High Efficiency Axial Wet Cyclone Air Sampler

Early detection of highly contagious viral diseases is particularly important for controlling and minimizing the spread of these viruses. Moreover, the sampling of airborne viruses is a crucial step towards an effective epidemic diagnosis. However, most air samplers require long sampling times, and real-time virus detection is consequently not possible. Therefore, in this study, an axial-type wet cyclone capable of obtaining highly concentrated samples in an outdoor environment in real time was developed. In the newly designed axial wet cyclone, the airborne particles are firstly collected by centrifugal force in cyclone vanes rotating several times and then captured by inertial force before the air enters the outlet. The collection efficiency at 500 L min was tested with polystyrene latex (PSL), and it was observed that the air sampler efficiency for 1.1 μm and 2 μm PSL was approximately 50% and 90%, respectively. For the large-sized PSL (> 3 μm), the observed collection efficiency was almost 100%. The Stokes number was used to design geometrically similar cyclones with air sampling rates of 100 L min and 300 L min. The results of the scaled sampler show a somewhat lower collection efficiency than for the 500 L min device but with lower liquid consumption and a lower pressure drop. For 100 L min and 300 L min samplers, the cutoff point for collection efficiency was approximately 1.5 μm and 3 μm PSL, respectively.


INTRODUCTION
Cyclonic separation is used for various purposes such as collecting bio-aerosols and indoor and outdoor particles of a certain size.Cyclone usage is increasing because they are geometrically simple, require relatively low-level operating power, and can be operated in harsh conditions.Inertial impaction collectors, either cyclonic or impactor, can experience the re-inflow of particles or overloading.Such problems, however, occur less frequently with cyclones compared to impactors (Hsiao et al., 2011;Seon and Ahn, 2016).
Typically, cyclones are divided based on their airflow flowing either tangentially or axially.A general cyclone has air inflowing tangentially, generating a rotational force using a vortex finder, and collecting particles based on rotational inertia.An axial wet cyclone has the air inflowing axially, generating rotational force using vanes inside the cyclone, and collecting particles based on rotational inertia.While there has been sufficient research on general cyclones in terms of design equations and computational fluid dynamics, there have been few studies that focused on axial wet cyclones.
Biological weapons can cause lethal bodily damage during a war or terrorist attack.Biological weapons can infect people through three paths: the respiratory tract, scarred skin, or the digestive tract.Infection through the respiratory tract is the most deadly, and this type of biological weapon is released into the air in aerosol form.Delays in response time exponentially increase the damage from biological weapons.Thus, the type of agent should be recognized quickly, and appropriate action should be taken accordingly.In addition, the quantity of a biological weapon is extremely small, and its concentration drops sharply upon release.In this regard, it is absolutely critical to have a system that can concentrate the biological weapon material for immediate and accurate identification and analysis; this is the first step to be taken in preparation for a biological war.
Concentrating the biological agent requires making an aqueous solution from the continuous extraction of aerosols using a particle collector.The aerodynamic diameters of most particles used as biological weapons range between 1-7 µm (McFarland et al., 2010).For this reason, a cyclone, which is useful for collecting particles of this aerodynamic diameter and could be established as a wet collector as illustrated in Fig. 1, is an appropriate mechanism for concentrating materials for biological weapons.In designing the system, an axial wet cyclone is particularly useful for miniaturizing and serializing the system.
Axial cyclones are widely used under special circumstances, such as air purification in express trains or at industrial sites.However, there have been few publications presenting research on axial wet cyclones.According to the axial cyclone literature, an axial separator with a two-pronged inlet was first suggested by Wedding et al. (1982) and created to be consistent with the transmission curve of the human thorax.An axial cyclone used in low pressure and low airflow conditions was created by Tsai et al. (2004) and has been used to remove very fine particles.Hsiao et al. (2011) performed an experiment with an axial cyclone at atmospheric pressure conditions and an airflow of 50 L min -1 ; the authors researched the axial wet cyclone characteristics depending on the geometrical features of the vortex finder and upside-down cup.Gong et al. (2012) performed numerical and experimental research on an axial cyclone at a high airflow of or above 1,000 L min -1 and examined the effect of the vane angle and distance between vane and inlet on the cut size of the axial cyclone.In this study, an axial flow wet cyclone was designed by spraying a liquid onto an existing axial cyclone form in order to rapidly concentrate the particles in the air.The suggested design enables the aerosol to be concentrated in the liquid faster than conventional bio-aerosol samplers.

Design of Axial Wet Cyclone
Axial Wet Cyclone Design Equations Maynard (2000) derived a mathematical model for the axial wet cyclone based on laminar flow.Tsai et al. (2004) used Maynard's approach and established a mathematical model that calculated the tangential velocity and residence time of the axial wet cyclone under low pressure and low inflow conditions.Hsiao et al. (2011) used a vane to design a cyclone using the empirical equation of Pui, Romay-Novas, and Liu on the 90°-bend particle penetration in turbulent flow (Pui et al., 1987).The vane had three rotations, rotated 1,080° in total, and showed the same effect as 12 different 90° bends.The empirical equation provided was modified as follows (Tsai et al., 2004): 10 and 1 8 where Pen v is the penetration in the vane, P B is the penetration in the 90° bend, Stk is the Stokes number, C c is slip correction factor, ρ p is particle density, d p is particle diameter, U is average velocity, µ is the viscosity coefficient of the fluid, and a h is the hydrodynamic diameter of the vane channel.To design the body of a cyclone, the cyclone's flow was simplified in accordance with Fraser et al. (1997).Using an integral equation, the following equation was derived: where Pen b is the penetration in the body, R 2 is the cyclone radius, u r is the particle radial velocity, and t b is the residence time of the particle in the cyclone body.The major variables for axial wet cyclone behavior were calculated using Eq. ( 2), and a cyclone was designed based on these calculated variables (Fraser et al., 1997).Multiplying Eqs. ( 1) and ( 2) resulted in the following final penetration equation for an axial wet cyclone: where η t is the theoretical collection efficiency in the entire cyclone, D e is the diameter of the outlet, and t res is the residence time of the particle in the cyclone.An axial wet cyclone can be designed using Eq. ( 3) (Grinshpun et al., 2005).Zhou and Soo (1990) explained the limitations of using an ordinary k-epsilon model to analyze flows with a rotational component.Fraser et al. (1997) used a modified k-epsilon model to perform a three-dimensional numerical analysis of the cyclone; the modified k-epsilon model provided numerical analysis results more consistent with experimental results than the previous k-epsilon model.Hoekstra et al. (1999) compared the ordinary k-epsilon model and renormalization group (RNG) k-epsilon model, and found that the assumption of an isotropic turbulent flow component in the k-epsilon model did not agree with the results from a cyclone with a non-isotropic turbulent flow component.Instead, the Reynolds stress model fit better with the analysis of a cyclone with a non-isotropic turbulent flow component (Wang et al., 2006).Later, Gong et al. (2012) used the RNG k-epsilon model to obtain number convergence, and subsequently used the Reynolds stress model to perform a numerical analysis.

Numerical Analysis of the Cyclone
The RNG k-epsilon model was established by Yakhot et al. (1992).Although the model is similar to the ordinary k-epsilon model, the epsilon equation was modified with a mathematical technique from the RNG approach, which allowed the k-epsilon model to satisfy more variable turbulent flow conditions.The RNG k-epsilon model is expressed in Eqs. ( 4) and ( 5): where t is time, ρ is air density, k is turbulent flow kinetic energy, x i is the location component of the i th order term, u i is the velocity component of the i th order term, α k is the effective reverse Prandtl number related to k, µ eff is the effective viscosity coefficient, G k the rate of turbulent flow kinetic energy generation, ϵ is the rate of turbulent flow dissipation, α ϵ is the effective reverse Prandtl number related to ϵ, and C is a constant in the RNG model.The Reynolds stress model (RSM) is the most complicated turbulent flow model and is expressed as a second-order tensor.This model is calculated using Eq. ( 6) and the software FLUENT: In this equation, the two mathematical expressions on the left are the partial derivatives for stress and convection with respect to time, D ij is a tensor related to the spread of stress, P ij is a tensor related to generation, ∏ ij is a tensor related to pressure strain, and ϵ ij is a tensor related to the loss rate.

Design of Axial Wet Cyclone
As shown in Fig. 2, the design of the vane was based on the research presented in Hsiao et al. (2011) and was similar to that of Tsai et al. (2004).The axial wet cyclone was designed using Eq. ( 3), and the dimensions for each part are as follows: the inlet inner diameter is 80 mm, outer diameter of outlet is 40 mm, and inner diameter is 34 mm.The vane begins 50 mm from the inlet and spirals down in one, three, and five rotations with a pitch of 15 mm.The total length of the cyclone is 340 mm, and is shaped such that the bottom area narrows to encourage flow to hit the floor and move smoothly toward the outlet.To increase the standing water inside, the dust box was designed to better collect a solution.These dimensions are also listed in Table 1.

Numerical Analysis Method
The commercial code FLUENT 16.1 was used for numerical analysis, and the analytical conditions were a normal turbulent flow model using the RNG k-epsilon and   4) and (5), the final numerical computation was performed using the RSM that utilizes Eq. ( 6).
To identify the particle trajectory, the discrete phase model (DPM) was used to solve Eq. ( 7): where m p is the particle mass, ν is particle velocity, and F d is particle drag.Approximately 5,000 particles with aerodynamic diameters ranging between 0.7-2 µm were injected into the inlet at a constant dispersion.To compute numerical collection efficiency, trap conditions were set in the vane and wall of the body.The collection efficiency was calculated using Eq. ( 8): where η n is the collection efficiency using the numerical method, C trap is the number of trapped particles computed from numerical analysis, and C injection is the number of injected particles.
For VOF analysis, the following equations were applied: where α is the volume fraction, ṁ ba is the mass transfer rate from a to b, and ṁ ab is the mass transfer rate from b to a.By analyzing those equations numerically, the path of water movement was predicted and identified.Due to the limitations of the analysis tool, which cannot apply VOF and DPM models at the same time, two models were applied separately in this study.

Experimental Method
Two types of experiments were performed using the method described in Fig. 4 to measure the (a) number concentration, and (b) experimental collection efficiency.The design shown in Fig. 4(a) uses polystyrene latex (PSL) particles with 0.7 µm, 0.8 µm, 1 µm, and 2 µm aerodynamic diameters, generated from the atomizer.These particles pass through a diffusion dryer where water is removed, and then the particle number concentration is measured using the dust spectrometer (Grimm, Model 1.109).The design shown in Fig. 4(b) has a 1 m 3 chamber that maintains a constant PSL particle concentration, and the airflow inside the cyclone is maintained at 500 L min -1 using a ring blower (DJM, 2RB 210-7AV11) and rotor meter.The water injector puts 3 mL min -1 of water inside the chamber.PSL particles generated from the atomizer pass through the  diffusion dryer, which removes water, and then are injected into a wet axial wet cyclone.Finally, 1-1.5 mL min -1 of solution containing PSL particles is acquired using the vacuum pump.The resulting experimental collection efficiency was calculated using Eq. ( 11).The PSL particle concentration in the solution that did not go through the cyclone was used as a baseline concentration for comparison with the solution concentration collected by the cyclone.where η e is the experimental collection efficiency, C post is the PSL particle number concentration of the post-collection solution, and C pre is the PSL particle number concentration of the pre-collection solution.Both collection efficiencies were measured using a dust spectrometer as shown in Fig. 4(a), and C post was measured by re-aerosolization of hydrosols collected in the wet cyclone.As demonstrated in Fig. 5, as the number of vane rotations increased, the absolute minimum tangential velocity increased proportionally from 49.9 m s -1 to 52.1 m s -1 , and then to 59.7 m s -1 ; the rotational component strengthened as the number of vane rotations increased.As shown in Fig. 5, the absolute minimum axial velocity was 42.1 m s -1 , 34.0 m s -1 and 19.7 m s -1 , which demonstrates that it decreased proportionally as the number of vane rotations increased.The tangential velocity increased as the number of vane rotations increased, thereby enhancing the particle collection efficiency.In addition, as the axial velocity decreased with increasing vane rotations, the average particle residence time increased.In summary, as the number of vane rotations went up, the average residence time of particles and pressure drop increased; both factors improved collection efficiency.Using the axial wet cyclone with a three-rotation vane, flow characteristics were studied for changing airflow, from 100 L min -1 to 300 L min -1 , and then to 500 L min -1 .To simulate each of these flows, the inlet flow rate was set at 0.442 m s -1 , 1.336 m s -1 and 2.210 m s -1 , and a numerical analysis was performed.The pressure drop for each flow rate was 105 Pa, 852 Pa and 2430 Pa.The efficiencies for each of the airflows are illustrated in Fig. 6.The efficiency was higher in order of higher airflow rates: 500 > 300 > 100 L min -1 .efficiency was better at higher airflows, there was also a greater loss in pressure.Therefore, it is important to consider this relationship when designing an axial wet cyclone.

Axial Wet Cyclone Characteristics Based on the Number of Vane Rotations
Using FLUENT's VOF model, the path of water movement was predicted for the axial wet cyclone with a three-rotation vane.By adjusting the volume fraction cutoff in the multiphase flow setup box, the water mass flows going in and out were consistent, which ensured continuity.An inaccurate volume fraction cutoff could lead to physical discrepancies, where the outflow is numerically larger than the inflow.In this regard, this study input an accurate volume fraction cutoff to perform the computations.Fig. 7 describes the results of a VOF analysis: The total amount of water released from the three inlets in the numerical analysis was 1.5 mL min -1 (0.5 mL min -1 per inlet).Fig. 7(a) shows that water from the inlet was diluted with air, and the VOF gradually decreased; (b) shows that water was tilted towards the edge, not the center, because of the centrifugal force of the flow; and (c) shows that water was also collected near the edge of the suction pipe when it flowed out because of the centrifugal force from rotational flow.
As shown in Fig. 8(a), water flowed down in a spiral in the cyclone body.In addition, water droplets remained on the edge of the dust box as water gathered there.Accordingly, a better placement for the solution suction inlet was on the edge of the dust box, not the center.In this regard, two types of axial wet cyclones were manufactured to perform the experiment; both are shown in Fig. 8(b).When a center outlet was used alone to suction water, backflow occurred   because a small swirl was created in the center, regardless of the flow rate of water injected.However, when two edge outlets were used to suction water, backflow did not occur, as long as enough water was injected; 100% of the water was suctioned, except for the amount that evaporated.
The relationship between pressure drop and number of vane rotations was investigated; there was almost no difference in pressure drop between one and three vane rotations, but there was a 10% increase from three to five vane rotations.The cut size declined from 1.97 µm to 1.24 µm when the number of vane rotations increased from one to three.However, when the number of rotations increased from three to five, the cut size decreased more, from 1.24 µm to 1.08 µm.The increase to five rotations provided a low increase in efficiency; therefore, the threerotation vane, which showed stable efficiency in existing research, was manufactured to perform the physical experiment.Particle efficiency results from the axial wet cyclone with a three-rotation vane were obtained and compared to the results from the numerical simulation, shown in Fig. 9.Although the results were not identical, they all showed a similar trend.The efficiency was higher in the numerical analysis than in the experimental results for all diameters of the particles.The numerical analysis showed better efficiency likely owing to the particles leaning towards the vane wall during the numerical analysis.The first and second experiments were performed on different days; however, their results were almost identical within the error of the measurement, which demonstrates the high repeatability of the experiment.While the particles with smaller diameters, specifically, 0.7 µm and 0.8 µm, had a low margin of error of approximately 1%, the particles with larger diameters such as 1.3 µm and 1.8 µm had a high margin of error of approximately 10%.This difference is due to atomizer limitations: There is difficulty generating high concentrations of large particles, but it can generate sufficiently high concentrations of small particles.In addition, the small particles were well dispersed, while the large particles stuck together more often, which led to an additional decrease in the concentration of particles with larger diameters.

SUMMARY
This study experimentally verified the performance of an axial wet cyclone design based on analytical and numerical predictions under high flow rate conditions.The following conclusions were drawn based on the analyses of the numerical simulations and physical experiments: 1.The analytical design was based on the equations provided in Hsiao et al. (2011), which numerically identified trends in physical parameters based on the number of vane rotations.The numerical method indicated that as the number of vane rotations decreased, the axial velocity increased, tangential velocity decreased, and cyclone collection efficiency decreased.In addition, the numerical predictions were better than the analytical results because of the tendency of the particles to lean towards the vane wall when they entered the vane during the numerical simulations.2. Numerical and experimental results identified the path of the water movement inside the cyclone.Using VOF, the numerical results indicated that water tilted toward the edge of the dust box due to centrifugal force.The experimental results confirmed that water droplets remained on the edge of the box.The experiments indicated that 60-70% of the water could be suctioned from the center, while nearly 100% of the water was collected, ignoring the amount lost to evaporation, when water was suctioned from the edge of the box at an injection flow rate at or above 3 mL min -1 .3. Ultimately, this research compared numerically predicted efficiency and experimentally measured efficiency and found a margin of error of up to 10%.The analytical, numerical, and experimental processes presented in this paper will serve as a reference for the design of an axial wet cyclone, for which there has been a lack of research.Ultimately, wet cyclone technology could be applied to concentrator designs that can detect infectious airborne viruses in real time.

Fig. 4 .
Fig. 4. Schematic of experimental set-up of (a) measurement of number concentration and (b) set-up for measuring experimental collection efficiency.
Hsiao et al. (2011) manufactured a three-rotation vane and performed experiments.To analyze the change in cyclone characteristics based on the number of vane rotations, this study investigated the differences obtained when increasing vane rotations from one to three to five while all other dimensions remained unchanged.Streamlines according to the number of vane rotations were obtained, as illustrated in Fig.5.As each of the streamlines indicates, a drastic change in velocity occurred in the area where the air flowing from the inlet first entered the vane, and the maximum velocity was observed around this area.An instantaneous change in pressure was the largest in the area where the air entered the vane.As the number of vane rotations increased, pressure drop between inlet and outlet increased from 2,370 Pa to 2,430 Pa, and finally to 2,680 Pa.

Fig. 8 .
Fig. 8. Manufactured axial flow wet cyclone; (a) Streamlines of water through experimental method, and (b) Dust box section: One outlet (left) and three outlets (right).

Table 1 .
Geometrical dimensions for axial wet cyclone.