Particle Bounce During Filtration of Particles on Wet and Dry Filters

Transcription

1 Aerosol Science and Technology ISSN: (Print) (Online) Journal homepage: Particle Bounce During Filtration of Particles on Wet and Dry Filters Benjamin J. Mullins, Igor E. Agranovski & Roger D. Braddock To cite this article: Benjamin J. Mullins, Igor E. Agranovski & Roger D. Braddock (2003) Particle Bounce During Filtration of Particles on Wet and Dry Filters, Aerosol Science and Technology, 37:7, , DOI: / To link to this article: Published online: 30 Nov Submit your article to this journal Article views: 463 View related articles Citing articles: 21 View citing articles Full Terms & Conditions of access and use can be found at Download by: [ ] Date: 09 January 2018, At: 21:21

2 Aerosol Science and Technology 37: (2003) c 2003 American Association for Aerosol Research Published by Taylor and Francis /03/$ DOI: / Particle Bounce During Filtration of Particles on Wet and Dry Filters Benjamin J. Mullins, Igor E. Agranovski, and Roger D. Braddock School of Environmental Engineering, Faculty of Environmental Sciences, Griffith University, Nathan, Australia This paper experimentally examines the bounce and immediate re-entrainment of liquid and solid monodisperse aerosols under a stable filtration regime (precake formation) by wet and dry fibrous filters. PSL and DEHS were the solid and liquid aerosols, respectively, used in four monodisperse sizes of 0.52, 0.83, 1.50, and 3.00 µm. Three different fibrous filters were used to filter the aerosol streams, and the efficiency of the filtration process for each aerosol type under dry and wet regimes was measured. It was found that the solid particles generally exhibited a lower fractional filtration efficiency than liquid particles, although this difference decreased in the smaller size fractions. The difference between solid and liquid efficiencies was found to be greatest in the 1.5 µm size range. As particle sizes of liquid/solid aerosols and filtration parameters were similar, this difference is most likely to be due to the effect of particle bounce and or immediate re-entrainment occurring inside the filter, with the greater efficiency of filtration of the liquid particles being due to their greater capacity to plastically/elastically deform in order to absorb the impact forces. However, for the wet filtration regime (each fibre of the filter was coated by a film of water), no significant difference in filtration efficiency was detectable between solid and liquid aerosols. Therefore, the conclusion can be drawn that the either the bounce effect of the particles is inhibited by the liquid film, or the filtration conditions in the wet filter are so different that the aerosol properties are less significant with respect to capture. INTRODUCTION Currently, most filtration models assume complete capture of all particles contacting a fiber; however, this is not necessarily the case. It is possible for particles to be re-entrained after capture, or strike the fiber but immediately bounce off again. Received 4 April 2002; accepted 22 January The authors would like to thank the staff of the QUTGU SEM Facility for producing the SEM images and Mr. Scott Byrnes for his laboratory assistance. Address correspondence to Ben Mullins, School of Environmental Engineering, Faculty of Environmental Sciences, Griffith University, Nathan QLD, Australia b.mullins@mailbox.gu.edu.au When a particle strikes a fiber, there is a complex system of plastic and elastic deformation as the fiber absorbs the force imparted by the particle. The principles which govern the mechanics of particle bounce are quite complex. Many of the previous attempts to quantify and describe particle bounce have focused on aerosols contacting flat plates rather than fibers. Gillespie and Rideal (1955) examined the impact and bounce of liquid (dioctylphthallate DOP) particles on glass slides and attempted to quantify bounce by examining the traces left on the impact slide and droplets which fell onto another slide. Other experiments have been conducted using more advanced techniques such as laser Doppler velocimetry to measure the bounce of polystyrene latex (PSL) particles on polished quartz and stainless steel surfaces (Dahneke 1973, 1975; Li et al. 1999). Techniques involving impaction of aerosols on flat plates to compare bounce properties have also been developed (Wang and Walter 1987; Xu and Willeke 1993). Other work has examined the effect of particle deformation on adhesion to a flat surface (Tsai et al. 1991; Rimai et al. 1994). If particles are able to deform to absorb the impact force they will be less likely to bounce. While such work is crucial for understanding of the general principles involved, and very useful for impactor design/research where the particles are collected on flat surfaces, it is not as useful in the fibrous filtration field. Since many of the above authors have emphasised that the impaction surface has a significant effect on particle bounce, then impact on fibers (which may be able to deform to absorb impact) will be different from that on a solid surface. Also, the flow field around a filter fiber will be completely different to the flow field across a flat plate. There are classically two approaches to describe particle bounce. The first defines a critical velocity V c, above which bounce will occur, of the form (Cheng and Yeh 1979; Brown 1993) V c = β = 1 ( ) 1 e 2 1/2 pl d p d p e 2 pl A, [1] πx 2 (6p pl ρ) 1/2 where β is a constant for a particular impaction surface, d p is the particle diameter, e pl is the coefficient of restitution (for 587

3 588 B. J. MULLINS ET AL. plastic deformation only), A is the Hamaker constant, p pl is the microscopic yield pressure, ρ is the particle density, and x is the separation distance of the particle mass center and surface. Hamaker constants are given in the literature for a limited number of elements and compounds, including PSL ( ergs) (Tsai et al. 1991). However values for liquid aerosols such as DEHS are not available. The other method involves the kinetic energy (KE b ) required for bounce to occur when a particle (d p ) collides with a surface (Dahneke 1971), KE b = d p A(1 e 2 ), [2] 2xe 2 where e is the coefficient of restitution (plastic and elastic deformation), which is equal to the ratio of the rebound velocity to the approach velocity. The value of e is reported to range from 0.73 to 0.81, although these values were typically derived using hard impaction surfaces such as glass and metal (Wall et al. 1990). A and e depend only on the material of the particle and the surface. It is reported that these constants must be determined experimentally, as it is very difficult to determine them theoretically. Some experimental results for irregularly shaped fly ash particles (mean diameter 0.14 µm) give the probability of bounce as (Ellenbecker et al. 1980) P b = (KE) 0.233, [3] where KE is expressed in Joules. These values, however will not be representative, since it is reported that spherical particles are much more likely to bounce than irregular particles (Brown 1993). One of the main factors governing adhesion of particles to fibers is the particle surface area, thus flakes are the most likely to adhere and least likely to bounce (Mullins et al. 1992). For liquid particles, the most significant force is the surface tension acting on the liquid droplet. As particle size diminishes, the surface tension force increases to the point where the particle is so rigid its behavior is identical to that of solid particles. Liquid particles are better able to absorb energy than solid particles, principally through deformation (Brown 1993). It is reported that liquid particles are far more likely to break up on impact than solid particles. This is principally through shear when the particle impact is not normal to the fiber (Gillespie and Rideal 1955). The continuing work of researchers such as Dahneke (1995) shows that there is still work to accomplish in describing simple aerosol/plate collision processes, which have yet to be adapted to complex processes inside air filters. To this end there has been very little work which has examined the practical effects of particle bounce in real filter systems, principally due to the complex particle and fiber interactions involved, making theoretical calculations difficult. Maus and Umhauer (1997) have, however, examined the fractional filtration efficiencies of biological aerosols and compared these to nonbiological aerosols with equivalent aerodynamic diameter, such as PSL and limestone dust. Bounce of the nonbiological aerosol was reported during filtration tests. There has been some research into the effect that coating filter fibers with a liquid has on particle adhesion/bounce. Walkenhorst (1974) examined the effect of coating model wire filters with vegetable oils, mineral oils, and vaseline, and reported that the former two substances increased filtration efficiency (particle adhesion) while the latter did not. Although oil-coated filters are suitable for laboratory scale processes, using such liquids in industry would not be advantageous. A process has been developed previously (Agranovski et al. 1999; Agranovski et al. 2001) in which the filter is coated with a thin layer of water, allowing collection of aerosols on the water film rather than directly on to the fiber. It has been stated (Agranovski and Braddock 1998) Figure 1. Experimental apparatus used to examine filtration efficiencies for liquid and solid particles on dry filters and filters coated with H 2 O.

4 PARTICLE BOUNCE DURING FILTRATION 589 that such liquid films coating the fibers will reduce the likelihood of particle bounce, although this has yet to be proven experimentally. These technologies are industrially applicable, as quantities of water for filter irrigation are more readily available and recyclable in industry than oils. A trial plant using such technology has been used for particulate removal in a steel pipe galvanizing plant with great success (Agranovski and Whitcombe 2001). The current project aims to examine the effect of particle bounce in fibrous filter systems by comparing the filtration efficiencies of a solid particle (PSL) with a liquid particle (diethylhexyl sebacate DEHS) of the same size and shape factor under identical filtration conditions. These two particles types have been chosen since they are readily obtained in precise monodisperse sizes, and they are both completely spherical in shape, although the density of DEHS is slightly less than that of PSL (910/1060 kg/m 3, respectively). The experiments were initially conducted in dry filter systems, then in wet systems to determine if particle bounce is reduced when water is used to coat the filter. METHODS Experimental Setup Figure 1 shows the experimental apparatus, which consisted of: a HEPA filter to remove all extraneous aerosols entering the system and thus ensure clean incoming air; flow meters to measure incoming air flow rates; a filter chamber; a condensation monodisperse aerosol generator (CMAG TSI, USA) and a collison nebulizer to generate the test aerosol; a process aerosol monitor (PAM TSI, USA) and Malvern Mastersizer (Malvern Instruments, UK) to ensure accuracy of test aerosols; and a Dust- Trak (TSI, USA), which was used to measure aerosol concentrations upstream and downstream of the test filter. The equipment was verified to be working before each experimental run, with no detectable aerosols passing the HEPA filter. The aerosol (either DEHS or PSL) was injected into the center of the air stream, downstream from the HEPA filter. The aerosols were generated using a CMAG (TSI, USA) and a three-jet collison-type nebuliser respectively. The cylindrical filter chamber held circular filters with a useable diameter of 14.5 cm. The pipework upstream and downstream of the filter contained isokinetic sampling points. The filters were aligned horizontally and clamped between two rings, with the airflow passing vertically upwards through the filter. The experimental chamber and all pipework was grounded, so as to minimize the effect of any electrical charges. Filters Three filters with a range of fiber size and packing density (Figures 2a 2c) were used for the experiments. The filters were selected as representatives of low, medium, and high efficiency types commonly used in industries such as those mentioned previously (Agranovski and Whitcombe 2001). The Scanning (a) (b) (c) Figure 2. SEM images of the three filters used. (a) SEM image of filter H (highest efficiency polyester). (b) SEM image of filter M (medium efficiency polypropylene). (c) SEM image of filter L (lowest efficiency polypropylene image not representative of packing density).

5 590 B. J. MULLINS ET AL. Table 1 Filter parameters Fiber size Thickness Packing density Filter (mean ± SD) (µm) (mm) (mean ± SD) (%) Material High (H) 16 ± ± 1.7 Polyester Med (M) 20 ± ± 1.3 Polypropylene Low (L) 60 ± ± 2.2 Polypropylene Note: All filters needle felts (filter H woven), all sintered one side. Electron Microscope (SEM) images show the structure of the filter, fiber orientation, and relative packing density. The parameters of the three filters are given in Table 1. It will be noted that the SEM image of filter L (Figure 2c) is not representative of the overall packing density. This image was of the outer extremity of the nonsintered side of the filter, where the packing density is lower than the average packing density. Parameters of the filters were determined analytically. Sample filters of known dimensions (10 for each type H, M, L) were weighed on an A&D (Japan) balance, accurate to five significant figures, to determine the overall density of the filter. Fiber sizes were determined by examining a number of randomly sampled fibers (100 per filter) using a Zeiss (Germany) Standard 25 polarizing microscope with a 10 objective lens and a graduated eyepiece. Acid digestion of 1 sample of each filter type was undertaken to determine the chemical composition of the fibers. Particles The PSL particles were obtained in 4 precisely monodisperse sizes in aqueous suspension. The sizes were 0.52, 0.83, 1.50, and 3.00 µm. Liquid DEHS particles were generated in aerodynamic sizes exactly corresponding to the above sizes using the CMAG. Although PSL sizes are geometric and DEHS were aerodynamic, since both particle types were completely spherical (and the density of PSL is very close to 1 g/cm 3 ), the difference will be insignificant. The PSL particle sizes and lack of agglomeration were verified using a microscope (Zeiss std. 25, Germany) with a graduated eyepiece, with an aqueous PSL sample placed on a slide. Table 2 shows the size distributions of the DEHS and PSL particles and their standard deviations. Procedures For the liquid aerosols, the sizes generated by the CMAG were verified to correspond to the PSL sizes using a Malvern Table 2 Particle sizes (NMD) and standard deviation (GSD) for the aerosols considered NMD ± GSD for each particle size class of each aerosol type used DEHS 0.52 ± ± ± ± 0.08 PSL 0.52 ± ± ± ± 0.04 Mastersizer prior to commencement of the measurement process (to ensure that the particle size/oven temperature calibration data supplied in the CMAG manual was accurate). During measurement, the input size and concentration of both liquid and solid aerosols were monitored continually using a PAM (TSI, USA). For the PSL, clean, dry compressed air was also injected to dry any water film on the particles, and the procedure for this was carefully developed to ensure that the particles would be completely dry before reaching the sampling points or the filter. This entailed adding compressed air until the aerosol size and number measured by the PAM did not change, then adding a small amount of surplus air. The system was operating at low relative humidity (<40%). In the first stage the filters were operated dry and aerosol was supplied by a CMAG to generate liquid DEHS particles or a collison-type nebulizer to aerosolize PSL particles from an aqueous suspension. Three different face velocities were used 0.57, 0.45, and 0.30 m/s for each particle size of liquid and solid particles, on each filter. The face velocities were chosen as typical of the range used for such filters in industry. Also, with lower velocities it is less likely for bounce to occur, and with higher velocities it is more likely that the water will be removed from the wet filter. The filters were only operated for short periods of time during experimental runs for each particle size (less than 5 min per aerosol size per filter for each of the three flow rates). The filters were replaced frequently (multiples of each filter were cut from the same sheet of filter material before experimentation commenced). For each particle size and flow velocity in each filtration regime a new filter was used (i.e., 12 identical filters of each type for the dry regime and the same for wet). The influent aerosol mass concentration was kept approximately the same (0.5 mg/m 3 ) for all particle sizes. This was necessary to ensure that the filter was not coated with a cake during the short sampling time (for industrial use such filters may receive hundreds of mg/m 3 ). The equipment thus attains a pseudo steady state, since time scales for clogging are large and thus can be neglected due to the short time and low aerosol concentration. The difference in pressure drop across each filter before and after the experiment was 2 Pa or less. To determine the filtration efficiency, the mass-based aerosol concentration was measured before and after the filter using isokinetic sampling points, with measurements taken using a

6 PARTICLE BOUNCE DURING FILTRATION 591 DustTrak (TSI, USA). The DustTrak is capable of measuring the total mass-based concentration of aerosols from µm with an accuracy of ± 0.1% of the measured value or ± mg/m 3, whichever is greater (TSI 2000). To ensure the accuracy of the DustTrak for the particle types, sizes, and concentrations used, a calibration was performed against the gravimetric method. The two measurements were in agreement to ± 5%. During measurement the upstream/downstream aerosol concentrations were read from the display of the DustTrak, using valves to control the source of the sample the DustTrak was measuring. The sampling lines from the sampling points were of equal length and kept as short as possible, so any line losses can be assumed to be equal. Upon commencement of each experimental measurement with the DustTrak the aerosol concentration to be measured stabilized within 10 s, at which point a measurement was taken. At least 3 runs were taken for each particle size/type/filter/velocity combination to ensure the accuracy of results. The efficiency of the filter E was calculated using the classical equation (Brown 1993), E = ( 1 C ) A 100, [4] C B where C B and C A are the mass-based particle concentrations before and after (upstream/downstream) the filter, respectively. The next stage of the research was the operation of the filters in a wet regime for the same sizes of solid and liquid aerosols. The horizontal filter was irrigated with distilled water in an amount sufficient to just coat the fibers. Due to the short sampling time it was not necessary to continually irrigate the filter. The wetting process was developed prior to the commencement of this stage of the experiments. The filter was irrigated with water and the fibers examined using a confocal microscope (Zeiss, Germany), and a polarizing micoscope (as previously) to ensure sufficient coverage. The filter was then operated at the maximum face velocity necessary for a period of time sufficient to (a) Figure 3. (a) Filter H measured efficiencies for solid (PSL) and liquid (DEHS) aerosols for the dry regime (conventional filtration). (b) Filter H measured efficiencies for solid (PSL) and liquid (DEHS) aerosols for the wet regime (H 2 O coating filter fibers). (Continued)

7 592 B. J. MULLINS ET AL. Figure 3. take aerosol measurements under normal circumstances. The filter was then removed and examined again to ensure that a sufficient coating remained on the fibers. The liquid on the fibers forms into droplets attached to the fiber (usually with a film between the droplets). The wet filter then behaves as a filter with thicker fibers. A simple approximation of this structure would be to consider a wet filter as a filter with fibers possessing a large standard deviation of fiber sizes. The same face velocities as for the dry regime were used (0.57, 0.45, 0.30 m/s). The upstream and downstream aerosol concentrations were measured as with the dry filters. The filter was initially operated in the wet regime before aerosol injection and the amount of aerosolized water measured (usually mg/m 3 or less). This was again verified after aerosol measurements and the average of the two values were subtracted from downstream aerosol measurements. Note that there were no detectable aerosols passing the HEPA filter (therefore upstream aerosol counts were always zero when the aerosol injection was not running). Thus any aerosol measured downstream from the filter during this time could only be aerosolized water (although the amount was negligible as is evident from the above value). (b) (Continued) RESULTS AND DISCUSSION Figures 3 5 show the filtration efficiency for filters H, M, and L. The a and b figures show the results for the dry and wet regimes, respectively, for all 3 face velocities. All data points on figures are the average of 3 or more values. Error bars are shown (giving the mean ± standard deviation (SD) for the data); however, the SD is usually quite low, therefore most error bars are not visible in the figures. The lines with the same symbol in each figure represent the DEHS and PSL efficiencies for the same velocity. The curves through the data points were fitted through the mean values using MS Excel. For the conventional dry filtration regime, it will be observed that there is a significant difference in filtration efficiency between the solid and liquid particles, which is greatest for the 1.5 µm size range. The difference between solid and liquid aerosols generally decreases with decreasing face velocity, as does the overall filtration efficiency for both aerosols. The overall efficiency generally decreases with decreasing particle size, due to the lessening effect of inertial capture forces. A decreasing difference (between solid and liquid efficiency) is evident for the 3.0 µm particles, which is unusual because

8 PARTICLE BOUNCE DURING FILTRATION 593 (a) (b) Figure 4. (a) Filter M measured efficiencies for solid (PSL) and liquid (DEHS) aerosols for the dry regime (conventional filtration). (b) Filter M measured efficiencies for solid (PSL) and liquid (DEHS) aerosols for the wet regime (H 2 O coating filter fibers).

9 594 B. J. MULLINS ET AL. (a) (b) Figure 5. (a) Filter L measured efficiencies for solid (PSL) and liquid (DEHS) aerosols for the dry regime (conventional filtration). (b) Filter L measured efficiencies for solid (PSL) and liquid (DEHS) aerosols for the wet regime (H 2 O coating filter fibers).

10 PARTICLE BOUNCE DURING FILTRATION 595 Figure 6. Comparison of single-fiber efficiency theory with experimental results for filter H at V = 0.57 m/s in the dry filtration regime. All other results showed a similar or better correlation with the theory and experiment. larger particles have greater inertia and are thus more likely to bounce. This, however, must be countered by the far greater number of fiber/particle collisions which could be predicted for a 3.0 µm particle over a 1.5 µm particle when traversing a filter. The number and nature of particle/fiber interactions in a filter are complex, and further research would be needed to determine the exact cause of this behavior. There is a visible difference between the means of the PSL/DEHS efficiencies for each particle size/velocity/filter combination, although the difference is not always significant. All means for the 1.5 µm size were significantly different (between DEHS and PSL for each particle size/velocity/filter combination), with decreasing numbers of significantly different means for the 0.83, 3.00, and 0.52 µm sizes, respectively. Although the error bars for most of the means in Figures 3a, 4a, and 5a are quite low, some of the error bars for the wet results are relatively large. This could possibly be accounted for by the fluid liquid coating moving during the experimental procedure, thus affecting efficiency. For the wet regime (Figures 3b, 4b, and 5b), it is observed that there is negligible difference between the solid and liquid Table 3 Comparison of differential bounce effect between solid and liquid aerosols for dry filter only 0.52 µm 0.83 µm 1.50 µm 3.00 µm V (m/s) Nr EX-PSL % Nr EX-PSL % Nr EX-PSL % Nr EX-PSL % H M L Refer to Equations (4) and (5) for method of calculation of parameters.

11 596 B. J. MULLINS ET AL. efficiencies, the slight difference most likely due to experimental error or the density difference between the substances. All differences between solid/liquid means are not significant. The efficiencies for the wet filters were far greater than those for the same filter operated in the dry regime. The results in Figures 3a, 4a, and 5a were compared to the classical single fiber efficiency theory calculations (Hinds 1999) for particles with the same size and density as those used in the experiments under the same conditions, and they were found to approximately correlate (usually to within ±15% or better) as would be expected. Since the single-fiber efficiency does not effectively account for the occurrence of particle bounce or some properties of particle type (e.g., solid/liquid, hardness, elasticity), the theory cannot be expected to give a completely (a) (b) Figure 7. Plot of kinetic energy versus efficiency for all filters, for all particle sizes and flow rates. (a) (c) show the results for the dry regime (H L, respectively), and figure (d) (f) show the results for the wet regime (H L, respectively). (Continued)

12 PARTICLE BOUNCE DURING FILTRATION 597 (c) (d) Figure 7. accurate correlation. However, Figure 6 shows the correlation between the single-fiber efficiency (calculated using the equations in Hinds (1999)) and the experimental results for filter H (at V = 0.57 m/s). Figure 6 is typical of the correlation between theory and experiment for the dry filters, with the theory not being applicable to wet filtration. For the dry filters only, Table 3 shows the additional number of PSL particles (compared to DEHS) of each size passing (Continued) through a particular filter (Nr EX-PSL ). The number results have been obtained using Nr EX-PSL = Nr PSL Nr DEHS, [5] Where Nr EX-PSL is the number of additional PSL particles passing the filter, Nr PSL and Nr DEHS are the total numbers of PSL and DEHS particles passing completely through the filter,

13 598 B. J. MULLINS ET AL. (e) (f) Figure 7. (Continued) respectively. The percentage results (%) are % = ( NrEX-PSL Nr DEHS ) 100. [6] These results show the differential efficiencies for solid and liquid aerosols in terms of the number of actual particles bouncing completely through the filter. All input particle masses were corrected to exactly 0.5 mg/nm 3 to obtain these results. The percent values are the percentage of extra PSL particles passing the filter compared with the total percentage of DEHS particles passing the filter. It will be noted that there is a general trend (especially in the larger size fractions) for the percentage of PSL particles bouncing to reduce with reducing face velocity. This would be expected, since the particle kinetic energy will also be decreasing as velocity decreases. Furthermore, a trend will

14 PARTICLE BOUNCE DURING FILTRATION 599 be noticed for the percentage bounce to reduce with reducing filter efficiency/type (H, M, L). This feature can be accounted for by the fact that this corresponds to a decrease in packing density, meaning that the average velocity inside the filter will be decreasing with filters of decreasing efficiency (for the same face velocity). The results for the dry regime clearly show that the efficiencies for DEHS are significantly greater than those for PSL, and since the fibers with which the aerosols are impacting are of the same type, this implies that the DEHS is better able to absorb the collision force than the PSL. This differential efficiency between the two particle types decreases with decreasing particle size and face velocity, which could be expected, since both these factors will reduce the kinetic energy which must be dissipated by the particle/fiber on contact. Figure 7 shows the particle size versus efficiency results converted to kinetic energy (KE) versus efficiency. These graphs combine the four particle sizes and three velocities from each filter into one curve for each filter for each particle type (for dry and wet regimes separately). For the dry regime Figures (7a 7c) it is evident that the liquid particles exhibit greater efficiency for the same KE. This difference decreases with decreasing KE as would be expected. For the wet regime Figures (7d 7f) there is no difference in efficiency between solid and liquid particles of the same KE. For the analysis of bounce of aerosols on filter fibers, Equations (1) and (2) are of little use at this stage, since the Hamaker constant (A) of liquids is not reported in the literature, or even if it is possible to determine A values for liquids, neither are e values given in the literature for substances such as polyester and polypropylene. In addition to this, e values for plates of the aforementioned polymers may differ from those of fibers of the same substances. Note that Equation (3) indicates that an increase in KE leads to an increase in probability of bounce by flakes. Note also that Equations (1) and (2) suggest minimum KE levels for bounce to occur and support the concept of increasing KE leading to greater probability of bounce. However, in Figure 7, increasing KE gives greater capture due to the increased inertial forces, thus a bounce probability equation cannot be determined. If it was possible to determine how many particles of each size were not contacting fibers, then it would be possible to determine the probability of bounce along the lines of Equation (3), if the filters were approximated to a uniform fiber spacing and orientation. For the wet filtration regime (Figures 3b, 4b, 5b, 7d 7f) the efficiencies do not significantly differ with changing particle type. Any discrepancy present is most likely due to experimental error or the expected lower efficiency of DEHS due to its lower particle density. This lack of, or greatly reduced bounce in, such wet filter systems is significant as it implies that the water film must act to inhibit bounce either by aiding energy dissipation or preventing the particle from being repelled from the fiber. The significantly improved efficiency is also an important factor of the wet filtration technology. It is possible that the greatly increased efficiency of the wet filtration process is masking the detection of aerosol bounce. Although this is possible, it is unlikely, since a general trend of the PSL particles to bounce more than the DEHS should be noticeable in the results (even if not large enough to be significant). CONCLUSION It is evident from the above work that the filtration efficiency in dry fibrous filters can be significantly altered by the ability of the particle to absorb the impact forces when contacting a fiber. Although PSL is reported to be a soft particle and thus able to deform more than other solid particles, the DEHS is evidently able to deform to a greater degree to limit the bounce/reentrainment effect. Therefore liquid DEHS particles must be better able to dissipate the energy involved in the impaction process than solid PSL particles. It has also been shown, however, that in wet filter systems this effect is completely removed or at least significantly reduced, most likely by the influence of the water coating in dissipating the impact energy. This points to the further applicability for the technology, not only due to its self-cleaning nature and high efficiency, but also for the ability to efficiently remove aerosols with advanced bounce properties. Future work examining the bounce of individual aerosols on filter fibers using a model filter at microscopic scale would be advantageous. This would allow adaptation of the theories developed for flat plates to be used to determine and predict the bounce properties of aerosols on filter fibers. REFERENCES Agranovski, I. E., and Braddock, R. D. (1998). Filtration of Mists on Wettable Fibrous Filters, American Institute of Chemical Engineering 4(12):2775. Agranovski, I. E., Braddock, R. D., and Myojo, T. (1999). Removal of Aerosols by Bubbling Through Porous Media, Aerosol Science and Technology 31(4): Agranovski, I. E., Braddock, R. D., and Myojo, T. (2001). Comparative Study of the Performance of Nine Filters Utilised in Filtration of Aerosols by Bubbling, Aerosol Science and Technology 35(4): Agranovski, I. E., and Whitcombe, J. M. (2001). Case Study of the Practical Use of Wettable Filters in the Removal of Sub-Micron Particles, Chem. Eng. Tech 24(5): Brown, R. C. (1993). Air Filtration: An Integrated Approach to the Theory and Applications of Fibrous Filters. Pergamon Press, Oxford. Cheng, Y. S., and Yeh, Y. C. (1979). Particle Bounce in Cascade Impactors, Environmental Science and Technology 13(11): Dahneke, B. (1971). The Capture of Aerosol Particles by Surfaces, Journal of Colloid and Interface Science 37: Dahneke, B. (1973). Measurements of the Bouncing of Small Latex Spheres, Journal of Colloid and Interface Science 45(3): Dahneke, B. (1975). Further Measurements of the Bouncing of Small Latex Spheres, Journal of Colloid and Interface Science 51(1): Dahneke, B. (1995). Particle Bounce or Capture Search for an Adequate Theory: I Conservation-of-Energy Model for a Simple Collision Process, Aerosol Science and Technology 23: Ellenbecker, M. J., Leith, D., and Price, J. M. (1980). Impaction and Particle Bounce at High Stokes Numbers, Journal of the Air Pollution Control Association 30:

15 600 B. J. MULLINS ET AL. Gillespie, T., and Rideal, E. (1955). On the Adhesion of Drops and Particles on Impact at Solid Surfaces, Journal of Colloid Science 10: Hinds, W. C. (1999). Aerosol Technology: Properties, Behaviour, and Measurement of Airborne Particles. John Wiley and Sons, New York. Li, X., Dunn, P. F., and Brach, R. M. (1999). Experimental and Numerical Studies on the Normal Impact of Microspheres with Surfaces, Journal of Aerosol Science 30(4): Maus, R., and Umhauer, H. (1997). Collection Efficiencies of Coarse and Fine Dust Filter Media for Airborne Biological Particles, Journal of Aerosol Science 28(3): Mullins, M. E., Michaels, L. P., Menon, V., Locke, B., and Ranade, M. B. (1992). Effect of Geometry on Particle Adhesion, Aerosol Science and Technology 17: Rimai, D. S., Demejo, L. P., and Bowen, R. C. (1994). Mechanics of Particle Adhesion, Journal of Adhesion Science and Technology 8(11): Tsai, C.-J., Pui, D. Y. H., and Liu, B. Y. H. (1991). Elastic Flattening and Particle Adhesion, Aerosol Science and Technology 15: TSI (2000). DustTrak Aerosol Monitor Operation and Service Manual. TSI Incorporated, St. Paul, MN. Walkenhorst, W. (1974). Investigations on the Degree of Adhesion of Dust Particles, Staub Reinhaltung der Luft 34: Wall, S., John, W., Wang, H.-C., and Goren, S. (1990). Measurements of Kinetic Energy Loss for Particles Impacting Surfaces, Aerosol Science and Technology 12: Wang, H.-C., and Walter, J. (1987). Comparative Bounce Properties of Particle Materials, Aerosol Science and Technology 7(3): Xu, M., and Willeke, K. (1993). Technique Development for Particle Bounce Monitoring of Unknown Aerosol Particles, Aerosol Science and Technology 18: