Scale-up Design of a Spacer-Filled Disk-Type Membrane Module Using CFD

Transcription

1 Journal of Water Sustainability, Volume 1, Issue 1, June 2011, University of Technology Sydney & Xi an University of Architecture and Technology Scale-up Design of a Spacer-Filled Disk-Type Membrane Module Using CFD Yu-Ling Li, Kuo-Lun Tung * R&D Center for Membrane Technology and Dept. of Chemical Engineering Chung Yuan University, Chung-Li, Taoyuan 320, Taiwan ABSTRACT The scale-up design of a spacer-filled disk-type membrane module has been conducted by using computational fluid dynamic (CFD) technique. A three-dimensional CFD technique was used to analyze fluid flow in the membrane module, permeate flux, permeate volumetric flow rate and the distribution of permeation rates on the membrane surfaces. The numerical results showed that there were variations in the volumetric flow rate and permeation rates with different membrane module sizes. As radius ratio of the collection tube to radius of the spacer-filled disk-type membrane module was smaller than 0.162, a series of membrane module could be considered instead of large membrane module sizes based on the same total membrane areas. Microscopic understanding derived from the CFD analysis can improve the design of collection tubes, spacer thicknesses and membrane module sizes to enhance module performance and assist the construction of new designs. Keywords: Membrane module, Membrane bioreactor, CFD, Spacer-filled disk, Wastewater treatment 1.0 INTRODUCTION Flat-sheet membranes and tubular membranes are fabricated in several membrane modules. Plate and frame, rotating disk, spiral-wound configurations, annular-gap dynamic membrane modules and pleated membrane cartridges are made with flat-sheet membranes; tubular, capillary and hollow-fiber geometries are made with tubular-membrane modules. Hence, membrane modules are chosen based on various treated targets. Computational fluid dynamics (CFD) has been widely used to understand the hydrodynamic behavior of membrane processes, including membrane modules (Schwinge et al., 2004; Ghidossi et al., 2006a). Therefore, the optimum design of * Corresponding to: kuolun@cycu.edu.tw membrane modules is an important topic when using the CFD technology. A schematic drawing of a plate-and-frame module is described in Figure 1. Two membrane and feed spacers are placed in a sandwich with their feed sides facing each other. Flat and frame membrane modules were one of the earliest types of membrane modules and were widely used in the separation process. It has been also used widely in a membrane bioreactor system. The optimization of the flat sheet module design and operating conditions despond on the geometry (membrane plate spacing) and aeration intensity (bubble size) was obtained from the experimental and CFD investigations results (Drews et al., 2010). The two-phase flow in a submerged flat sheet membrane module system was identified as the most effective flow profiles for fouling mitigation by using a gas-liquid two-phase CFD model (Ndinisa et al., 2006).

2 60 Y. L. Li, K. L. Tung / Journal of Water Sustainability 1 (2011) Spacer Membrane Permeate Permeate Membrane Figure 1 A schematic drawing of a plate-and-frame membrane module. Spiral wound modules are constructed by using two flat membranes placing together with their active sides facing away from each other as shown in Figure 2. They are separated by a sheet of permeate collection material. Another feed channel spacer is placed on either side of the envelope. Then, this whole assembly is rolled around in a perforated center tube in a spiral or jelly-roll assembly. The feed solution is then pumped in from one side along the tube. The permeate and the concentrate drained from the other side. Substantial literature describes the complex structure of the spiral-wound membrane module (Li et al., 2002; Schwinge et al., 2004). Li and Tung (2008a) pointed that the appropriate cell types and periodic boundary conditions have been suggested based on the three-dimensional CFD analysis of spacer-filled membrane module designs with various spacer arrangements. Li and Tung (2008b) noted that the curvature of a spacer-filled channel affected the flow field in spiral-wound membrane modules. They also found that the shear stress at the inner wall was greater than that at the outer wall in a curved, spacer-filled channel by using a two-dimensional numerical scheme. The particulate deposition on the membrane surface in a spacer-filled channel was investigated by a CFD technique (Li et al., 2006). In addition, Li et al. (2009) showed that a three-dimensional CFD technique and an experimental setup with a curved channel filled with a two-layer-filament spacer were used to understand the fluid flow in the channel. A spacer with unequal filament diameters between the inner layer and outer layers was adopted owing to mitigating the curvature effect of the spacer-filled channel in a spiral-wound membrane module. This type of spacer could be considered to reduce the imbalance in shear stress between the inner and outer walls so as to extend the life of a membrane module. Tubular modules are usually made with ce-

3 Y. L. Li, K. L. Tung / Journal of Water Sustainability 1 (2011) ramic membranes. The skin layer is cast from the tubular supporter of stainless steel or ceramic (Figure 3 (a)) and the channels of a monolith (Figure 3 (b)). Several configurations by changing parameters, including the diameter, membrane geometry and the form of the channels (cylindrical, square-section, triangular, hexagonal, etc.), were studied by with CFD to efficiently increase the area of the membrane unit containing the ceramic membranes (Ghidossi et al., 2010). The hollow-fiber module has been widely used in separation processes. A schematic drawing of a hollow-fiber module is depicted in Figure 4. Base on different operation conditions, there are two kinds of hollow-fiber modules: Inside-out (Tube-side feed) (Figure 4 (a).) and Outside-In (Shell-side feed) (Figure 4 (b).). Controlling production quantities, mitigating fouling and clogging phenomena depended on selecting a suitable operating pressure and fiber locations within the module. The hydraulic-path variation within the connection box for the hollow-fiber module was caused by the geometry of the permeate outlet (Glucina et al., 2009). The CFD results found vortex zones at the side where is opposite to the feed inlet. Owing to the increasing recirculation velocity in these vortex zones, the risk of clogging should be reduced. Ghidossi et al. (2006b) determined the degree of clogged hollow fibers and estimated modular energy consumption by using a CFD technique to calculate the pressure drop with adjustments to the inlet velocities, inlet pressures, internal diameters and permeabilities of the hollow fibers. However, little emphasis has been placed on studying the membrane module size and its effect on the performance of a membrane module. In this study, a spacer-filled disk-type membrane module was optimized by considering the collection-tube size, spacer thickness and membrane module sizes. Fluid flow in the membrane module and permeate flux, permeate volumetric flow rate and the distribution of permeation rates on the membrane surfaces were analyzed by a three-dimensional CFD technique. The purpose was to obtain the optimum conditions and configurations to yield a module design with maximum performance. Figure 2 A schematic drawing of a spiral-wound membrane module (Li et al., 2006).

4 62 Y. L. Li, K. L. Tung / Journal of Water Sustainability 1 (2011) Retentate Permeate Feed (a) stainless steel or ceramic supporter (b) monolith supporte Figure 3 A schematic drawing of a tubular membrane module.

5 Y. L. Li, K. L. Tung / Journal of Water Sustainability 1 (2011) (a) Inside-out or Tube-side feed. (b) Outside-in or shell-side feed. Figure 4 A schematic drawing of a hollow fiber membrane module. 2.0 THEORETICAL STUDY 2.1 The Simulated System Two membrane layers, one spacer layer and a collection tube were established in the spacer-filled disk-type membrane module, as shown in Figure 5 (a). The membrane structure contained a skin layer and a support layer. The selective skin layer was a laminated polyacrylonitrile (PAN) nanofiber membrane with an average pore size of approximately 0.27 μm. A nonwoven polyethylene terephthalate (PET) fabric was regarded as the support layer. This geometry configuration was symmetrical, as shown in Figure 5 (b). Faces ABHG, BCIH, JKED, KLFE and CILF were defined as symmetrical planes. Faces GHKJ, ABED and BCFE were set as the walls. Faces ADJG and HKLI were regarded as the pressure inlet and outlet boundaries, respectively. In addition, Zones AGJDEBHK and BCIHKLFE were the membrane layer and spacer layer individually. In order to minimize the large computational time, this simplified system would be adopted. The geometric parameters of the spacer-filled disk-type membrane module (i.e., membrane radius, membrane thickness, collection-tube radius and spacer thickness) used in this study are listed in Table 1. The body-fitted structure grids were used for the CFD analysis in this study. The computational grids with 81, ,500 cells for the different cases were used as shown in Figure 5 (c).

6 64 Y. L. Li, K. L. Tung / Journal of Water Sustainability 1 (2011) Symmetry Symmetry Collection tube Spacer Top view (a) Cross sectional view (b) (c) Figure 5 Schematic diagrams of the spacer-filled disk-type membrane module: (a) construction, (b) simulated system and (c) computational grids.

7 Y. L. Li, K. L. Tung / Journal of Water Sustainability 1 (2011) Table 1 The geometric parameters of the spacer-filled disk-type membrane module. Parameter Value (mm) Disk-type membrane radius (R) 092.5, 185.0, Membrane thickness (h m ) , 10, 15, 20, 25, 30 (R = 092.5) Collection-tube radius (R tube ) 10, 20, 30, 40, 50, 60 (R = 185.0) 15, 30, 45, 60, 75, 90 (R = 277.5) Spacer thickness (h sp ) 0.50, 0.75, 1.00, Governing Equations and Numerical Calculations The simulation model was assumed to be at steady state and isothermal; the wall was assumed to have a no-slip boundary condition. The flow field was analyzed by solving the continuity equation and the momentum-balance equations of the system. The governing equations for the steady-state fluid flow in the spacer-filled disk-type membrane module are the continuity and momentum equations, i.e.: (I) Continuity Equation u = 0 (1) (II) Momentum Equation ρ Du = p + μ 2 u (2) Dt The membrane-permeation effect was investigated to study the fluid flow through the membrane module. The transmembrane pressure (TMP) gradient across the membrane was regarded as the driving force, and the permeate flux J was defined by Darcy s equation as follows: TMP J = μ ( R m + R c ) (3) where R m is the resistance of the clean membrane. Because a single phase was used in this study, the resistance of the cake, R c, would be neglected. In addition, the resistance of the spacer was also neglected since its value was far smaller than that of membrane. The SIMPLEC (semi-implicit method for pressure-linked equations, consistent) algorithm and the QUICK differencing scheme with the velocity and pressure components were used to analyze the flow field. Water was regarded as the fluid in this simulation study. The sum of the normalized residuals of all variables converged to The flow field in the membrane module was calculated by using the commercially available CFD software FLUENT. 3.0 RESULTS AND DISCUSSION Three spacer-filled disk-type membrane module sizes (R = 92.5, 185 and mm), six values of the collection-tube size (R * = 0.054, 0.108, 0.162, 0.216, 0.27 and 0.324) and four spacer thicknesses (h sp = 0.50, 0.75, 1.00 and 1.50 mm) were chosen in the analysis. Based on the simulation results obtained from the module, the effects of the collection-tube size, spacer thickness, membrane module sizes on the permeate flux, permeate volumetric flow rate, permeation-rate distributions and dimensionless permeation rate are discussed comprehensively in the following sections. 3.1 The Effect of Membrane Module Configurations on the Performance of The Membrane Module In order to obtain the R m value in advance for this simulation work, it would be estimated from a preliminary dead-end filtration experiment (Tung et al., 2010). According to the

8 66 Y. L. Li, K. L. Tung / Journal of Water Sustainability 1 (2011) flux and TMP values using Eq. (3), the R m values were obtained. In addition, the flux was calculated by CFD with the experimental R m value. Figure 6 describes that the CFD simulation data agreed well with the experimental results. It also confirmed the validity of the simulation assumptions. Hence, the R m value of /m was used to calculate the flow field in this work at a TMP values of 1.0 bar. CFD Exp. data J.10 3 (m 3 / m 2. s) TMP (bar) Figure 6 The pure-water flux at various TMP values J.10 5 (m 3 / m 2. s) Q (m 3 / day) R tube (mm) Figure 7 The effect of collection-tube size on the permeate flux and permeate volumetric flow rate for pure-water at a TMP of 1.0 bar and an h sp of 0.75 mm.

9 Y. L. Li, K. L. Tung / Journal of Water Sustainability 1 (2011) The simulated system of the spacer-filled disk-type membrane module is indicated in Figure 5 (b). The effect of six values of the collection-tube size on the permeate flux and permeate volumetric flow rate at a TMP of 1.0 bar and a spacer thicknesses of 0.75 mm is shown in Figure 7. As the collection-tube size increased, the permeate flux also increased. There was an apparent increase in the permeate flux when increasing the tube size up to 15 mm. Only slight flux increased when the size of the collection-tube was more than 15mm. One spacer-filled channel and two membrane layers on each side were fabricated in the disk-type membrane module. Because of different collection-tube sizes and spacer thicknesses, the filtration area and the total collection volume differed among the modules. In addition to the permeate flux, the permeate volumetric flow rate should be considered owing to the difference in the filtration area and spacer thickness. Figure 7 illustrates the effect of collection-tube size on the permeate volumetric flow rate at a TMP of bar and a spacer thickness of 0.75 mm. There was a maximum value for the permeate volumetric flow rate with a collection-tube size of 15 mm. As the collection-tube size increased, the permeate flux would be enhanced; however, the filtration area decreased. Consequently, the permeate volumetric flow rate decreased after a sharp increase in the smaller collection-tube sizes. Four spacer thicknesses ranging from 0.5 mm to 1.5 mm were simulated in this study. The effect of the spacer thicknesses on the permeate flux and permeate volumetric flow rate at a TMP of 1.0 bar and a collection-tube size of 15 mm is shown in Figure 8. As the spacer thickness was increased, the permeate flux and the permeate volumetric flow rate also increased. However, the cost of pumping power and spacer thickness should be considered. The pumping power and spacer cost are increased as the spacer thickness is increased. Therefore, the last resort for improving the performance of a spacer-filled disk-type membrane module should be increasing the spacer thickness. 24 J.10 5 (m 3 / m 2. s) Q (m 3 / day) h sp (mm) Figure 8 The effect of spacer thickness on the permeate flux and permeate volumetric flow rate for pure-water at a TMP of 1.0 bar and an Rtube of 15 mm.

10 68 Y. L. Li, K. L. Tung / Journal of Water Sustainability 1 (2011) The Effect of Membrane Module Sizes on the Performance of The Membrane Module Many membrane module sizes depended on various requirements; three values of membrane module size were considered in Table 1. Table 2 shows the effect of membrane module sizes with the permeate volumetric flow rate and different spacer thicknesses. As the spacer-filled disk-type membrane module sizes (R) increased to and mm, the membrane area would be enhanced to fourfold and ninefold membrane areas based on the R value of 92.5 mm individually. As the membrane areas increased, the permeable volumetric flow rate would be enhanced. However, the permeable volumetric flow rate would not rise to fourfold and ninefold quantity even with an increase of spacer thicknesses. With the same spacer thickness, the relationship between membrane areas and Q i / Q 1 would be closed to the straight line if the collection-tube size increased as shown in Figure 9. As the line was not straight, a series of membrane module could be considered rather than large membrane module sizes based on the same total membrane areas. Table 2 The permeate volumetric flow rate at various spacer thicknesses and membrane module sizes. h sp (mm) R (mm) Q (m 3 /Day) ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ In addition to the pure water flux, the distribution of permeation rates on the membrane was estimated for the performance of the spacer-filled disk-type membrane module. Effects of the collection-tube sizes and membrane module size on the distribution of the permeation rates on the membrane at an h sp of 0.75 mm and a TMP of 1.0 bar are shown in Figure 10. The permeation rates ranged from to m/s and are shown here in color-contoured deciles. The ratius radio of the collection tube to radius of the spacer-filled disk-type membrane module was defined as: R R * tube = (4) R There was a wide distribution of permeation rates; the lowest permeation rate was at an R * of 0.054; especially for an R of 277.5

11 Y. L. Li, K. L. Tung / Journal of Water Sustainability 1 (2011) mm (Figure 10 (c)). The distribution of the permeation rates on the membrane was not uniform. The widest range of permeation rates ( m/s) on the membrane was at an R of and an R of mm (Figure 10 (c)). The narrowest range of permeation rates ( m/s; a single color contour) on the membrane were at cases of Figs. 10 (d), (g), (j), (m), (n), (p), (q). Therefore, the distribution of the permeation rates on the membrane was non uniform and was improved by increasing the collection-tube size. In addition, the distribution of the permeation rates on the membrane was more deteriorated by enhancing the membrane module size. In order to quantify the variation of the permeation rates, the dimensionless radius and the dimensionless permeation rate, ξ and ζ, respectively, were defined as: r Rtube ξ = (5) R Rtube u ζ = (6) u max Here, a ξ value close to 1 reveals a position near the exterior of the membrane module, 12 and a ξ value close to 0 is a position near the collection tube. Figure 11 shows the effect of the dimensionless radius on the dimensionless permeation rate at an h sp of 0.75 mm and a TMP of 1.0 bar with various collection-tube sizes and membrane module sizes. The ζ value would decrease as the ξ value increased. The ζ value kept a constant value at a ξ value of 0.5 to 1.0. In other words, the variation of the permeation rates was not evident in this range. In addition, the variation of permeation rate is greater at lower ζ values. The highest permeation rate occurred near the collection tube because the suction force was enhanced when the distance from the collection tube was reduced. The variation of the permeation rates would be more obvious with the enhancement of membrane module sizes rather than collection-tube sizes. However, the ζ value was maintained at 0.8 when the R * value was larger than 0.162; even if an R of mm. Therefore, a series of membrane module could be considered rather than large membrane module sizes based on the same total membrane areas as the R * value was smaller than Q i / Q 92.5 (-) 6 3 Figure A (m 2 ) The relationship between membrane areas and Qi / Q92.5 at a TMP of 1.0 bar.

12 70 Y. L. Li, K. L. Tung / Journal of Water Sustainability 1 (2011) R * = R * = (a) R tube = 5 mm, R = 92.5 mm (d) R tube = 10 mm, R = 92.5 mm (b) R tube = 10 mm, R = 185 mm (e) R tube = 20 mm, R = 185 mm (c) R tube = 15 mm, R = mm (f) R tube = 30 mm, R = mm Figure 10 The effect of membrane module size with various collection-tube sizes on the permeation rate at a TMP of 1.0 bar and an h sp of 0.75 mm

13 Y. L. Li, K. L. Tung / Journal of Water Sustainability 1 (2011) R * = R * = (g) R tube = 15 mm, R = 92.5 mm (j) R tube = 20 mm, R = 92.5 mm (h) R tube = 30 mm, R = 185 mm (k) R tube = 40 mm, R = 185 mm (i) R tube = 45 mm, R = mm (l) R tube = 60 mm, R = mm Figure 10 Continued

14 72 Y. L. Li, K. L. Tung / Journal of Water Sustainability 1 (2011) R * = R * = (m) R tube = 25 mm, R = 92.5 mm (p) R tube = 30 mm, R = 92.5 mm (n) R tube = 50 mm, R = 185 mm (q) R tube = 60 mm, R = 185 mm (o) R tube = 75 mm, R = mm (r) R tube = 90 mm, R = mm Figure 10 Continued

15 Y. L. Li, K. L. Tung / Journal of Water Sustainability 1 (2011) ζ (-) ξ (-) Figure 11 The effect of the dimensionless radius on the dimensionless permeation rate at a TMP of 1.0 bar and an h sp of 0.75 mm with various membrane module sizes and R*. 4.0 CONCLUSIONS Effects of varying membrane module configurations and sizes on the fluid flow through a spacer-filled disk-type membrane module were analyzed by the CFD technique. Various structural configurations and module sizes including R * values, spacer thicknesses and membrane module sizes were modeled in this study. The numerical results of the permeate flux, the permeate volumetric flow rate, the distribution of permeation rates and the variation of permeation rates showed that variations in the structural configuration and membrane module size caused inherent changes in the hydrodynamic behavior. The enhancement of permeate volumetric flow rate could use a series of membrane modules rather than increasing membrane module sizes as the R * value was smaller than The study also showed that the CFD tool could enhance the performance of a membrane module and promoted new module designs. ACKNOWLEDGEMENTS We thank the Center-of-Excellence (COE) Program on Membrane Technology with the Ministry of Education (MOE), R.O.C., the Taiwan Textile Research Institute project, and the National Science Council (NSC) for their financial support. J p Q Q i NOMENCLATURE permeate flux, m 3 /m 2 day pressure, Pa permeate volumetric flow rate, m 3 /day permeate volumetric flow rate based

16 74 Y. L. Li, K. L. Tung / Journal of Water Sustainability 1 (2011) on an R of i mm, with i (92.5, and 277.5) Q 92.5 permeate volumetric flow rate based on an R of 92.5 mm. r radial position of the disk-type membrane module, mm R radius of the spacer-filled disk-type membrane module, mm R c resistance of the cake, 1/m R m resistance of the clean membrane, 1/m R tube radius of the collection tube, mm R * the ratio of R tube to R t time, s u permeation rate, m/s u max maximum permeation rate, m/s Greek letters ζ dimensionless permeation rate ξ dimensionless radius ρ fluid density, kg/m 3 μ fluid viscosity, kg/m-s REFERENCES Drews A., Prieske H., Meyer E.-L., Senger G. and Kraume M. (2010). Advantageous and detrimental effects of air sparging in membrane filtration: Bubble movement, exerted shear and particle classification. Desalination, 250, Ghidossi R., Veyret D. and Moulin P. (2006a). Computational fluid dynamics applied to membranes: State of the art and opportunities. Chem. Eng. Process., 45, Ghidossi R., Daurelle J.V., Veyret D. and Moulin P. (2006b). Simplified CFD approach of a hollow fiber ultrafiltration system. Chem. Eng. J., 123, Ghidossi R., Carretier E., Veyret D., Dhalerd D. and Moulinb P. (2010). Optimizing the compacity of ceramic membranes. J. Membr. Sci., 360, Glucina K., Derekx Q., Langlais C. and Laine J.-M. (2009). Use of advanced CFD tool to characterize hydrodynamic of commercial UF membrane module. Desalination Water Treat., 9, Li F., Meindersma W., de Haan A.B. and Reith T. (2002). Optimization of commercial net spacers in spiral wound membrane modules, J. Membr. Sci., 208 (1-2), Li Y.L., Chang T.H., Wu C.Y., Chuang C.J. and Tung K.L. (2006). CFD analysis of particle deposition in the spacer-filled membrane module. J. Water Supply Res. Technol.-Aqua, 55 (7-8), Li Y.L. and Tung K.L. (2008a). CFD simulation of fluid flow through spacer-filled membrane module: selecting suitable cell types for periodic boundary conditions. Desalination, 233, Li Y.L. and Tung K.L. (2008b). The effect of curvature of a spacer-filled channel on fluid flow in spiral-wound membrane modules, J. Membr. Sci., 319 (1-2), Li Y.L., Tung K.L., Lu M.Y. and Huang S.H. (2009). Mitigating the curvature effect of the spacer-filled channel in a spiral-wound membrane module. J. Membr. Sci., 329 (1-2), Ndinisa N.V., Fane A.G., Wiley D.E. and Fletcher D.F. (2006). Fouling control in a submerged flat sheet membrane system: Part II-Two-phase flow vharacterization and CFD simulations. Sep. Sci. Technol., 41 (7), Schwinge J., Neal P.R., Wiley D.E., Fletcher D.F. and Fane A.G. (2004). Spiral wound modules and spacers. Review and analysis. J. Membr. Sci., 242, Tung K.L., Li Y.L., Wang S.J., Nanda D., Hu C.C., Li C.L., Lai J.Y. and Huang J. (2010) Performance and effects of polymeric membranes on the dead-end microfiltration of protein solution during filtration cycles. J. Membr. Sci., 352 (1-2),