Computational Fluid Dynamics Simulation of Open-Channel Flows Over Bridge-Decks Under Various Flooding Conditions The 6th WSEAS International Conference on FLUID MECHANICS (WSEAS - FLUIDS'09) Ningbo, China, January 10-12, 2009 S. Patil, M. Kostic and P. Majumdar Department of Mechanical Engineering NORTHERN ILLINOIS UNIVERSITY 1 2009 January 10-12 www.kostic.niu.edu Motivation: Bridges are crucial constituents of the nation’s transportation systems Bridge construction is critical issue as it involves great amount of money and risk Bridge structures under various flood conditions are studied for bridge stability analysis Such analyses are carried out by scaled experiments to calculate drag and lift coefficients on the bridge Scaled experiments are limited to few design variations and flooded conditions due to high cost and time associated with them Advanced commercial Computational Fluid Dynamics (CFD) software and parallel computers can be used to overcome such limitations 2 2009 January 10-12 www.kostic.niu.edu CFD is the branch of fluid mechanics which uses numerical methods to solve fluid flow problems In spite of having simplified equations and high speed computers, CFD can achieve only approximate solutions CFD is a versatile tool having flexibility is design with an ability to impose and simulate real time phenomena CFD simulations if properly integrated can complement real time scaled experiments Available CFD features and powerful parallel computers allow to study wide range of design variations and flooding conditions with different flow characteristics and different flow rates CFD simulation is a tool for through analysis by providing better insight of what is virtually happening inside the particular design 3 2009 January 10-12 www.kostic.niu.edu Literature Review: Ramamurthy, Qu and Vo, conducted simulation of three dimensional free surface flows using VOF method and found good agreement between simulation and experimental results Maronnier, Picasso and Rappaz, conducted simulation of 3D and 2D free surface flows using VOF method and found close agreement between simulation and experimental results. Harlow, and Welch, wrote Navier stokes equations in finite difference forms with fine step advancement to simulate transient viscous incompressible flow with free surface. This technique is successfully applicable to wide variety of two and three dimensional applications for free surface Koshizuka, Tamako and Oka, presented particle method for transient incompressible viscous flow with fluid fragmentation of free surfaces. Simulation of fluid fragmentation for collapse of liquid column against an obstacle was carried. A good agreement was found between numerical simulation and experimental data 4 2009 January 10-12 www.kostic.niu.edu Objectives: The objective of the present study is to validate commercial code STAR-CD for hydraulic research The experimental data conducted by Turner Fairbank Highway Research Center (TFHRC) at their own laboratories will be simulated using STAR-CD The base case of Fr = 0.22 and flooding height ratio, h*=1.5 is simulated with appropriate boundary conditions corresponding to experimental testing The open channel turbulent flow will be simulated using two different methods First by transient Volume of Fluid (VOF) methodology and other as a steady state closed channel flow with top surface as slip wall Drag and lift coefficients on the bridge is calculated using six linear eddy viscosity turbulence model and simulation outcome will be compared with experimental results 5 2009 January 10-12 www.kostic.niu.edu The suitable turbulence model will be identified which predicts close to drag and lift coefficients The parametric study will be performed for time step, mesh density and convergence criteria to identify optimum computational parameters The suitable turbulence model will be used to simulate 13 different flooding height ratio from h*=0.3 to 3 for Fr =0.22 6 2009 January 10-12 www.kostic.niu.edu Experimental Data: Experiments are conducted for open channel turbulent flow over six girder bridge deck for different flooding height ratios (h*) and with various flow conditions (Fr) ΔWSimulation=0.00254 S=0.058 m LBridge =0.34 m LFlow = 0.26 m Flow Direction Schematic of experimental six girder bridge deck model 7 2009 January 10-12 www.kostic.niu.edu Theory Dimensions of experimental six girder bridge deck model W Fr Vavg Fr Vavg L Flow Froude Number Y Flooding Ratio h* ghu gW hu hb S X Nomenclature for bridge dimensions and flooding ratios 2009 January 10-12 www.kostic.niu.edu 8 Experimental data consists of drag and lift coefficients as the function of Froude number, Fr and dimensionless flooding height ratio h* Experimental data consists of five different sets of experiments for Froude numbers from Fr =0.12 to 0.40 and upstream average velocity 0.20 m/s to 0.65 m/s The experiments for the Froude number, Fr=0.22 are repeated four times with an average velocity of 0.35 m/s for h*=0.3 to 3 The lift coefficient is calculated by excluding buoyancy forces in Y (vertical) direction 9 2009 January 10-12 www.kostic.niu.edu Drag Coefficients vs h* for Fr = 0.22 3.00 Drag Coefficient - C D 2.50 2.00 1.50 1.00 0.50 0.00 -0.10 0.40 0.90 12-29-06_2 01-03-07_1 01-31-07_3 AVG Drag 1.40 1.90 2.40 01-29-07_1 2.90 3.40 h* 10 2009 January 10-12 www.kostic.niu.edu Lift Coefficient vs h* for Fr = 0.22 0.50 Lift Coefficient - C L 0.00 -0.10 0.40 0.90 1.40 1.90 2.40 2.90 3.40 -0.50 -1.00 -1.50 -2.00 12-29-06_2 01-03-07_1 01-31-07_3 AVG Lift 01-29-07_1 -2.50 h* 11 2009 January 10-12 www.kostic.niu.edu Governing Equations for fluid flow: Mass conservation equation . V 0 t Momentum conservation equation DV P 2 .V g Dt Energy conservation equation Ein Eout dECV dt 12 2009 January 10-12 www.kostic.niu.edu Dimensionless parameters for open channel flow: Reynolds Number Vavg Rh Re Rh AC yb p b 2y For 2D open channel flow b , Rh y y b 13 2009 January 10-12 www.kostic.niu.edu Froude Number: Froude number is dimensionless number which governs character of open channel flow Vavg Fr gLC The flow is classified on Froude number Fr 1 Fr 1 Subcritical or tranquil flow Critical Flow Fr 1 Supercritical or rapid flow InertiaForce Fr GravityForce 2 Open channel flow is dominated by inertial forces for rapid flow and by gravity forces for tranquil flow 14 2009 January 10-12 www.kostic.niu.edu Froude number is also given by Fr Vavg C0 Vavg gy Where C 0 Wave speed (m/s) y = Flow depth (m) 15 2009 January 10-12 www.kostic.niu.edu Force Coefficients: The component of resultant pressure and shear forces in direction of flow is called drag force and component that acts normal to flow direction is called lift force Drag force coefficient is CD FD 0.5 Vavg AD 2 Lift force coefficient is FL CL 2 0.5Vavg AL In the experimental testing, the drag reference area is the frontal area normal to the flow direction. The lift reference area is the bridge area perpendicular to Y direction. 16 2009 January 10-12 www.kostic.niu.edu Drag and lift reference areas for experimental data: For drag, if h* 1 ,then drag area is (hu hb ) * LBridge if h* 1 ,then drag area is S * LBridge For lift, for all h * ,lift area is LFlow * LBridge 17 2009 January 10-12 www.kostic.niu.edu Turbulent Flow: Turbulent flow is complex phenomena dominated by rapid and random fluctuations Turbulent flow is highly unsteady and all the formulae for the turbulent flow are based on experiments or empirical and semi – empirical correlations Turbulent Intensity u' TI Vavg Turbulence mixing length (m) lm C 0.75 k 1.5 Turbulent kinetic energy (m2/s2) k 1.5Vavg TI 2 2 18 2009 January 10-12 www.kostic.niu.edu Turbulence dissipation rate (m2/s3) C 0.75 1.5 k lm Specific dissipation rate (1/s) C k 19 2009 January 10-12 www.kostic.niu.edu Turbulence Models: Six eddy viscosity turbulence models are studied from STAR-CD turbulence options Two major groups of turbulence models k-ε and k-ω are studied The k- ε turbulence model The k-ω turbulence models a. Standard High Reynolds a. Standard High Reynolds b. Renormalization Group b. Standard Low Reynolds c. SST High Reynolds d. SST Low Reynolds 20 2009 January 10-12 www.kostic.niu.edu The k-ε High Reynolds turbulence model: Most widely used turbulent transport model First two equation model to be used in CFD This model uses transport equations for k and ε in conjunction with the law-of-the wall representation of the boundary layer The k-ε RNG turbulence model: This turbulence model is obtained after modifying k-ε standard turbulence model using normalization group method to renormalize Navier Stokes equations This model takes into account effects of different scales of motions on turbulent diffusion 21 2009 January 10-12 www.kostic.niu.edu k-ω turbulence model: The k-ω turbulence models are obtained as an alternative to the k-ε model which have some difficulty for near wall treatment The k-ω turbulence models Standard k-ω model High Reynolds Shear stress transport (SST) model Low Reynolds High Reynolds Low Reynolds 22 2009 January 10-12 www.kostic.niu.edu SST k-ω turbulence model: SST turbulence model is obtained after combining best features of k-ε and k-ω turbulence model SST turbulence model is the result of blending of k-ω model near the wall and k-ε model near the wall 23 2009 January 10-12 www.kostic.niu.edu Computational Model: STAR-CD (Simulation of Turbulent flow in Arbitrary Regions Computational Dynamics) is CFD analysis software STAR-CD is finite volume code which solves governing equations for steady state or transient problem The first method used in STAR-CD to simulate open channel turbulent flow is free surface method which makes use of Volume of Fluid (VOF) methodology VOF methodology simulates air and water domain VOF methodology uses volume of fraction variable to capture airwater interface 24 2009 January 10-12 www.kostic.niu.edu VOF technique: VOF technique is a transient scheme which captures free surface. VOF deals with light and heavy fluids VOF is the ratio of volume of heavy fluid to the total control volume Volume of fraction is given by i Vi V Transport equation for volume of fraction i .( i u) 0 t 2 Volume fraction of the remaining component is given by i 1 i 1 25 2009 January 10-12 www.kostic.niu.edu The properties at the free surface vary according to volume fraction of each component 2 i . i i 1 26 2009 January 10-12 www.kostic.niu.edu Free Surface method: 0.30 0.08 0.06 Y 0 X Z -0.15 -1.50 0 0.26 1.78 3 Dimensions for computational model h*=1.5 generated in STAR-CD (Dimensions not to scale and in SI units) 27 2009 January 10-12 www.kostic.niu.edu Computational Mesh: Y Y X Full computational domain with non uniform mesh and 2 cells thick in Z direction for =1.5 28 2009 January 10-12 www.kostic.niu.edu Boundary Conditions: Top wall (slip) Symmetry Plane w 0 Air Inlet w 1 Water Inlet Y Y X Z Bottom Wall (No Slip) Outlet 29 2009 January 10-12 www.kostic.niu.edu Computational parameters for VOF methodology: Inlet velocity, U 0.35 m/s Turbulent kinetic energy, k 0.00125 m2/s2 Turbulent Dissipation Rate, ε 0.000175 m2/s3 Solution method Transient Solver method Algebraic Multigrid approach (AMG) Solution algorithm SIMPLE Relaxation factor Pressure - 0.3 Momentum, Turbulence, Viscosity - 0.7 Differencing scheme MARS Convergence Criteria 10-2 Time Step (Δt) 0.01 s 30 2009 January 10-12 www.kostic.niu.edu Water slip top wall method: 0.08 Y 0.06 0 X Z -0.15 -1.5 0 0.26 1.78 3 Dimensions for computational model h*=1.5 for water slip –top-wall method (Dimensions not to scale and in SI units) 31 2009 January 10-12 www.kostic.niu.edu Boundary conditions: Top wall (slip) Symmetry Plane Water Inlet Y Y X X Bottom wall (No slip) Outlet (Standard) Computational domain with boundary surfaces and boundary conditions for water slip-top-wall method 32 2009 January 10-12 www.kostic.niu.edu Computational parameters for water slip-top-wall method: 0.35 m/s Inlet velocity, U 0.00125 m2/s2 Turbulent kinetic energy, k 0.000175 m2/s3 Turbulent Dissipation Rate, Solution Method Steady State Solver Method Algebraic Multigrid approach (AMG) Solution Algorithm SIMPLE Relaxation factor Pressure - 0.3 Momentum, turbulence, Viscosity - 0.7 Differencing scheme UD Convergence Criteria 10-6 33 2009 January 10-12 www.kostic.niu.edu STAR-CD simulation Validation with basics of fluid mechanics : Fully developed velocity profile for laminar pipe flow 6.00E-02 Y Coordinate (m) 4.00E-02 2.00E-02 0.00E+00 0 0.0005 0.001 0.0015 0.002 0.0025 0.003 0.0035 0.004 0.0045 -2.00E-02 -4.00E-02 -6.00E-02 W Velocity (m/s) Velocity Profile for Laminar Pipe Flow Average velocity profile Fully developed velocity profile for laminar pipe flow after STAR-CD simulation 34 2009 January 10-12 www.kostic.niu.edu Fully developed velocity profile for turbulent pipe flow 0.06 Y coordinate (m) 0.04 0.02 0 0 0.05 0.1 0.15 0.2 0.25 0.3 -0.02 -0.04 -0.06 W velocity (m/s) Velocity Profile for turbulent pipe flow Average Velocity Profile Fully developed velocity profile for the turbulent pipe flow after STAR-CD simulation 35 2009 January 10-12 www.kostic.niu.edu Comparison between theoretical and simulated friction factor : Flow type Wall Roughness (m) Theoretical friction factor (Reference) Simulation friction factor Absolute Difference Percentage Difference Laminar Smooth 0.2844 0.2865 0.0021 0.74 Turbulent Smooth 0.0121 0.0116 0.0005 4.13 Turbulent 0.005 0.053 0.048 0.005 9.43 Turbulent 0.015 0.0872 0.0756 0.0116 13.30 Turbulent 0.075 0.2529 0.2019 0.051 20.17 36 2009 January 10-12 www.kostic.niu.edu Calculation of entrance length: Lh 0.05 Re D Re 500 Shear stress at bottom wall in flow direction Wall shear stress (N/m2) 2.00E-05 Dh 2m 1.80E-05 1.60E-05 1.40E-05 Lh 50m 1.20E-05 1.00E-05 8.00E-06 6.00E-06 4.00E-06 2.00E-06 0.00E+00 0 20 40 60 80 100 120 X distance (m) Shear stress at bottom wall Continued on next page 37 2009 January 10-12 www.kostic.niu.edu Developement of velocity profile in laminar duct flow 1.00E+00 9.00E-01 Y distance (m) 8.00E-01 7.00E-01 6.00E-01 5.00E-01 4.00E-01 3.00E-01 2.00E-01 1.00E-01 0.00E+00 0.00E+ 1.00E- 2.00E- 3.00E- 4.00E- 5.00E- 6.00E- 7.00E- 8.00E- 9.00E- 1.00E00 03 03 03 03 03 03 03 03 03 02 U velocity (m/s) At 20 M At 40 M At 50 M At 60 M At 75 M At 90 M 38 2009 January 10-12 www.kostic.niu.edu Verification of power law velocity profile: Comparision for power law velocity profile from theory and after simulation 1.00E+00 8.00E-01 6.00E-01 r/Rh 4.00E-01 2.00E-01 0.00E+00 -2.00E-01 0 0.2 0.4 0.6 0.8 1 1.2 -4.00E-01 -6.00E-01 -8.00E-01 -1.00E+00 U/Umax Theoretical velocity profile Velocity profile from simulation 39 2009 January 10-12 www.kostic.niu.edu Comparison between Fluent and STAR-CD for same geometry: 0 0.254 1.016 0.504 0.127 0.097 Y 0 X Operating Condition Variables Inlet Velocity U = 2 m/s Inlet turbulence intensity 10 % Inlet turbulence mixing length 0.1 m Outlet gauge pressure 0 Pa Walls No Slip Convergence 0.001 40 2009 January 10-12 www.kostic.niu.edu 2009 January 10-12 Comparison for velocity contours between STAR-CD and Fluent www.kostic.niu.edu 41 Comparison for velocity vectors between STAR-CD and Fluent 42 2009 January 10-12 www.kostic.niu.edu Comparison for X velocities between Fluent and STAR-CD 43 2009 January 10-12 www.kostic.niu.edu Pressure difference (Pa) Parameter Fluent STAR-CD (Reference Data) Absolute Difference Percentage Difference ΔP STAT 1120 1161 41 3.53 % ΔP TOT 1083 1120 37 3.30 % Force Coefficients Force Coefficients Fluent STAR-CD (Reference Data) Absolute Difference Percentage Difference CD 1.89 2.00 0.11 5.5 % CL -6.77 -7.05 0.28 3.97 % 44 2009 January 10-12 www.kostic.niu.edu VOF simulation of experimental data: Effect of time steps on drag coefficients Effect of time steps for k- High Re TM on drag coefficients 3.60000 3.40000 3.20000 3.00000 CD 2.80000 2.60000 2.40000 2.20000 -40 10 60 110 160 210 260 310 time (sec) time step 0.01 sec time step 0.05 sec time step 0.1 sec 45 2009 January 10-12 www.kostic.niu.edu Effect of time steps on lift coefficients: Effect of time steps for k- High Re TM on lift coefficients 0.00000 -40 -0.20000 10 60 110 160 210 260 310 -0.40000 -0.60000 CL -0.80000 -1.00000 -1.20000 -1.40000 time (sec) time step 0.01 sec time step 0.05 sec time step 0.1 sec 46 2009 January 10-12 www.kostic.niu.edu Effect of decreased downstream length on force coefficients Comparision for C L between full computational domain and computational model with decreased downstream length for k- High Re TM for 0.05 s time step Comparision for C D between full computational model and computational model with decreased downstream length for k- High Re TM for 0.05 s time step 4.00000 2.50000 2.00000 3.50000 1.50000 1.00000 3.00000 0.50000 CD 2.50000 CL 2.00000 0.00000 -0.50000 0 50 100 150 200 250 -1.00000 -1.50000 1.50000 -2.00000 -2.50000 1.00000 0 50 100 150 200 250 -3.00000 time (sec) time (sec) Full Computational model Computational model with decreased downstream length Full computational domain Computatioanl domain with decreased downstream length 47 2009 January 10-12 www.kostic.niu.edu Effect of decrease in under bridge water depth Comparision for C D between full computational model and computational model with decrease in under bridge water depth for k- High Re TM for time step 0.05 s Comparision for C L between full computational model and computational model with decrease in under bridge water depth for k- High Re TM for time step of 0.05 s 12.00000 0.00000 10.00000 -1.00000 8.00000 -2.00000 0 CD 6.00000 50 100 150 200 250 CL -3.00000 4.00000 -4.00000 2.00000 -5.00000 0.00000 0 50 100 150 200 250 -6.00000 time (sec) time (step) Full Cvomputational model Compuatational model with decrease in under bridge water depth Full computational model Computational model with decrease in under bridge water depth 48 2009 January 10-12 www.kostic.niu.edu Effect of top boundary condition at top as slip wall and symmetry Comparision between top wall as a slip and symmetry for k- High Reynolds turbulence model Comparision between top wall as a slip and symmetry for k- High Reynolds turbulence model 3.45000 0.00000 0 50 100 150 200 250 300 350 -0.20000 3.20000 -0.40000 CD CL -0.60000 2.95000 -0.80000 -1.00000 2.70000 0 50 100 150 200 250 300 350 -1.20000 time (sec) time (sec) Slip wall Slip wall symmetry wall Symmetry wall 49 2009 January 10-12 www.kostic.niu.edu Free Surface Development: Nomenclature for VOF contour plot Free surface, w 0 0.01 w 0.99 w 1 w Volume fraction for water 50 2009 January 10-12 www.kostic.niu.edu Effect of k-ε standard turbulence model on free surface development: t=10sec t=50 sec t=30sec t=100se c t=150 sec t=200 sec sec t =250 sec t=300 sec 51 2009 January 10-12 www.kostic.niu.edu Effect of different turbulence models on drag coefficients: Effect of different turbulence models on drag coefficient 13.00000 11.00000 9.00000 CD 7.00000 5.00000 3.00000 1.00000 -1.00000 0 50 100 150 200 250 300 350 time (sec) k-epsilon High Re k-epsilon RNG k-omega STD High Re k-omega STD Low Re k-omega SST High Re k-omega SST Low Re Experimenal Results 52 2009 January 10-12 www.kostic.niu.edu Effect of different turbulence models on lift coefficients: Effect of different turbulence models on lift coefficient 0.00000 -0.20000 0 50 100 150 200 250 300 350 -0.40000 -0.60000 -0.80000 CL -1.00000 -1.20000 -1.40000 -1.60000 -1.80000 -2.00000 Time (sec) k-epsilon High Re k-epsilon RNG k-omega STD High Re k-omega STD Low Re k-omega SST High Re k-omega SST Low Re Experimental Results 53 2009 January 10-12 www.kostic.niu.edu Comparison between simulation results for different turbulence model and experimental results: Turbulence Models h*up h*dw h*avg CD avg CD exp CL avg CL exp k-ε High Re 1.40 1.30 1.35 3.17 1.98 -0.83 -1.04 k-ε RNG 1.45 1.45 1.45 2.77 2.02 -1.39 -0.73 k-ω STD High Re 1.15 1.30 1.38 4.69 1.99 -0.55 -1.00 k-ω STD Low Re 1.84 1.50 1.67 10.91 1.97 -0.29 -0.60 k-ω SST High Re 1.30 1.20 1.25 3.03 1.98 -1.15 -1.10 k-ω SST Low Re 1.35 1.20 1.28 4.03 1.96 -0.91 -1.07 h*up h*dw h*avg CD avg CD exp CL avg CL exp Count 6.00 6.00 6.00 6.00 6.00 6.00 6.00 Maximum 1.84 1.50 1.67 10.91 2.02 -0.29 -0.60 Average 1.41 1.33 1.40 4.77 1.98 -0.85 -0.92 Std. Dev. 0.23 0.13 0.15 3.09 0.02 0.40 0.21 Minimum 1.15 1.20 1.25 2.77 1.96 -1.39 -1.10 54 2009 January 10-12 www.kostic.niu.edu Water slip-top-wall method: (b) Refined near bridge (a) Basic Coarse mesh Mesh Density (c) Fully refined model CD % Difference CL % Difference Basic coarse grid 2.96061 1% -1.39188 0.54% Refined near bridge 2.93367 0.08 % -1.38328 0.08 % Fully refined model 2.93109 0 % (Ref) -1.38439 0 % (Ref) 55 2009 January 10-12 www.kostic.niu.edu Effect of convergence criteria on final solution: Convergence criteria CD % difference CL % difference 10-6 2.96061 0 % (ref) -1.39188 0 % (ref) 10-5 2.96062 0.00033 % -1.39185 0.0022 % 10-4 2.95445 0.2 % -1.37877 0.94 % 56 2009 January 10-12 www.kostic.niu.edu Comparison between VOF and Water slip-top-wall method with experimental results: Comparision between VOF and steady state simulation for different turbulence models for base case of Fr=0.22 and h*=1.5 Comparision between VOF and steady state simulation for different turbulence models for base case of Fr= 0.22 and h*=1.5 -1.80 12.00 -1.60 10.00 -1.40 -1.20 8.00 -1.00 CL CD 6.00 -0.80 ` -0.60 4.00 -0.40 2.00 -0.20 0.00 0.00 k-epsilon High Re k-epsilon RNG k-omega k-omega k-omega SST k-omega SST STD High Re STD Low Re High Re Low Re k-epsilon High Re k-epsilon RNG Turbulence models Turbulence models VOF simulation Steady state simulation k-omega k-omega k-omega SST k-omega SST STD High Re STD Low Re High Re Low Re Experimental data VOF Simulation Steady State Simulation Experimental data 57 2009 January 10-12 www.kostic.niu.edu Drag coefficient, CD Lift Coefficient, CL VOF Exp. Water slip-topwall VOF Exp. Water sliptop-wall k-ε High Re 3.17 2.02 2.96 -0.83 -0.70 -1.39 k-ε RNG 2.77 2.02 2.57 -1.39 -0.70 -1.08 4.69 2.02 3.19 -0.55 -0.70 -1.43 k-ω STD Low Re 10.91 2.02 10.59 -0.29 -0.70 -1.35 k-ω SST High Re 3.03 2.02 2.78 -1.15 -0.70 -1.26 k-ω SST Low Re 4.03 2.02 4.03 -0.91 -0.70 -1.63 Turbulence model k-ω STD High Re The k-ε RNG predicts closet drag and lift coefficients 58 2009 January 10-12 www.kostic.niu.edu Effect of inlet turbulence on drag and lift coefficients: Effect of inlet turbulence intensity on force coefficients when mixing length is 1 mm 3 2.5 CD 2 1.5 1 0.5 0 -0.5 0% CL -1 -1.5 -2 Effect of inlet turbulence intensity on force coefficients when mixing length is 41.5 mm 4 CD 3 2 1 5% 10% 15% 20% 25% 30% 0 0% CL 5% 10% 15% 20% 25% 30% -1 -2 Inlet turbulence intensity Inlet turbulence intensity Effect of inlet turbulence intensity on drag coefficient Effect of inlet turbulence intensity on drag coefficients Effect of inlet turbulence intensity on lift coefficient Effect of inlet turbulence intensity on lift coeffcients 59 2009 January 10-12 www.kostic.niu.edu Fully developed velocity profile after selected runs: Development of velocity profile for open channel flow for selected runs 4 Y coordinates (m) 2 0 0.2 0.22 0.24 0.26 0.28 0.3 0.32 0.34 0.36 0.38 0.4 -2 -4 -6 -8 U velocity (m/s) 1st Run 3rd Run 5th Run 9th Run 13th Run 15th Run 60 2009 January 10-12 www.kostic.niu.edu Fully developed turbulence kinetic energy after selected runs: Development of turbulence kinetic energy for open channel flow for selected runs 4 Y coordinate (m) 2 0 0.00E+00 1.00E-04 2.00E-04 3.00E-04 4.00E-04 5.00E-04 6.00E-04 7.00E-04 -2 -4 -6 -8 Kinetic energy per unit mass (m2/s2) 1st Run 3rd Run 5th Run 9th Run 13th Run 15th Run 61 2009 January 10-12 www.kostic.niu.edu Fully developed turbulence dissipation rate after selected runs: Development of turbulence dissipation rate for open channel flow for selected runs 4 Y coordinate (m) 2 0 0.00E+0 4.10E04 -2 0 8.20E04 1.23E03 1.64E03 2.05E03 2.46E03 2.87E03 3.28E03 3.69E03 4.10E03 -4 -6 -8 Turbulence dissipation rate (m2/s3) 1st Run 3rd Run 5th Run 9th Run 13th Run 15th Run 62 2009 January 10-12 www.kostic.niu.edu Comparison between CFD simulations and experimental data for Fr=0.22 for drag coefficients: h* CFD Simulation Experimental (Reference) Absolute Difference Percentage Difference 0.289 1.63 1.92 0.29 15.10 0.493 1.78 1.21 0.57 47.10 0.68 1.92 1.57 0.35 22.29 0.972 2.29 1.37 0.92 67.15 1.281 2.68 1.98 0.7 35.35 1.500 2.66 2.02 0.64 31.68 1.709 2.62 1.95 0.67 34.35 2.015 2.51 1.89 0.62 32.80 2.309 2.39 1.82 0.57 31.31 2.517 2.33 1.79 0.54 30.16 2.706 2.28 1.73 0.55 31.79 3.008 2.19 1.71 0.48 28.07 3.097 2.17 1.69 0.48 28.40 63 2009 January 10-12 www.kostic.niu.edu Comparison between CFD simulation and experimental data for Fr=0.22 for lift coefficients: h* CFD Simulation Experimental (Reference) Absolute Difference Percentage Difference 0.289 -0.42 -1.70 1.28 75.29 0.493 -0.77 -1.28 0.51 39.84 0.68 -1.00 -1.76 0.76 43.18 0.972 -1.44 -1.75 0.31 17.71 1.281 -1.53 -1.13 0.40 35.40 1.500 -1.01 -0.70 0.31 44.29 1.709 -0.81 -0.53 0.28 52.83 2.015 -0.46 -0.29 0.17 58.62 2.309 -0.10 -0.14 0.04 28.57 2.517 -0.12 -0.04 0.08 233.33 2.706 -0.05 0.03 0.08 275.00 3.008 -0.06 0.06 0.13 201.59 3.097 -0.10 0.10 0.19 198.97 64 2009 January 10-12 www.kostic.niu.edu Comparision between CFD simulations and experimental results for drag coefficients for case of Fr=0.22 3.00 2.50 CD 2.00 1.50 1.00 0.50 0.00 0 0.5 1 1.5 2 2.5 3 3.5 h* CFD simulation Experimental results 65 2009 January 10-12 www.kostic.niu.edu Comparision between CFD simulation and experimental results for lift coefficients for case of Fr=0.22 0.50 0.00 0 0.5 1 1.5 2 2.5 3 3.5 CL -0.50 -1.00 -1.50 -2.00 h* CFD Simulation Experimental results 66 2009 January 10-12 www.kostic.niu.edu Conclusion: CFD simulations by STAR-CD for Fr=0.22 case , predicts more drag than experimental drag except for h*=0.289 The percentage difference if the experimental data is taken as reference, is maximum of 67% for h*=0.972 and minimum of 15% for h* =0.289 For lift predictions, for cases of h*<1, CFD simulations predict more lift than experimental . For h*>1, CFD simulations predict lower lift than experimental 67 2009 January 10-12 www.kostic.niu.edu Recommendations for future work: VOF simulations are run for convergence criterion of 0.01. VOF should be run for more convergence criterion and that is only available with large computing power. VOF simulations should be run for lower time step than 0.01 sec and for longer simulation time up to 500 sec. In this study only linear eddy viscosity turbulence models are used. The effect of Large Eddy Simulation, Reynolds stress models and non linear eddy viscosity turbulence models should be tested on force coefficients 68 2009 January 10-12 www.kostic.niu.edu Acknowledgments: The authors like to acknowledge support by Dean Promod Vohra, College of Engineering and Engineering Technology of Northern Illinois University (NIU), and Dr. David P. Weber of Argonne National Laboratory (ANL); and especially the contributions by Dr. Tanju Sofu, and Dr. Steven A. Lottes of ANL, as well as financial support by U.S. Department of Transportation (USDOT) and computational support by ANL’s Transportation Research and Analysis Computing Center (TRACC). 69 2009 January 10-12 www.kostic.niu.edu QUESTIONS ??? More information at: www.kostic.niu.edu 70 2009 January 10-12 www.kostic.niu.edu