United States
Environmental Protection
Agency
-------
-------
EPA/600/R-08/012 January 2008 www.epa.gov/ord
FINAL REPORT ON
Evaluation of In-Room Air Cleaners
for Building Protection
Contract No. GS-10F-0275K
Task Order 1105
Prepared for
Joseph Wood and Les Sparks, Project Officers
U.S. ENVIRONMENTAL PROTECTION AGENCY
Research Triangle Park, NC
January 2008
Prepared by
Vladimir Kogan (614)424-7970
Chris Harto (614) 424-3025
David J. Hesse (614) 424-5610
Kent C. Hofacre (614) 424-5639
BATTELLE COLUMBUS OPERATIONS
505 King Avenue
Columbus, Ohio 43201-269
Office of Research and Development
National Homeland Security Research Center, Decontamination and Consequence Management Division
-------
Disclaimer
The U.S. Environmental Protection Agency through its Office of Research and Development funded this research.
It has been subject to an administrative review but does not necessarily reflect the views of the Agency. No official
endorsement should be inferred. EPA does not endorse the purchase or sale of any commercial products or services.
-------
Table of Contents
EXECUTIVE SUMMARY viii
1.0 INTRODUCTION 1
2.0 AIR CLEANER SELECTION 1
3.0 EXPERIMENTAL STUDY 3
3.1 Single-Pass Testing 3
3.1.1 lest Setup 3
3.1.2 lest Procedure 8
3.1.3 Data Analysis 9
3.1.4 Test Results and Discussion 10
3.2 In-Room Testing of Air Cleaners 16
3.2.1 Test Setup 16
3.2.2 Test Procedure 18
3.2.3 Data Analysis 19
3.2.4 Results and Discussion 19
4.0 MODELING STUDY 23
4.1 FLUENT CFD Modeling 23
4.1.1 Results 23
4.2 Perfectly-Mixed Zone Analysis 25
5.0 CONCLUSIONS AND RECOMMENDATIONS 27
6.0 REFERENCES 29
APPENDIX A: AIR CLEANER SELECTION A-l
APPENDIX B: CFD MODELING, DETAILED METHODOLOGY, AND RESULTS B-l
APPENDIX C: QUALITY ASSURANCE C-l
-------
List of Tables
Table 1. CADR Characteristics of Selected Air Cleaners 1
Table 2. Air Cleaner Specifications 2
Table 3. Test Matrix for Single-Pass Evaluation of Air Cleaners 3
Table 4. SMPS Size Channels 5
Tables. Velocity Characteristics of the ESP Air Cleaner 10
Table 6. ESP Air Cleaner Flow Rates 11
Table 7. Flow Characteristics of the HEPAFilter 11
Table 8. In-Room Test Matrix 16
Table 9. Calculated Decay Constants 21
Table 10. Collected Filter Masses 22
List of Figures
Figure 1. Electrostatic Precipitator and HEPA Air Cleaner 2
Figure 2. Schematic of the Single-Pass Efficiency Test Apparatus 3
Figure 3. Airflow Measurements for ESP Air Cleaner 4
Figure 4. Aerosol Sampling Instruments, TSI SMPS and Climet CI-500 5
Figure 5. Representative Diameter Distributions of Test Aerosols Obtained Using TSI SMPS and Climet CI-500 7
Figure 6. Plots of Bias Correction Factor (R) and Sampling Variability (V) 9
Figure 7. Local Velocities of the ESP Air Cleaner 10
Figures. Local Velocities of the HEPA Filter 11
Figure 9. Filtration Efficiency for the ESP Air Cleaner 12
Figure 10. Filtration Efficiency for the HEPA Filter 12
Figure 11. ESP Bioaerosol Filtration Efficiency 14
Figure 12. HEPA Bioaerosol Filtration Efficiency 14
Figure 13. Downstream Particle Counts From ESP Bioaerosol Test, Run 2 15
Figure 14. Upstream Particle Counts From ESP Bioaerosol Test, Run 2 15
Figure 15. Test Chamber for In-Room Experiments 16
Figure 16. In-Room Test Configuration A 17
Figure 17. In-Room Test Configuration B 17
Figure 18. In-Room Test Configuration C 18
Figure 19. Concentration vs. Time Plots, Individual Configurations 20
Figure 20. Concentration vs. Time Plot, Averaged Data 20
Figure 21. Size-Resolved Concentration vs. Time 21
Figure 22. Collected Filter Masses 22
Figure 23. CFD Model Results 24
Figure 24. Filter Masses From CFD Calculations 24
Figure 25. Well-Mixed Zone Model Results 25
-------
List of Acronyms
AHAM Association of Home Appliance Manufacturers
Bg Bacillus globigii
CADR Clean Air Delivery Rate
CFD computational fluid dynamics
CV coefficient of variation (computed as the standard deviation divided by the mean)
ESP electrostatic precipitator
HEPA high-efficiency paniculate air
HVAC heating, ventilation, and air conditioning
KC1 potassium chloride
LES large eddy simulation
PM paniculate matter
QAPP Quality Assurance Project Plan
SMPS scanning mobility particle sizer
-------
Executive Summary
This report describes the experiments and modeling
conducted to determine the effectiveness of commercially
available in-room air cleaners (i.e., paniculate matter [PM]
filtration devices) in mitigating the impact of an aerosolized
biological threat agent attack on a building. In one set
of experiments, two air cleaners were evaluated for their
single-pass filtration efficiency as a function of airflow rate,
particle diameter (ranging from 0.03 jjum to 10 jjum), and
type of particle (an inert aerosol and a bioaerosol). One of
the air cleaners tested was a high-efficiency paniculate air
(HEPA) filtration device, and the other was an electrostatic
precipitator (ESP) air cleaner. Following the single-pass
experiments, the HEPA filter was further evaluated to verify
its effectiveness in reducing in-room PM levels.
For the single-pass testing, the ESP air cleaner displayed a
pronounced minimum in filtration efficiency for particles
~ 0.2 (jum diameter, consistent with the principles of
electrostatic precipitation. Also, the single-pass efficiency of
the ESP air cleaner was found to decrease with increasing
flow rate through the unit, most likely due to the decreasing
residence time of the particles in the charging and deposition
zones of the collector.
For the HEPA filter, no noticeable effect of flow rate on
the filtration efficiency of the unit was observed, but an
unexpected drop in efficiency was observed for particles
below 0.3 (Jim in diameter. This observation could be
explained by the probability that some leaks developed
around the filter due to its relatively loose fit in the single-
pass test apparatus.
Both air cleaners' filtration efficiencies for particles with
diameters smaller than approximately 0.04 jjum were lower
than expected. No difference in the air cleaners' filtration
efficiencies was observed for the biological and inert aerosols
having similar particle diameters.
For the in-room experiments, the HEPA filter was
evaluated in a test chamber under four configurations.
In these experiments, the PM concentration decay rate
was determined by continuously monitoring the aerosol
concentration at a particular location in the room, using a
real-time particle counter. Also, average PM levels occurring
in the room were determined using filter samples taken at five
different locations.
The effectiveness of an in-room air cleaner under typical
operating settings depends on three principal characteristics:
1) the single-pass filtration efficiency, 2) the airflow rate
through the filter, and 3) the airflow pattern that the cleaner
induces in the room. While the first two characteristics can be
determined from some straightforward measurements, such
as those used in this study, the airflow pattern in the room
is also dependent upon other factors, such as room size and
shape; heating, ventilation, and air conditioning (HVAC)
characteristics; furnishings; leak patterns; and the presence
of mixing fans. In this research, some of these factors were
investigated using the HEPA filter.
The HEPA filter tested was found to provide significant
protection with respect to the contaminant concentration
profile in the test room, when compared to the case when the
air cleaner was not operating. This observation was expected
and illustrates the usefulness of in-room air cleaners to reduce
ambient PM levels following an attack with an aerosolized
threat agent. The location of the air cleaner relative to the
aerosol source was found to have a minimal impact. The
addition of an office desk and a chair in the test chamber also
did not appear to noticeably alter the performance of the air
cleaner. Overall, the HEPA filter provided reasonable mixing
conditions in the test room, although some variability in the
PM levels measured at different locations within the chamber
was observed.
Following the completion of the experimental phase of
the project, model calculations were performed using
computational fluid dynamics (CFD) for one of the specific
in-room test configurations investigated in this work. In
addition, non-CFD calculations were performed for the
test conditions, using the perfectly-mixed zone modeling
approach.
CFD model simulations can offer a viable alternative to field
tests because of the demonstrated ability of this technique to
predict the general trends of contaminant behavior in various
indoor settings. The CFD model did a reasonable job of
modeling the variability of the concentration of PM within
the chamber, although it overestimated the concentration
decay rate at the specified location. The main issue associated
with this application of CFD is the broad spectrum of flow
regimes evolving within the room, ranging from laminar
to fully turbulent conditions. This requires specification
of different turbulence closure models, such as large eddy
simulations (LES) for describing large flow recirculation
patterns. The model also requires more refined schemes
to adequately describe the dispersion of PM in the aerosol
generation and air cleaner exhaust zones.
-------
1.0
Introduction
This report describes experimental and modeling efforts to
determine the effectiveness of commercially available in-
room air cleaners in mitigating the impact of an aerosolized
biological agent attack on a building. The project consisted
of four principal phases. In the first phase, two representative
air cleaners were selected for experimental evaluation, one
based on the high- efficiency paniculate air (HEPA) filtration
mechanism and one using electrostatic precipitation (ESP). In
the second phase of the project, the selected air cleaners were
evaluated for their single-pass filtration efficiency for both an
inert aerosol and a bioaerosol. In the third phase, one of the
air cleaners was selected for evaluation of its effectiveness
in reducing PM levels in a test chamber designed to simulate
"in-room" conditions. Finally, in the fourth phase, both a
computational fluid dynamics (CFD) model and calculations
based on a "well-mixed volume" model were used for
theoretically evaluating the effectiveness of in-room air
cleaners. This experimental portion of this project was
performed according to a test/quality assurance project plan
(QAPP; see Appendix C).
2.0
Air Cleaner Selection
The first step in the overall project was to select the air
cleaners for testing. A list of candidate air cleaners was
obtained by performing a brief market survey through the
Internet. Based on the Internet search, the manufacturers
of the candidate air cleaners were contacted to collect the
detailed performance and specification data on their products.
Two air cleaners were then selected, based on the following
selection criteria:
Operating principle: one HEPA filter type air cleaner
and one ESP type air cleaner
Off-the-shelf commercial availability of a high-quality
brand
High aerosol collection efficiency, as specified by the
manufacturer, preferably higher than 99% efficiency for
0.3 (jum aerosol particles
Airflow rate capacity on the order of 300 fWmin (0.142
mVs), or 7 to 10 air exchanges per hour for a single-
room of 8 ft x 16 ft x 8 ft (2.4 m x 4.9 m x 2.4 m)
the size of the test facility available at Battelle
Having at least three-stage adjustable airflow capability
Most room air cleaners on the market are certified under the
Room Air Cleaner Certification Program, which is sponsored
by the Association of Home Appliance Manufacturers
(AHAM). Under this program, room air cleaners are
characterized using the Clean Air Delivery Rate (CADR),
which determines how effectively they remove different
paniculate pollutants such as tobacco smoke, dust, and
pollen. One hundred seventy-four different models of room
air cleaners from 18 manufacturers are currently certified
under the program. Among them, only one was an ESP-type
air cleaner, so it was selected for testing under this program.
Another air cleaner, the HEPA filter-type air cleaner, was
selected for testing because it had a similar CADR rating to
the ESP. The CADR values for the two air cleaners are shown
in Table 1; additional specification parameters are listed in
Table 2. It should be noted that these two air cleaners are also
the top two room air cleaners recommended by Consumer
Reports. Pictures of the two selected air cleaners are shown
in Figure 1.
Table 1. CADR Characteristics of Selected Air Cleaners
HEPA
320 (.151)
330 (.156)
330 (.156)
-------
Table 2. Air Cleaners Specifications
Type
Fan Flow Rate
Dimensions
Initial Cost
Operating Costs
(energy use)
Maintenance Costs
ESP
Low 225 cfm (0. 106 mVs)
Medium 275 cfm (0.130 mVs)
High 365 cfm (0.172 mVs)
(flows specified by manufacturers)
0.48 m H x 0.38 m L x 0.55 m W
$490
$62/yr (based on 90W power consumption,
assuming 24-hr operation on high flow at a
costof7.770/kwhr)
None specified
HEPA
Low 275 cfm (0. 130 mVs)
Medium 340 cfm (0.160 mVs)
High 440 cfm (0.208 mVs)
(flows measured)
0.56 m H x 0.46 m L x 0.28 m W
$260
Energy consumption not specified
$80/yr for replacement filter
Figure 1. Electrostatic Precipitator (left) and HEPA Filter (right)
The selected air cleaners were tested as received, without
conditioning. It is realized that performance of air cleaners
containing an ESP element will degrade with accumulation of
certain aerosols, such as oil aerosols, silicon oxide particles,
etc. However, investigating the effect of potential degradation
of the performance of air cleaners due to aerosol loading was
beyond the scope of this study. More detail on the selection
of the air cleaners, the AHAM standard, and the definition of
CADR is provided in Appendix A.
-------
3.0
Experimental Study
3.1 Single-Pass Testing
The purpose of the single-pass testing was to
characterize the filtration efficiencies of the two
selected air cleaners for a single pass of aerosol
through the unit. The goal was to characterize
the efficiency of the units over a range of flow
rates and across an aerosol diameter range from
0.03 to 10 (jum. This was achieved by placing
the unit within a sealed flow duct and measuring
the size-resolved aerosol concentration
upstream and downstream of the unit with a
particle counter. The units were characterized
for both an inert potassium chloride (KC1)
aerosol and a bioaerosol consisting of Bacillus
globigii (Bg) spores.
Table 3. Test Matrix for Single-Pass Evaluation of Air Cleaners
Type of Aerosol
Inert Aerosol
(KC1 particles)
Bioaerosol
(Bg spores)
Size
(Diameter)
Range (|jim)
0.03 to 0.5
0.3 to 10
~ 1
Air
Capacity
high
medium
low
high
medium
low
medium
HEPA
(# of tests)
2
2
2
2
2
2
3
ESP
(# of tests)
2
2
2
2
2
2
3
3.1.1 Test Setup
3.1.1.1 Inert Aerosol In order to adequately characterize
the air cleaners over the full range of particle diameters, the
inert aerosol testing was completed in two stages. During
the first stage, the filtration efficiency was measured for
particles ranging from 0.03 to 0.5 jjum. During the second
stage, the filtration efficiency was characterized for particles
ranging from 0.3 to 10 jjum . For each size range, the single-
pass efficiency was determined at three rates of airflow
through the air cleaners (low, medium, and high - see Table
2). Duplicate tests were conducted for each experimental
condition. The complete test matrix, including bioaerosol
tests performed in triplicate, is shown in Table 3.
A test system, conforming to the dimensions of the air
cleaners, was constructed out of Plexiglas to test the single-
pass efficiency of the air cleaners. A schematic of the test
system is illustrated in Figure 2. This system was designed
for testing in-room air cleaners that contain their own air-
handling means, which requires that air pressure be close to
ambient in the vicinity of both the inlet and outlet locations
of the units so that the air cleaner design flow rates are not
affected by the test system. It was also designed to facilitate
testing air cleaners using not only inert aerosols but also
biological aerosols. Custom duct sections were fabricated for
each air cleaner to ensure a tight fit with the flow section of
the test system.
Figure 2. Schematic of the Single-Pass Efficiency Test Apparatus
-------
Before performing the single-pass filtration efficiency testing,
airflow rates were estimated for each of the three settings
for both test air cleaners. This was done using a hot-wire
anemometer to measure air velocity at nine points that were
identified in the middle of nine equal, representative areas
across the airflow grilles. This was done for both the inlet and
outlet locations of both air cleaners. The mean flow velocity
was calculated using the nine velocity values obtained
for both the inlet and outlet flows. The flow rate was then
calculated by multiplying the mean velocity by the flow area.
Figure 3 shows the setup used for determining the flow rate.
Figure 3. Airflow Measurements for ESP Air Cleaner
To ensure that the test rig components did not alter the
natural airflow of the room air cleaner, a boosting blower was
installed upstream of the test section. During a test, the output
of the boosting blower was adjusted so that the air velocity
measured for the test air cleaner was maintained close to that
obtained during the air cleaner flow characterization phase
(± 10%). For the ESP, the velocity was matched upstream
in the middle of the unit. For the HEPA filter, however, the
air velocity was found to be too variable across the inlet to
be accurately represented by a single location. Therefore, its
mean flow velocity was determined by taking the average of
measurements at each of the nine outlet locations used for the
airflow characterization. The blower was adjusted until this
average was ± 10% of the average obtained for the free-
standing unit.
The challenge aerosol was generated using a Baxter Co.
Airlife nebulizer. A 5% KC1 solution was used to generate
the smaller aerosol, and a 20% solution was used to generate
the larger aerosol. Dry, clean air was supplied to the
nebulizer. The airflow to the nebulizer was controlled with a
needle valve, and its pressure and flow rate were monitored.
The nebulizer was connected to a large drying chamber to
allow the droplets to dry before they reached the test duct.
An additional flow of dry air was supplied to the chamber to
assist in drying the particles and to carry them into the test
duct. Before entering the test duct, the particles were passed
through an aerosol neutralizer (containing Kr-85, 10 mCi)
to reduce their charge level. This was necessary as aerosol
particles have a tendency to collect static charge, which may
influence their filtration characteristics.
One of the requirements of this testing was to establish not
only undisturbed, natural airflows through the air cleaners,
but also to provide uniform aerosol concentration across
their inlets. In order to achieve efficient mixing of particles
across the test duct, the aerosol was injected into the duct
against the airflow, as illustrated in Figure 2. In addition,
four small mixing fans and a turbulence enhancement plate
were installed upstream of the air cleaners to improve mixing
efficiency. However, due to the high air velocity in the duct,
combined with the short duct length, an additional dispersion
means was required. An auxiliary aerosol disperser was
fabricated using quarter-inch diameter copper tubing,
which was placed coaxial to the aerosol delivery tube in the
proximity of its injection point. The tube was plugged, and
a number of small holes were drilled radially near the end of
the tube. Pressurized air was passed through the tube and out
through the holes at high velocity. The resulting flow
was designed to entrain the aerosol as it entered the duct
and carry it away from the center. This setup allowed for a
uniform challenge concentration to be achieved at the air
cleaner test location.
Before the tests were conducted, the uniformity of aerosol
concentration was confirmed. To achieve this, aerosol
measurements were performed upstream of the air cleaner,
at a cross-sectional plane perpendicular to the flow. The
cross-section was divided into nine equal representative
areas, and concentration was measured at the center of each
area. The mean concentration and the coefficient of variation
(CV, computed as the standard deviation divided by the
mean) of the nine corresponding grid point concentration
values was then calculated. The maximum acceptable CV
value was set at 15%. If the measured CV exceeded 15%,
the mixing measures were adjusted, and the uniformity was
recharacterized until the requirement of CV less than 15%
was met. This uniformity test was performed for one flow
rate for both aerosol diameter ranges.
These tests were performed before the air cleaner was
installed in the duct, with the airflow generated by the booster
blower only in order to test the effectiveness of mixing in the
duct independent of the individual air cleaners. The apparatus
was designed to minimize its effect on the performance
of the air cleaner, and mixing was designed to take place
in a separate section upstream of the air cleaner location.
Introduction of the air cleaner into the duct was assumed to
have no negative impact on the uniformity in the challenge
concentration. The justification for this assumption was that
if the concentration was uniform upstream of the air cleaner
location then it was thought to be unlikely that the air cleaner
would measurably affect the concentration profile based on
basic mass transport principles.
During a test, the upstream and downstream concentrations
were measured at two fixed locations along the duct
centerline. This was considered adequate for the purposes of
this study because that concentration CV in the cross-sections
was required to be less than 15%. The spatial variability
-------
in flow velocities introduced by the air cleaners was
assumed to have a minimal effect on the results of this study
because a uniform concentration remains uniform across
noncompressible flow regardless of velocity. The aerosol
was sampled from two identical sample probes. These probes
were fabricated from quarter-inch stainless steel tubing bent
90 degrees and tapered at the inlet and were inserted into the
test duct with the inlet facing the flow. The particle counter
was then attached to the appropriate sample probe when a
measurement was to be taken. The aerosol concentration was
measured by using two different instruments, one for each
aerosol diameter range.
The small aerosol was sampled using a TSI scanning
mobility particle sizer (SMPS). This instrument consists of
an electrostatic classifier, which is used to separate particles
by size, and a condensation particle counter, which counts the
particles. The detection range of the instrument is 20 to 107
particles/cm3. In the test configuration, this instrument had a
sample rate of 0.3 L/min and a sample time of two minutes.
The design of the instrument is such that the particles are
counted one size channel at a time, thus each size channel
is sampled for only a fraction of the two-minute sampling
time. The effective range of particle diameters measured
by the instrument was 0.02 to 0.56 microns (only values
from 0.0237 to 0.316 microns were used due to low particle
concentrations of test aerosol). Table 4 shows the breakdown
of the instrument size channels and their upper and lower
limits. All values are in nanometers. The SMPS was
controlled by a computer, and all data were collected using
the TSFs "Aerosol Instrument Manager" software. The data
collected were then transferred to Excel using the "cut and
paste" function for further analysis.
The large aerosol was sampled using a Climet CI-500 laser
particle counter. This unit is designed to detect light scattered
by aerosol particles as they pass through the measuring
volume defined by the width of the instrument's laser beam.
In order to ensure that only one particle passes through
the measuring volume at a time, the CI-500 has an upper
detection limit of up to 107 particles/ft3 (-350 particles/
cm3). This, however, did not introduce an aerosol counting
problem because the instrument samples the aerosol at a
relatively high airflow rate of 2.83 L/min, and its sampling
time was set to one minute. The size range of the instrument
is 0.3 to 10 microns, which is broken down into five size
channels. Unlike the SMPS, the CI-500 measures all particle
sizes simultaneously. During the single-pass tests, the unit
was operated from the control panel on the front of the
instrument. The data collected were stored in the unit's
internal memory during the test, after which they were
downloaded into Excel, using the software provided with
the instrument.
Figure 4 shows photographs of the two particle counters
used in this work. Figure 5 shows typical distributions of
the fine and "coarse" test aerosols obtained with the SMPS
and CI-500 particle counters, respectively. The two selected
instruments measure particles based upon different physical
properties: electrical mobility in the case of the SMPS and
light scattering in the case of the Climet. This can lead to a
small difference in the particle size measured for a specific
particle measured by both instruments; however, this error
is reduced because particles sizes are binned over a range
of particle sizes. Also, this will not affect the efficiency
measurements, which compare concentration in the same
size bin.
Table 4. SMPS Size Channels (all values in nanometers)
Size Channel
Lower Limit
Upper Limit
20.5
17.8
23.7
27.4
23.7
31.6
36.5
31.6
42.2
48.7
42.2
56.2
64.9
56.2
75
86.6
75.0
100
115
100
133
154
133
178
205
178
237
274
237
316
365
316
422
487
422
562
Figure 4. Aerosol Sampling Instruments, TSI SMPS (left) and Climet CI-500 (right)
-------
3.1.1.2 Bioaerosol The single-pass nitration efficiency of
the two test air cleaners was also evaluated using a biological
aerosol. The same test apparatus described above was used
with the bioaerosol challenge. A suspension of Bg spores,
which are elongated particles with approximate dimensions
of (0.7-0.8) (jum x (1.0-1.5) um, were used to generate
airborne microorganisms in the bioaerosol tests. The tests
were performed for each air cleaner at the medium flow rate.
The purpose of this testing was to determine whether there
was a noticeable difference in the effectiveness of the air
cleaners in handling biological particles versus equally sized
inert particles.
The Bg spores were selected for this testing because they
are a well-accepted surrogate for anthrax spores, having
similar dimensions and viability characteristics. They
can remain viable under a variety of harsh environmental
conditions. The Bg spores have a mass median aerodynamic
diameter of approximately 1 (Jim and therefore have a
relatively high probability of penetration through furnace
filters. The Bg slurry was prepared by adding a dry powder
of Bg spores to high-purity water. The target concentration
of spores in the slurry was approximately 107 CPUs (colony
forming units)/mL in order to achieve a sufficiently high
challenge concentration of Bg in the air. At the same time,
this concentration of microorganisms in liquid suspension
also provided favorable conditions for single-spore
aerosolization. Considering, for example, that the nebulizer
generates liquid droplets with diameters on the order of 10
|jjn, aerosolization of a suspension of 107 CFU/mL results
in an aerosol containing approximately 0.01 spores/drop,
suggesting a very low probability of producing multiple-
spore particles. The actual concentration of spores in the
test slurry was determined by plating an aliquot of the slurry
and counting the resulting colonies, which resulted in a
concentration of 4.4 x 107 CFU/mL. The Bg spores were
aerosolized using a Baxter Airlife nebulizer.
The overall aerosol concentration was measured upstream
and downstream of the air cleaner using the Climet CI-500
particle counter, while concentration of Bg spores in the air
was determined using water-soluble gelatin filters (Sartorius,
Edgewood, NY). These filters were placed in standard
47-mm filter housings, which had been autoclaved prior to
testing, and connected to the sampling probes. A vacuum
pump was used to sample through the filters at a rate of
10 L/min. The Sartorius gelatin filters function in the same
manner as standard membrane filters, but since they are
soluble, the collected microorganisms can be plated and
counted for determining their concentration in the air.
Each bioaerosol test consisted of simultaneously sampling
upstream and downstream of the air cleaner with a single
filter at each location. Upon completion of the test, the gelatin
filters were dissolved in 10 mL of high-purity water and then
diluted to an appropriate concentration before being plated
on tryptone soy agar (TSA). Three plates were made from
each filter, and the organisms were allowed to incubate at
36 °C for 24 hours. After the incubation period, the organisms
were counted using the Qcount automatic plate counter
(Spiral Biotech, Inc.), and the organism counts were used to
calculate the filtration efficiency of the test air cleaner.
-------
Figure 5. Representative Diameter Distributions of Test Aerosols Obtained Using TSI SMPS (a) and Climet CI-500 (b)
Small Aerosol Challenge Measured With SMPS
00574 Q03&5 D04B7 Q0649 0 0866 0115 0154 Q9Q5 0374
Geometric Mean Particle Diameter [microns}
(a)
Large Aerosol Challenge Measured With Climet CI-500
1 E+CE
| 1.E+04 -
o
*
|
,
1
0.30? O/O/ 1.6B 364 S Qf
Geometric Mean Particle Diameter (microns)
(b)
-------
3.1.2 Test Procedure
3.1.2.1 Inert Aerosol The procedure developed for the
single-pass inert aerosol testing can be broken down into two
distinct stages: system startup and data collection. The startup
stage consisted of turning on the equipment, establishing
the correct flow within the test duct, and achieving a stable
challenge aerosol concentration. The data collection stage
consisted of taking three alternating measurements at both the
upstream and downstream locations. This was done in order
to minimize the potential effect of any temporal variability
in the challenge concentration on the measurement results,
as well as to obtain better statistics in the particle counts.
The total particle counts from the three measurements were
summed, and the ratio of total downstream to total upstream
particle counts was used to calculate the fractional filtration
efficiency.
The detailed test procedure was as follows:
Turn on the aerosol counting instrument.
Open the air supply to the aerosol disperser and turn
on the mixing fans in the test duct and in the aerosol
generation box.
Turn on both the booster blower and the air cleaner and
set the air cleaner to the appropriate setting.
Set the air cleaner flow rate. Measure the air velocity
using a hot-wire anemometer. If the velocity is within
10% of the velocity measured for the same unit in the
free-standing configuration, record the value on the data
sheet and continue. If the velocity is not within
the specified range, adjust the blower setting and
measure again. Repeat this process until the correct
flow is achieved.
Record the test duct pressure and the ambient
temperature and humidity on the data sheet.
Attach a cartridge HEPA filter to the inlet of the aerosol
sampling instrument to make a zero count measurement
and check for leaks. If the total number of particles
detected is above 10, check for leaks, tighten fittings,
and measure again.
Detach the HEPA filter and attach the instrument's
sampling tube to the upstream sample probe. Take an
upstream background measurement. If the background
is higher than 1% of the target challenge concentration,
let the system run for a period of time and take the
background measurements again. Then switch to the
downstream sample probe and take a downstream
background measurement. Use the final background
measurement taken before the test in data analysis.
Fill the Airlife nebulizer with the appropriate pre-
prepared KC1 solution and attach it to the aerosol
generation box. Record the solution concentration on
the data sheet.
Turn on the nebulizer and dilution air to the mixing
chamber. Record the flow rate of both the air to the
nebulizer and the air to the mixing chamber, as well as
the nebulizer supply pressure.
Allow the nebulizer to run for at least 10 minutes
to allow the aerosol concentration to approach
steady-state.
Attach the instrument's sample tube to the upstream
sample probe and take the first measurement.
If the large aerosol is being tested, attach the HEPA
cartridge filter to the instrument and take a measurement
to clear the sample line. This should be done after
each upstream measurement using the Climet CI-500
because this instrument does not sample continuously.
This is not necessary when using the SMPS because
it does sample continuously. However, a wait time
of 15 seconds should be allowed between attaching
the sample tube to the sample probe and starting the
measurement to allow the sample tube to clear.
Move the sample tube to the downstream sample probe
and take another measurement.
Repeat the previous three steps two more times for a
total of three measurements at each location.
Turn off the aerosol generation system and allow the
blower and air cleaner to run for a period of time before
starting another test.
This procedure was followed for all inert aerosol tests for
both the large and small aerosol size ranges. Test information
was recorded on the test-specific data sheets kept in a three-
ring binder. Data from the SMPS and Climet particle counters
were transferred to Microsoft Excel datasheets where all
subsequent data analyses were performed.
3.1.2.2 Bioaerosol The procedure for the bioaerosol tests
was similar to that used in the inert aerosol testing, with
the exception of the sampling process. The bioaerosol tests
were performed in three sequential runs, each involving
simultaneously taking an upstream and downstream filter
sample. The filter samples were collected by attaching
the inlet of the filter housing to the appropriate sampling
probe and the outlet to a vacuum pump. The pump was then
turned on and run at a flow rate of 10 L/min for a period of
time estimated for each air cleaner. The sampling time was
calculated based upon the measured single-pass efficiency
of the air cleaner obtained using the inert aerosol and the
expected challenge concentration of Bg spores in the air.
These sampling times were 20 minutes for the HEPA filter
and 2 minutes for the ESP. Real-time aerosol measurements
were also performed using the CI-500 before each set of filter
samples was taken. The real-time measurements were taken
to ensure the system was operating correctly, as well as to
collect additional data points.
Mean upstream and downstream concentrations of the
microorganism were calculated from these replicate filter
samples. Particle penetration was determined in the same
manner as for the inert aerosol, by taking the ratio of the
-------
downstream to upstream CPU concentrations in the air.
Background CPU counts, obtained while running the air
cleaners without bioaerosol generation, were accounted for.
To eliminate possible system bias, the measured air cleaner
efficiency was corrected by subtracting the background
counts from the test data.
3.1.3 Data Analysis
3.1.3.1 Inert Aerosol The computation of inert-aerosol
filtration efficiency was based on the ratio of the downstream
to upstream particle concentrations, corrected on a channel-
by-channel basis for background counts (i.e., upstream and
downstream counts observed when the aerosol generator was
turned off) and accounting for potential system bias by using
a correction factor measured at the start of a test sequence. A
minimum of one background measurement was taken at the
upstream and downstream locations, and three alternating
upstream and downstream measurements were taken during
each test. These measurements were used for determining the
air cleaner filtration efficiency by computing the observed
penetration fraction (Pobserved) for a given particle size:
number of the instrument's size bins. This condition was used
in order to eliminate the effect of random noise counts that
may appear arbitrarily in some size bins of the instruments.
As mentioned above, to remove any potential system-
level sampling bias between the upstream and downstream
sampling locations, the observed penetration was adjusted by
the correction factor (R):
p
corrected
P / R
observed
(2)
(i)
where:
D = Downstream particle concentration,
Db = Downstream background concentration,
U = Upstream particle concentration, and
Ub = Upstream background concentration.
The background concentration measurements were not used
in filtration efficiency determinations if the total number of
downstream background particles counted was lower than the
The correction factor was determined from measuring
particle concentrations at the two sampling locations of
the test apparatus but without the air cleaner present. Two
measurements were taken at each location for each flow
condition. From these measurements, the bias correction
factor^?, defined as the ratio of downstream particle counts to
the upstream counts, was calculated for each size bin, and a
linear trend line was fitted to these data.
In order to ensure that the calculated R value represented
actual system bias, it was examined with respect to
the variability of test data obtained during separate
measurements of the upstream concentrations. This was
done using upstream variability ratios, V which were
calculated by dividing the upstream counts, x, obtained
during one measurement by the upstream counts, y, obtained
during another measurement under the same test conditions.
There were a total of six upstream samples taken for each
air cleaner at each flow rate during their filtration efficiency
measurements; Run 1 consisted of samples 1, 2, and 3, and
Run 2 consisted of samples 4, 5, and 6. A total of six V
values were then calculated for each size bin: V21, V32, V13,
V54, V65, and V46. A scatter plot of all the variability ratios
V was created along with the bias correction factor R, an
example of which is shown in Figure 6.
Figure 6. Plots of Bias Correction Factor (R) and Sampling Variability (V)
ESP, Medium Flow Rate
T fl
1 .;
1 1?
w
* 1 -
c
arm -
OJI
S D-fl
| °-4'
n i
U i
-Ti- : : *
i \ * I 5Ti*T1
* VZ1
vxz
A V13
vw
i VB4
* V46
UpMrKXB**u
UhwBSVBeuna
ft
Ui«r|R)
10 100 1000
Particle Diameter (urn)
-------
A standard deviation was then determined for the sets of
sampling variability data (F. ), and the upper and lower
95% confidence intervals were calculated and plotted on the
graph. If the trend line of R versus particle diameter (shown
in Figure 6 plotted on a logarithmic axis) was found to lie
within the confidence bounds of the V data, no correction
xy 7
was made to the original efficiency measurements because
the measurements bias, if any, would be obscured by the
variability of challenge concentration. If this was not the
case, the filtration efficiency measurements were corrected by
using the measured value of R in Equation 2. This procedure
was followed for each combination of air cleaner, flow rate,
and aerosol size range. In all cases, however, the trend line
was fully within the confidence bounds, so no corrections
were made to the data.
The filtration efficiency is then computed as the following:
Filtration Efficiency
(%)=100(1-P ,)
^ ' ^ corrected'
(3)
3.1.3.2 Bioaerosol Consistent with the inert aerosol testing,
calculation of the penetration of viable microorganisms
through the test air cleaner was based on the ratio of the
downstream to upstream concentrations of Bg spores
determined from the culturable counts on the plates. The
equations used for calculating the penetration and filtration
efficiency of the air cleaner are the same as described above
for inert aerosol. Since no sampling bias was observed in the
tests using the inert particles, no correction was applied in the
determinations of bioaerosol penetration of the air cleaner.
3.1.4 Test Results and Discussion
3.1.4.1 Flow Rate Measurements The flow rates of the
two test air cleaners were measured in their free-standing
configurations before they were tested for their filtration
performance characteristics. The flow rates were determined
by using a hot-wire anemometer to measure the air velocity
and calculating the flow rate based upon the flow area.
The results of the air velocity measurements for the ESP
air cleaner are shown graphically in Figure 7. Each square
represents one of the nine imaginary flow areas chosen for
this analysis. The velocity measured at the inlet is shown
at the top of each square, and the velocity at the outlet is
shown at the bottom. These results indicate that, by design,
the flow is nonuniform across both the inlet and outlet of
the air cleaner. Table 5 summarizes the test results for the
average velocity upstream and downstream of the air cleaner
as well as the overall average velocity for each of its three
flow settings. The airflow rates were then calculated using
the overall average air velocity and the flow area, which was
the same for both the inlet and outlet. These values are shown
in Table 6. The measured flows were considered acceptable,
relative to the manufacturer's specification, for proceeding
with testing.
Figure 7. Local Velocities of the ESP Air Cleaner (top inlet, bottom outlet)
Local Velocity (m/s)
High Medium Low
1.38
2.17
1.43
1.55
1.48
2.00
1.32
1.83
1.43
0.21
1.47
1.70
1.52
2.00
1.83
1.67
1.84
2.13
1.07
1.74
1.17
1.25
1.20
1.85
1.04
1.41
1.18
0.22
1.16
1.25
1.17
1.62
1.49
1.15
1.41
2.02
0.88
1.43
0.92
0.97
0.94
1.21
0.86
1.13
0.84
0.16
0.93
0.92
0.97
1.33
1.15
0.90
1.14
1.56
Table 5. Velocity Characteristics of the ESP Air Cleaner
Flow Setting
Low
Medium
High
Average Upstream
Velocity
(m/s)
0.96
1.21
1.52
Average Downstream
Velocity
(m/s)
1.07
1.39
1.70
Overall Average
Velocity
(m/s)
1.02
1.30
1.61
-------
Table 6. ESP Air Cleaner Flow Rates
Flow Setting
Low
Medium
High
Specification
cfm (m3/s)
225(0.106)
275(0.130)
365(0.172)
Measured
cfm (m3/s)
219(0.103)
280(0.132)
347(0.164)
Relative Error
(%)
2.7
1.8
4.9
The velocity results obtained for the HEPA filter are shown
in Figure 8 in the same format as above. Because the inlet
and outlet flow areas are not the same in this model, no direct
comparison can be made between the velocities at the inlet
and outlet of this unit. It can be seen, however, that there is
a large degree of variability in the flow velocities across the
outlet of the unit. This is due to the unit design. The average
velocities obtained separately for the inlet and outlet were
used with the corresponding flow areas to calculate flow
rates. The average between the inlet and outlet flow rates
was taken to be the overall flow rate at each flow setting.
Table 7 summarizes these flow measurement results. No
specifications were provided by the manufacturer of the air
cleaner for its flow rates, so no comparisons could be made.
Figure 8. Local Velocities of the HEPA Filter (top inlet, bottom outlet)
Local Velocity (m/s)
High Medium Low
1.07
6.32
1.01
5.73
1.17
1.89
1.05
5.41
0.82
6.24
0.82
1.27
0.88
2.05
0.84
1.69
0.98
1.47
0.82
4.67
0.82
4.51
0.86
1.50
0.69
4.52
0.62
5.03
0.64
0.98
0.72
1.453
0.72
1.34
0.78
1.13
0.71
3.45
0.68
4.83
0.76
0.98
0.59
3.67
0.49
2.85
0.53
0.79
0.60
1.21
0.56
1.13
0.62
0.97
Table 7. Flow Characteristics of the HEPA Filter
Flow Setting
Low
Medium
High
Upstream Flow Rate
cfm (m3/s)
271 (0.128)
326(0.154)
422(0.199)
Downstream Flow Rate
cfm (m3/s)
278(0.131)
352(0.166)
449 (0.212)
Overall Flow Rate
cfm (m3/s)
275(0.130)
339(0.160)
436 (0.206)
-------
3.1.4.2 Filtration Efficiency - Inert Aerosol The single-
pass testing of the air cleaners' filtration efficiencies was
completed in two stages. The first stage covered the range of
particles with diameters between 0.03 jjum and 0.3 jjum. The
second stage covered the range from 0.3 jjum to 10 jjum. The
results from the two stages were combined, and Figures 9 and
10 show the filtration efficiencies obtained for the air cleaners
over the entire range of particle diameters. All efficiencies are
plotted at the geometric mean of the measured size bin.
Figure 9. Filtration Efficiency for the ESP Air Cleaner
1.00
0.90
CD
O
O
CD
0.80
- 0.70
CD
c=
o
o
CD
0.60
0.50
0.01
ESP Air Cleaner Filtration Efficiency
High 1
High 2
High Avg
Med 1
Med 2
- Med Avg
Low 1
Low 2
Low Avg
0.1 1
Particle Diameter (microns)
10
Figure 10. Filtration Efficiency for the HEPA Filter
0.01
HEPA Air Cleaner Filtration Efficiency
- Med Avg
Low 1
O Low 2
- - - - Low Avg
0.1 1
Particle Diameter (microns)
10
-------
It can be observed from the two graphs that the filtration
efficiency data obtained using two different aerosols and
two different instruments show good agreement. The vertical
red lines shown in the figures denote locations where the
two branches of filtration efficiency curves are connected,
for both air cleaners. Also, the graphs indicate good
agreement between the duplicate runs performed for all test
conditions. There was somewhat more data scatter observed
for the smallest particles tested. This is mostly due to the
lower concentration of these particles in the challenge
aerosol and relatively high nitration efficiency of the air
cleaners, which results in lower counts of these particles at
the downstream location and, accordingly, leads to their poor
counting statistics.
The results obtained for the electrostatic precipitator show
that there is a minimum in the nitration efficiency curves,
associated with particle diameters in the neighborhood of
approximately 0.2 jjum. This dip in efficiency is a well-
recognized phenomenon for electrostatic precipitators; it
is related to the two charging mechanisms present in such
devices field charging and diffusion charging. Field
charging results from distortions in the electrical field lines,
which are caused by particles greater than approximately
one micron. These distortions cause charged ions traveling
along the field lines to impact on the particle and charge
it. Diffusion charging is dominant for particles less than
approximately 0.1 microns. It results from random collisions
between small particles and charged ions due to Brownian
motion. Between 0.1 and 1 micron, neither mechanism is
dominant and a minimum in collection efficiency is typically
seen. Zukeran et al. cited observations of poor particle
collection efficiencies of ultrafine particles (0.01-0. Ijjum) for
electrostatic precipitators. They proposed poor charging and
flow instabilities as possible causes for this observation.
It can also be seen in Figure 9 that there is a clear trend
of increasing efficiency with decreasing flow rate for the
electrostatic precipitator. This is a result of increasing
particles' residence time within the unit at lower flow rates,
which allows more time for their charging and transport to
the collector surface.
The filtration efficiency curves obtained for the HEPA filter
unit do not show any clear trend with respect to the flow rate.
The data show that the efficiency decreases with particle
diameter smaller than approximately 0.3 jjum. For particles
above 0.3 jjum in diameter, however, the efficiencies were
found to approach HEPA specifications. The cause for the
unexpected but systematic decrease of filtration efficiency
for particles smaller than 0.3 um in diameter is not known
at this point; it may simply be due to some leaks associated
with relatively loose fitting of the filter media in the unit. It
should also be noted that the particle-counting statistics from
the SMPS were much poorer than from the Climet due to
the instrument design and operating principle, even though
it represents the state of the art in nanoparticle measurement.
Nevertheless, the decrease in efficiency for smaller particles
is also consistent with the CADR value decreasing from
330 cfm (0.156 mVs) for larger dust and pollen particles to
320 cfm (0.151 mVs) for smaller smoke particles, as shown
earlier in Table 1.
It should also be noted that both air cleaners display the same
significant drop-off in filtration efficiency for nanoparticles
with diameters smaller than -0.04 jjum, the exact cause of
which is also not known, especially for HEPA filters, and
may represent an area of further investigation.
3.1.4.3 Filtration Efficiency- Bioaerosol During the
single-pass testing of the air cleaners using bioaerosol,
concentrations were measured with both the Climet CI-
500 instrument and with water soluble gelatin filters.
Three sets of filter samples were obtained for each air
cleaner, simultaneously taking aerosol samples at the
upstream and downstream locations of the test units. The
Climet measurements were taken between each set of filter
samples. The gelatin filters were sampled simultaneously
for a predetermined period of time, each sampling at a
10 L/min flow rate. As specified in the test matrix, both
air cleaners were tested at their medium flow rates only,
and the results were compared to those obtained for the
inert aerosol. Figures 11 and 12 show the results obtained
from the bioaerosol testing for the ESP and HEPA filters,
respectively. In these figures, the green line represents the
average filtration efficiency observed using the Climet CI-
500, and the yellow line represents the efficiency measured
earlier during the inert aerosol testing. Good agreement is
observed between these Climet CI-500 measurement results.
The results from the gelatin filters are plotted in the graphs
using the blue symbols, assuming that the microorganisms
are detected as 1-um particles, although, as mentioned above,
the bacteria are actually elongated particles with approximate
dimensions of (0.7-0.8) jjumx (1.0-1.5) jjum.
-------
Figure 11. ESP Bioaerosol Filtration Efficiency
1 -
o
o>
o 0.95 -
LU
O
| 0.9 -
LZ
ro
1 0.85 -
o
O3
LJ_
0.8 -
c
ESP Medium Flow Rate, Bioaerosol
>
I*
/
i
^
r
A
A
Bio 1
° Bio 2
* Bio 3
X Gel Filters
Inert 1
Inert 2
Inert Avg
1123
4
Particle Diameter (microns)
Figure 12. HEPA Bioaerosol Filtration Efficiency
1.000 -
>-.
o
E: n Q QW
o
b=
LU
Q u.yyo
CO
CD
c
o
ro U.99^
LJ_
0.990 -
C
HEPA, H
J 9
y2i
^ A
1
ledium Flow Ra
( A
A
:e, Bioaerosol
A
Bio 1
° Bio 2
» Bio 3
Bio Avg
X Gel Filters
Inert 1
Inert 2
- Inert Avg
1234
Particle Diameter (microns)
In Figure 11, Climet CI-500 measurements are plotted for
the bioaerosol for only the first two particle size ranges of
the instrument, since the downstream counts were too low
for the larger particles with respect to the background counts
to allow for accurate measurements. This is illustrated in
Figure 13, which shows the downstream particle counts
for the second run with the electrostatic precipitator, along
with the corresponding background counts. It can be seen
from this graph that the background is on the same order
of magnitude as the measurements for the larger particles,
indicating that the efficiencies calculated from those size
bins cannot be considered accurate with any certainty. This
was a problem unique to the bioaerosol tests due to the
relatively low challenge concentration compared to the inert
aerosol. Because of this, the gel filters were a more effective
measurement of bioaerosol penetration.
Figure 14 shows the upstream particle counts obtained
during the same run considered in Figure 13. From the
graph no clear peak can be resolved in the size distribution
of aerosol particles that can be identified with the airborne
microorganisms. The lack of a peak is likely due to the
impurities in the biological powder that was used to prepare
the slurry. Therefore, an order-of-magnitude analysis was
performed for the purpose of determining what portion
of the challenge aerosol best represents the spores. It was
determined, based upon the measured concentration of the
initial slurry and the results obtained from the gelatin filters,
-------
that the upstream spore counts should have been on the order
of 7,000 to 10,000 during each run. Returning to Figure 14,
it can be seen that particle counts obtained during this run
for the 0.5 jjum to 1 |jjn size bin are on the same order of
magnitude as expected. Therefore, it was concluded that the
efficiencies measured for the 0.5 to 1 (Jim size bin should give
a close representation of the efficiency that can be expected
for the biological particles, which also agrees well with the
actual size of the spores as discussed above.
Returning to the efficiency curves shown in Figures 11 and
12, very strong agreement can be seen between the results
obtained for both air cleaners using the biological and inert
aerosols in the 0.5 to 1 um size bin. Also, the efficiencies
measured using the gelatin filters, which are plotted as blue
symbols corresponding to the 1 um particles, fit well to the
curves obtained using the CI-500 particle counter. From the
above analysis, it can be concluded that the performance
of both units was consistent for both inert and biological
aerosols.
Figure 13. Downstream Particle Counts From ESP Bioaerosol Test, Run 2
ESP Bioaerosol Background
0.3-0.5 0.5-1.0 1.0-2.5 2.5-5.0
Particle Diameter (microns)
a Downstream O Background
Figure 14. Upstream Particle Counts From ESP Bioaerosol Test, Run 2
ESP Bioaerosol Challenge
ZOUUU
ZUUUU
d
^
o 15000"
_03
o
t: 10000
co
Q_
5000
o
0.3-0.5 0.5-1.0 1.0-2.5 2.5-5.0
Particle Diameter (microns)
-------
3.2 In-Room Testing of Air Cleaners
The purpose of the in-room testing was to investigate
the effectiveness of a room air cleaner in an operational
setting. There are two essential characteristics of a room
cleaner that are independent of operational setting, i.e.,
its single-pass filtration efficiency and airflow capacity.
However, the effectiveness of the air cleaner in removing
airborne pollutants in a room is also dependent upon such
characteristics as the flow pattern and degree of mixing that
the air cleaner induces in the room. An air cleaner will not be
effective if a flow patterns establishes in the room that causes
a high level of clean air recirculation from its outflow back
into the inlet. Based on the results of the single-pass testing,
the HEPA filter was selected as a more efficient room air
cleaner for testing.
In order to test the overall effectiveness of the air cleaner
under particular operating conditions, concentration decay
profiles were obtained by measuring aerosol concentration
as a function of time within an enclosed chamber after
generating a KC1 aerosol. In addition, the degree of mixing
in the chamber was assessed by collecting filter samples at
several locations, which can be related to the variability of
exposure dosages in the chamber and used as a measure of
the mixing ability of the air cleaner.
3.2.1 Test Setup
The in-room testing took place in an 8 ft x 16 ft x 8 ft (2.4 m
x 4.9 m x 2.4 m) chamber located at Battelle's West Jefferson
facility. The chamber was sealed as tightly as possible with
silicon caulking. The only air exchange with the surrounding
room was diffusion through any remaining tiny cracks and
four static HEPA filters installed in the walls to provide over-
pressure relief. Two blowers were attached to the chamber
Figure 15. Test Chamber for In-Room Experiments
to provide recirculation but were not used in this study. A
photograph of the chamber is shown in Figure 15.
A KC1 aerosol was generated within the chamber using a
nebulizer similar to that used in the single-pass testing. The
concentration within the chamber was measured continuously
using the Climet CI-500 laser particle counter. A total of five
open-face filters were used during each test. The filters were
used to simulate the cumulative exposure levels of theoretical
occupants during the test period at the five strategic locations
within the chamber. They also provided a method for
characterizing the mixing conditions within the chamber. The
test matrix developed for the in-room testing is summarized
in Table 8.
Table 8. In-Room Test Matrix
Type of Aerosol
Inert aerosol
(0.3 to 10 (jim)
Location of
Source
Center
Near wall
Location of Air Cleaner
No air cleaner
Near source
Remote from source
Remote from source (Desk)
Air Cleaner
Capacity
Zero
High
High
Low
Number of
Tests
2
2
2
2
Configuration
Code
D
A
B
C
-------
Figures 16 through 18 show the three principal setup
configurations used in this testing (Configuration D is
the same as Configuration A but with the air cleaner not
operating). All the sampling locations were fixed in the
chamber, while locations of the air cleaner and the nebulizer
were varied. The Climet CI-500 laser particle counter
was placed in the middle of the chamber at a height of
approximately 5.5 feet (1.7 m). One open-face filter was
placed next to the CI-500 at the same height. The remaining
four filters were placed approximately one foot from the
walls in each of the four corners. The heights of these filters
were alternated between 5.5 feet (1.7 m) and 3.5 feet (1.1 m),
as shown in the diagrams. The two heights were chosen to
represent a person standing and sitting.
Figure 16. In-Room Test Configuration A
Simple calculations were performed to determine an
appropriate aerosol generation rate for the measurements.
The manufacturer of the Climet CI-500 specifies an upper
limit to aerosol concentration of 107 particles per cubic
foot (-350 particles/cm3). However, a sufficient aerosol
concentration must be present in the chamber to allow a
quantifiable mass to be collected on the filters. The most
accurate balance available for this project could record up
to six significant digits, i.e., up to 1 microgram. In order to
minimize any measurement error, collecting a mass on the
order of hundreds to thousands of micrograms was desired.
/ Filter 2
5.5' High
Fitter 3
3.5' High
Climet
CI-500
Nebulizer
4' High
Filter 5 ^
5.5' High
Air Cleaner
\
Filter 1
3.5' High
Filter 4 t
5.5J High
Figure 17. In-Room Test Configuration B
Fitter 2
5.5r High
Air Cteaner
Climet
Ci-500
Filters' 5.5'High
5.5' High
Filter 3 1
3.5J High
Nebulize*
6' High
Filter 1
3.5' High
Filter 4
5.5 High
-------
Figure 18. In-Room Test Configuration C
/V Filter 2
/> y 5.5' High
\X
Air Cleaner
Oiat
Fittec 1
3.5' High
Desk
FHtef 3 *
3.5' High
Clirrat Nebulizer
CI-500 6' High
Fitters" 5.5rHigh
5.5' High
Filter A
5.5' High
It was estimated that the mass collected on the niters would
be on the order of only tens of micrograms if aerosol
concentration was maintained in the chamber within the CI-
500 detection limit, assuming a constant concentration of 107
particles per cubic foot, 0.3 micron particle diameter, and 10
L/min sampling flow rate for 30 min. Considering also that
the concentration would be lower at some sampling locations,
and decreasing in time, it was determined that it was not
desirable to run both the Climet and the filter samplers under
the same conditions.
Thus, two separate nebulizers were used to achieve the
required concentrations in the test chamber, and the tests
were performed in two stages. During the first stage, a
low concentration aerosol was generated and real-time
concentration measurements were performed using
the Climet. During the second stage, a higher aerosol
concentration was achieved under otherwise similar
conditions, and the filter samples were collected. Climet
measurements were also taken during the second phase,
but any readings that exceeded the detection limit of the
instrument were considered invalid. Both stages of the test
were performed without changing the test configuration.
3.2.2 Test Procedure
As described above, the in-room tests were performed in
two stages. The procedure for these tests was as follows.
At the beginning of the test, the background within the
chamber was brought down to a level of no more than 105
particles per cubic foot (i.e., not to exceed 1% of Climet
CI-500 max concentration limit). This was done, using a
second in-room air cleaner (used in the single-pass efficiency
testing), because preliminary testing showed that the chamber
ventilation system was not effective in decreasing the
background aerosol to the desired concentration level.
After the desired background was achieved, the auxiliary
air cleaner was turned off and the test air cleaner turned on.
At this point (time zero) the low concentration nebulizer
was switched on. The nebulizer was run for 10 minutes,
after which it was turned off, while the air cleaner continued
running. This stage of the test was considered complete when
the concentration was returned to the background level or
after one hour (which only occurred in the tests configured
with no air cleaner running). In the latter case, the auxiliary
air cleaner was then used to return to the background level.
The second stage of the test began once the background
concentration was achieved. This stage was run similarly to
the first stage with respect to the test configuration. At the
beginning of stage two, the filter sampling pumps and the
high concentration nebulizer were simultaneously turned
on. The nebulizer was run for 10 minutes, after which it
was turned off, while the filter sampling pumps and the air
cleaner continued running for another 50 minutes. A constant
time of one hour was selected for running the sample filters
to allow for dosage comparisons between the different test
configurations.
It should be noted that the air cleaners were running
throughout the test, including during the aerosol generation
period, in order to simulate a realistic attack scenario. In such
a scenario, the air cleaner would be running continuously
and would offer some protection during as well as after the
release. This also allowed for some assessment of how well
the air cleaner mixed the air within the chamber and for
assessing the effect of room configuration on the performance
of the air cleaner.
-------
3.2.3 Data Analysis
For each in-room test configuration, two principal data sets
were obtained, real-time concentration data from the Climet,
which was used to construct concentration vs. time profiles,
and the cumulative exposure data obtained using sample
filters. The concentration vs. time plots were used to illustrate
the mitigation capabilities of the air cleaner.
The total decay constant, "k", characterizes the rate at which
particles are removed particles from the air in the chamber;
it combines both the effect of the air cleaner and deposition
within the chamber. Due to the small size of the particles
generated in this study, deposition is assumed to have a minor
effect on the measured value of k. "K" is therefore indicative
of the air cleaner effectiveness. "K" was determined by
fitting an exponential function, Equation 4, to the data points
between the peak concentration (C0) and the point when the
concentration dropped to less than two times the background.
c.
(4)
Where:
C = aerosol concentration at time t, particles/ft3
(particles/m3),
Co =peak aerosol concentration, particles/ft3 (particles/m3),
k = overall rate constant of concentration decay, 1/min (1/s),
and
t = time, min (s).
The second set of data consists of the five open-face filter
samples. All filters were run for one hour as discussed above.
In addition to assessing the effect of an air cleaner on the
cumulative exposure level, comparing the mass collected
on each of the five filters from the same run allowed for a
characterization of the mixing within the chamber. This can
be illustrated by calculating the CV for the five sampling
locations.
3.2.4 Results and Discussion
3.2.4.1 Decay Rate For each test configuration, the real-
time concentration vs. time results are plotted in Figure 19. In
addition, Figure 20 shows the configuration-average profiles
determined for each of the test Configurations A, B, and C.
Figures 19 and 20 show graphs of total particle number
concentrations plotted as functions of time. Since particle
deposition rate in the chamber is also dependent on particle
diameter, Figure 21 shows particle number concentrations
plotted as functions of time for the six particle diameter
bins of the Climet CI-500 instrument used in the study. The
particle concentrations in Figure 21 were normalized by
the respective peak concentrations of the size bins in order
to illustrate the relative decay rate for the different particle
sizes. According to this figure, some increase can be seen in
the decay rate with increasing particle diameter. The decay
constants calculated from this plot ranged from 0.315 for the
smallest particles to 0.398 for the largest particles. Based
upon the results of the single-pass efficiency measurements,
which showed high filtration efficiency values for particles
between 0.3 and 10 microns, the difference in decay
constants is attributed to the increased settling velocity of the
larger particles.
Concentration decay constants were determined for
each of the curves shown in Figure 19. The total particle
concentration was chosen for determination of k, instead of
size-dependent decay, for ease in comparing the various test
configurations. These constants were calculated by fitting
exponential functions to the data points obtained for each
curve between the peak concentration and the point when
the concentration decreased to less than two times the initial
background level. The average lvalues were then determined
for each of the test configurations, which are shown in
Table 9. The R2 value for the fit of all the curves was greater
than 0.99, indicating a very good fit to Equation 4.
-------
Figure 19. Concentration vs. Time Plots, Individual Configurations
Configuration A, Concentration vs. Time
+06 r
20
Time (min)
-Testl
-Test4
-Test2
Tests
-Test3
-Test6
Configuration B, Concentration vs. Time
-Testl
Test2
-Test3
Test 4
Configuration C, Concentration vs. Time
Testl
-Test 2
Configuration D, Concentration vs. Time
Testl
-Test2
Figure 20. Concentration vs. Time Plot, Averaged Data
Averaged Data, Concentration vs. Time
Configuration A
- Configuration B
A -Configuration C
-------
Figure 21. Size Resolved Concentration vs. Time
Cc
I 1'2
j 0.8-
T5
S 0.6
1
? 0.4 -
H
1 0.2
u
2
0, .
^figuration A. Test 2, Normal
V'
jf.! ,
i !
izedSIze Data
*0.1-O5«ncrofi5
D & t QmKICKIS
* 1 l> i! inHamiK
ii Ij-L 0 mcrons
> 5 0 tO OlTICfOrri
MO.OnUoranft
\
;l«**(
*V|t**»__>_. __
000 10.00 20.00 30.00 4000
Time (min)
Table 9. Calculated Decay Constants (1/min)
Run #
1
2
3
4
5
6
Average
Scenario A
.318
.316
.275
.283
.279
.262
.289
Scenario B
.285
.296
.286
.283
No test performed
No test performed
.287
Scenario C
.188
.191
No test performed
No test performed
No test performed
No test performed
.189
Scenario D
.0058
.0055
No test performed
No test performed
No test performed
No test performed
.0057
The chamber tests were run in series, with different scenarios
tested at different times. There were a total of six tests
performed in Configuration A, as additional Configuration A
tests were run during each test day to ensure that consistent
conditions were used in all trials. As shown in Figure 19,
the variability in the aerosol test conditions maintained
during subsequent tests in the same configuration was low.
Considering the decay constants in Table 9, however, some
variability was observed in their individual values calculated
for Scenario A. Runs 1 and 2 were performed during the
first test period, Runs 3-5 were performed during the second
test period, and Run 6 was performed during the final test
period. The decreasing decay constant was attributed to the
increasing leak of ambient particles into the chamber, caused
by its expansion and contraction over the summer test period,
as suggested by the gradually increasing level of achievable
background counts. Simple "well-mixed model" calculations
were also performed, which also suggested that this could
account for the slight decrease in the observed decay constant
for test Configuration A. Nevertheless, the background never
exceeded 5% of the peak aerosol concentration in this testing,
and the CV in the decay constant was on the order of 6%.
A number of observations can be made from these data. The
most obvious observation is that the concentration within the
chamber decayed much more slowly when the air cleaner
was not turned on (Configuration D), indicating that the
presence of the air cleaner had a significant mitigating effect.
As expected, there also was a clear increase in the mitigation
ability of the air cleaner with increasing flow rate, which
can be seen by comparing the decay constant determined for
Configuration C to those of Configurations A and B.
It can also be observed from both these plots and the
calculated decay constants for Configurations A and B that
no significant difference was observed in the performance of
the air cleaner in these different test settings. In Configuration
B, when the air cleaner was positioned farther from the
nebulizer, it may have been slightly more effective than
in Configuration A. This may be due to the fact that in
Configuration B the nebulizer was pointed in the direction
of the air cleaner inlet, whereas in Scenario A the nebulizer
was pointed in the opposite direction. When the nebulizer
is pointed in the direction of the air cleaner inlet, the flow
of aerosol toward it is enhanced in comparison with the
nebulizer pointing in the opposite direction. This effect is
-------
the likely explanation for the higher peak concentration of
aerosol observed at the Climet location for Configuration
A than for Configuration B. However, the decay constants
were very similar for both cases, indicating that after some
adequate mixing time, the effectiveness of the air cleaner was
practically equivalent between these two configurations.
3.2.4.2 Mixing Efficiency and Dosage In addition to the
real-time monitoring of aerosol concentration in the test
chamber, cumulative samples of aerosol were collected using
standard 47-mm filters for each of the four test configuration.
Average mass and CV between the different filter locations
were calculated. A summary of the results is shown in Table
10, and the individual filter masses are graphed in Figure 22.
The average mass values can be compared between different
test configurations to give a relative exposure dosage
received in the room over a one-hour period. The CV gives a
relative indication of the mixing conditions developed within
the chamber.
As expected, the masses collected on the individual filters
varied to some extent between Configurations A and B;
however, the average mass collected and the CV were found
to be very close. This supports the conclusions made from
the concentration decay data that, although the flow patterns
within the chamber were different between these two test
configurations, the air cleaner induced sufficiently high
mixing conditions in both cases to result in an overall similar
effectiveness of the air cleaner.
The dosage observed in Configuration D (air cleaner not
operating) was found to be more than an order of magnitude
greater than those obtained for Configurations A and B,
indicating a significant protection factor provided by the air
cleaner. Also, the CV in Configuration D was much higher,
indicating a much slower mixing process.
For test Configuration C (low flow setting, desk in the
room), the dosage was higher than in Configurations A and
Figure 22. Collected Filter Masses
B, as expected, due to the lower concentration decay rate.
However, according to the CV values, no decrease in the
mixing efficiency was observed in this case.
Table 10. Collected Filter Masses
Test
Scenario A 1
Scenario A 2
Scenario B 1
Scenario C 1
Scenario C 2
Scenario D 1
Average Mass (|jig)
667
550
564
1086
989
7571
CV
35%
17%
28%
19%
18%
42%
The variability in the filter masses collected at different
sampling locations, observed during the in-room testing of
the air cleaner, is also illustrated as a diagram in Figure 22.
This diagram shows that despite the fact that the air cleaner
has high airflow capacity, relative to the size of the test room,
the level of mixing it induces, although high, is not "perfect"
(referring to the perfectly-mixed zone concept frequently
used in model calculations). In fact, while the air cleaner
promotes convective mixing in the room, it also creates and
maintains a low-concentration zone in its outflow region,
as well as a point of very high aerosol concentration at the
aerosol source during generation. Therefore, the effectiveness
of a room air cleaner in mitigating the effect of an
aerosolized agent attack on a building will strongly depend
upon the agent dissemination scenario. Nevertheless, based
on the results of this investigation, it can be concluded that
a room air cleaner may lead to a very significant reduction
in the level of indoor exposure to the agent, which may be
approximately evaluated using the well-mixed zone concept.
Collected Filter Masses
r.-r.
E 03
i r. .0
M
J BKO-t
1 6 t-d-i
B
o
4 ECU -W
: r -14
Ub-JU
rT
Fieri RUST 2 FOE-:- FUer4 Filer? Average
Filter Number
DAlnA2~BlnClDC2
-------
4.0
Modeling Study
Mathematical modeling is widely used in countless
applications, especially when experimental investigations
are impractical for one reason or another (for example, cost
of experimental trials may be prohibitively high). In the
context of this work, as mentioned above, the effectiveness
of an indoor air filtration system against an aerosolized
agent attack on a building is highly scenario-dependent, and
its detailed exploration would require considering a great
number of cases. A computational fluid dynamics (CFD)
model was therefore used in this project with the purpose
of evaluating the accuracy and efficiency with which it can
predict concentration evolution of contaminant in the room
air. A CFD package, FLUENT, was used to implement
the model and generate the numerical simulations. The
model geometry and flow conditions selected represented
Configuration A of the in-roomtest scenarios discussed
above. In addition, simplified calculations were performed
based upon the perfectly-mixed zone assumption. This model
was implemented in Microsoft Excel, and the calculations
were compared to the CFD model predictions and
experimental results.
The ultimate goal of this task was to gain an understanding
of how to correctly develop and use computational models
to assess the effectiveness of an in-room air cleaner or
other HVAC equipment in minimizing the impact of an
agent dissemination attack on a building. There are two
principal effects that an indoor air cleaner may have on the
evolution of pollutant concentration in a room: enhanced
mixing of the room air and cleaning the air by some aerosol
removal mechanism. Both of these effects will result in
the development of a unique scenario-, time-, and space-
dependent concentration profile in the room of interest.
The cost associated with testing all the potentially viable
HVAC configurations would be prohibitive and logistically
complex. Modeling, if properly applied, offers potential
savings in identifying the main phenomena that control the
effectiveness of in-room air cleaners in reducing the impact
of an aerosolized agent dissemination event.
4.1 FLUENT CFD Modeling
A single simulation was performed under this task, using
commercially available software to resolve the effect of
air cleaner location on its effectiveness, addressing both
the mixing and filtration aspects of the overall effect.
Consideration of both these aspects is important because
while air filtration acts to reduce pollutant concentration
in a room, mixing tends to decrease a pollutant's spatial
nonuniformity in the room, thus reducing its concentration at
some locations while increasing it at others.
The CFD modeling approach was based on the Eulerian
treatment, whereby the contaminant is treated as a continuum
fluid dispersing in the air by advection and diffusion
processes. The model geometry consisted of a rectangular
room with a single air cleaner located near the center
of the room, a nebulizer injecting an airborne challenge
simulant for a portion of the simulation, one real-time
aerosol concentration monitor, and five individual points
for predicting the cumulative mass collected on the filter
samples. The indoor air cleaner was treated as a stand-alone
interior unit with specified dimensions, flow capacity, and
contaminant removal efficiency.
The computer code FLUENT, a well-validated industry
standard code for CFD calculations, was applied in this
analysis. Using the model geometry and the boundary
conditions, a steady-state three-dimensional solution
was obtained for the in-room flow pattern, which was
subsequently used to predict a time-dependent contaminant
concentration field in the room. A highly resolved spatial
and temporal map of the contaminant concentration profile
was obtained and compared to experimental data. A more
detailed discussion on the modeling approach and results
can be found in Appendix B. A detailed description of the
mathematical approach to CFD is available in the FLUENT
User's Guide, which can be accessed online at FLUENT'S
Online Support Resources (FLUENT, 2007).
4.1.1 Results
From the CFD model calculations, aerosol concentration at
each time step was determined at the locations of each of the
five filter samples and the Climet, as used in the experiments.
From these data, concentration vs. time plots were
obtained, decay constants calculated, and the cumulative
concentration was determined for each of the filter locations.
The concentration vs. time curves predicted for each of the
sample locations are shown in Figure 23. The decay constants
were very similar for each of these curves, so the average was
taken. The average decay constant for the CFD model was
then found to be 0.535 1/min, which is much higher than that
obtained from the experimental results (0.318 1/min).
As mentioned above, by integrating the curves in Figure
23, masses of aerosol particles collected on the simulated
filters, or simulated exposure dosages, were determined
and compared to the experimental results. In general,
the predicted dosages were found for all locations to be
somewhat higher than the experimental dosages, although
the overall trends were captured. One of the reasons for this
disagreement was associated with the uncertainty in the rate
at which aerosol was introduced into the room. Since the
experimental spray rate was not well known (the spray rate
used in the model was based on an estimate obtained from
a nebulizer characterization test spraying pure water), the
masses from the CFD model predictions were normalized
to give the same average filter mass as in the experiment,
which was done for the results comparison purposes.
-------
Figure 24 shows both the normalized and nonnormalized
masses, predicted from the CFD calculations, as well as the
average of the two experimental runs performed for this test
configuration. The CV for the model filters is 23%, which is
Figure 23. CFD Model Results
within the range of the experimental CVs shown in Table 10.
As a result of this analysis, the CFD model appears to offer
a reasonable potential for replicating the general trend of the
experimental results, with the possible exception of Filter 5.
CFD Model Results, Room Concentration vs. Time
0.014
200
400
600
800
1000 1200 1400 1600
1800
Climet Filter 1
Filter2 Filters Filter4
Filters
Figure 24. Filter Masses from CFD Calculations
CFD Filter Masses
Filter 1 Filter2 Filter 3 Filter 4 Filter 5 A/erage
Filter Number
Q CFD a Experimental B CFD Normalized to ExprA/g
-------
4.2 Perfectly-Mixed Zone Analysis
For comparison purposes, a series of calculations was also
performed based upon the assumption that the chamber
was instantaneously and perfectly mixed. The exact model
is shown by Equation 5 (note an alternative derivation
of the equation using SI units is shown in parentheses
and would lead to k in units of 1/s). These calculations
assumed that 100% of the particles that entered the filter
unit were removed from the air and that deposition was
insignificant. The calculations were performed using both the
experimentally measured flow rate of the air cleaner, 450 cfm
(0.212 mVs), and the certified clean air delivery rate (CADR),
330 cfm (0.156 mVs). The spray rate, expressed in particles/
min, used in the calculations was determined by extrapolating
back the decay curves obtained during the Configuration D
tests (air cleaner not operating) to time zero. Figure 25 shows
the results of these calculations plotted along with the test
data obtained in one of the experimental runs. The decay
constants are also shown on the graph, according to which
their values obtained using the well-mixed zone calculations
are appreciably higher than those obtained using the CADR
value, and both are higher than the experimental value.
Jj sPmy zt filter ' ^
(5)
Where:
V = the volume of the chamber, ft3 (m3),
C = the particle concentration, particles/ft3 (particles/m3),
N = the spray rate, particles/min (particles/s), and
Qfllter = the flow rate of the air cleaner, cfm (m3/s).
The CADR is determined experimentally, according to the
AHAM procedure, reflecting the flow rate and filtration
efficiency of the air cleaner, as well as its contribution to the
degree of mixing established in the CADR test chamber. It
is determined by measuring the concentration decay in an
isolated test chamber that has been uniformly mixed prior to
the test. Also, the air cleaners are rated for CADR using a
test chamber smaller than that used in this study; besides,
in this work, the challenge aerosol was not mixed prior to
testing. Therefore, it was not unexpected that the HEPA filter
would demonstrate a lower concentration decay rate during
this testing as compared to the decay rate based on the
CADR value.
Figure 25. Well-Mixed Zone Model Results
2.5E+06
CO
I 2.0E+06
1§
1 1.5E+06
E
O
s 1.0E+06
.a
E
^ 5.0E+05
Well-Mixed Room Calculations
O.OE+00
0.00 5.00 10.00 15.00 20.00 25.00 30.00
Tim e (m in)
Experimental
330 CFM '
1 ' 450 CFM
-------
-------
5.0
Conclusions
This report describes the investigations conducted to
quantitatively verify the ability of room air cleaners
(specifically, filters that remove PM) to mitigate the effect
of an aerosolized biological agent attack on a building. Two
filter systems were evaluated with regard to their building
protection effectiveness. One of the air cleaners selected for
this project used HEPA filtration technology, and the other
was based on the principle of electrostatic precipitation. This
work comprised both an experimental investigation and a
modeling study.
The test air cleaners were experimentally evaluated for their
single-pass filtration efficiencies as a function of particle
diameter (ranging from 0.03 jjum to 10 jjum) and airflow rates,
using both an inert aerosol and a bioaerosol. The HEPA filter
was then selected for further evaluation, in a test chamber
under various room configurations, to verify its effectiveness
in reducing ambient levels of PM. In the test chamber
experiments, the in-room particle concentration decay rate
was determined from data obtained for a particular location
in the chamber using a real-time particle counter.
Following the completion of the experimental phase of
the project, model calculations were performed using
computational fluid dynamics for one of the specific in-room
test configurations. In addition, simple calculations were
performed for the test conditions using the perfectly-mixed
zone modeling approach.
During the single-pass efficiency testing, two replicate test
runs were performed for each test condition to demonstrate
the precision between them. The ESP-based air cleaner
displayed a pronounced minimum in filtration efficiency for
particles of ~ 0.2 jjum diameter, which is consistent with the
principles of electrostatic precipitation. Also, the single-pass
efficiency of the ESP air cleaner was found to decrease with
increasing flow rate through the unit, due to the decreasing
residence time of the particles in the charging and deposition
zones of the collector.
For the HEPA filter, no noticeable effect of flow rate on
the filtration efficiency of the unit was observed, but an
unexpected drop-off in efficiency was observed for particles
below 0.3 (jum in diameter. This observation could be
explained by some leaks that probably developed aroundthe
filter because of its relatively loose fit in the single-pass test
unit. However, the consistent tendency of both air cleaners
to have reduced efficiency for particles with diameters
smaller than 0.04 jjum warrants further investigation. No
difference was observed between the air cleaners' filtration
efficiencies for biological and inert aerosols having similar
particle diameters.
The effectiveness of an in-room air cleaner in reducing a
room's aerosol level under typical operating settings depends
on three principal characteristics: 1) single-pass filtration
efficiency, 2) filtration airflow rate, and 3) the airflow pattern
that the cleaner induces in the room. While the first two
characteristics can be obtained from some straightforward
measurements, such as those used in this study, the airflow
pattern in the room is also dependent upon such other factors
as room size and shape, HVAC characteristics, furnishing,
leak patterns, presence of mixing fans, etc. Some of these
factors were investigated in this project using the HEPA filter.
The HEPA filter was found to provide significant protection
(in terms of reduced in-room air PM concentration) when
compared to the case when the air cleaner was not operating.
This observation was not unexpected and illustrates the
potential usefulness of in-room air cleaners in the event of
an aerosolized biological agent attack on a building. The
location of the air cleaner relative to the aerosol source was
found to have a minimal effect on reducing the PM level.
The addition of an office desk and a chair in the test chamber
also did not appear to noticeably alter the performance of
the air cleaner. Overall, the HEPA filter provided reasonable
mixing conditions in the test room, although some variability
in the PM levels was observed for different locations inside
the chamber.
The CFD model simulations performed under this study
demonstrate the ability of this technique to predict aerosol
levels in various indoor settings, with the caveat that it
overestimated the PM concentration decay rate. The main
issue associated with this application of CFD is the broad
spectrum of flow regimes evolving within the room, ranging
from laminar to fully turbulent conditions. This requires
specification of different turbulence closure models, such as
large eddy simulation for describing large flow recirculation
patterns, and more refined schemes for considering the
dispersion of contaminant in the aerosol generation and air
cleaner "jet" exhaust zones. Depending on the particular
scenarios of interest, CFD simulations can also be expensive
to perform. In this regard, it is also recommended that
an alternative modeling methodology be developed for
evaluating the effectiveness of in-room air cleaners in
real situations; this type of model would be capable of
applying the CADR-type of characteristics to various room
configurations while accounting for the different degrees of
mixing the air cleaner may induce under different settings.
-------
-------
6.0
References
Association of Home Appliance Manufacturers. Directory of Certified Room Air Cleaners. Edition No. 3, 2004.
Association of Home Appliance Manufacturers. Method for Measuring Performance of Portable Household Electric
Cord-Connected Room Air Cleaners. ANSI/AHAM AC-1-2002.
Association of Home Appliance Manufacturers. Directory of Certified Room Air Cleaners. Edition No. 3, 2005.
Consumer Reports. Air Cleaners, CR Quick Recommendations. (October, 2003).
FLUENT Online Support Resources, https://secure.fluent.com/sso2/login.htm (6/20/2007).
FLUENT User's Guide. Lebanon, NH, 2005.
Pope, S.B. Turbulent Flows. Cambridge University Press, 2000.
Reid, R.C., J.M. Prausnitz, and B.E. Poling. The Properties of Gases & Liquids. 4th Edition. McGraw-Hill, 1987.
U.S. Environmental Protection Agency. "Lesson 1: Electrostatic Precipitator Operation." APTI Virtual
Classroom SL412B. 2002. http://yosemite.epa.gov/oaqps/EOGtrain.nsf/fabbfcfe2fc93dac85256afe00483cc4/
ca9ael7f9567495885256b66004e7985/$FILE/12blesl.pdf (9/20/05).
Zukeran, Akinori, et al. "Collection Efficiency of Ultrafine Particles by an Electrostatic Precipitator Under DC and
Pulse Operating Modes." IEEE Transactions on Industry Applications 35, (5): 1184-1190 (1999).
-------
-------
Appendix A
Air Cleaner Selection
The objective of the project described in this report was
to conduct experiments and mathematical modeling
to determine the effectiveness of room air cleaners in
minimizing the impact of an aerosolized biological agent
attack on a building. Two types of room air cleaners, a HEPA
filter and an electrostatic precipitator (ESP), were tested.
A brief market survey was conducted through the Internet
to select representative room air cleaners for this project.
It was found that most room air cleaners in the market are
certified under the Room Air Cleaner Certification Program,
which is sponsored by the Association of Home Appliance
Manufacturers (AHAM, ANSI/AH AM AC-1-2002). This
standardized measurement procedure was designed to
determine the Clean Air Delivery Rate (CADR), indicating
how effective a room air cleaner is in reducing concentrations
of such paniculate pollutants as tobacco smoke, household
dust, and pollen.
Approximately 174 different models of room air cleaners
from 18 manufacturers are currently certified under the
CADR program (AHAM Directory of Certified Room Air
Cleaners. Edition No. 3, 2004). The specifications of the
certified air cleaners were reviewed from the manufacturers'
(or venders') Web sites. It was found that among the 174
certified air cleaners, only one was an ESP type air cleaner.
Another air cleaner, which has a similar CADR rating as the
ESP, was suggested for testing as the filter-type air cleaner.
Note that these two air cleaners are also the top two room
air cleaners recommended by Consumer Reports (2003).
The certified CADRs for the two selected air cleaners are
summarized in Table A-1.
Table A-l. CADR Values of Selected Air Cleaners
330 (.156) 330 (.156)
According to the ANSI/AHAM Standard AC-1, the CADR is
the rate of contaminant reduction in a standard test chamber
when the test air cleaner is operating, minus the rate of
natural decay when the air cleaner is not operating, times
the volume of the test chamber. During a certification test, a
given quantity of aerosol is generated into the test chamber
followed by one minute of mixing with the mixing fan. The
test air cleaner is then turned on and the real-time aerosol
concentration in the chamber is recorded. The concentration
decay inside the chamber is characterized using the following
exponential equation:
Where:
C = aerosol concentration at time t, particles/ft3
(particles/m3),
Co = initial aerosol concentration, particles/ft3 (particles/m3),
K = overall rate constant of concentration decay, 1/min (1/s),
and
t = time, min (s).
The CADR is then calculated as:
(2)
Where:
CADR = clean air delivery rate, cfm (m3/s),
V = volume of test chamber, ft3 (m3), and
Kn = the rate constant of the natural concentration decay,
without the air cleaner operating, 1/min (1/s).
The following calculations were performed before testing to
provide initial estimates based upon available specifications.
Since the single-pass efficiency data of the selected air
cleaners were not available from the manufacturers,
the following analysis was performed to estimate their
efficiencies using the reported CADR values. In an initially
well-mixed test chamber, the material balance of the test
aerosol can be expressed using the following equation:
dt
- + uA\
WO )
(3)
Where:
Q = the fan speed of the test air cleaner, cfm (m3/s),
T| = single-pass efficiency, %,
u = the average aerosol terminal-settling velocity, ft/min
(m/s), and
A= test chamber area, ft2 (m2).
According to Equation (3), concentration decay in the test
chamber is controlled by both the intrinsic characteristics of
the air cleaner and its test configuration. Integrating Equation
(3), one obtains:
100
C=Ce *
(4)
-------
Where:
K=(QT|/100+uA)/Vand
Kn= uA/V
Combining Equations (4) and (2), the following equation is
obtained:
r)=-
CADR
xlOO
(5)
As shown in Equation (5), the single-pass efficiency can be
estimated from both the CADR value and the air cleaner flow
rate. It should be noted that this analysis ignores the effect
of imperfect mixing induced by the air cleaner. According to
the (ANSI/AHAM AC-1-20023) standard, room air cleaners
with multi-level fan speeds are tested at the highest setting.
The single-pass efficiencies of the ESP for the three different
types of aerosols were estimated using Equation (5). The
results are summarized in Table A-2. Note that the estimated
single-pass efficiency of pollen is slightly over 100%, which
is believed to be due to some uncertainties associated with
the specified airflow rate.
Table A-2. Estimated Single-Pass Efficiencies for ESP
Air Cleaner (Model C-90A)
Tobacco Smoke
(0.09 to 1.0 |jum)
Dust
(0.5 to 3|jim)
Pollen
(5 to lljjim)
82
89
101
The single-pass efficiencies for the HEPA filter were not
estimated because its flow rate specifications were not
available from the manufacturer. However, since it is a HEPA
filter type, the single-pass efficiency was assumed to be 99%.
The maximum flow rate of the air cleaner was then estimated
to be 330 cfm (0.156 mVs), based on Equation (5).
The specifications of the selected air cleaners, including their
flow rates are summarized in Table A-3.
Table A-3. Specifications of Selected Air Cleaners
ESP
HEPA
Three levels
Low 225cfm(0.106m3/s)
Medium 275 cfm (0.130 mVs)
High 365 cfm (0.172 mVs)
Three levels
High 330 cfma (0.156 mVs)
0.48m Hx 0.38m
Lx 0.55m W
0.56m Hx 0.46m
Lx 0.28m W
"Estimated by Battelle using Equation (5), by assuming a 99% single-pass
efficiency for dust and pollen.
The flow rates of the selected air cleaners were 1.8 to 2.9
times the typical airflow rate of a ventilation system relevant
to the Battelle test room, based on the typical 1 cfm/ft2
(0.0051 m3/s/m2) standard for an all-air constant-volume
ventilation system with ducted returns, and the 128 ft2
(11.9 m2) of areaforthe test facility.
-------
Appendix B
CFD Modeling,
Detailed Methodology and Results
B.I. Summary
A computational model was developed and used to evaluate
the accuracy and efficacy with which a fluid flow model
could successfully predict the concentration evolution of
an aerosol injected into a room. The computational fluid
dynamics (CFD) package FLUENT was used to implement
the model and generate the numerical simulations. The model
geometry and flow conditions selected represent one of the
test conditions in the test chamber characterization runs
performed at Battelle's West Jefferson Facility.
A comparison of numerical predictions to experimental
measurements was completed and is documented below in
the results section. The agreement is judged sufficiently good
to potentially provide guidance on relative trends such as
under which conditions select areas in a room will experience
higher concentrations than others and the duration of time
under which those conditions exist. The results also exhibit
good agreement with experiments in terms of the associated
decay constant in airborne concentration after the simulant
source has been turned off and equipment such as an air
cleaner is allowed to continue to remove contaminant from
the air. However, in terms of an absolute, highly precise
predictor of air concentration, the requisite precision in
modeling inputs and the computational expense remain as
challenges to using CFD as a general design tool for in-room
air contaminant modeling.
B.2. Objective
The ultimate goal of this task is to gain an understanding
of how to correctly develop and use computational models
to assess the effectiveness of an in-room air cleaner or
other HVAC equipment in minimizing the impact of an
aerosolized agent attack on a building. There are two
principal effects that an indoor air cleaner may have on the
evolution of pollutant concentration in a room: enhanced
mixing of the room air and cleaning the air by some aerosol
removal mechanism. Both of these effects will result in
the development of a unique, scenario-, time-, and space-
dependent concentration profile in the room of interest. The
cost associated with testing all potentially viable HVAC
configurations would be prohibitive and logistically complex.
Modeling, if properly applied, offers potential savings in
identifying the main phenomena that control the effectiveness
of in-room air cleaners in reducing the impact of a biological
agent dissemination event.
A single simulation was performed under this project, using
commercially available software to resolve the effect of
air cleaner location on its effectiveness, addressing both
the mixing and filtration aspects of the overall effect. This
is important because while air filtration acts to reduce
pollutant concentration in the room, mixing tends to decrease
the pollutant's spatial nonuniformity in the room, thus
reducing its concentration at some locations while increasing
it at others.
The CFD modeling approach was based on the Eulerian
treatment, whereby the contaminant is treated as a continuum
fluid dispersing in the air by advection and diffusion
processes. The model geometry consisted of a rectangular
room with a single air cleaner located near the center
of the room, a nebulizer injecting an airborne challenge
simulant for a portion of the simulation, one real-time
aerosol concentration monitor, and five individual points
for monitoring the aggregate mass collected on the filter
samples. The indoor air cleaner was treated as a stand-alone
interior unit with specified dimensions, flow capacity, and
contaminant removal efficiency.
The computer code FLUENT, a well-validated industry
standard code for CFD calculations, was applied in this
analysis. Using the model geometry and the boundary
conditions described in the approach section, a pseudo
steady-state three-dimensional solution was obtained for
the in-room flow pattern, which was subsequently used to
predict a time-dependent contaminant concentration field
in the room. A highly resolved spatial and temporal map
of the contaminant concentration profile was obtained and
compared to experimental data.
-------
B.3. Approach
The steps taken in completing the in-room simulation are as
follows:
1. Model Construction - Create a computational mesh that
represents all the major features of the experimental
configuration.
An illustration of the three-dimensional model geometry is
provided in Figure B-l. The complete geometry of the air
cleaner, Climet CI-500, the nebulizer, and the room walls
were included since their precise features were judged to
have the largest impact on subsequent fluid flow patterns.
The filter monitoring locations are indicated as open circles,
but the physical descriptions themselves were not included
in the model.
Dimensions of the room, offset locations of the equipment
modeled, and dimensions of the air cleaner intake and
exhaust as well as the nebulizer outlet orifice are included in
a plan view in Figure B-2a and side or profile view in Figure
B-2b. Note that the monitor points representing Filters 1
through 4 were assumed to be 12 inches (0.3m) from each of
the two walls in their respective corners and placed at one of
the two elevations indicated, namely 3 '6" (1.1 m) or 5 '6"
(1.7 m) from the floor. The Filter 5 monitoring point was
located just above and to one side of the Climet, while the
Climet concentration monitor was located 6 inches (0.15 m)
above the face center of the top of the unit itself.
2. Boundary Condition Assignments - Specify flow rates
for the air cleaner intake and exhaust, the nebulizer mass
outflow (including mass fraction of simulant and the time
period of operation), and fate indicators on the walls (i.e.,
whether contaminant that strikes the wall reflects off or is
trapped against the surface).
Figure B-l. Model Geometry
-------
Figure B-2a. Model Dimensions (Plan View)
12'
FT I X 4 Vfi' W
insane
18-H* 15'L
Air Cteaner
22' HX IB' LX11" W
aur
r-a*
7-a*
Figure B-2b. Model Dimensions (Side View)
Uir-We-U-
* i"*!*
ir -
s
T-
1
/B=
V^-
/8"
J'6'
3'fi*
-------
No active HVAC was present, exchanging air inside the
room with another compartment or the outside ambience.
Therefore, flow rates were assigned to only the air cleaner
and the nebulizer, which was assumed to be on for the first
10 minutes of the total 30-minute simulation. The total
flow rate assigned to the air cleaner intake was set equal
to the sum of the air cleaner exhaust flow rate and the flow
rate assigned to the nebulizer (when in operation). The air
cleaner was assumed to have perfect efficiency in removing
air contaminant, so the intake could be modeled as a domain
flow outlet and the exhaust could be represented by an inflow
boundary condition.
There are a number of options within FLUENT for specifying
the outflow from the air cleaner exhaust and nebulizer
orifice as well as the flow rate for the air cleaner intake. The
boundary condition specifications that consistently gave
the best computational results in terms of mass balance and
stability of the solution algorithms are as follows:
Mass outflow from the air cleaner exhaust
corresponding to a flow rate of 450 SCFM (0.212 mVs)
Pressure outlet condition for the air cleaner with a
specified target mass flow rate corresponding to 450
SCFM (0.212 mVs) with the additional flow attributed
to the nebulizer when it is in operation
Mass outflow from the nebulizer (when in operation)
with a specified mass fraction of solids (KC1) content
3. Solver Control Specification - Select for turbulence
model, conservation equation closure methods, primitive
variable relaxation factors, and solution residual criterion
for advancement to the next time step.
A crucial element of the overall success of the model
depended upon the choice for turbulence model. Exploratory
calculations showed that the flow regime within the room
covered all three major regimes, namely laminar, transitional,
and fully turbulent. Application of a single turbulence model
therefore yielded poor results for both the flow structure as
well as the simulant transport. After some trial-and-error
application of FLUENT, it was determined that the detached
eddy simulation (DBS) model gave the physically most
realistic results and also preserved numerical stability best for
the transient-state calculations. The DBS model is a hybrid
of the large eddy simulation (LES) and Spallart-Almarus
(S-A) models of turbulence. The objective of this type of
model is to use LES in the "far field" regions away from flow
inlets/outlets and domain boundaries, where the unsteady
turbulent motions are directly computed and the smaller scale
motions are approximated coupled with S-A near boundaries,
a Reynolds-Averaged Navier-Stokes (RANS) one-equation
version for kinematic eddy viscosity. Essentially LES is most
applicable where the large-scale structure of turbulence is
most prevalent while the RANS solver is most applicable to
the wall-bounded, small-scale turbulent flow where viscous
effects dominate the flow development.
4. Flow Solution - Establish a fully conjugate pseudo
steady-state solution followed by a species conservation
solution during the injection and cleaning phases of the
simulation.
Prior to the transient simulation of the experiment, a steady-
state flow solution was obtained to establish the initial
conditions under which the test was performed. The steady-
state flow field consisted of passage of air through both the
air cleaner and nebulizer. When the transient phase of the
analysis was conducted, the outlet stream of the nebulizer
was replaced with the actual simulant composition. During
the process of obtaining the steady-state solution, the base
mesh was further refined based on adaptation on velocity
magnitude gradient to improve both the fidelity of the
solution as well as the convergence properties of the solution.
The final mesh consisted of a total of over 1 million cells (see
Figures B-3 and B-4).
The steady-state velocity vector flow field is illustrated in
Figure B-5. The smaller-scale structure in the flow field
is evident from these results. This small-scale structure
enhances diffusion mixing in addition to mixing by
advection, which in turn promotes more uniform mixing
globally. The pathlines of tracer particles released from the
nebulizer outlet are given in Figure B-6. Note the persistent
tendency for particles to initially travel to Filters 1 and 5.
This particle streaming will be reflected in the final results
as correspondingly higher aggregate mass recordings as
compared to the other filters.
Finally, before initiating the species transport analysis, a
separate analysis was conducted to estimate the diffusion
coefficient. Turbulent diffusion will dominate in the fully
turbulent portion of the flow and advection will dominate in
the laminar flow regions; nonetheless, molecular diffusion
will play an important role in transport for the transitional
flow regime. The transitional flow (roughly on the order
of 0.1 to 0.25 m/s in velocity magnitude for this problem)
comprises a significant portion of the entire flow spectrum,
as can be seen in Figure B-5, and should not be neglected.
The value calculated for the diffusion coefficient from
the Chapman-Enskog equation agreed very well with the
diffusion coefficient generated from the kinetic model
in FLUENT. For simplicity, water was chosen as the
surrogate for the solids species in the simulation injection
stream because its diffusion parameters are much better
characterized. Since water was the solvent for the KC1
simulant, and the correct mass fraction was used, little error
is expected to be incurred due to this simplification.
5. Monitor time-dependent concentration predictions for the
Climet and filters as well as a time-integrated collected
mass estimates for the filters.
The simulant concentration was recorded at the end of
each time step for the locations in the model domain
corresponding the Climet sample and Filters 1 through 5. The
concentration data were then integrated with respect to time
to get the cumulative collected mass on each of the filters.
-------
Figure B-3. Model Mesh with Outlines of Face Cells Illustrated
Nebulizer
Air Cleaner
Air Cleaner Intake
Grid (Time=6.0000e+02)
Jun21,2005
FLUENT 6.2 (3d, segregated, spe, DES, unsteady)
Figure B-4. Cells From the Base Mesh Marked for Adaptive Refinement
-------
Figure B-5. Velocity Vector Flow Field
'
1.02e+01
9.77e+00
9.46e+00
9.16e+00
8.85e+00
8.55e+00
8.24e+00
7.94e+00
7.63e+00
7.33e+00
7.02e+00
6.72e+00
6.41 e+00
6.11e+00
5.80e+00
5.50e+00
5.19e+00
4.89e+00
4.58e+00
4.27e+00
3.97e+00
3.66e+00
3.36e+00
3.05e+00
2.75e+00
2.44e+00
2.14e+00
1.83e+00
1.53e+00
1.22e+00
9.17-01
6.12e-01
3.06e-01
1.268-03
Velocity Vectors Colored By Velocity Magnitude (m/s) (Time=7.3011e+01)
Jul 25, 2005
FLUENT 6.2 (3d, segregated, DES, unsteady)
Figure B-6. Updated Pathlines (Colored by Total Residence Time)
"
1.43e+02
1.37e+02
1.33e+02
1.29e+02
1.24e+02
1.20e+02
1.16e+02
1.11e+02
1.07e+02
1.03e+02
9.866+01
9.436+01
9.006+01
8.576+01
8.156+01
7.726+01
7.296+01
6.866+01
6.436+01
6.006+01
5.576+01
5.146+01
4.726+01
4.296+01
3.866+01
3.436+01
3.006+01
2.576+01
2.146+01
1.716+01
1.296+01
8.576+00
4.296+00
O.OOe+00
Filters
Filter 2
o
Filters
Filter 1
Path Lines Colored by time (s) (Time=7.5101e+01)
Jul 25, 2005
FLUENT 6.2 (3d, segregated, spe, DES, unsteady)
-------
B.4. Results
In order to accelerate the computations, the flow field
established in the steady-state solution was held constant and
the solution proceeded by iterating on the species balance
equations for the water and solids content of the injected
simulant. The simulation began with a 600-second phase
during which the nebulizer was continuously injecting
simulant into the room air space at a constant flow rate,
followed by a 20-minute period during which the nebulizer
was turned off and the air cleaner operated at 450 SCFM
(0.212 m3/s), assuming 100% removal efficiency.
The results of the species concentration predictions are
graphically presented in Figure B-7 (half-plane passing
through the nebulizer) and Figure B-8 (planes passing
through the two filter elevations) at the very end of the
600-second period just prior to turning off the nebulizer.
The concentration field 20 minutes later with the nebulizer
turned off and the air cleaner continuously running is
provided in Figures B-9 and B-10 (nebulizer half-plane and
filter elevation planes, respectively). Note that in the time
interval following shut-down of the nebulizer and continued
operation of the air cleaner, the concentration field of
simulant becomes more homogeneous.
The predicted concentration of simulant as a function of
time for the Climet and the five filter locations are given in
Figure B-ll. From the data presented in Figure B-11, the
decay constants predicted from the model were computed
and compared to experimentally observed values as well as
to the value derived from an analytical model assuming
perfect mixing within the room. Table B-l provides a
comparison of the experimental, model, and well-mixed
approximation results.
Figure B-7. KCI Concentration With Nebulizer at 10 Minutes of Operation
'
3.00e-01
2.88e-01
2.806-01
2.716-01
2.626-01
2.546-01
2.456-01
2.366-01
2.286-01
2.196-01
2.106-01
2.016-01
1.93e-01
1.84e-01
1.75e-01
1.67e-01
1.58e-01
1.49e-01
1.416-01
1.32e-01
1.23e-01
1.146-01
1.06e-01
9.706-02
8.836-02
7.966-02
7.096-02
6.226-02
5.356-02
4.486-02
3.616-02
2.746-02
1.87e-02
1 .OOe-02
Filter 1
Filters
Filter 4
o
solids airborne concentration (mg/m"3)
Contours of kc1-concentration (Time=6.7500e+02)
Jul 25, 2005
FLUENT 6.2 (3d, segregated, spe, DES, unsteady)
-------
Figure B-8. KCI Concentration With Nebulizer at 10 Minutes of Operation (Cont'd)
'
3.00e-01
2.88e-01
2.80e-01
2.71e-01
2.62e-01
2.546-01
2.456-01
2.366-01
2.286-01
2.196-01
2.106-01
2.016-01
1.93e-01
1.84e-01
1.75e-01
1.67e-01
1.58e-01
1.49e-01
1.416-01
1.32e-01
1.23e-01
1.146-01
1.06e-01
9.706-02
8.836-02
7.966-02
7.096-02
6.226-02
5.356-02
4.486-02
3.616-02
2.746-02
1.87e-02
1 .OOe-02
Filter 2
Filter 4
solids airborne concentration (mg/m"3)
Contours of kc1-concentration (Time=6.7500e+02)
Jul 25, 2005
FLUENT 6.2 (3d, segregated, spe, DES, unsteady)
Figure B-9. KCI Concentration after 20 Minutes of Air Cleaner Operation (No Nebulizer)
I
9.006-04
8.646-04
8.376-04
8.106-04
7.836-04
7.566-04
7.296-04
7.026-04
6.756-04
6.486-04
6.216-04
5.946-04
5.676-04
5.406-04
5.136-04
4.866-04
4.596-04
4.336-04
4.066-04
3.796-04
3.526-04
3.256-04
2.986-04
2.716-04
2.446-04
2.176-04
1.90e-04
1.63e-04
1.36e-04
1.09e-04
8.196-05
5.496-05
2.806-05
1 .OOe-06
Filter 1
Filters
Filter 4
o
solids airborne concentration (mg/m"3)
Contours of kc1-concentration (Time=1.3140e+03)
Jul 25, 2005
FLUENT 6.2 (3d, segregated, spe, DES, unsteady)
-------
Figure B-10. KCI Concentration after 20 Minutes of Air Cleaner Operation (No Nebulizer)
"
9.00e
8.64e
8.37e
8.1 Oe-
7.83e
7.56e
7.29e
7.026'
6.75e
6.48e
6.21e
5.94e
5.67e
5.40e
5.13e
4.86e
4.59e
4.33e
4.06e
3.79e
3.52e
3.25e
2.98e
2.716'
2.446'
2.17e.
1.90e
1.63e
1.36e
1.09e
8.19e
5.49e
2.80e
1 .OOe.
Filter 2
Filter 4
solids airborne concentration (mg/m"3)
Contours of kd-concentration (Time=1.3140e+03)
Jul 25, 2005
FLUENT 6.2 (3d, segregated, spe, DES, unsteady)
The collected mass predicted for each filter was computed
and compared to the collected data samples from the
experiment in Figure B-12. Note that there are two sets of
data compared to the experimental results: the unaltered
results from the CFD predictions and the results from CFD
renormalized to yield the same average collected mass as
reported in the experiment (6X10"4 g). The normalization
process is somewhat justified by the fact that (1) there are
some uncertainties in the stated flow rate of solids from the
nebulizer, (2) the species equation was completely decoupled,
and (3) without a comprehensive turbulence model for the
class of flow problem, a trends analysis is the most valuable
contribution from this type of simulation. A numerical
comparison of the experimental to normalized collected mass
values is given in Table B-2. The only large deviation is
recorded for Filter 5, located near the Climet.
Table B-l. Comparison of Decay Constants for Experiment versus Model Predictions.
Average Decay
Constant, k (1/min)
Experiment
0.289
Model
0.535
Well -Mixed
Assumption
0.501
Table B-2. Comparison of Experimental to Predicted Values for Collected Mass Values.
Filter 1
Filter 2
Filter 3
Filter 4
Filter 5
Average
Experimental (g)
8.20E-04
7.00E-04
5.50E-04
5.25E-04
4.10E-04
6.01E-04
Model (g)
1.20E-03
8.10E-04
6.87E-04
7.92E-04
1.03E-03
9.03E-04
Model
Normalized (g)
7.98E-04
5.39E-04
4.57E-04
5.27E-04
6.84E-04
6.01E-04
% Deviation
Experimental vs.
Normalized (%)
o
3
23
17
0
-67
0
-------
Figure B-ll. KCI Concentration versus Time
.U1 4
.012
co~
£ 0.01
3
E 0.008 ~
0
nj
i 0.006
E
0)
o
Q 0.004
O
.002
0 "
(
CFD Model Results, Room Concentration vs. Time
---- i
i
/' i
|
! /£c*-'- -:"~-"~\
/ J/^ \
?# \
1 \^_
) 200 400 600 800 1000 1200 1400 1600 18
Time (s)
./iimet '-liter 1 r liter L r liter o r liter 4 r liter o
00
Figure B-12. Collected Mass of KCI in Filters
14Q&C3 i
2 1 OQE-03 4-
* 6 00&04 4-
= 4 DC1F.Q4 - -
O
P.OO&04 4-
CFD Filter Masses
Filter 3 Filter 4
Filter Number
u .-^
/'.-L :,-:-rr j s: 1 to !_<: "^
-------
B.5. Conclusions and Recommendations
CFD simulation has been demonstrated here as a potentially
viable means of obtaining spatially and temporally highly
resolved estimates of the concentration field for an in-room
release scenario. The normalized predictions agreed well
with experimentally measured results ranging from 3% to
23% difference with the exception of one filter sample, which
recorded a 67% difference. The predictions also generally
demonstrated the correct trends in terms of which filters
recorded the greatest amount of accumulated mass (again
with the exception of one outlier filter).
The model exhibited a large difference in the average decay
constant as compared to the experimentally determined value
while comparing very favorably to the value obtained from
the well-mixed approximation. There could be a number of
reasons why the removal rate is more compatible with the
well-mixed theory than the experimental results:
Source rate of KC1 Simulant. There is the potential for
vaporization of water inside the nebulizer, resulting
in a lower actual emission rate of aerosolized KC1
in solution as compared to the assumed value in the
model. Recognition of this fact was one of the primary
motivations for renormalizing the model results to
reflect the average collected mass from the experiments.
However, renormalization will have relatively little
impact on the calculated value for the decay constant.
Uncertainty in the solids source rate would have a
significant influence on decay constant if the rate varied
appreciably with time.
Air Cleaner Flow Rate. The rated flow rate of the air
cleaner is 330 CFM, whereas the model assumed a
rate of 450 cfrn (0.212 mVs). Periodic measurement,
however, consistently indicated a flow rate in the
range of 425 to 450 cfm (0.201 to 0.212 mVs) in the
experiments. Nonetheless, if the flow rate varied
with time, or the actual cleaner collection efficiency
(assumed to be 100% in the model) was significantly
less at the high flow rate as compared to the rated flow
rate of the device, then significant variations between
predicted and measured concentration as a function of
time could be observed.
Limitations in the Turbulence Model. Only the large-
scale turbulence is explicitly calculated, whereas the
small-scale structure is modeled using an empirical
one-equation expression for closure. If the small-scale
turbulence is over-estimated with this high-Reynolds
number model, then predictions will reflect a higher
degree of mixing than actually takes place.
Decoupling the Species and Flow Field Solutions. It
is not immediately obvious what the magnitude of
the effect is by first solving the flow field and then
the species balance equations separately. Presumably
the largest impact would be the decoupling of the
species balance and the equation of turbulent viscosity.
However, this should result in a global change in
concentration level and therefore is expected to have
little impact on the average decay constant.
Due to the large computational expense incurred by this type
of modeling effort, CFD simulation of in-room contaminant
transport when the flow is transitional (neither laminar nor
fully turbulent) should be used judiciously. Currently, the
explicit computation and resolution of large-scale turbulence
can only be done time-dependent. Until such time as there are
good flow-averaging methods such as RANS for the Navier-
Stokes equations with closure expressions (otherwise known
as models) for turbulence that can accurately capture large-
scale turbulence in the transition region, problems of this
class will be computationally expensive to simulate.
Nonetheless, CFD offers a high degree of modeling flexibility
and as many monitor locations as desired can be specified
with negligible additional computational expense. As the
turbulence modeling capabilities improve, CFD will offer
the potential for generating spatially resolved simulations of
contaminant transport that cannot be replicated by lumped-
parameter or zonal models that assume homogeneously
mixed compartments. Therefore, application of CFD can be
a viable alternative to conducting experiments with either
complex and expensive HVAC requirements or a large
number of measurements with high record frequencies.
-------
-------
Appendix C
Quality Assurance
Work under this project was completed in accordance with
the EPA-approved quality assurance test plan (QAPP)
entitled "Research on Air Cleaning and HVAC Systems for
Protecting Buildings From Terrorist Attacks; Test/Quality
Assurance Plan for Task 4: Evaluation of In-Room Air
Cleaners." The QAPP describes the test procedures, and data
handling and analysis procedures. These procedures are also
described throughout this report. The QAPP also describes
the quality objectives and quality assurance (QA) procedures,
some of which are listed below:
Before performing the single-pass efficiency tests,
the concentration was confirmed to be uniform with a
coefficient of variation (CV) of 15% or less across the
air cleaner inlet.
Duplicate tests were performed with individual
measurements not deviating from the average by more
than 15%.
Filters were allowed to equilibrate in a humidity-
controlled room for 24 hours before weighing, blank
(untested) filters were carried with all sample filters,
and all filters were weighed at least three times.
As an example, the data from the concentration uniformity
measurement is shown in Table C-l. The total particle
concentration was measured with the scanning mobility
particle sizer (SMPS) three times each at nine different
locations. The sampling location was varied over the course
of the one-hour test period in order to determine both
temporal and spatial variability in concentration. As shown
in Table C-l, the temporal variability in concentration was
low, with a CV of less than 3% at all sampling locations. The
spatial variability was a bit higher, however, with an average
concentration of 19,700 particles per cm3 and a standard
deviation of 2,600 particles per cm3 leading to a CV of
13.2%. Additional QA calculations can be found throughout
the body of the report.
Table C-l. Concentration Uniformity: Average of Three
Measurements at Each Location
1.93E+04
CV-2.1%
2.21E+04
CV - 2.8%
2.39E+04
CV - 2.8%
1.75E+04
CV-1.0%
1.88E+04
CV - 0.8%
2.16E+04
CV - 2.9%
1.56E+04
CV-1.0%
1.78E+04
CV-0.9%
2.05E+04
CV-0.5%
As outlined in the QAPP, an internal QA audit of laboratory
procedures and data was performed by a Battelle QA officer.
The results of this audit were communicated to the Battelle
quality assurance manager and project manager as well as the
EPA project manager. No significant corrective actions were
required from this audit. At the completion of this project, all
quality objectives had been achieved.
-------
&EPA
United States
Environmental Protection
Agency
PRESORTED STANDARD
POSTAGES FEES PAID
EPA
PERMIT NO. G-35
Office of Research and Development
National Homeland Security Research Center
Cincinnati, OH 45268
Official Business
Penalty for Private Use
$300
Recycled/Recyclable
Printed with vegetable-based ink on
paper that contains a minimum of
50% post-consumer fiber content
processed chlorine free
------- |