CROSS-CODE VALIDATION OF UN
Transcript
CROSS-CODE VALIDATION OF UN
CROSS-CODE VALIDATION OF UN-STEADY, THREE-DIMENSIONAL FLOWS PAST BLUFF BODIES USING UNSTRUCTURED GRIDS A. Bonfiglioli,† R. Paciorri¶ † Università della Basilicata, Scuola di Ingegneria, Potenza, Italy ¶ Università di Roma La Sapienza, Dip. di Ing. Meccanica e Aerospaziale, Rome, Italy EFMC9 9th European Fluid Mechanics Conference 9-13 Sept 2012 / Rome, Italy A. Bonfiglioli, R. Paciorri Aim of the study and Outline of the talk • Perform a code-to-code (same grid) validation exercise for a non-trivial 3D, un-steady (laminar) flow configuration – We have chosen one of the test-cases belonging to the iCFD db hosted at CINECA: 3D un-steady flow around a confined square cylinder (contributed by Camarri et al. )1 • Outline – – – – 1 Test-case definition. Features of the two codes. Cross-comparison. Physical flow features. Camarri et al. XVII Congresso di Meccanica Teorica (2005) EFMC9 9-13 Sept 2012 / Rome, Italy 2 A. Bonfiglioli, R. Paciorri 3D channel flow past a square obstacle EFMC9 9-13 Sept 2012 / Rome, Italy 3 A. Bonfiglioli, R. Paciorri Mathematical & numerical models • Model equations: – Compressible (low Mach), laminar (ReD = 200) Navier-Stokes equations. • Numerical models confronted: 1. The AERO code2 (provides the reference calculation3 available in the iCFD db) 2. The in-house developed eulfs code Both codes are vertex-centred and used tetrahedral meshes; therefore the nof degrees of freedom is the same in both calculations (approx. 32.5M dofs per physical time step). 2 C. Farhat, B. Koobus, H. Tran - Computational Methods for Fluid-Structure Interaction (1999) 3 Camarri et al. XVII Congresso di Meccanica Teorica (2005) EFMC9 9-13 Sept 2012 / Rome, Italy 4 A. Bonfiglioli, R. Paciorri Distinctive features of the two numerical models The AERO code The eulfs code • Mixed FEM-FVM space discretization (O(h2) accurate) (distance-2 neighbours stencil) • Fluctuation Splitting space discretization (O (h2) accurate) (distance-1 neighbours stencil) • Turkel’s low preconditioning • Merkle’s low preconditioning speed • Implicit time integration (O(∆t2) accurate) EFMC9 9-13 Sept 2012 / Rome, Italy speed • Implicit (3 time levels) time integration (O(∆t2) accurate) 5 A. Bonfiglioli, R. Paciorri Initial/boundary conditions Inflow/outflow boundary conditions are not trivial due to the need of prescribing a Poiseuille inflow profile with maximum (centreline) Mach number equal to 0.1. The AERO code • Initial conditions correspond to an impulsive start EFMC9 9-13 Sept 2012 / Rome, Italy The eulfs code • Initial conditions are given by a snapshot of the 2D calculation (with w = 0) 6 A. Bonfiglioli, R. Paciorri Time evolution of the aerodynamic coefficients 2 Aerodynamic coefficients 1.5 CD AERO code CL AERO code CL eulfs code CD eulfs code 1 0.5 0 -0.5 0 100 200 300 400 500 t Um/D EFMC9 9-13 Sept 2012 / Rome, Italy 7 A. Bonfiglioli, R. Paciorri Time evolution of the aerodynamic coefficients 2 Aerodynamic coefficients 1.5 CD AERO code CL AERO code CL eulfs code CD eulfs code 1 0.5 0 -0.5 300 EFMC9 9-13 Sept 2012 / Rome, Italy 400 t Um/D 500 8 A. Bonfiglioli, R. Paciorri Onset of three-dimensionality The eulfs code The AERO code eulfs code AERO code time evolution of the transverse velocity component Time evolution of the crossflow (w) velocity component 0.6 0.6 0.5 0.5 max(w) -min(w) mean(w) x 10 w/Um 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 100 150 EFMC9 9-13 Sept 2012 / Rome, Italy 200 t Um/D 250 300 max(w) -min(w) avg(w) x 10 350 0 200 250 300 350 t Um/D 400 450 500 9 A. Bonfiglioli, R. Paciorri Three-dimensionality instabilities (un-bounded flow)4 2D flow (Re=155) 4 Mode A instability (Re=188) Mode B instability (Re=244) Experimental observations from Luo, Thong, Khoo J. of Fluids and Structures 23 (2007) EFMC9 9-13 Sept 2012 / Rome, Italy 10 A. Bonfiglioli, R. Paciorri Mode A or mode B ? The above figure shows mode A instability for the un-bounded flow past a square cylinder at Re = 210, from Sheard et al. J. Fluid Mech. (2009) EFMC9 9-13 Sept 2012 / Rome, Italy 11 A. Bonfiglioli, R. Paciorri Movies Here we show movies .......... EFMC9 9-13 Sept 2012 / Rome, Italy 12 A. Bonfiglioli, R. Paciorri Conclusions • Numerics: – differences btw. the results (e.g. time evolution of the aerodynamic coefficients) obtained using the two codes have been observed, despite the use of the same spatial grid. – 2D grid convergence studies (un-feasible in 3D) have shown that the spatial resolution (mesh size, preconditioning parameter) has a larger impact on the results than the temporal one. • Flow physics: – are we observing both mode A and mode B instabilities? Or what else? EFMC9 9-13 Sept 2012 / Rome, Italy 13 A. Bonfiglioli, R. Paciorri Aknowledgments Computational resources on the Matrix cluster have been made available by CASPUR (Consorzio interuniversitario per le Applicazioni di Supercalcolo per Università e Ricerca) under grants Standard HPC 2010 and Standard HPC 2011. EFMC9 9-13 Sept 2012 / Rome, Italy 14