Experimental and numerical evaluation of the performance of a novel compound demister

Desalination 409 (2017) 115–127

Contents lists available at ScienceDirect

Desalination journal homepage: www.elsevier.com/locate/desal

Experimental and numerical evaluation of the performance of a novel compound demister Yilin Liu a, Dunxi Yu a,⁎, Jingkai Jiang b, Xin Yu a, Hong Yao a, Minghou Xu a,⁎ a b

State Key Laboratory of Coal Combustion, School of Energy and Power Engineering, Huazhong University of Science and Technology, 1037 Luoyu Road, Wuhan 430074, China School of Energy and Power Engineering, Huazhong University of Science and Technology, 1037 Luoyu Road, Wuhan 430074, China

H I G H L I G H T S • • • •

A novel compound demister with some advantages was proposed. The performance was evaluated by experimental and numerical methods. At low gas velocities or for small droplets, the compound demister is superior to the wave-plate demister. At high gas velocities, the compound demister shows higher resistance to droplet re-entrainment.

a r t i c l e

i n f o

Article history: Received 3 August 2016 Received in revised form 20 November 2016 Accepted 4 January 2017 Available online xxxx Keywords: Compound demister Separation efficiency Droplet re-entrainment Dry pressure drop

a b s t r a c t A novel compound demister that combines an upstream tube bank and downstream wave plates was proposed in this work for application in multistage flash (MSF) desalination process. Its performance was evaluated by experimental and numerical methods. Compared with the individual tube-bank and wave-plate demisters, the compound demister is found to have the highest separation efficiency (N 95%) with much less fluctuation for a wide range of gas velocities. At low gas velocities (≤4 m/s) or for removing small droplets (b20 μm), the separation efficiency of the compound demister is much higher than that of the wave-plate demister mainly because of the large separation capability of the tube bank. At high gas velocities (N 4 m/s), the compound demister shows higher resistance to droplet re-entrainment that occurs at inlet gas velocity of approximately 7 m/s compared with the tube-bank demister. This is due to the compensation from the wave plates in the compound demister that separate secondary droplets generated by tubes. The compound demister possesses higher dry pressure drops than either the tube-bank or wave-plate demister, but is acceptable for industrial application. All these advantages make the compound demister a promising candidate for droplet removal in the desalination process. © 2017 Elsevier B.V. All rights reserved.

1. Introduction Desalination is one of the major processes to generate fresh water for various daily human usages and industrial applications, requiring different levels of quality and quantity [1,2]. Among the various desalination technologies, the microfiltration (MF) and multistage flash (MSF) are the most widely applied ones [3]. The MF technology is increasingly popular with the development of the membrane technology. It would be more competitive with the use of cheap membranes made of raw materials like kaolin and CaCO3 [4], anorthite [5], ZrO2 [6], natural alumino-silicates [7], and natural hydroxyapatite obtained from cortical bovine bones [8]. Despite the popularity of the MF technology, the MSF process still occupies a large portion of the global installed capacity as it is deemed as the most reliable thermal desalination technology [3]. It ⁎ Corresponding author. E-mail addresses: [email protected] (D. Yu), [email protected] (M. Xu).

http://dx.doi.org/10.1016/j.desal.2017.01.022 0011-9164/© 2017 Elsevier B.V. All rights reserved.

also has the advantage of larger daily production capacity than other technologies [9], and possesses great economic potential if driven by renewable energy [10]. In the MSF process, demisters are adopted to remove the entrained saline droplets from the fresh water vapor in order to maintain the level of salinity in the generated fresh water. Those demisters mainly include filters [1], vanes [11], wave plates [12] and wire mesh [13]. Among these devices, wave-type demisters [11, 12,14–17] and wire mesh demisters [13,18] are the most widely adopted. Their performance is, however, largely dependent on demister configurations and operating conditions such as gas velocity and droplet size. For wave-type demisters, increasing droplet size or gas velocity leads to higher separation efficiency [19–22]. Nevertheless, they cannot effectively remove inlet mist under conditions of low gas velocities or small droplets [21–23]. For example, the separation efficiency of a wave-plate demister was higher than 90% for 30 μm droplets when the inlet gas velocity was over 5 m/s; but it dropped significantly to

116

Y. Liu et al. / Desalination 409 (2017) 115–127

Nomenclature Asn CD Dd Dt Eu H1 H2 H3 h k L l Min Mout Mnoz Map Md p Qsn S1 S2 St T t Ug,cr,1 Ug,cr,2 Ug,cr,max Ug,in Ug,max u ug,sn Wed wi Y

area of the inlet of the sampling nozzle, m2 drag coefficient droplet diameter, m or μm diameter of tubes, mm Euler number length of the wave-plate part, mm length of the tube-bank part, mm depth of the horizontal demisters, mm length of certain part of the wave plate, mm turbulent kinetic energy, m2/s2 channel width, mm characteristics dimension of the flow, m mass flow rate of liquid sampled at the inlet of demisters, kg/s mass flow rate of liquid sampled at the outlet of demisters, kg/s mass flow rate of liquid from the spray nozzle, kg/s mass flow rate of liquid collected by the air passage before demisters, kg/s. mass flow rate of liquid collected by the demisters, kg/s pressure, Pa or MPa volume flow rate at the inlet of sampling nozzle, kg/s tube spacing, mm row spacing, mm Stokes number temperature, K time, s minimum gas velocity for possible detachment of liquid film, m/s minimum gas velocity to suspend a droplet, m/s maximum critical gas velocity, m/s gas velocity at the inlet of demisters, m/s predicted maximum gas velocity in the demisters, m/s flow velocity, m/s gas velocity at the inlet of the sampling nozzle, m/s droplet Weber number mass fraction of a small group of droplets relative distance, mm

Greek symbols α bend angle, ° ΔP pressure drop, Pa ε turbulent dissipation rate, m2/s3 η separation efficiency separation efficiency for a small group of droplets ηi μ dynamic viscosity, Pa·s or μPa·s ρ density, kg/m3 σ surface tension of water, N·m−1 or mN·m−1 Subscript d droplet g gas phase

b50% at the inlet gas velocity of 3 m/s [21]. When the droplets' diameter was above 37.5 μm, the separation efficiency of the wave-plate demister was over 90% at an inlet gas velocity of 4 m/s; but it dropped sharply to b40% when the droplets' size decreased to b 25 μm [21]. Similar phenomena have also been observed in the literature [15,20,24,25]. Compared with wave-type demisters, wire mesh demisters generally have much higher separation efficiency at low gas velocities and for fine droplets [23], despite similar dependence of the droplet separation efficiency on operating conditions (e.g. gas velocity and droplet size)

[13,18,26–28]. For example, for 10 μm droplets, the separation efficiency of the wave-plate demisters in [20] was b 20% at inlet gas velocity of 2 m/s; but the separation efficiency of the wire mesh demister in [29] could achieve N 90% under the same conditions. For both types of demisters, it is also noticed that increasing the gas velocity to a certain higher value may result in the decline of separation efficiency because of droplet re-entrainment [13,19,30–32]. For example, increasing the gas velocity to above 7 m/s [33] would lead to a decrease in the separation efficiency of the wave-plate demisters due to the apparent occurrence of re-entrainment. For some wave-type demisters with multiple bends in [19,34], the re-entrainment will not occur until the gas velocity increases to as large as over 8 m/s. In contrast, wire mesh demisters are more vulnerable to flooding and reentrainment [13]. It was reported that droplet re-entrainment might occur at a gas velocity of as low as 4 m/s for a wire mesh demister, causing a sharp increase of pressure drop to above 1000 Pa [13]. From the above literature review, although most of the wave-type demisters have relatively higher resistance to droplet re-entrainment, they have difficulties in achieving high separation efficiency under conditions of fine droplets and/or small gas velocities. By contrast, the wire mesh demisters can obtain high efficiency under these conditions, but are much more vulnerable to droplet re-entrainment than the wavetype demisters. In summary, both types of demisters could hardly operate with high stable efficiencies under a wide range of industrial conditions. In this work, a novel compound demister was proposed to take advantage of the merits of the wave-type and wire mesh demisters, which has not been reported in the literature. A staggered tube bank, simplified from the wire mesh demisters, was adopted in the compound demister aiming at removing fine droplets at higher efficiencies. Wave plates were arranged downstream just after the tube bank to mitigate potential flooding and re-entrainment. Both experiments and numerical simulation were carried out to evaluate the performance of the proposed compound demister. The effects of gas velocity and droplet size on the separation efficiency and pressure drops of the compound demister were investigated. Its performance was compared with that of the individual tube-bank demister and wave-plate demister. The mechanisms of droplet re-entrainment in the compound demister were also discussed. The results show that, compared with the individual demisters, the compound demister has higher and more stable separation efficiency under the conditions investigated. Further comparisons with findings in the literature reveal that, under the same conditions of fine droplets and low gas velocities, the compound demister has higher efficiency than most of the wave-type and wire mesh demisters [18,20,21]. The results also show that the compound demister has higher resistance to droplet re-entrainment than the wire mesh demisters in the literature [13,35]. 2. Experimental and simulation methods 2.1. Experimental 2.1.1. Demisters The compound demister tested in this work is presented in Fig. 1. It consists of an upstream tube bank and downstream wave plates. The tube bank consists of 4 lines of staggered tubes that are made of stainless steel, while the wave-plate part includes six plexiglass wave plates that form five channels. The tubes are vertically embedded into the plexiglass channel. The two-dimensional sketches of single units of the three demisters investigated are presented in Fig. 2. To make the comparison between different demisters reasonable, the individual tube-bank demister and wave-plate demister were designed to have the same geometric parameters as their counterparts in the compound demister. The configurative parameters of single units of the investigated demisters are given in Table 1. The geometric parameters of the wave plates are obtained from the optimal configurative results reported in

Y. Liu et al. / Desalination 409 (2017) 115–127

117

Table 1 Configurative parameters of single units of the demisters.

Fig. 1. Three-dimensional diagram of the compound demister.

the literature [33] and are also shown in Table 1. For all the demisters investigated in experiments, their height (H3) is 150 mm. The diameter of tubes (Dt) is 4 mm and the inlet width (L) of the wave-plate channels is 30 mm.

Configurative parameters

Value

Dt (mm) h (mm) H1 (mm) H2 (mm) H3(mm) L (mm) S1 (mm) S2 (mm) Y (mm) α (°)

4 38 190 40 150 30 7.5 6 15 50

2.1.2. Experimental apparatus and testing conditions An experimental system was purposely designed to evaluate the performance of the three demisters under various operating conditions. The schematic diagram of the system is shown in Fig. 3. The system consisted of four major parts: demister section, air-blow subsystem, water loop, and measurement subsystem. Different types of demisters were separately arranged in the demister section for performance evaluation. The air flow with various velocities was fed by the airblow subsystem while the mist to be entrained by the air flow into demisters was generated by the water loop. In the demister section, the wave-plate part (A in Fig. 3) and the tube-bank part (B in Fig. 3) were both dismountable. For tests of the compound demister, both parts (A and B in Fig. 3) were assembled into the apparatus. For tests of the individual tube-bank or wave-plate demister, only one of them (A or B in Fig. 3) was installed into the system with the other being removed. In the air-blow subsystem, the essential components included a centrifugal blower (≤ 4580 Pa,

Fig. 2. Two-dimensional sketches of single units of the three types of demisters: (a) compound demister, (b) wave-plate demister, (c) tube-bank demister.

118

Y. Liu et al. / Desalination 409 (2017) 115–127

Fig. 3. Schematic diagram of the experimental system.

1839 m3/h), a venturi tube, air-distribution plates and an air passage with detecting apertures on both sides. A frequency converter (0– 50 Hz) was used to control the rotating speed and the electrical power of the centrifugal blower so that different gas velocities could be obtained for various conditions. The plexiglass air passage had an interior cross section of 170 mm (width) × 150 mm (height) and an overall length of 2000 mm. The air-distribution plates were specially devised to homogenize the sectional gas velocity in the air passage. The water loop was mainly composed of a single-fluid spray nozzle, a hydraulic pressure gauge, a water circulating pump (≤ 4 MPa, 6.36 L/min) and a water tank (about 600 L). The required mist was supplied from the spray nozzle fed by the high-pressure (1.8 MPa) water from the water circulating pump. The measurement subsystem mainly consisted of a handheld anemometer, a sampling tube, two gas drying towers (about 1.75 L), a gas rotameter, a vacuum pump (≤0.5 mbar, ≤ 22.5 m3/h) and an electrical balance. To gain a better view of the compound demister's application value, the experimental conditions in this work were chosen within the range of working conditions for the demisters in MSF plants. In those plants, the droplet size mainly ranges from several to thousands micron milligrams and the gas velocity is approximately between 1 and 12 m/s [13,18]. The operating conditions and fluid properties in the experiments are summarized in Table 2. The particle size distribution of the mist fed from the spray nozzle was characterized by a Laser Diffraction Particle Size Analyzer (LDPSA, declared accuracy: ≤±1%) before tests. The results are presented in Fig. 4, which shows that the Sauter Mean Diameter (SMD) is 39.35 μm (± 7.7%). The gas velocities in the air passage were measured by the calibrated handheld anemometer. The range and accuracy of the meters used in the experiments are summarized in Table 3.

2.1.3. Sampling and measurements During the tests, the separation efficiency and dry pressure drop of the demisters were evaluated by the measurement subsystem. To obtain the separation efficiency, the isokinetic sampling method and grid sampling method [36] were employed to sample the air-liquid flow before and after the demisters via sampling path (Fig. 3). The gas velocity at the inlet of the sampling nozzle (ug , sn), which was required to be equivalent to the average gas velocity in the air passage, was calculated by

ug;sn ¼

Q sn Asn

ð1Þ

Table 2 Operating conditions and fluid properties in experiments. Parameters

Value

Demisters Fluid Inlet gas velocity (m/s) Flow pattern Diameter range of droplets (μm) SMD of droplets (μm) pg (MPa) Tg (K) ρg (kg/m3) ρd (kg/m3) μg (μPa/s) μd (μPa/s) σ (mN/m) Wetness fraction (volume ratio)

Compound, tube bank and wave-plate Air-water 3, 4, 5, 6, 7 Dispersed flow 2–997 39.35 0.1013 About 293.15 1.204 998.2 18.25 1002 72.7 6.5E-05–9.8E-05

Y. Liu et al. / Desalination 409 (2017) 115–127

119

of experimental data follows the t-distribution. For the two-sided confidence interval of 90%, the relative composite uncertainty of the experimental separation efficiency is in the range of ±1.1–4.0%. The relative uncertainty of the experimental dry pressure drop is from ±1.4–12.0%. 2.2. Simulation 2.2.1. Governing equations and setup Simulation work was mainly carried out to study the effect of droplet size on separation efficiency using commercial software ANSYS Fluent 16.1. The simulation calculation was carried out based on the following assumptions:

Fig. 4. Diameter distribution of the droplets entrained into demisters in experiments.

where Asn and Qsn are the inlet area and inlet volume flow rate of the sampling nozzle, respectively. The value of Qsn was given by the gas rotameter and modulated by a valve at the inlet of the vacuum pump. The gas rotameter was calibrated using a gas mass flow meter within the full scale. Their range and accuracy are presented in the Table 3. From calculation, the accuracy of the meters will cause b 2.6% undulation of the velocity at the inlet of the sampling nozzle. The sampled two-phase flow was desiccated through two gas drying towers. The weight of upstream components before the rotameter in the sampling path (Fig. 3) was measured together before and after sampling by a sensitive electronic balance. And the separation efficiency is defined as: η¼

Min −M out  100% Min

ð2Þ

It was noticed that the isokinetic sampling method could not precisely measure the separation efficiency when obvious droplet reentrainment occurred. Alternatively, another method was employed by measuring the liquid mass flow of the mist that was fed by the spray nozzle (Mnoz), removed by demisters (Md) and collected by the air passage before demisters (Map). The separation efficiency is calculated as [33]: η¼

Md  100% Mnoz −Map

ð3Þ

The dry pressure drop between the inlet and outlet of the demisters was measured by the calibrated handheld anemometer. All experimental measurements were conducted under steady state conditions and each group of experiments was repeated at least 3 times. According to the statistical analysis methods [37,38], the uncertainty analysis was carried out by taking the accuracy of meters and the unsteadiness of the system operation into account. It was assumed that the distribution Table 3 Range and accuracy of meters. Meters

Range

Declared accuracy

Electrical balance Gas mass flow meter Gas rotameter Anemometer Gas velocity Pressure Vernier caliper Hydraulic pressure gauge

0–3000 g 0–10 L/min 40–400 L/h 1.27–78.7 m/s -3735–3735 Pa 0–300 mm 0–2.5 MPa

±0.01 g (resolution) ±0.035 L/min ±1.5% of reading ±1.5% of reading ±1% of reading ±1 Pa ±0.05 mm ±0.05 MPa

(1) Since the height of demisters in experiments was much larger than the channel width or tube pitch, the two-phase flow was simplified as two-dimensional [39]. (2) The gas velocity and Mach number were small enough to suppose that the flow was incompressible [20]. (3) The droplets were modeled as hard spherical particles and were collected and removed immediately when they impinged walls [15]. (4) Since the simulation work mainly intended to investigate the separation efficiency for fine droplets at low gas velocities without apparent re-entrainment, the re-entrainment of droplets and the effect of water film on the gas flow were ignored [39]. The Euler-Lagrangian method was adopted to model the two-phase flow. The Realizable k-ε turbulence model [40] was employed to simulate the gas phase field with enhanced wall treatment. The equations governing the continuity, momentum and turbulent quantities are defined as ∂ui ∂u j þ ¼0 ∂xi ∂x j

ð4Þ

" 2 # 2 ∂ui ∂u ∂u 1 ∂p ∂ ui ∂ ui þν þ þ ui i þ u j i ¼ f i − ρ ∂xi ∂t ∂xi ∂x j ∂xi 2 ∂x j 2

ð5Þ

" 2 # 2 ∂u j ∂u j ∂u j ∂ uj ∂ uj 1 ∂p þ uj ¼ f j− þν þ þ ui ρ ∂x j ∂t ∂xi ∂x j ∂xi 2 ∂x j 2

ð6Þ

 ∂ ∂  ∂ ρku j ¼ ðρkÞ þ ∂t ∂x j ∂x j  ∂ ∂  ∂ ρεu j ¼ ðρεÞ þ ∂t ∂x j ∂x j

"

#  ∂u j μ t ∂k −ρu0i u0j μþ −ρε σ k ∂x j ∂xi

" μþ

#  μ t ∂ε ε2 pffiffiffiffiffiffi þ ρC 1 Sε−ρC2 σ ε ∂x j k þ νε

ð7Þ

ð8Þ

The model constants and parameters were set as default values and are well documented elsewhere (ANSYS, Fluent) [40]. The equations were solved by SIMPLE algorithm to achieve a relatively quick convergence. The second-order upwind approximation method was adopted to discretize these equations so as to obtain a relatively high accuracy. The converging criterion used for all variables was set as 1.0E-5 to terminate the iteration solution of gas flow equations. When the computation of the continuous flow field was converged, the motion of droplets was calculated using the Lagrangian method. The effect of droplets on the gas phase was considered by a two-way coupling method. In the horizontal demisters, the deposition of droplets due to gravity can be ignored since droplets' horizontal movement is much larger than their vertical one. The equation governing the droplets' motion is defined as: dud ¼ F D ðu−ud Þ þ F dt

ð9Þ

120

Y. Liu et al. / Desalination 409 (2017) 115–127

FD(u−ud) is the drag force per unit droplet mass and F is the additional force including pressure gradient force, virtual mass force, lift force, Brownian force and thermophoretic force. Experience in previous works [15,41] shows that the effect of these forces can be neglected in this study. The Discrete Random Walk (DRW) Model [40] was used to predict the dispersion of particles due to turbulence in the continuous phase. The time scale was set constant for calculating the eddy lifetime. Definition of the DRW Model and other parameters can be referred to elsewhere [40]. The computational domains and boundary conditions of the single unit of the demisters are presented in Fig. 2. The geometric parameters of the domains are the same as the entities used in experiments, as shown in Table 1. At the inlet, the gas velocity was uniform along the width and the inlet droplets' velocity was equal to it. The investigated diameters of the droplets in the simulation were in the range of 2–80 μm. The diameter range was sufficient to have the simulated results fit with the experimental data. The liquid mass flow rate at the inlet was 0.00475 kg/s and the interaction between droplets were ignored. The outlet pressure was set as the atmospheric pressure. Other operating parameters were set according to the experimental conditions that are shown in Table 2. The grids were created using the software Gambit. All cells were quadrilateral and the mesh near tubes and wave plates was locally refined. The grid analysis was conducted to figure out the optimal computational cells to achieve relatively high accuracy and rapid convergence. The results showed that the grids generated by the following size function were optimal for computation: the minimum interval size near walls was 0.02 mm while the maximum in the flow field was 0.4 mm with a size growth rate of 1.025. Fig. 5 shows the expansion of the mesh around tubes. The total grid numbers for the compound, wave-plate and tube-bank demisters were 964934, 662435and 299369, respectively. 2.2.2. Model validation Model validation is demonstrated in Fig. 6 and Table 4 by comparing the simulation results (denoted as cal.) with experimental data (denoted as exp.) from the perspective of pressure drop and separation efficiency. As shown in Fig. 6, the predicted dry pressure drop agrees well with the experimental data for the three demisters. The largest discrepancy is only 21.9 Pa (experimental result is 284.9 Pa) for the compound demister, appearing at the highest inlet gas velocity of 7 m/s. It is suggested that the air flow in demisters can be well described by the models adopted. The predicted overall separation efficiency for the mist in experiments is calculated by:

η¼

Fig. 6. Comparison of simulated dry vapor pressure with experimental data. (Compound: the compound demister; Tube-bank: the individual tube-bank demister; Wave-plate: the individual wave-plate demister).

and droplet size is in accordance with the size distribution curve in Fig. 4. As shown in Table 4, the simulated separation efficiency of the wave-plate and compound demisters are in good agreement with the experimental data with error b3.6%. As for the tube-bank demister, the simulated separation efficiency agrees well with experimental results at inlet gas velocities of 3 m/s and 4 m/s (error b 6.5%). However, the discrepancy increases at higher gas velocities. This is mainly because the droplet re-entrainment was not considered in the simulation model but took place in the tests. Therefore, the CFD models adopted in this work are appropriate for predicting two-phase flow with inlet gas velocities no N 4 m/s. They are mainly used to investigate the effects of droplet size on separation efficiency at low gas velocities (detailed in Subsection 3.2). 3. Results and discussion 3.1. Separation efficiency at low gas velocities

 N  ∑i¼1 wi ηi

ð10Þ

N

∑i¼1 wi

where wi and ηi are the mass fraction and separation efficiency of droplets with certain diameter, respectively. The relationship between wi

This section puts emphasis on the evaluation of the separation efficiency of the compound demister at low gas velocities (3 m/s and 4 m/s), under which conditions no significant droplet re-entrainment takes place (shown in Table 4). Subsection 3.3 is dedicated to the Table 4 Comparison of simulated separation efficiency with experimental data. Demister

Inlet gas velocity

Separation efficiency (%)

type

(m/s)

Predicted

Experimental

Relative discrepancy

Compound

3 4 5 6 7 3 4 5 6 7 3 4 5 6 7

96.99 98.07 98.42 98.95 99.09 70.67 74.63 77.53 79.77 81.69 97.20 98.09 98.59 98.86 99.11

95.85 96.12 96.66 97.03 95.66 73.08 73.76 78.71 82.50 80.99 92.83 92.25 89.63 80.04 79.51

1.18 2.03 1.83 1.99 3.59 -3.30 1.17 -1.51 -3.31 0.86 4.71 6.33 9.99 23.52 24.66

Wave-plate

Tube-bank

Fig. 5. Generated mesh for computation.

Y. Liu et al. / Desalination 409 (2017) 115–127

Fig. 7. Experimental separation efficiency at low inlet gas velocities.

121

Fig. 9. Predicted maximum gas velocity in the fluid domain of the demisters.

separation efficiency under a wide range of gas velocities so that appreciable droplet re-entrainment can be observed (shown in Table 4). The measured separation efficiencies of the compound demister at gas velocities of 3 m/s and 4 m/s are compared with those of the individual tube-bank demister and wave-plate demister in Fig. 7. As can be seen, the demisters with a tube bank have higher separation efficiency than the wave-plate demister. Generally, the compound demister has the highest separation efficiency of above 95% at these inlet gas velocities.

It is observed that the separation efficiency of the tube-bank demister is approaching that of the compound demister. Since the tube bundle and wave plates in the compound demister share the same configurative parameters as the individual demisters, it suggests that the high separation efficiency of the compound demister at low gas velocities is mainly attributed to the tube bundle. Fig. 8 presents the velocity contours of gas flow in different demisters carried out by CFD simulation at inlet gas velocity (Ug,in) of

Fig. 8. Contours of the velocity magnitude in the demisters at Ug,in of 3 m/s.

Fig. 10. Gas flow streamlines in the demisters at Ug,in of 3 m/s.

122

Y. Liu et al. / Desalination 409 (2017) 115–127

3 m/s. It is shown that demisters with a tube bank generally have higher velocity magnitude than the wave-plate demister. This is mainly due to the throttling effect of the tube bundle. Specifically, in Fig. 9, the maximum gas velocity (Ug,max) in demisters with the tube bundle is considerably larger than that in the wave-plate demister. The Ug,max in demisters with a tube bank is almost 1.5 times of that in the waveplate demister. When flowing through the tube bundle, the droplets will be accelerated to a much higher speed because of the drag force from the ambient gas flow. And droplets with higher velocity possess larger inertia and better ability to overcome the tendency of following streamlines of gas flow. In Fig. 10, it is noted that, under similar conditions, the streamlines through the tube bank are more sinuous than the wave-plate channel. Compared with droplets between wave plates, those droplets flowing through the tube bundle will have a larger centrifugal force, thus creating a better chance to escape the gas streamlines and impinge on tubes. Therefore, at low gas velocities, the tube bank is superior to remove droplets than wave plates, which significantly contributes to the high separation efficiency of the compound demister. Additionally, the collision rate between droplets in these demisters largely depends on the rate of turbulent energy dissipation (ε), under the similar mist and air-water flow condition with similar dimensions of droplets and kinematic viscosity [42]. And a higher level of ε can increase the chance of droplet collision [42]. Fig. 11 compares the turbulence in different demisters when Ug,in is 3 m/s. It is shown that demisters with the tube bank have greater turbulent kinetic and dissipation rate in the fluid field compared with the wave-plate demisters. It is more likely for droplets flowing through the tube bank to collide with each other and grow into larger droplets with greater inertia than those through wave-plate

Fig. 12. Predicted separation efficiency for small droplets (Dd ≤ 20 μm).

channels. Consequently, the compound demister with the tube bank has a greater capability of removing inlet mist even at low gas velocities. 3.2. Separation efficiency for fine droplets at low gas velocities Subsection 3.1 shows that the compound demister is capable of overcoming the disadvantage of the wave-plate demister, i.e., inefficiency to

Fig. 11. Contours of turbulent kinetic energy and dissipation rate in the demisters at Ug,in of 3 m/s.

Y. Liu et al. / Desalination 409 (2017) 115–127

123

Fig. 15. Comparison of the separation efficiency for 10 μm droplets at low gas velocities between the compound demister in this work and those demisters in some literature.

remove mist at low gas velocities. This section is to find out whether the compound demister could mitigate another disadvantage of the waveplate demister, i.e., inefficiency to remove small droplets. Simulation work was carried out to investigate the compound demister's capability to remove fine droplets (Dd b 20 μm) with Ug,in no higher than 4 m/s. The results are compared in Fig. 12. It is shown that the separation efficiencies of all demisters generally increase with the droplet diameter due to greater inertia. Compared with the wave-plate demister, the compound demister and the tube-bank demister have much higher separation

efficiency for small droplets under all the conditions investigated. The compound demister can capture almost all droplets above 15 μm, while the wave-plate demister's separation efficiency for the 15 μm droplets is no higher than 30%. As to smaller droplets b15 μm, the compound demister's separation efficiency is also much higher than the wave-plate demister. Significant differences in separation efficiency between the two demisters are apparent when the droplet diameter is between 6 and 20 μm. When compared with the tube-bank demister, however, the compound demister shares similar separation efficiency under the same conditions. The results suggest that the compound demister's high efficiency to remove fine mist is largely dependent on the tube bundle which is more competitive to remove small droplets than the wave plates. Fig. 13 compares the predicted instantaneous positions of droplets of 6 μm and 10 μm in the compound and wave-plate demisters with the same inlet gas velocity (Ug,in = 3 m/s) and liquid mass flow rate. It is noticed that the tube bank in the compound demister is predominant to collect small inlet droplets while the wave plates have limited ability to remove these fine mist. Moreover, in the compound demister, if the initial fine droplets cannot be removed by the tube bank, they could hardly be collected by the downstream wave plates as well.

Fig. 14. Stokes number versus droplet diameter at Ug,in of 3 m/s.

Fig. 16. Experimental separation efficiency at variable inlet gas velocities.

Fig. 13. Droplet tracking in the compound (a, c) and wave-plate (b, d) demisters at Ug,in of 3 m/s: (a, b) Dd = 6 μm, (c, d) Dd = 10 μm.

124

Y. Liu et al. / Desalination 409 (2017) 115–127

demister to effectively remove fine droplets at low gas velocities. However, the compound demisters' efficiency is slightly smaller than that of the wave-plate demister in [15], which has its channel width as small as 8 mm and the number of bends as large as 7. It indicates that optimization of configurative parameters is of great significance to enhancing the separation ability of demisters. To further improve the separation efficiency of the compound demister, parametric optimization will be included in the follow-up work. In general, compared with the findings in literature, the compound demister could work stably with high efficiency under particular conditions of fine mist and low gas velocities. 3.3. Re-entrainment characteristics

Fig. 17. Predicted maximum gas velocity in the fluid domain of the three demisters.

The throttling effect of the tube bank will accelerate the gas flow, thus increasing the speed and inertia of fine inlet droplets. So the fine mist is more likely to be removed via tube bank than the wave plates. In addition, the deposition of fine droplets is highly dependent on the turbulent dispersion. Higher turbulent dispersion will contribute to larger separation efficiency [23,41]. As discussed above, demisters with the tube bank have larger turbulent kinetic energy (k) and dissipation rate (ε) than the wave-plate demister. Thus the tube bank is better at removing the fine inlet mist. Further, the Stokes number, representing the ratio of the droplets' response time to the residence time of the flow or obstacles [43], can be a good indicator of the droplets' ability to detach from the air stream. It is defined as: St ¼

ρd D2d U g;in 18μ g l

ð11Þ

where the characteristic dimension (l) of the flow or the obstacle in the flow is the channel width for the wave-plate demister and the tube diameter for the compound demister. Fig. 14 compares the Stokes number (St) for fine droplets in different demisters at Ug,in of 3 m/s. As can be seen, under the same operating conditions, the Stokes number for the similar inlet droplets in the compound demisters is much higher than that in the wave-plate demister. It suggests that the fine mist in the compound demister is more likely to detach from the gas flow stream and impinge the solid surface. Therefore, the compound demister has larger separation efficiency for fine mist than the wave-plate demister. To further understand the advantages of the compound demister, its separation efficiency for 10 μm droplets at low gas velocities is compared with different demisters that are reported in the open literature, as shown in Fig. 15. It can be seen that, at gas velocities from 2 to 4 m/s, the compound demister generally has higher separation efficiency for 10 μm droplets than most of the common wave-plate and wire mesh demisters. It suggests the obvious advantages of the compound

The results presented above show that, because of the tube bank, the compound demister has high separation efficiency at low gas velocities or for small droplets. Since the tube bank is similar to the wire mesh demister that is vulnerable to droplet re-entrainment [13,23], the re-entrainment characteristic of the compound demister should be further explored. This characteristic is illustrated in Fig. 16 which compares the measured separation efficiencies of the three demisters at a wide range of gas velocities. To better indicate the gas velocity above which obvious droplet re-entrainment takes place, Fig. 16 also includes the separation efficiencies at low gas velocities (3 m/s and 4 m/s). As shown in Fig. 16, the separation efficiencies of the compound and wave-plate demisters increase with rising inlet gas velocity but slightly drop when Ug,in surpasses 6 m/s. For the tube-bank demister, its separation efficiency slightly decreases when Ug,in rises from 3 m/s to 4 m/s but sharply declines with further increase of Ug,in until a bottom value. According to experimental observation, the initial decline of separation efficiency mainly took place along with the onset of droplet re-entrainment which was visible and perceivable at the outlet of demisters. It is indicated that, in the tube-bank demister, the re-entrainment occurs when Ug,in is approximately 4 m/s while for the wave-plate demister, it happens when Ug,in is near 7 m/s. It suggests that the tube bundle is more inclined to re-entrainment than wave plates. However, when it comes to the compound demister, the re-entrainment does not take place until Ug,in reaches about 7 m/s even though it incorporates the same tube bundle. In order to have a deeper insight into the special performance of the compound demister, the causes of re-entrainment and the comparison between different configurations have to be discussed in detail. According to previous work [30,44], re-entrainment may result from (1) the breakup of droplets due to interaction with the gas phase, (2) the impingement on water film, and (3) the detachment of liquid film caused by the synthetic action of body force and interfacial shear stress. When interacting with turbulent air flow, the stability of droplets can be estimated by the droplet Weber number [44]:  2 D d Wed ¼ ρg ug −ud σd

ð12Þ

The critical value of Wed for the breakup of droplets is 13 [45]. However, in this work, the possible maximum Wed for the droplets between wave plates and tubes are 2.356 and 4.867, respectively. Therefore, it is unlikely for the droplets to breakup due to gas turbulence in this study.

Fig. 18. Mechanism for the compound demister to remove droplets in experiments.

Y. Liu et al. / Desalination 409 (2017) 115–127

When colliding with liquid surface, the droplets might coalesce, bounce and splash [46]. According to previous research [47], increasing either droplet size or the impact velocity will improve the chance of splashing. Since the tube bundle is more competitive to accelerate suspending droplets, the breakup of droplets is more likely to take place on the surface of tubes than that of wave plates. The breakup of liquid film happens when the droplets are sheared off from the crest of film wave and get re-entrained by ambient air flow [30]. The minimum gas velocity, determining the possible generation of secondary droplets, can be described as [48]:

U g;cr;1

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2ðρd σ d gÞ0:5 ¼ ρg

U g;cr;2

Table 5 Comparison of droplet re-entrainment characteristics between demisters in this work and in literature. Reference Demisters

Approximate gas velocity of initial re-entrainment (m/s)

[21] [19] [34] [13] [35] This work

7 8 8 4 ≤3 7

Wave-plate Wave-plate Wave-plate Wire mesh Wire mesh Compound

3.4. Dry pressure drop and economic analysis ð13Þ

The minimum gas velocity for suspending a droplet is defined as [30]:  10:25 0 σ d g ρd −ρg A ¼ 2:46@ ρ2g

125

ð14Þ

For the air-water flow in this work, the Ug,cr,1 and Ug,cr,2 are 6.66 m/s and 11.57 m/s, respectively. The predicted maximum gas flow velocity (Ug,max) in the demisters is summarized in Fig. 17. It is shown that when inlet gas velocity is 4 m/s, the Ug,max in the compound demister and the tube-bank demister (about 11.5 m/s) approaches the maximum critical gas velocity (Ug,cr,max). But it does not occur in the wave-plate demister until the Ug,in gets to 6 m/s. Moreover, the region of maximum gas velocity appears near the surface of tubes (shown in Fig. 8). It suggests that the re-entrainment is more likely to take place at the space between tubes than wave plates. Further, compared with wave plates, the liquid film is more likely to form and grow on the tube surface due to its larger capability to collect droplets and relatively smaller superficial area. And the rapid flow of the liquid film and large liquid Reynolds number may occur on the tube walls, thus causing the breakup of water film [30]. The above analysis justifies that the tube bank is more prone to droplet re-entrainment than wave plates. However, with regard to the compound demister, the inlet gas velocity for the initial occurrence of re-entrainment is larger than the individual tube-bank demister. It is mainly because of the arrangement of downstream wave plates after the tube bank. According to literature [30,49], the maximum possible diameter of secondary droplets generated by liquid film breakup is 0.0054 m. Also, based on experimental observation, the droplets at the outlet of the tube-bank demister appeared much larger than the mist from spray nozzle when re-entrainment occurred. And the larger secondary droplets could be removed by the wave plates especially at large gas velocity [20]. Therefore, it is highly possible that, in the compound demister, the tube bank mainly collects the original fine mist from inlet while the wave plates help separate the larger secondary droplets re-generated by the tube bank as well as part of the initial mist, as shown in Fig. 18. Further, under the experimental conditions investigated, the compound demister possesses the highest separation efficiency of above 0.95 with the least fluctuation among all the three demisters, because it combines the advantages of tube bundle and wave plates (Fig. 16). Table 5 summarizes the comparison of the compound demister's droplet re-entrainment characteristics with other demisters in the open literature. As can be seen, the wave-plate demisters generally have higher gas velocity for the onset of droplet re-entrainment than the wire mesh demisters. However, due to the downstream wave plates, the compound demister still possesses high gas velocities for the initial re-entrainment, which are similar to the wave-plate demisters studied in the literature. Thus, the compound configuration is generally effective to resist droplet re-entrainment at relatively high gas velocities.

For practical applications, the pressure drop of the compound demister should also be evaluated in addition to the separation efficiency. As shown in Fig. 6, it is noticeable that the dry pressure drops of all demisters increase with inlet gas velocity. Under the same operating conditions, the compound demister has the highest dry pressure drop, followed by the tube-bank demister and wave-plate demister. In addition, the dry pressure drop for the compound demister is almost the sum of those of the two individual demisters. Fig. 19 compares the average dimensionless pressure loss—Euler number (Eu ¼ 1ρΔpv 2 )—of the compound demister in this work with 2 g g

some demisters reported in the open literature [12,50]. The smaller the Euler number is, the lower dry pressure drop the demister will have under the same working conditions. It is shown that the compound demister proposed in this work has smaller Euler number than the wave-plate demisters with multiple bends and drainage channels [12] or the wire mesh demister with pad thickness of 200 mm [50]. In this regard, Fig. 19 suggests that the compound demister is also acceptable to the desalination process from the point of view of dry pressure drop. Table 6 compares the dry pressure drops of the compound and individual wave-plate demisters with the same separation efficiency. It indicates that the wave-plate demister must contain multiple stages so as to achieve the same separation efficiency as the compound demister. With multiple stages, the pressure drop of the wave-plate demister will also increase sharply which can be significantly higher than the compound demister. It is suggested that the compound configuration is more economical to remove droplets than the individual wave-plate demister. In

Fig. 19. Comparison of average Euler number of the compound demister in this work with those of some demisters in literature (WP: wave-plate demister; WM: wire mesh demister).

126

Y. Liu et al. / Desalination 409 (2017) 115–127

Table 6 Comparison of the dry pressure drop between the compound and individual wave-plate demister with the same separation efficiency at Ug,in of 3 m/s. Droplet size (μm)

Separation efficiency (%)

Number of stages

10 15 20 30

85.88 100 100 100

1 1 1 1

Pressure drop (Pa)

Compound Wave-plate Compound Wave-plate 203 129 30 11

57.5 57.5 57.5 57.5

1847.3 1173.9 273 100.1

addition, the multistage wave-plate demister would consume a larger amount of materials for construction and larger space for installment than the compound demister. The multiple stages may also induce increasing complexity of maintenance such as fouling combating. Therefore, the compound demister is generally cost effective compared with the individual wave-plate demister. 4. Conclusion A novel compound demister to be applied in the MSF desalination plant was proposed by combining an upstream tube bank and downstream wave plates. Both experiments and simulation were carried out to evaluate its performance in terms of separation efficiency at low inlet gas velocities or for small droplets, re-entrainment characteristics and the dry pressure drop. The performance of the compound demister was compared with that of the individual wave-plate demister and tube-bank demister, as well as those demisters in the open literature. The mechanics for the droplet re-entrainment and high separation capability of the compound demister were also discussed. The obtained results are as follows. (1) The compound demister has the highest and most stable separation efficiency (N95%) under the experimental conditions investigated when compared with the individual tube-bank demister and wave-plate demister. (2) The compound demister has much higher separation efficiency at low inlet gas velocities or for small droplets than the individual wave-plate demister mainly due to the large separation capability of the tube bundle. (3) Different from the individual tube-bank demister, the compound demister shows higher resistance to re-entrainment which initially occurs at inlet gas velocity of about 7 m/s, mainly because the downstream wave plates can help remove most upstream large secondary droplets.

Acknowledgements The authors gratefully acknowledge the financial supports from the National Key Research and Development Program of China (No. 2016YFB0600601) and the National Natural Science Foundation of China (Nos. 51676075 and 51661125011). References [1] H. Ettouney, Brine entrainment in multistage flash desalination, Desalination 182 (2005) 87–97. [2] N. Ghaffour, T.M. Missimer, G.L. Amy, Technical review and evaluation of the economics of water desalination: current and future challenges for better water supply sustainability, Desalination 309 (2013) 197–207. [3] T. Mezher, H. Fath, Z. Abbas, A. Khaled, Techno-economic assessment and environmental impacts of desalination technologies, Desalination 266 (2011) 263–273. [4] A. Harabi, F. Zenikheri, B. Boudaira, F. Bouzerara, A. Guechi, L. Foughali, A new and economic approach to fabricate resistant porous membrane supports using kaolin and CaCO3, J. Eur. Ceram. Soc. 34 (2014) 1329–1340. [5] A. Harabi, F. Bouzerara, L. Foughali, B. Boudaira, A. Guechi, N. Brihi, Elaboration and characterization of low cost ceramics microfiltration membranes applied to the sterilization of plant tissue culture media, J. Taiwan Inst. Chem. Eng. 59 (2016) 79–85.

[6] F. Bouzerara, S. Boulanacer, A. Harabi, Shaping of microfiltration (MF) ZrO2 membranes using a centrifugal casting method, Ceram. Int. 41 (2015) 5159–5163. [7] B. Ghouil, A. Harabi, F. Bouzerara, B. Boudaira, A. Guechi, M.M. Demir, A. Figoli, Development and characterization of tubular composite ceramic membranes using natural alumino-silicates for microfiltration applications, Mater. Charact. 103 (2015) 18–27. [8] A. Harabi, E. Harabi, S. Chehlatt, S. Zouai, N.-E. Karboua, L. Foughali, Effect of B2O3 on mechanical properties of porous natural hydroxyapatite derived from cortical bovine bones sintered at 1,050 °C, Desalin. Water Treat. 57 (2016) 5303–5309. [9] I.C. Karagiannis, P.G. Soldatos, Water desalination cost literature: review and assessment, Desalination 223 (2008) 448–456. [10] N. Ghaffour, J. Bundschuh, H. Mahmoudi, M.F.A. Goosen, Renewable energy-driven desalination technologies: a comprehensive review on challenges and potential applications of integrated systems, Desalination 356 (2015) 94–114. [11] G. Venkatesan, N. Kulasekharan, S. Iniyan, Numerical analysis of curved vane demisters in estimating water droplet separation efficiency, Desalination 339 (2014) 40–53. [12] M.H. Estakhrsar, R. Rafee, Effects of wavelength and number of bends on the performance of zigzag demisters with drainage channels, Appl. Math. Model. 40 (2016) 685–699. [13] H.T. El-Dessouky, I.M. Alatiqi, H.M. Ettouney, N.S. Al-Deffeeri, Performance of wire mesh mist eliminator, Chem. Eng. Process. Process Intensif. 39 (2000) 129–139. [14] G. Venkatesan, N. Kulasekharan, S. Iniyan, Influence of turbulence models on the performance prediction of flow through curved vane demisters, Desalination 329 (2013) 19–28. [15] R. Rafee, H. Rahimzadeh, G. Ahmadi, Numerical simulations of airflow and droplet transport in a wave-plate mist eliminator, Chem. Eng. Res. Des. 88 (2010) 1393–1404. [16] G. Venkatesan, N. Kulasekharan, S. Iniyan, Design and selection of curved vane demisters using Taguchi based CFD analysis, Desalination 354 (2014) 39–52. [17] G. Venkatesan, N. Kulasekharan, V. Muthukumar, S. Iniyan, Regression analysis of a curved vane demister with Taguchi based optimization, Desalination 370 (2015) 33–43. [18] H. Al-Fulaij, A. Cipollina, G. Micale, H. Ettouney, D. Bogle, Eulerian–Lagrangian modeling and computational fluid dynamics simulation of wire mesh demisters in MSF plants, Desalination 385 (2016) 148–157. [19] J. Li, S. Huang, X. Wang, Numerical study of steam-water separators with wave-type vanes, Chin. J. Chem. Eng. 15 (2007) 492–498. [20] Y. Hao, J. Liu, Z. Yuan, L. Yang, Movement characteristics of droplets and demisting efficiency of mist eliminator, J. Chem. Ind. Eng. 65 (2014) 4669–4677. [21] G. Lei, Z. Yuan, L. Yang, Z. Gu, Numerical study on the penetration of droplets in a zigzag demister, Environmental Engineering Science 33 (2016). [22] J. Zhao, B. Jin, Z. Zhong, Study of the separation efficiency of a demister vane with response surface methodology, J. Hazard. Mater. 147 (2007) 363–369. [23] C. Galletti, E. Brunazzi, L. Tognotti, A numerical model for gas flow and droplet motion in wave-plate mist eliminators with drainage channels, Chem. Eng. Sci. 63 (2008) 5639–5652. [24] Y. Wang, P.W. James, Assessment of an eddy-interaction model and its refinements using predictions of droplet deposition in a wave-plate demister, Chem. Eng. Res. Des. 77 (1999) 692–698. [25] Y. Wang, P.W. James, The calculation of wave-plate demister efficiencies using numerical simulation of the flow field and droplet motion, Chem. Eng. Res. Des. 76 (1998) 980–985. [26] A.S. Al-Dughaither, A.A. Ibrahim, W.A. Al-Masry, Investigating droplet separation efficiency in wire-mesh mist eliminators in bubble column, J. Saudi Chem. Soc. 14 (2010) 331–339. [27] R. Kouhikamali, S.M.A. Noori Rahim Abadi, M. Hassani, Numerical study of performance of wire mesh mist eliminator, Appl. Therm. Eng. 67 (2014) 214–222. [28] I. Janajreh, A. Hasania, H. Fath, Numerical simulation of vapor flow and pressure drop across the demister of MSF desalination plant, Energy Convers. Manag. 65 (2013) 793–800. [29] E. Brunazzi, A. Paglianti, Design of wire mesh mist eliminators, AICHE J. 44 (1998) 505–512. [30] C.C.J. Verlaan, Performance of Novel Mist Eliminators, TU Delft, Delft University of Technology, 1991. [31] B.J. Azzopardi, K.S. Sanaullah, Re-entrainment in wave-plate mist eliminators, Chem. Eng. Sci. 57 (2002) 3557–3563. [32] P.W. James, B.J. Azzopardi, Y. Wang, J.P. Hughes, A model for liquid film flow and separation in a wave-plate mist eliminator, Chem. Eng. Res. Des. 83 (2005) 469–477. [33] X. Huang, Experimental Study and Optimal Modification of the Performance of Desulfurization Mist Eliminator, North China Electric Power University, Beijing, 2010. [34] E. Narimani, S. Shahhoseini, Optimization of vane mist eliminators, Appl. Therm. Eng. 31 (2011) 188–193. [35] Y. Yao, A. Pavlenko, O. Volodin, Effects of layers and holes on performance of wire mesh packing, J. Eng. Thermophys. 24 (2015) 222–236. [36] C. Wang, X. Liu, D. Li, J. Si, B. Zhao, M. Xu, Measurement of particulate matter and trace elements from a coal-fired power plant with electrostatic precipitators equipped the low temperature economizer, Proc. Combust. Inst. 35 (2015) 2793–2800. [37] W.J. Dixon, FJ Massey Introduction to Statistical Analysis, in, McGraw-Hill, New York, 1957. [38] R.J. Moffat, Describing the uncertainties in experimental results, Exp. Thermal Fluid Sci. 1 (1988) 3–17. [39] F. Kavousi, Y. Behjat, S. Shahhosseini, Optimal design of drainage channel geometry parameters in vane demister liquid–gas separators, Chem. Eng. Res. Des. 91 (2013) 1212–1222.

Y. Liu et al. / Desalination 409 (2017) 115–127 [40] ANSYS, ANSYS FLUENT Theory Guide, Release 16.1, 2015. [41] K. Cen, J. Fan, Theory and Calculation of the Industrial Gas-Solid Mutiphase Flow, Zhejiang University Press, Hangzhou, 1990. [42] P.G. Saffman, J.S. Turner, On the collision of drops in turbulent clouds, J. Fluid Mech. 1 (1956) 16–30. [43] J.A. Schetz, A.E. Fuhs, Fundamentals of Fluid Mechanics, John Wiley & Sons, 1999. [44] T. Nakao, M. Nagase, G. Aoyama, M. Murase, Development of simplified wave-type vane in BWR steam dryer and assessment of vane droplet removal characteristics, J. Nucl. Sci. Technol. 36 (1999) 424–432. [45] T. Ueda, Two-Phase Flow-Flow and Heat Transfer, Yokendo, Japan(in Japanese) 1981.

127

[46] M. Rein, Phenomena of liquid drop impact on solid and liquid surfaces, Fluid Dyn. Res. 12 (1993) 61. [47] A. Wang, C. Chen, Splashing impact of a single drop onto very thin liquid films, Phys. Fluids 12 (2000) 2155–2158. [48] G. Hetsroni, Handbook of Multiphase Systems, McGraw Hill Book Company, 1982. [49] Y. Taitel, D. Bornea, A. Dukler, Modelling flow pattern transitions for steady upward gas-liquid flow in vertical tubes, AICHE J. 26 (1980) 345–354. [50] R. Rahimi, D. Abbaspour, Determination of pressure drop in wire mesh mist eliminator by CFD, Chem. Eng. Process. Process Intensif. 47 (2008) 1504–1508.