
10. PROGRAM ELEMENT NO.
C9HA1A
11. CONTRACT/GRANT NO
68-03-3038, Task 10
12. SPONSORING AGENCY NAME AND ADDRESS
U. S. Environmental Protection Agency
Hazardous Waste Engineering Laboratory
Cincinnati, OH 45268
13. TYPE OF REPORT AND PERIOD COVE RE O
Final (2/82-8/84)
14. SPONSORING AGENCY CODE
EPA/600/10
15. SUPPLEMENTARY NOTES
16. ABSTRACT
,This report reviews the feasibility considerations regarding each of several potential
alternate approaches for determining capture efficiency and experimental testing of
one approach, the liquid/gas-phase material balance. Two phases of experimental test-
ing were conducted: laboratory and field investigations. The measured mean capture
efficiencies for the the field test were higher than expected (106.7% with a coeffi-
cient of variation 7.4%). At this time, there is no material balance methodology whicV
is superior to the gas-phase approach using a total enclosure around the regulated
coating operation.,
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b. IDENTIFIERS/OPEN ENDED TERMS
cosati Field'Group
V0C Capture Efficiency
Volatile Organic
Compounds
Capture Efficiency
Material Balances
Surface Coating
138
19 SECURITY Class .This Report/
UNCLASSIFIED
18. DISTRIBUTION STATEMENT
RELEASE
21 NO. OP PAGES
119
20. SECURITY CLASS iTHis page,
UNCLASSIFIED
22 PRiCE
EPA Form 2220.1 (IUv. 4.77)
PREVIOUS EDITION 11 OBSOLETE
------- |