Flow Analysis around a Dimpled Cylinder Using Detached
Transcription
Flow Analysis around a Dimpled Cylinder Using Detached
Flow Analysis around a Dimpled Cylinder Using Detached-Eddy Simulation Symposium on Hybrid RANS-LES Methods Rica City Hotel, Stockholm, 14-15 July, 2005 Hyoung-Chol KIM, Kazuhiro NAKAHASHI, Hyoung-Jin KIM Dept. of Aerospace Eng., Tohoku Univ. Masaya TSUNODA SRI Research and Development LTD. Takuma KATO Inst. of Fluid Science, Tohoku University Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Contents I. Background II. Objectives III. Numerical Method IV. Results V. Conclusions Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Background : DES 9 Computational Fluid Dynamics (CFD) can be applied to the complicated flow fields. 9 However, analysis of complicated flow fields with massive separation such as blunt body problems at high Reynolds number is still challenging. 9 One of the reasons is difficulty in adequate consideration of turbulence effects. 9 To deal with these problems, recently, Detached-Eddy Simulation (DES) based on Spalart-Allmaras one equation turbulence model is proposed by Spalart et al. Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Background : Gold ball 1.2 smooth cylinder (Wieselsberg) smooth sphere (Achenbach, 1974) golf ball (Bearman and Harvey, 1976) 1 9 Flow around a sphere has a very complicated and interesting physics. 9 A golf ball flies at about M=0.2 and the Reynolds Number of 105, but the drag coefficient is about half of the smooth sphere. CD 0.8 0.6 0.4 0.2 0 4 10 105 106 107 Re Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Background : Gold ball 9 Dimples on the golf ball surface play an important role to trigger the boundary layer transition and reduce the drag. 9 However, the mechanism of the boundary layer transition by dimples has not been fully understood and the design of the dimples around a golf ball still highly depends on the experiences and experiments. 9 The CFD application to the golf ball is still challenging, but is becoming a powerful tool to investigate the effect of the geometry of dimples to the flows and the drag reduction. Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Objectives 9 Main objective is to understand the effectiveness of the dimple shape on the turbulent flow around a dimpled sphere. 9 As a first step, flows around dimpled cylinders rather than dimpled spheres are simulated for simplicity and reduced computational cost. 9 This talk mainly focuses on the effectiveness of the DES for simulations of dimpled cylinders. <3-D view> <top view> On the cylinder surface, there are three lines of dimples, each line having thirty dimples in the circumferential direction. Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Numerical Method : Flow solver & Conditions TAS code (Tohoku univ. Aerodynamic Simulation code) i. ii. Governing Eq. : Compressible Navier-Stokes Eq. Spatial Discritization : Cell-Vertex, Finite Volume Method iii. Numerical Flux Evaluation : HLLEW Riemann Solver iv. Time Integration : LU-SGS Implicit Method Flow conditions (shot off) i. ii. iii. iv. v. vi. vii. Ma. Number Re. Number Far boundary Wall boundary Side boundary Time step Newton subiterations : 0.17 : 1.65 x 105 : Uniform Flow : No slip condition : Symmetry : UΔt=D/500 : 4 times Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Numerical Method : Turbulence models Applied turbulence models i. Without turbulence model (LAMINAR) ii. Goldberg-Ramakrishnan one equation model (G-R) iii. Spalart-Allmaras one equation model (S-A) iv. Detached-Eddy Simulation based on S-A model (DES) Assumption of fully turbulent boundary layer Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Numerical Method : Grid 9 Hybrid volume grid information 1) nodes : 1,045,951 2) edges : 4,695,576 3) tetrahedra : 1,132,589 4) prisms : 1,620,868 5) pyramids : 8,817 9 Grid density in the boundary layer region is increased by prismatic grid layer. (# of prism layers = 30, minimum spacing = 2.2 x 10-5) 9 Outer boundary is located at 50D from the cylinder surface (D is the cylinder diameter). Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Result : Separation regions (time averaged) (U_velocity contours ) LAMINAR S-A G-R DES In turbulent flows, separation regions move downward resulting in drag reduction. Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Result : Vorticity magnitude contours (time averaged) LAMINAR S-A G-R DES The difference with the high vorticity magnitude regions in the wake. Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Result : Vorticity magnitude contours (time averaged) LAMINAR S-A edges. The vortices take place at dimple G-R DES Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Result : Vorticity magnitude contours (instantaneous, CL=0) LAMINAR S-A G-R DES The difference in vortex structure at wake. Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Result : LAMINAR Velocity vectors (time averaged) S-A Secondary vortex G-R DES Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Result : Section CP distributions (I) y=0 section 2 2 LAMINAR G-R Experiment 1 S-A DES Experiment 1 0 -1 -1 CP CP 0 -2 -2 -3 -3 dimple geometry -4 -5 0 30 60 90 degree 120 150 180 dimple geometry -4 -5 0 30 60 90 120 150 180 degree Peaky CP occurs between dimples Experiments : Institute of Fluid Science, Tohoku Univ. Japan (2004). In the front of the cylinder, CP distributions show good agreement with numerical and experimental results. Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Result : Section CP distributions (II) y=0.4 section 2 2 LAMINAR G-R Experiment 1 S-A DES Experiment 1 0 CP CP 0 -1 -1 -2 -2 -3 0 30 60 90 degree 120 150 180 -3 0 30 60 90 120 150 180 degree Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Result : Velocity gradients distributions (I) y=0 section 1200 1200 LAMINAR G-R 600 300 Separation pointpoint Separation 0 -600 30 60 90 degree 120 300 G-R -300 97.0° S-A -600 96.4° 150 DES 180 Separation point (y=0 section) 73.0° dimple geometry 0 600 0 LAMI -300 S-A DES 900 (du/dy)y=surface (du/dy)y=surface 900 0 30 85.0° dimple geometry 60 90 120 150 180 degree At between dimples, the velocity gradients have a sudden peak, due to the dimpled surface geometry. This corresponds to the lower peak of CP. Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Result : Velocity gradients distributions (II) y=0.4 section 500 500 LAMINAR G-R 400 (du/dy)y=surface (du/dy)y=surface 400 300 200 300 200 Separation point (y=0.4 100 section) Separation point 83.7° LAMI 100 0 -100 S-A DES 0 G-R 0 30 60 90 degree 120 S-A 150 DES Separation point 106.1° 180 -100105.4° 0 30 106.1° 60 90 120 150 180 degree In the smooth region of the dimpled cylinder, the separation takes place around 106°from the foremost stagnation point. Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan RESULT : CD history 2 2 LAMINAR G-R 1.8 S-A DES 1.8 1.6 1.4 1.4 CD CD 1.6 1.2 1.2 Time averaged CD 1 1.298 LAMI 1 0.8 0.6 40 G-R 50 60 70 80 tU/D 90 100 S-A120 110 DES 0.8 0.6 40 0.942 0.763 50 60 0.903 70 80 90 100 110 120 tU/D In only S-A model, the CD history shows the periodic oscillation manner. Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Result 1.6 Comparisons of CD smooth cylinder (Wieselsberg) smooth sphere (Achenbach, 1974) golf ball cd-comparison (Bearman and Harvey, 1976) LAMINAR G-R S-A DES 1.5 1.4 1.2 1.3 1 1.2 CD CD 1.4 : 0.8 1.1 0.6 1 0.4 0.9 0.2 0.8 0 4 10 105 106 Re 107 Experiment (Tohoku Univ.) LAMINAR G-R S-A DES 0.7 2 104 4 104 6 104 8 104 105 Re At this flow conditions and grid, S-A model predicted the most accurate CD with experimental results Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Conclusions i. Turbulence effects around a dimpled cylinder using DES were numerically simulated. ii. For the comparison with DES, G-R and S-A one equation turbulence models and without turbulence model were applied. iii. The results plotted with the time averaged vorticity magnitude and velocity vectors showed the difference in the high vorticity regions in the wake of cylinder. iv. The time averaged section CP and velocity gradients distributions show the sudden peaks, due to the dimpled surface geometry. v. For drag comparisons, S-A model predicted the most accurate results with experimental data. Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan Thank you for your attention. Hyoung-Chol KIM, Dep’t of Aerospace Engineering, Tohoku Univ., Japan