Journal of Scientific & Industrial Research Vol. 77, June 2018, pp. 353-358 Ejector Pump CFD Model Validation and Performance Improvement Studies B H Arun 1, J Vyas 2 *, B M Gopalsamy 2 and M Chimmat 2 1 Department of Mechanical Engineering, UVCE, Bangalore, India 2 Structural Technologies Division, NAL, CSIR, Bangalore, India Received 31 May 2017; revised 18 December 2017; accepted 04 April 2018 The ejector pump is the noiseless operating and the simplest of all pumping devices with no moving parts involved. Ejector pumps are used in many pumping applications owing to their low first cost, simplicity in operation and the ability to mix fluids thoroughly. Though the ejector pump has advantages, more efforts need to be done to improve the efficiency. In the present study, an experimental test rig is set up for a single-phase (incompressible fluid) ejector pump. A CFD simulation model has been developed on ANSYS-FLUENT platform for the ejector pump and results are validated with the established test rig data. Further, the validated CFD model is applied for geometry optimization studies to maximize the efficiency. Critical parameters affecting the efficiency are identified and reported. Keywords: Ejector Pump, Optimization, Computational Fluid Dynamics, Design of Experiments, Efficiency Introduction The ejector/jet pump is a device for pumping fluids by means of a high-velocity jet of the same or a different kind of fluids. The principle of operation of a jet pump is the transfer of energy and momentum from primary to the secondary fluid through the process of turbulent mixing. The high pressure primary fluid is accelerated to high velocity through a nozzle as a result primary fluid static pressure decreases and creates a suction environment at the nozzle exit. There by, the secondary fluid gets entrained and mixed with the primary fluid in the constant diameter throat section. The mixed fluids then pass through the diffuser in which a portion of the velocity head is converted to static pressure. Jet pumps are being used in many applications such as Aeration, Cleaning and Reducing Turbidity, Domestic Water Supply, Dredging, Lubrication Systems, Mixing, Multi-pressure Systems, Nuclear Industry, Priming Devices, Pumping Sludge's, Solids Transport, Thrust Augmenters and Aircraft Fuel System 1. In Aircraft fuel systems, the ejector pumps are used as scavenge pumps. Scavenge pumps are ejector devices that use the high-pressure discharge from the main feed pumps to scavenge fuel from the corners of the tank. The performance of the jet pump is expressed in terms of flow ratio (M) and the head ratio (N), and efficiency (η). *Author for Correspondence E-mail: jaidev.vyas@nal.res.in η = M X N Flow ratio (M) is defined as the ratio of flow rate of the secondary fluid (Q 2 ) to the flow rate of the primary fluid (Q 1 ). Head ratio (N) is defined as the ratio of the increase in head of the secondary fluid (P 3 -P 2 ) to the decrease in head of the primary fluid (P 1 -P 3 ), where P 1, P 2 and P 3 are the total pressure values at the primary fluid inlet, secondary fluid inlet and the discharge outlet respectively. The performance of jet pumps depends on the geometric parameters like area ratio, setback distance, mixing tube length, and shape of the driving nozzle. These parameters affect the flow ratio in jet pumps, which in turn have the influence on their efficiency 2-4. Improving the mode of entry of the suction fluid to the pump and the use of stepped diffuser will result in increased efficiency of the ejector pump 5. CFD studies can be performed to analyze the ejector pump performance but selection of the appropriate turbulence model is of prime importance. Among the available alternatives (the Standard, RNG and Realizable k-epsilon models, the standard and SST, k- omega models, RSM model and two wall treatment methods) the realizable k-epsilon model with standard wall function gives better performance prediction 6. Experimental setup The test rig has been setup to evaluate the ejector pump single phase model. The tests have been done under different input pressures in order to characterize
354 J SCI IND RES VOL 77 JUNE 2018 ejector pump performance. The test setup comprises of a boost pump, ejector, tanks, three pressure gauges and two flow meters to read the data. After establishing test rig, the experiments are being carried out under varying pressure, and their results are recorded. Figure 1 shows the experimental setup. The boost pump provides motive flow to ejector pump from Tank 1 and suction/secondary flow is derived from Tank 2. The pressure gauges and rotameters measure the pressure and the flow rate respectively of both motive flow and the secondary flow. The flow-regulating valve on a motive side is used to control the amount of motive flow. The two tanks are interconnected to make sure that water from Tank 2 does not overflow and ejector pump operates in a closed cycle. The Table 1 gives the calibrated specimen readings obtained from the test rig. The recorded pressures are gauge. Numerical methodology Numerical simulations have been carried using ANSYS-FLUENT CFD package. The realizable k ε turbulence model is used to model the turbulence effects. Governing equations Governing equations solved are the mass and momentum conservation equations. The governing equations for a steady, three-dimensional, incompressible viscous flow is given below 7-15. The dissipative, mass diffusion and thermal conductivity are neglected.... (1)... (2)... (3)... (4) Where is the velocity vector. f x, f y and f z are the x, y and z components of body force per unit mass acting on the fluid element respectively. xx, yy, zz are the normal stresses acting along x, y and z-directions respectively. xy, yz, zx are the shear stresses acting on xy, yz and zx plane respectively. Equation (1) represents the continuity and (2) to (4) represent the fluid momentum balance in x, y and z directions respectively. Since operating fluids are at room temperature and there is no heat addition involved, the energy equation is not solved for temperature. For this study, the realizable k-ε turbulence model with enhanced wall function is used. The equations for turbulent kinetic energy (k) and dissipation ( rate are given below... (5)... (6) Geometry To accomplish an exact simulation, the same dimensions of the tested jet pump are used. The major dimensions of the ejector pump are: Jet inlet diameter = 10.85mm, Sump inlet diameter = 24 mm, Jet outlet diameter = 24mm, Diameter of nozzle exit (Dn) = 5.05mm, Diameter of throat (Dt) = 12.18mm, Spacing between the nozzle and throat (S) = 14.982 mm, Length of the throat (L) = 41.501 mm, Diffuser angle = 12.94deg SL.NO. Fig. 1 Experimental setup Table 1 Calibrated specimen readings obtained from the test rig Grid details Initially, a grid of 342019 nodes and 1.09 million elements are generated. Grid independence study is carried out to ensure the minimum number of grid nodes is used to get a convergent solution. Three Motive Fluid Side Suction Side Discharge Side Pressure (Kgf/cm 2 ) Flow Rate (LPM) Pressure (Kgf/cm 2 ) Flow Rate (LPM) Pressure (Kgf/cm 2 ) Flow Rate (LPM) 1 2.69271 25.47-0.2511 17.5202 0.251 42.9902 2 2.29379 23.4324-0.220 968 15.9743 0.2008 39.4067 3 1.49595 19.3572-0.190836 12.8825 0.1004 32.2397
VYAS et al.: EJECTOR PUMP CFD MODEL VALIDATION AND PERFORMANCE STUDIES 355 different grids (342019, 485516 and 645472) are checked for the entire jet pump. The grid size of 485516 grid nodes is found to be optimum grid size and further analysis has been done considering this grid size. Boundary conditions The boundary conditions of fluid flow are the same as that of the experimental work. The pressure gauge data of the primary and secondary streams are converted into total pressure boundary by adding the dynamic pressure and specified at the corresponding inlets. At the diffuser outlet, static pressure boundary condition is assigned based on the test reading. Turbulence intensities of 4%, 0.25% and 6% have been applied to jet inlet, sump inlet and jet outlet boundaries respectively. For one of the specimen cases, Total pressure at jet inlet is 274572.7 Pa, sump inlet total pressure is -24416.8 Pa, outlet static pressure of 24614.69 Pa is assigned and no slip wall condition with zero wall roughness is considered. Solver settings In solver settings, Pressure correction algorithm used is SIMPLE. Spatial Discretization factors chosen for numerical simulations for u, v and p are Standard, while for turbulent quantity it is Power law. Under- Relaxation factors chosen for both u and v are 0.7 and 0.3 for p. All turbulent quantities (TKE, TDR, TV) are 0.8. The obtained results are assumed to be converged for the residual level of 10-6 for continuity, x- velocity and y-velocity. Validation of numerical simulation The simulation results obtained are compared against the data generated from the test rig. Figure 2 shows the plot of the results obtained from test rig and CFD simulations. The results are being validated at three data points of pressure 2.7 bar, 2.3 bar and 1.5 bar. From the graph shown in Figure 2, it can be Fig. 2 Comparison of experimental and CFD predicted mass flow rate observed that the CFD simulations carried out are good enough to replicate the test rig conditions and percentage errors are below 10% for suction inlet flow and below 1% for Jet inlet flow. Turbulence model study The choice of turbulence model has been reviewed with the combination of two wall functions namely standard wall function and enhanced wall function. Totally, seven combinations are obtained for standard k-ε, Realizable k-ε, RNG k-ε and SST turbulence model with two wall functions. From the study of various turbulence models, it can be concluded that the percentage error in the mass flow rate is very less when the Realizable K- ε model with standard wall function is used. The percentage error in the jet inlet mass flow rate with different turbulence models is as follows: RKE-EWF = 1.01%, RKE-SWF = 1.17%, SKE-SWF = 1.68%, SKE-EWF = 1.01%, RNG-KE- SWF = 1.17%, RNG-KE-EWF = 0.84% and SST = 1.01%. The percentage error in the sump inlet mass flow rate with different turbulence models are as follows: RKE-EWF = 9.43%, RKE-SWF = 0.67%, SKE-SWF = 2.52%, SKE-EWF = 7.75%, RNG-KE- SWF = 26.7%, RNG-KE-EWF = 27.13% and SST = 1.68%. Ejector pump geometry optimization based on existing operating conditions In this case study, the existing operating conditions from the test rig have been taken and an optimization study is carried out to optimize the geometry for maximum efficiency. The operating conditions selected for this DoE study are jet inlet total pressure of 274572.7 Pa, suction inlet total pressure of - 24416.8 Pa and jet outlet static pressure of 24614.69 Pa. The values for the upper bound and lower bound are specified according to the data available in the literature, lower bound and upper bound values for diffuser angle in degrees are 6 and 7 respectively. Lower bound values for jet inlet radius, nozzle exit radius,throat radius and throat length in mm are 4.5, 2, 4 and 30 respectively. Upper bound values for jet inlet radius, nozzle exit radius,throat radius and throat length in mm are 6, 3, 6 and 45 respectively. Table 2 shows the generated design points for the given input parameters. These design points are generated using Box-Behnken Design approach 8 in ANSYS. The design space is later optimized using response surface methodology. The DoE and optimization tools are part of ANSYS package. The generated design has
356 J SCI IND RES VOL 77 JUNE 2018 SL. NO. Diffuser Angle (degree) Jet Inlet Radius (mm) Table 2 Matrix of design experiments Nozzle Exit Radius Throat Radius Throat Length (mm) (mm) (mm) Sump Inlet (kg/s) Jet Inlet (kg/s) Efficiency (%) 1 6.5 5.25 2.5 5 37.5 0.19349 0.20722 18.86 2 6 4.5 2.5 5 37.5 0.18999 0.20572 18.66 3 7 4.5 2.5 5 37.5 0.19014 0.20667 18.58 4 6 6 2.5 5 37.5 0.19461 0.20683 19.01 5 7 6 2.5 5 37.5 0.19483 0.20812 18.91 6 6.5 5.25 2 4 37.5 0.12225 0.13052 18.92 7 6.5 5.25 3 4 37.5 0.15778 0.30254 10.53 8 6.5 5.25 2 6 37.5 0.028506 0.13095 4.4 9 6.5 5.25 3 6 37.5 0.24191 0.30534 16 10 6.5 4.5 2.5 5 30 0.18938 0.20648 18.53 11 6.5 6 2.5 5 30 0.19745 0.20822 19.16 12 6.5 4.5 2.5 5 45 0.18587 0.20465 18.35 13 6.5 6 2.5 5 45 0.19745 0.20822 19.16 14 6 5.25 2 5 37.5 0.10984 0.12971 17.11 15 7 5.25 2 5 37.5 0.10939 0.13021 16.97 16 6 5.25 3 5 37.5 0.23981 0.304 15.93 17 7 5.25 3 5 37.5 0.23441 0.30403 15.57 18 6.5 5.25 2.5 4 30 0.16046 0.20623 15.72 19 6.5 5.25 2.5 6 30 0.15593 0.20753 15.18 20 6.5 5.25 2.5 4 45 0.15146 0.20703 14.78 21 6.5 5.25 2.5 6 45 0.16102 0.20679 15.73 22 6.5 4.5 2 5 37.5 0.11004 0.13099 16.97 23 6.5 6 2 5 37.5 0.11145 0.13064 17.23 24 6.5 4.5 3 5 37.5 0.2338 0.30328 15.57 25 6.5 6 3 5 37.5 0.16361 0.20819 15.87 26 6 5.25 2.5 4 37.5 0.15819 0.20611 15.5 27 7 5.25 2.5 4 37.5 0.15509 0.20612 15.2 28 6 5.25 2.5 6 37.5 0.16003 0.20603 15.69 29 7 5.25 2.5 6 37.5 0.15974 0.20643 15.63 30 6.5 5.25 2 5 30 0.11248 0.13108 17.33 31 6.5 5.25 3 5 30 0.23471 0.30309 15.64 32 6.5 5.25 2 5 45 0.11134 0.13109 17.16 33 6.5 5.25 3 5 45 0.23549 0.30359 15.67 34 6 5.25 2.5 5 30 0.19472 0.2068 19.02 35 7 5.25 2.5 5 30 0.19338 0.20745 18.83 36 6 5.25 2.5 5 45 0.19337 0.20661 18.91 37 7 5.25 2.5 5 45 0.18929 0.20589 18.57 38 6.5 4.5 2.5 4 37.5 0.15598 0.20563 15.32 39 6.5 6 2.5 4 37.5 0.1561 0.20711 15.22 40 6.5 4.5 2.5 6 37.5 0.15297 0.20542 15.04 41 6.5 6 2.5 6 37.5 0.16361 0.20819 15.87 been solved and the graph of predicted versus observed values is obtained as shown in Figure 3, indicates the goodness of fit. The observed fit is almost linear and thus the optimization of responses has been carried out 9. Figure 4 shows the local sensitivity versus output parameters plot. From this plot, it can be concluded that response of sump inlet is sensitive to nozzle radius and throat radius whereas the response of jet inlet is sensitive to nozzle radius. In the optimization module, the objective is to maximize the sump inlet mass flow rate and minimize the motive inlet mass flow rate so that higher flow ratio for fixed operating conditions has been obtained which in turn maximizes the efficiency. The candidate
VYAS et al.: EJECTOR PUMP CFD MODEL VALIDATION AND PERFORMANCE STUDIES 357 Fig. 3 Predicted versus observed design points Table 3 Comparison of existing and optimized geometry Parameters Existing Test rig Geometry Optimized Geometry Diffuser Angle 12.94 deg 6.9275 deg Jet Inlet Radius 5.425 mm 5.9641mm Nozzle Exit Radius 2.525 mm 2.0586 mm Throat Radius 6.09 mm 4.84 mm Throat Length 41.501 mm 38.46 mm Motive Inlet (A) 0.21161 kg/s 0.13876 kg/s Suction Inlet(B) 0.14556 kg/s 0.14573 kg/s Flow Ratio(B/A),M 0.6878 1.05 Pressure Ratio, N 0.202 0.202 Efficiency, M*N*100 13.89% 21.21% inlet mass flow rates. The statistically predicted sump inlet and jet inlet mass flow rates for candidate A, are 0.14573 kg/sec and 0.13876 kg/sec respectively. The mass flow rates computed statistically are in good comparison with the actual (CFD) mass flow rates and hence candidate A, B and C are the best points to achieve the maximum efficiency. Fig. 4 Local sensitivity versus output parameters points generated after the optimization using the ANSYS screening method are as follows, For candidate point A Diffuser angle 6.9275 degree, jet inlet radius 5.9641 mm, nozzle exit radius 2.0586 mm, throat radius 4.84 mm, throat length 38.46 mm, sump inlet mass flow rate 0.14414 kg/sec and jet inlet mass flow rate 0.15102 kg/sec. For candidate point B Diffuser angle 6.1115 degree, jet inlet radius 5.9422 mm, nozzle exit radius 2.1281 mm, throat radius 4.625 mm, throat length 43.258 mm, sump inlet mass flow rate 0.15142 kg/sec and jet inlet mass flow rate 0.15775 kg/sec. For candidate point C Diffuser angle 6.3275 degree, jet inlet radius 5.8279 mm, nozzle exit radius 2.1171 mm, throat radius 4.8554 mm, throat length 42.209 mm, sump inlet mass flow rate 0.146 kg/sec and jet inlet mass flow rate 0.15372 kg/sec.once the candidate points are being generated, they are verified for any deviation from the predicted values. The optimized design configuration (Candidate A) has been run on ANSYS-FLUENT to verify the sump inlet and jet Results and Discussion CFD benchmarking is done for three sets of experimental reading for an accuracy of benchmarking. Comparison of the existing and optimized values is given in the Table 3. It can be observed that in the optimized geometry, the diffuser angle is less than that of existing geometry. The reason behind this is, if the diffuser angle is too large, the boundary layer separation will happen. This leads to poor pressure recovery at the exit. The throat radius and throat length are small compared to the existing geometry and help in reducing the wall friction for the current operating conditions. Figure 2 shows experimental and CFD predicted mass flow-rates at jet inlet (primary/motive flow) and sump inlet (secondary flow) for different motive (jet inlet) pressures. The mass flow rates are shown against primary axis and the percentage error ((mass flow CFD mass flow experimental)/ mass flow experimental*100) is shown against secondary axis. In Figure 2 the legends Jet_Inlet_Experimental, Sump_Inlet_Experimental represent the experimental mass flow rate readings for corresponding motive pressures. Similarly, Jet_Inlet_CFD and Sump_Inlet_CFD represent CFD predicted mass flow rates for corresponding motive pressures, % Error Jet_Inlet and % Error Sump_Inlet represent the percentage error between experimental and CFD
358 J SCI IND RES VOL 77 JUNE 2018 predicted mass flow rates at jet and sump inlets for corresponding motive pressures. Conclusion A CFD simulation model has been developed on ANSYS-FLUENT platform for the ejector pump and the results are validated with the established test rig data. Pressure boundary conditions for CFD model are taken from test data and predicted mass flow-rates are validated with the corresponding test data. The results are validated within 10% error margin. Realizable K-epsilon turbulence model with standard wall function predicts better validation of results. From the current optimization studies, for a case where the operating conditions were fixed an optimized geometry is obtained. It could be concluded from the study that the nozzle radius, diffuser angle and throat radius are critical parameters which affect the performance of ejector pump. The efficiency of ejector pump for a motive pressure of 2.7 bar has been increased from 13.9 % to 21.2%. Acknowledgement The authors would like to thank "Council of Scientific and Industrial Research (CSIR)", Government of India for funding this project under the 12 th Five Year Plan. The authors also like to thank Director, CSIR-National Aerospace Laboratories (NAL), India and Head, Structural Technologies Division, CSIR-NAL, for their kind support. Nomenclature RKE SKE RNG SWF EWF SST C 1 and C 2 DoE TDR TKE TV( ) S ij u v Realizable K-ε Model Standard K-ε Model Renormalized Group K-ε Model Standard Wall Function Enhanced Wall Function Shear Stress Transport Model Empirical constants Design Of Experiments Turbulence dissipation rate Turbulence kinetic energy Turbulence viscosity Mean strain rate Velocity in X direction Velocity in Y direction w p RSM S D th ε Velocity in Z direction Pressure Reynolds stress model Set back distance Diameter of throat Density Kinematic viscosity Turbulent Schmidt number References 1 Langton R, Clark C, Hewitt M & Richards L, Aircraft Fuel Systems, John Wiley & Sons Ltd, (2009). 2 Sanger N L, Noncavitating performance of two low area ratio water jet pumps having throat lengths of 7.25 diameters, NASA Technical Note, (1985). 3 Mallela R, Chatterjee D, Numerical investigations of the effect of geometry on the performance of jet pump, J Mech Eng Sci, 225 (2011) 1 12. 4 El-Sawaf A, Halawa M A, Younes M A & Teaima I R, Study of the Different Parameters that Influence on the Performance of Water Jet Pump, 15th Int Water Technol Conf, Alexandria (2011). 5 Karambirov & Chebaevskii, Possibilities of improving ejector pump characteristics, Chem Pet Eng, 41 (2005) 75 80. 6 X P Long & X L Xang, Numerical Investigation on the jet pump performance based on different turbulence models, 26th IAHR Symp on Hydraul Mach and Syst. 7 Versteeg H K & Malalasekara W, An Introduction to Computational Fluid Dynamics-The Finite Volume Method, Longman Group Ltd Publication, London, UK, (1995). 8 ANSYS Fluent Theory Guide. 9 Bradley N, The Response Surface Methodology, MS Thesis, Department of Mathematical Sciences, Indiana University of South Bend, (2007). 10 ESDU 85032, Ejectors and Jet Pumps, Design and performance for incompressible liquid flow, Eng Sci Data Unit (1984). 11 ANSYS Workbench Help. 12 Suresh Kumar J, Ganesan V, Mallikarjuna J M, Govidarajan S, Design and optimization of a throttle body assembly by CFD analysis, Indian J Eng Mater Sci, 20 (2013) 350-360. 13 Jatinder Kumar, Ajay Bansal, CFD simulations of immobilized-titanium dioxide based annular photocatalytic reactor: Model development and experimental validation, Indian J Chem Technol, 22 (2015) 95-104. 14 Suresh Kumar J & Ganesan V, Flow through S.I. engine air intake system using CFD at part throttle and full throttle, Indian J Eng Mater Sci, 11 (2004) 93-99. 15 Sujoy Chakraborty, Kishan Choudhuri, Prasenjit Dutta & Bishop Debbarma, Performance prediction of Centrifugal pumps with variations of blade number, J Sci Ind Res, 72 (2013) 373-378.