An experimental and modeling assessment of room air cleaners for building Protection.

ABSTRACT

An experimental study and model analysis were performed to assess the effectiveness of commercially available in-room air cleaners in minimizing the impact of a hazardous aerosol released in a building. Two air cleaners were evaluated: a HEPA-type air cleaner and an electrostatic precipitator.

The effectiveness of a stand-alone room air cleaner depends on three principal characteristics: single-pass filtration efficiency, air flow rate, and the airflow pattern that the cleaner induces in the room. Accordingly, the air cleaners were first experimentally evaluated for their single-pass filtration efficiencies, which were determined for aerosol particle ranging in diameter from 0.03 to 10 [micro]m under different airflow rates. A series of test chamber experiments was then performed in which concentration decay rates were determined from continuously monitoring aerosol concentration at a particular location in the room. The degree of variability in the cumulative exposure dosages occurring in the room was considered using filter samples taken at several different locations.

Additionally, model analyses were performed using computational fluid dynamics (CFD) calculations, which suggested that CFD methodology can offer a viable alternative to field testing. The model was reasonably effective at predicting trends in the relative variability of contaminant concentration in the test chamber, but overestimated the concentration decay rate. Challenges associated with performing CFD calculations for flow configurations encountered under such conditions are discussed.

As a general finding of this work, room air cleaners can offer a noticeable threat reduction potential by reducing both the peak concentration of aerosolized pollutant in the room and the overall level of occupants' exposure. Consequently, the room air cleaner can increase the time it takes to reach a particular level of threshold concentration in the room, or even prevent it. However, the extent of the threat reduction potential may depend greatly on many factors, including those related to the pathogen aerosolization scenario and room configuration.

INTRODUCTION

The purpose of this study was to perform experiments and mathematical modeling to investigate the effectiveness of commercially available room air cleaners in minimizing the impact of an indoor release of hazardous material. The effectiveness of a stand-alone room air cleaner depends on three principal factors. The first two factors, single-pass filtration efficiency and air flow rate, are design characteristics. The third factor relates to the mixing conditions established in the room, to which the air cleaner contributes by its own flow configuration. This factor is not easily quantifiable because it also depends on the aerosol release scenario occurred in the room. A room air cleaner will not be effective if a flow pattern that causes a high level of clean air recirculation from its outflow back into the inlet develops inside the room.

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). Under this program, room air cleaners are characterized using the clean air delivery rate (CADR), which determines how effectively they remove different particulate pollutants, such as tobacco smoke, dust, and pollen (AHAM 2005). The CADR purpose is to combine and quantify all of the air cleaner effectiveness factors for easy comparison between different units; it is determined in a standardized test chamber that is initially well mixed. This likely provides a good model for environmental contaminants, which are typically well dispersed, but a chemical or biological attack is more likely to come in the form of a point source.

EXPERIMENTAL STUDY

The experimental study consisted of two phases. In the first phase, two air cleaners were evaluated for their single-pass filtration efficiency, one based on the high-efficiency particulate absorbing (HEPA) filtration mechanism and one using electrostatic precipitation (ESP). In the second phase, the HEPA-based air cleaner was evaluated for its effectiveness under a set of in-room dissemination scenarios.

Single-Pass Efficiency

An independent evaluation of the single-pass filtration efficiency was performed for the two selected air cleaners using aerosol particles with diameters ranging from 0.03 to 10 [micro]m. This evaluation was conducted by placing the unit in a sealed flow duct and measuring the size-resolved aerosol concentration upstream and downstream of the unit with a particle counter. Flow in the duct was controlled to ensure the test apparatus did not alter the normal operating characteristics of the air cleaner when it operates in open air. Each test unit was characterized for both an inert KCl aerosol and a bioaerosol consisting of Bacillus globigii (Bg) spores.

The results of the single-pass efficiency measurements are plotted in Figure 1. These plots suggest a trend of decreasing collection efficiency with increasing flow rate for the electrostatic precipitator. This trend is expected because higher flow rates yield lower residence times of aerosol particles in the charging and deposition zones of the air cleaner. This unit also exhibited a minimum in the filtration efficiency for particles around 0.2 [micro]m. This minimum in electrostatic precipitation efficiency was also expected because of the nature of the two major charging mechanisms present in such devices (i.e., field charging and diffusion charging). Specifically, in the aerosol size range of 0.1 and 1 [micro]m, neither mechanism is efficient and, therefore, lower collection efficiency typically results (US EPA 2002). For example, Zukeran et al. (1999) found poor particle collection efficiencies for ultrafine particles (0.01-0.1[micro]m) in electrostatic precipitators. They proposed poor charging and flow instabilities as possible causes for this observation.

The filtration efficiency data obtained for the HEPA-based unitary cleaner do not suggest any clear trend in the flow rate. The data show that efficiency decreased with particle diameter smaller than 0.3 [micro]m. The removal efficiency of particles larger than 0.3 [micro]m in diameter, however, was found to approach HEPA specifications. The cause for the unexpected but systematic decrease of filtration efficiency for smaller particles is not well understood. Most HEPA-type filters are known to have a minimum in efficiency in this size range, but the efficiency typically increases for smaller particles. It is possible that this effect is simply due to some leaks associated with the relatively loose fitting of the filter media in the test unit.

In addition to testing the air cleaners with inert KCl aerosol, the units were also challenged with a bioaerosol consisting of Bg spores (having approximately 1 [micro]m aerodynamic diameter). Tests for both units displayed similar effectiveness in removing biological and inert aerosols with similar particle diameters.

In-Room Effectiveness

The HEPA-type air cleaner was tested for in-room effectiveness. These experiments were performed in a 900 [ft.sup.3] (25.5 [m.sup.3]) test chamber with no additional mixing or flow other than that introduced by the sampling equipment. To test the overall effectiveness of the air cleaner under specific operating conditions, a concentration decay profile was obtained by measuring the aerosol concentration with time at a fixed location after generating a fine KCl aerosol for a fixed time period. In addition, the degree of mixing in the chamber was assessed by collecting filter samples at several locations, which were used to determine the variability of exposure dosages in the chamber. These filter sample results were subsequently used as a measure of the mixing ability of the air cleaner.

[FIGURE 1 OMITTED]

The test matrix listed in Table 1 was designed to evaluate the effect of the location of the source relative to the air cleaner, flow rate, and flow obstructions on the removal effectiveness of the air cleaner. Figure 2 contains two diagrams representing the four test configurations. Test configurations A and D were essentially the same, except that in configuration D the air cleaner was not running and only the natural decay was measured. Test configurations B and C were also similar. The only differences were that a desk and chair were added to the test chamber and the low-flow setting was used for configuration C. Duplicate runs were performed for each Scenario A and C.

For comparison between the configurations, the concentration decay profiles obtained were fitted with an exponential decay correlation:

C = [C.sub.0][e.sup.-kt] (1)

where

C = aerosol concentration, particles/[ft.sup.3] (particles/[m.sup.3])

[C.sub.0] = initial aerosol concentration, particles/[ft.sup.3] (particles/[m.sup.3])

k = concentration decay rate constant, 1/min

t = time, min

[FIGURE 2 OMITTED]

The decay constant k can be defined as

k = [[Q.sub.e]/V] + [k.sub.n], (2)

where

[Q.sub.e] = effective air cleaning rate, [ft.sup.3]/min ([m.sup.3]/min);

V = volume of the test chamber, [ft.sup.3] ([m.sup.3]); and

[k.sub.n] = natural decay rate constant (determined from scenario D), 1/min.

It was assumed that concentration decay measurements performed at a fixed point were representative of the entire volume when the aerosol was not generated and an established flow pattern has developed in the room. In order to evaluate the degree of mixing the air cleaner provided in the test room, a mixing efficiency coefficient was introduced. As mentioned above, there are three main factors influencing the effectiveness of an air cleaner. Therefore, [Q.sub.e] was defined to include terms for each of these factors, as the following:

[Q.sub.e] = [Q.sub.f][[eta].sub.s][[eta].sub.m] (3)

where

[Q.sub.f] = the measured flow rate of the air cleaner, [ft.sup.3]/min ([m.sup.3]/min)

[[eta].sub.s] = fractional single-pass efficiency

[[eta].sub.m] = mixing efficiency coefficient

Combining Equations 2 and 3 and solving for the mixing efficiency coefficient results in the following equation:

[[eta].sub.m] = [V(k - [k.sub.n])]/[[Q.sub.f][[eta].sub.s]] (4)

Table 2 summarizes the results of the concentration decay measurements in the test room. The results from scenarios A and B indicate that, in this testing, the location of the air cleaner relative to the aerosol source did not affect air cleaner effectiveness. In scenarios B and C, as a result of increasing flow capacity of the air cleaner and introducing some flow obstructions in the room, lower values of the decay constant were obtained. The mixing conditions, however, somewhat improved, as indicated by a slightly larger value of the mixing coefficient. It should also be noted that these observations are specific to the test conditions used and should not be generalized for any test configurations.

Included in Table 2 is the mixing coefficient that was calculated from the AHAM dust-particle CADR value published for the test air cleaner (AHAM 2005). The CADR-based value of mixing coefficient was greater than that measured in this study. This difference is likely a result of premixing the aerosol in the AHAM test chamber and of the different chamber volumes used in the respective procedures.

Although an appreciable flow circulation can be induced by the air cleaner, pollutant concentration is always non-uniform in the room, in part because the air cleaner by itself is a source of concentration nonuniformity. Figure 3 contains a plot of the aerosol particle masses collected with each of the five filter samples located in the test chamber, as shown in Figure 2.

Samples collected on the filters over a time interval are indicative of the exposure dosage received in the room during the time. The results shown in Figure 3 suggest that while the air cleaner had relatively high airflow capacity for the size of the test room, aerosol concentration was not uniform. Therefore, even though a room air cleaner may lead to a significant reduction in the level of indoor exposure to agent, the dosage received may vary substantially for different room locations.

CFD MODEL ANALYSIS

In the final phase of this study, mathematical modeling of air cleaner effectiveness using computational fluid dynamics (CFD) was conducted to determine the accuracy and efficiency with which it could predict concentration evolution of airborne contaminant in a room. The motivation for the model analysis is:

1. Cost Reduction. Modeling, if properly applied, can serve as a cost-effective alternative to an expensive and complex experimental program for identifying the primary factors that control the effectiveness of various air-cleaning approaches to reducing airborne contaminant levels.

2. Flexibility. Modeling offers the potential for predicting concentration profiles on a spatial and temporal scale that may be technically infeasible in an experimental program.

[FIGURE 3 OMITTED]

3. Multi-Functionality. Once a CFD solution has been obtained for a particular room configuration and set of operating conditions, aerosol trajectory calculations can be run for different release scenarios, often without the need to repeat flow field calculations.

The potential benefits of modeling, however, depend on the level of accuracy provided in its predictions and the amount of time and resources required to obtain those predictions. This study will therefore present the quantitative comparison of a CFD simulation to one of the full-scale experiments described above, and discuss some of the modeling challenges encountered. The ultimate goal 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 effect of airborne contaminant in a building.

Model Setup and Assumptions

A single simulation was performed under this study using the computer code FLUENT (2004), a well-validated commercially available code for CFD calculations. 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 mirrored that of configuration A of the experimental study shown in Figure 2. The indoor air cleaner was treated as a stand-alone interior unit with specified dimensions and flow capacity; its aerosol filtration efficiency was set to unity.

The boundary conditions included the air cleaner outflow rate (450 [ft.sup.3]/min [12.7 [m.sup.3]/min]) and nebulizer outflow rate (including aerosol mass fraction and the time of operation), the sum of which was assigned to the air cleaner intake flow. No active HVAC equipment was considered, and the room was taken to be entirely sealed off from the outside environment.

A crucial factor in the successful prediction of contaminant transport is the choice of turbulence closure model. Exploratory calculations showed that the hydrodynamic flow regimes established in the room ranged from laminar to fully turbulent. As a result, application of a single turbulence model generally yielded poor results for both the flow structure and the simulant transport pattern.

The CFD code offered a number of different closure schemes for describing turbulent flows. The turbulence model that was found to provide the best representation of the room concentration field was the detached eddy simulation (DES) model, which is a hybrid of the large eddy simulation (LES) and Spallart-Almarus (SA) turbulence models. The objective of this type of model is to use LES in the "far field" regions, sufficiently far away from flow inlets/outlets and other boundaries, where the unsteady turbulent eddy motions are directly represented and the smaller-scale motions are approximated. The LES turbulence model is then coupled near the boundaries with SA, a one-equation model of turbulence using turbulent viscosity. Essentially, LES is most applicable where the large-scale structure of turbulence is most prevalent, whereas the SA model is most applicable in the case of wall-bounded, small-scale turbulent flows where viscous effects dominate the flow development. One disadvantage of the LES-based turbulence model, however, is that it must be run in the transient-flow manner and therefore may incur a high computational cost.

In this work, before the transient simulation of the test scenario, a steady-state flow solution using the k-[epsilon] turbulence model was obtained to establish the initial conditions throughout the room. Although at this point the flow field is not physically correct, it does provide a reasonable initialization data set from which to begin using the DES model. During the process of obtaining the flow solution, the mesh was further refined based on adaptation of velocity gradient to improve both the fidelity and convergence of the solution, yielding a total mesh size of approximately 1 million cells.

[FIGURE 4 OMITTED]

To accelerate the computations, the flow field obtained from the steady-state solution was held constant and calculations proceeded for the concentration evolution of the released species. Consistent with the experiment, the simulation began with a 10-minute phase where the nebulizer operated continuously, injecting aerosol into the room at a constant rate, followed by another 20-minute period where the nebulizer was turned off but the air cleaner continued running at a continuous flow rate of 450 [ft.sup.3]/min (12.7 [m.sup.3]/min) and 100% filtration efficiency.

As the solution iterated in time, the simulant concentration was recorded at the end of each time step for the locations in the model domain corresponding to the locations of the Climet monitor and filters 1 through 5. The concentration predictions were then integrated with respect to time to obtain the cumulative collected mass on each filter.

Results

The flow field was obtained by the means described above. The mixture of large eddy currents and smaller-scale structure in the flow field was evident from the solution obtained. This small-scale structure enhances diffusional mixing of the species, in addition to the mixing by advection. Pathlines illustrating the flow field are shown in Figure 4, which suggests some persistent tendency for the particles to initially travel to filters 1 and 5. This particle streaming is reflected in the final results in terms of correspondingly higher aggregate mass recordings compared to the other filters.

The results of the species concentration predictions are graphically presented in Figure 5 for planes passing through the two filter elevations at the end of the 10-minute time period, just before turning off the nebulizer. The concentration field 20 minutes later, with the nebulizer turned off and the air cleaner continuously running, is shown in Figure 6. Note that, in the time after shutdown of the nebulizer and continued operation of the air cleaner, the concentration field of simulant becomes more homogeneous.

Time-dependent concentrations of aerosol were calculated for each of the filter sampling locations, from which an average value of slightly higher than unity was calculated for the mixing coefficient [[eta].sub.m], defined in Equation 4 ([[eta].sub.m] = 1 corresponds to perfect mixing). A value of greater than unity indicates that CFD predictions were not sufficiently accurate, because it is unrealistic for an air cleaner to be receiving airflow from areas of high concentration, throughout the room and at all times.

The CFD-predicted and experimentally obtained filter masses are compared in Figure 7. Note that the results from CFD were normalized to yield the same average collected mass as reported in the experiment (6 x [10.sup.-4] g). The normalization factor was insignificant and needed because there was some variability in the stated aerosol generation rate. Nevertheless, correct trends were predicted with the CFD model calculations with respect to the nonhomogeneity of pollutant concentration in the room studied.

[FIGURE 5 OMITTED]

CONCLUSIONS

An experimental program and modeling study were conducted to examine the effectiveness of indoor air cleaners in minimizing the impact of a chemical or biological attack on a building. As a principal result of this work, room air cleaners can have a high threat reduction potential with respect to reducing both the peak concentration of aerosolized pollutant in the room and the overall level of occupants' exposure. The room air cleaner can increase the time it takes to reach a particular level of threshold concentration in the room, or even prevent it. However, the extent of the threat reduction potential may depend greatly on many factors, including those related to the pathogen aerosolization scenario and room configuration. To quantify these effects, a mixing efficiency coefficient was introduced.

The analysis suggests that, for the room configurations considered in this study, both the location of the air cleaner in the room relative to the aerosol source and the presence of flow obstructions have minimal effect on the air cleaner's effectiveness.

In the context of mixing, the agreement between CFD calculations and experimental data was found to be reasonably good, especially with respect to the trends of developing select areas in the room where higher concentrations and exposure levels can be expected, as well as the duration of time under which those conditions exist.

Several recent publications describe application of CFD to modeling of indoor air quality: Lee et al. (2002), Cheong et al. (2003), Beghein et al. (2005), and Bouilly et al. (2005). As turbulence modeling capabilities improve, CFD will offer the potential for generating spatially resolved simulations of contaminant transport that is difficult to represent by the lumped-parameter or zonal models that assume homogeneously mixed compartments. These promising results suggest 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. Major and still outstanding challenges to applying CFD to this type of analysis include the following:

[FIGURE 6 OMITTED]

[FIGURE 7 OMITTED]

* Limitations on the Turbulence Model. Only large-scale turbulence is explicitly calculated within DES, whereas the small-scale structure is modeled using the Spallart-Almarus (SA) empirical, one-equation expression for closure. The SA model is likely not ideal in this application, because it is most appropriate for very-high-Reynolds-number applications (e.g., flow past air-foils). If small-scale turbulence is overestimated with this high-Reynolds-number model, predictions will lead to 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 of first solving the flow field and then the species transport and diffusion equations separately. Presumably, the largest effect would be from decoupling the equation of turbulent viscosity.

* Mesh Quality. A hybrid mesh of unstructured hexahedral and tetrahedral elements was used for the computational cells. This approach was used to reveal any potential issues that may arise from a convergence or accuracy standpoint using unstructured meshes, recognizing that structured meshes are much more time consuming to construct. The main finding in this regard is that mesh refinement does have a significant effect on the numerical accuracy of the predictions when using an unstructured mesh.

REFERENCES

AHAM. 2005. Directory of Certified Room Air Cleaners, edition No. 3. Washington, DC: Association of Home Appliance Manufacturers.

Beghein, C., Y. Jiang, and Q.Y. Chen. 2005. Using large eddy simulation to study particle motions in a room. Indoor Air 15:281-90.

Bouilly, J., K. Limam, C. Beghein, and F. Allard. 2005. Effect of ventilation strategies on particle decay rates indoors: An experimental and modelling study. Atmospheric Environment 39:4885-92.

Cheong, K.W.D., E. Djunaedy, T.K. Poh, K.W. Tham, S.C. Sekhar, N.H. Wong, and M.B. Ullah. 2003. Measurements and computations of contaminant's distribution in an office environment. Building and Environment 38:135-45.

FLUENT. 2004. FLUENT[R] Computational Fluid Dynamics Package, Release 6.2.16 (three-dimensional, double-precision, segregated solver, laminar flow model with conjugate heat transfer), [C] 2004, Fluent Inc., Lebanon, NH.

Lee, E., C.E. Feigley, and J. Khan. 2002. An investigation of air inlet velocity in simulating the dispersion of indoor contaminants via computational fluid dynamics. Annals of Occupational Hygiene 46(8):701-12.

US EPA. 2005. Lesson 1: Electrostatic Precipitator Operation. APTI Virtual Classroom SI: 412B (2002). Available at http://yosemite.epa.gov/oaqps/EOGtrain.nsf/fabbfcfe2fc93dac85256afe00483cc4/ca9ae17f9567495885256b66004e7985/$FILE12bles1.pdf (9/20/2005). US Environmental Protection Agency.

Zukeran, A., P.C. Looy, A. Chakrabarti, A.A. Berezin, S. Jayaram, J.D. Cross, T. Ito, and J.-S. Chang. 1999. Collection efficiency of ultrafine particles by an electrostatic precipitator under DC and pulse operating modes. IEEE Transactions on Industry Applications 35(5):1184-90.

Vladimir Kogan

Christopher B. Harto

David J. Hesse

Leslie Sparks, PhD

Vladimir Kogan is a senior research scientist, Christopher B. Harto is a researcher, and David J. Hesse is a principal research scientist at Battelle Memorial Institute, Columbus, OH. Leslie Sparks is a senior chemical engineer at the U.S. Environmental Protection Agency's National Homeland Security Research Center, Research Triangle Park, NC.

Table 1. In-Room Test Configurations

Location of  Location of                Air Cleaner   Configuration
Source       Air Cleaner                Flow Setting  Code

Center       No air cleaner             Zero          D
             Near source                High          A
Near wall    Remote from source         High          B
             Remote from source (desk)  Low           C

Table 2. In-Room Decay and Mixing Results

             Flow Rate,                      Average k,  Average
Scenario     [ft.sup.3]/min ([m.sup.3]/min)  1/min       [[eta].sub.m]

A            435(12.3)                       0.289       0.587
B            435(12.3)                       0.287       0.583
C            275(7.8)                        0.189       0.601
D              0                             0.006       0.000
CADR (dust)  435(12.3)                       --          0.750