CHAPTER 2 LITERATURE REVIEW
2.2 Numerical model for predicting collection efficiency of ESPs
In ESPs, particles are collected by electrostatic forces, and thus the electric field strength and particle charge are very important factors influencing the collection efficiency of ESPs.
The conventional theory used to predict particle collection efficiency is based on the Deutsch-Anderson equation, which is shown as follows:
)
where VTE is particle migration velocity (m/s), A is the total collection area (m2), Q is the air flow rate (m/s), q is the number of elementary units of charge, e is the elementary electrical charge (1.6×10-19 C), Eave is the average electric field strength (V/m), Cc is the Cunningham slip correction factor, μair is the viscosity (kg/m-s), and dp is the particle diameter (m). In order to predict collection efficiency more accurately, Matt and Ohnfeldt (1964) and White (1982) presented a modified Deutsch-Anderson equation as follows:
c c
where VTE,k is the average migration velocity of particles (m), kc is a constant (0.4~0.6). The above semi-empirical equations can not predict collection efficiency precisely because the following assumptions were made: (1) Particles are assumed to reach their saturation charge during the collection process. (2) Particles migration velocity is not affected by air velocity and nonuniform distribution of the electric field. (3) The effects of particle reentrainment,
uneven gas flow distribution, and back corona on collection efficiency are not taken into consideration.
Many researchers developed numerical models based on charging model for continuum regime (Kn=2λi/dp1, λi : mean free path of ions, dp: particle diameter) to predict collection efficiency of the wire-in-plate dry ESPs for particles ranging from 0.3-10 μm in diameter.
Good agreement between experimental data and numerical results was obtained (Goo and Lee 1997; Park and Kim 2000, 2003; Park and Chun 2002; Talaie 2001, 2005). For example, a numerical model based on Lagrangian method with turbulent electrohydrodynamic (EHD) flow field was developed to predict particle collection efficiency and further validated with experimental data of Kihm (1987) (Goo and Lee 1997). The particle charging rate was calculated by using combined charging model of Fuchs model (1947) for diffusion charging and Pauthenier and Moreau-Hanot’s model (1932) for field charging (The combined charging model shown in section 3.2.4 was also used in the present study to predict the charge for particles with dp≧100 nm). The calculated results were found to be lower than the experimental data at the applied voltage of 6.5~13 kV for 4 μm particles. The deviation between the calculated values and the experimental data is probably due to the difficulty in estimating exact charging properties of particles, the inlet conditions of flow and particles, and the experimental error.
The effects of the EHD flow and turbulent diffusion on the collection efficiency of particles in a wire-cavity plate dry ESP was further studied by using a numerical model based on Eulerian method (Park and Kim, 2000; 2003). Particle charges were obtained by solving White’s equation (1967), which is given by
ave p d d
whitee d E
q 0 2
1
3
(2.4)
where κd is the dielectric constant of particles, ε0 is the permittivity of free air (F/m). Good agreement between experimental and computational data of collection efficiency was obtained for 4.5 μm Al2O3 particles when the applied voltage and air velocity were 7.5~15 kV and 0.5~1 m/s, respectively.
Park and Chun (2002) developed a numerical model to investigate the effect of turbulent dispersion on particle collection efficiency without considering the nonuniform distribution of electric field and ion concentration in a wire-in-plate dry ESP. The eddy diffusivity of a particle is given as follows:
h
Deddy p y , 0≦y≦h (2.5)
p
Deddy , h≦y≦sy (2.6)
h
p u 4 *
.
0
(2.7)
sy
h0.1 (Oron et al., 1988) (2.8)
where ξp is the turbulent dispersion coefficient within the turbulent core of the flow (m2/s), h is the distance from the wall to the interface between the region of a linearly increasing dispersion coefficient and that of a constant dispersion coefficient (m), sy is the wire to plate spacing (m), u* is the velocity distribution of a fully developed turbulent flow (m/s), which is defined as
u* 8f u 2 (2.9)
where uave is the average air velocity (m/s), f r is the friction factor, which is given by
f 0.0791Re0.25 (2800<Re<105) (2.10)
where Re is the Reynolds number based on the hydraulic diameter as follows:
calculating turbulent dispersion coefficient of particles was adopted in the present parametric study (in turbulence flow case). To calculate particle charges, Cochet equation (1961) was used, which is given by
ave
The predicted collection efficiency was validated with published experimental data of Riehle and Löffler (1993) and good agreement was found for limestone particles in the size range of 0.3~10 μm at an applied voltage of 25 kV, a mean air velocity of 1 m/s, and a turbulent dispersion coefficient of 50 cm2/s. The collection efficiency was found to increase with increasing turbulent dispersion coefficient for particles below 1 μm. For particles larger than 1μm, the increase of turbulent dispersion coefficient led to a decrease of collection efficiency due to the effect of dispersing the particles back to the flow. The author also noted that a complete model should necessarily include a realistic particle charging model, an electrostatic
field strength distribution, and a more realistic flow field.
Except for the effect of EHD and the turbulent dispersion coefficient, the effect of polydisperse particle loading on the particle collection efficiency was also taken into consideration in Talaie et al. (2001), who developed a mathematical model based on Eulerian method to predict the collection efficiency of a double-stage wire-in-plate dry ESP. The particle charging rate was calculated by using White’s equation (1967). The numerical values were in good agreement with the experimental data of Leonard (1982) for 3.5 μm oleic acid drops.
As discussed above, the numerical models for predicting collection efficiency for particles in the size range of 0.3-10 μm in wire-in-plate ESPs have been well-developed. For particles with 100≦dp≦200 nm, the numerical model of Yoo et al. (1997) based on Fuchs diffusion charging model (1947) and the field charging model of Pauthenier and Moreau-Hanot (1932) was able to predict the particle collection efficiency accurately.
However, the combined charging model used by Yoo et al. (1997) was found to over-predict particle charge for particles with dp≦100 nm in Lawless (1996), who concluded that 100 nm was the limit of applicability of the combined charging model.
In the transition regime (Kn≈1), Fuchs model (1963), which was used by Zhuang et al.
(2000) and Li and Christofides (2006) to predict particle charge in ESPs, was shown to be accurate for particles with dp>30 nm (Adachi et al. 1985; Pui et al. 1988). However, the numerical model could not predict the experimental collection efficiency accurately for 30<dp<400 nm because the flow field and the non-uniform ion concentration distribution were not calculated in Zhuang et al. (2000). In the work of Li and Christofides (2006), non-uniform electric field and ion concentration were not considered either and simulated particle collection efficiencies were not compared with experimental data. Therefore, the applicability of the previous models (Yoo et al. 1997; Zhaung et al. 2000; Li and Christofides 2006) for
For particles with dp<30 nm, experimental data (Huang and Chen 2002) and numerical results (Yoo et al. 1997; Zhaung et al. 2000; Li and Christofides 2006) showed that a fraction of particles was uncharged and penetrated through ESPs, resulting in decreasing collection efficiency as dp was decreased from 30 nm to 5 nm. This is called the partial charging effect.
Marlow and Brock’s model (1975) was shown to provide accurate prediction of particle charge for dp<30 nm (Pui et al. 1975). However, Marlow and Brock’s model has not been applied to examine the partial charging effect on the collection efficiency of the ESP. The combined charging model used in Yoo et al. (1997) over-predicted particle charge in the transition regime, as compared to the experimental data of Fjeld and MacFarland (1986), leading to an overestimation of collection efficiency for particles below 30 nm. The Fuchs model (1963) used in Zhuang et al. (2000) and Li and Christofides (2006) also over-predicted particle charge for dp<30 nm, which also led to an overestimation of the collection efficiency.
In the traditional dry ESPs, particle collection efficiency, especially for nanoparticle, decreases with increasing operation time due to particle deposition on discharge electrodes and collection electrodes (Huang and Chen 2003), back corona (Chang and Bai 1999), and particle re-entrainment (USEPA 2003). In order to solve typical problems associated with dry ESPs, wet ESPs were developed to control fine and nanoparticle effectively (Saiyasitpanich et al. 2006; Lin et al. 2010). Saiyasitpanich et al. (2006) compared measured and calculated collection efficiency by using Deutsch equation in a wire-in-tube wet ESP for particles in the size range of 20~700nm. The particle charging was calculated by solving Cochet equation (1961) and Robinson equation, which is given by
Robinson ln 1 2
2
is the ion number concentration (m-3), kb is the Boltzmann’s constant (N·m/K), and T is the temperature (K). The comparison revealed that the predicted values based on Deutsch equation underestimated the measured values.
Huang and Chen (2002) investigated aerosol penetration for particles in the size range of 7-100 nm by using a single-stage and a two-stage wire-in-plate dry ESP. A significant increase aerosol penetration was found for particles below 20 and 50 nm in the single- and two-stage dry ESP, respectively. Their experimental data could serve as a good benchmark for validating the simulation models for the nanoparticle collection efficiency in ESPs.
The above literature reviews for previous experimental and numerical studies were arranged in Table 2.1 and 2.2, respectively. In summary, according to the above discussions, an efficient wet ESP should be further designed and developed for controlling fine and nanosized particles. The optimal operation conditions for collecting nanoparticle in wet ESPs, and the collection efficiency of submicron particles under heavy particle loading conditions are needed to be investigated.
For predicting nanoparticle collection efficiency in ESPs by using numerical methods, the existing models can’t predict collection efficiency very well because the electric field and ion concentration distribution were not simulated, or charging models were not adopted appropriately to calculate particle charges. Thus, a numerical model using appropriate charging model to calculate particle charges requires to be developed for predicting nanoparticle collection efficiency accurately in ESPs.
Table 2.1 Literature reviews in experimental studies of wet ESPs.
Investigator ESP type Particle residence
Table 2.2 Literature reviews in numerical studies of particle collection efficiencies in wire-in-plate ESPs.
Researcher Numerical method
The calculated results were found to be lower than the
experimental data.
Good agreement between experimental and computational data of collection efficiency was
obtained.
Park and Chun Eulerian method
Cochet
equation 0.3-10μm
The calculated collection efficiencies match with the
experimental data.
The numerical values were in good agreement with the
experimental data.
Talaie (2005) Lagrangian method
The secondary flow and the high particle loading result in a decrease of collection efficiency.
Chang and Bai
The simulated values match with the experimental data.
Yoo et al.
The combined charging model is only valid for particles with dp
≧100 nm Lawless (1996).
Zhuang et al.
Ion concentration distribution was not simulated. Simulated collection efficiency didn’t match with experimental data.
Li and
Electric field and ion concentration distribution were
not simulated. Simulated collection efficiency wasn’t compared with experimental
data.
The comparison revealed that the predicted values based on
Deutsch equation underestimated the measured
values.
CHAPTER 3
METHODS 3.1 Experimental method
3.1.1 The present parallel-plate wet ESP
The collection electrodes of the present parallel plate wet ESP were designed based the parallel-plate wet denuder in Tsai et al. (2008). As shown in Figure 3.1, the wet ESP consists of two plexiglass plates (M) on which a copper plate (G) (100 mm in length, 75 mm in width and 3.0 mm in thickness) was attached to the inner surface. A copper plate was used as collection electrode because of its high conductivity and ease with sandblasting. In order to make scrubbing water film flow uniformly on the collection electrodes (G) at low flow rates, a frosted glass plate (FG, 70 mm in length, 75 mm in width and 3.0 mm in thickness) coated with TiO2 nanopowder (Degussa AEROXITE TiO2 P25, Anatase, 20 nm) is attached to the inner surface (M) above the copper plate. Between the collection electrodes, a center piece (C) is sandwiched to form a 9 mm gap between the electrodes. On the center piece, three gold discharge wires (GW) (99% purity, 100 μm in diameter, Surepure Chemetals Inc.) spaced at 16 mm in the flow direction are fixed. These gold wires were used as discharge wires due to their long lifetime of more than 6 months (Asbach 2004). Two overflowing (OR) and collection reservoirs (CR) for continuous scrubbing water flow are installed at the top and bottom of the ESP, respectively. A pulse jet valve (Bag Filter Enterprise Co. Ltd., Taiwan) was used to generate pulse jet passing through 3 rows of small holes (diameter: 1.0 mm, 24 holes on each row) on the collection plates. The discharge wires were cleaned every 5 minutes with an instantaneous pressure of 2.95 atm (3.05 kg/cm2) and the instantaneous air flow rate passing through all 24 holes per wire was calculated to be 11 L/s. Pulse duration was about 0.5 sec. During pulse jet cleaning, the scrubbing water flow was stopped for two minutes to prevent sparking over due to the water mist generated by the pulse jet.
LI LI
Figure 3.1 Schematic diagram of the parallel-plate wet ESP. Plexiglass plates (M), enter piece (C), frosted glass plate (FG), sand-blasted copper plate (G), overflowing reservoir (OR), collecting reservoir (CR), golden wire (GW), liquid inlet (LI), liquid outlet (LO), aerosol inlet (AI), aerosol outlet (AO), pulse jet valve (PJ), air hole (AH).
In order to increase the hydrophilicity of the collection plates, which will enhance the uniformity of the water film and reduce the scrubbing water flow rate, the copper plate surfaces were first sand blasted to an average depth of 61.0 μm and then coated with TiO2
nanopowder (Degussa P25) based on the method described in Tsai et al. (2008). The surfaces were first sonicated with DI water (18.2 MΩ-cm) and then dried by purging with compressed air. 0.5 g of TiO2 nanopowder was well dispersed in 50 ml DI water by ultrasonication and then applied onto the sand blasted surfaces. 30 min later, the excess solution was removed, and the plates were calcinated at 300 °C for 90 min to facilitate thermal bonding of the TiO2
nanopowder onto the copper plate surfaces.
Tap water was used as the scrubbing liquid for the collection electrodes. Two peristaltic pumps (Model MP-1000, Eyela Co., Japan) were used to continuously pump water into and out of the overflow and collection reservoirs. The appropriate scrubbing water flow rate, important to water film thickness and uniformity (Tsai et al. 2008), was determined in this study. The goal was to minimize the flow rate in order to decrease the water waste and film thickness (which decreases the electric field due to water’s resistivity), while also ensuring the film is uniform.
To determine the hydrophilicity of the coated surfaces, the morphology of the TiO2
nanopowder coated on the frosted copper plate was first determined by SEM, and then the water contact angle was measured by using a Contact Angle System (FTA125, First Ten Angstroms, VA). As shown in Figure 3.2, there is a porous surface in our case which is not the same as the water-repellent leaves with epicuticular wax crystals in combination with papillose epidermal observed by Barthlott and Neinhuis (1997). When the water contact angle was measured, we found the frosted surface coated with TiO2 nanopowder sucked in the water droplet forcing the droplet to spread out quickly. The contact angle was measured to be only 6.0±4.2°, which is very small compared to 104.03±1.73° and 117.17±2.83° for the smooth copper plates and uncoated sand-blasted copper plates, respectively, as shown in Figure 3.3. It
is certain that hydrophilicity can be achieved by this porous morphology to enhance the uniformity of the scrubbing water film.
Particle collection efficiency experiments were conducted at different aerosol flow rates and applied voltages under initially clean conditions to determine the optimum operation conditions. TiO2 nanopowder was then used to create heavy loading conditions for comparing the collection efficiency of the dry and present wet ESP.
(a)
(b)
Figure 3.2 SEM image of micromorphological characteristic of the frosted copper plate coated with TiO2 nanopowder. (a) X 70,000, (b) X 100,000.
(a)
(b)
(c)
Figure 3.3 Water contact angle on three copper plate surfaces. (a) smooth surface, (b) sand-blasted surface, and (c) sand-blasted surface coated with TiO2 nanopowder.
3.1.2 Particle collection efficiency experiment and loading test
The particle collection efficiency experiments under initially clean conditions were conducted at aerosol flow rates of 5 and 10 L/min (corresponding to residence times of 0.39 and 0.19 seconds, respectively, over the total precipitation length of 48 mm, which is assumed to be three times the wire spacing) with applied voltages ranging from 3.8 to 4.3 kV. The experiment under each operation condition was tested for 6 to 8 hours. The experimental setup is shown in Figure 3.4. Polydisperse liquid corn oil particles (density of 0.9 g/cm3) and silver particles (density of 10.49 g/cm3) were generated by the evaporation-condensation technique (Scheibel et al. 1982) at an oven temperature of 260 and 1050 °C, respectively. Two rotameters were used to control carrier gas and cooling air flow rates to obtain the desired particle diameter and concentration. Before being introduced into the present wet ESP, all charged particles were removed using a wire-in-tube ESP allowing only zero-charged particles to enter the wet ESP. To generate an electric field and corona ions, high positive voltage was supplied to the corona wires using a high voltage D.C. power supplier (Model SL150, Spellman High Voltage Electronics Corporation, NY, USA). The voltage and corresponding corona current were read directly from the power supplier. A scanning mobility particle sizer (SMPS, model 3081, TSI Inc.) coupled with a condensation particle counter (CPC, model 3022, TSI Inc.) was used to measure the size distribution of the polydisperse particles. The characteristics and number size distribution of the test aerosols are shown in Table 3.1, respectively. The particle collection efficiency of the wet ESP, ηtotal(dp), was calculated by the following equation:
%
where Cin(dp)is the particle inlet concentration (cm-3) and Cout(dp) is the outlet concentration (cm-3) for particle of diameter dp.
The particle collection efficiency experiment for monodisperse NaCl particle (particle density=2200 kg/m3) was also conducted in this study. The experimental setup is shown in Figure 3.5. Test particles were generated by the evaporation-condensation technique in an oven temperature at temperature of 650 to 700 ℃. The generated particles were then passed through a nano-DMA (differential mobility analyzer, TSI Model 3085) to obtain monodisperse particles with dp of 10 and 50 nm. The experiment was conducted at an aerosol flow rate of 5 L/min and at an applied voltage of +3.6~+4.3 kV. A condensation particle counter (CPC, model 3022, TSI Inc.) was used to measure particle number concentration. For 10 and 50 nm particles, the inlet concentration of the ESP was measured to be 1.2×109 (m-3) and 6.6×109 (m-3), respectively.
To study particle loading effect on the collection efficiency and possible particle reentrainment after particle loading, Degussa P25 TiO2 nanopowder (resistivity of 0.75×106 Ω·cm) (Salah et al. 2004) was dispersed by a Jet-O-Mizer (Model 000, Fluid Energy Processing and Equipment Co., Hatfield, UK) at an aerosol flow rate of 5.67~56.62 L/min corresponding to a mass loading rate of 0.1-100 g/hr as shown in Figure 3.4. The particles that penetrated through the ESP (Wout, g) and the particles in the excess flow (Wex, g) were measured by weighing the filters at the outlet of the wet ESP and the excess flow. After loading for two hours, the quantity of particles in the ESP per plate, Wloaded (g), was then calculated to be 1.2±0.06 g/plate by the following equation:
2
excess outlet
total loaded
W W
W W
(3.2)
where Wtotal (g) is the total particle dispersed by the Jet-O-Mizer. In order to examine the collection efficiency of the present wet ESP for nanosized particles after 0.5~2 hr of heavy TiO2 particle loading, corn oil particles were used instead of TiO2 nanopowders (P25) because
very few particles below 100 nm were generated by standard dispersive technique as found in Tsai et al. (2009). A small-scale powder disperser (SSPD, model 3433, TSI Inc.) was also used to disperse the TiO2 nanopowder for conducting the loading test (size distribution is shown in Table 3.1). However, the mass concentration was too low to generate the desired heavy loading condition. Therefore, corn oil particles generated from the evaporation-condensation technique were used to test the collection efficiency after heavy TiO2 particle loading.
To examine possible redispersion of particles into gas stream after heavy TiO2
nanoparticle loading in the dry ESP, a particle reentrainment test was conducted. After two hours of loading, clean air at a flow rate of 5 L/min was introduced into the dry ESP when the applied voltage was 4.3 kV. The scanning mobility particle sizer (SMPS, model 3081, TSI Inc.) was used to measure the reentrained particle size distributions at 135 second sampling interval.
Table 3.1 Characteristics of the test particles
Size range (nm)
CMD (nm) GSD (nm)
Total number concentration
(#/cm3) Corn oil,
for efficiency test
16.8~615 103.18 1.64 2.38×105
Silver, for efficiency
test
5.23~107.5 12.9 1.53 7.51×106
TiO2 *, for creating heavy particle condition
16.8~615 232.75 2.22 2.14×104
* Generated by a small-scale powder disperser (SSPD, model 3433, TSI Inc.)
Controller HEPA
(Polydisperse Aerosol)
ESP HV
Excess air filter
High voltage power supplier Peristaltic
pump Dry
compressed air
Rotameter Tube Furnace
High pressure air
Rotameter
Jet-O-Mizer (for particle loading only )
SMPS
CPC
Water tank
Wet ESP
Wet ESP