Shock/boundary layer interactions Turbulent compressible channel flows F.S. Godeferd Laboratoire de Mécanique des Fluides et d Acoustique Ecole Centrale de Lyon, France Journée Calcul Intensif en Rhône Alpes 1
Contents Presentation of the UFAST project Numerical method Shock in a curved channel Shock reflection Perspectives 2
An example of normal shock/boundary layer interaction [Experiment at the University of Cambridge - Bruce & Babinsky] 3
An example of normal shock/boundary layer interaction [Experiment at the University of Cambridge - Bruce & Babinsky] 3-a
The UFAST project 2006-2009 Unsteady effects of shock wave induced SeparaTion European STREP project 18 academic and industrial partners Experiments RANS, URANS LES Control http://www.ufast.gda.pl 4
Motivation of the study unsteady shock wave boundary layer interaction Aeronautical industry shock waves on wings/profiles, nozzle flows and inlet flows Interaction unsteadiness initiated and/or generated by SWBLI; often destabilized by the outer flow field; response of shock wave and separation to periodic excitations Control methods: synthetic jets, electro-hydrodynamic actuators, stream-wise vortex generators and transpiration flow Can we reproduce the unsteady interaction with URANS? Need of costly LES? 5
LMFA: URANS using platform elsa (ONERA) Unsteady Reynolds-Averaged Navier-Stokes Simulations Physical model mean flow Reynolds-averaged Navier-Stokes equations turbulence two-equation model k-l (turbulent kinetic energy, mixing length); or one-equation model Spalart-Allmaras for turbulent viscosity ν t. ( Sutherland viscosity law ν(t) = ν 0 T 0 +C T+C Adiabatic walls ) 3/2 T T 0 6
Numerical method Conservative finite volume method Roe fluxes with limiters Implicit Euler timestepping Multi-blocks structured mesh Parallel resolution 7
Initialization - Example: Channel flow with bump [expe. in Queens U. Belfast] Uniform initialization Euler laminar Navier-Stokes turbulent Navier-Stokes Subsonic freestream: Ma = 0.783 Peak: Ma = 1.365 Normal shock Animation 8
Performance Location Architecture # proc. # cores cpu/pt/ite Mem/pt Static max seconds bytes speedup LMFA Opteron 280 2Ghz 2 4 2.310 6 347 3.9 P2CHPD Opteron 252 2.6Ghz x3550 16 6.510 7 510 10 a IDRIS Nec SX8 1 1 6.710 7 394 1 ÉCL α-server EV7 1.15Ghz 1 1 9.910 6 495 1 9
URANS of curved channel flow at Ma = 1.45 Polish Institute of Mechanics experiment Inlet conditions: P=101kPa; T=290K. Turbulence: Tu=1%; L=1% channel height. Outlet conditions: pressure ratio specified in URANS to match experimental shock location 10
Splitted geometry with N = 8 procs; 4.5 10 6 grid points Bottom wall boundary layer resolution: y + 2 Side walls: y + 20 Need to test the dependence of the solution on various elements of the simulation 11
Details of flow in shock zone The simulation allows a detailed investigation of the flow Mach contours, showing Ma = 1.45 upstream the shock 12
Oil flow visualization URANS streaklines Detachment/re-attachment length 13
URANS streamlines access to 3D flow structure 14
Adjustment of shock position: dependence on geometry 1 degree opening of channel at the outlet Mach number contours in a transverse wall for choked and unchoked geometries. The curve shows the Mach number at mid-section in this plane. 15
Dependence on expression of fluxes and outlet pressure Effect of counterpressure. Effect of fluxes scheme: Roe or Jameson. 16
Flow unsteadiness Main shock oscillations in the experiment: power spectrum No unsteadiness in the URANS need to introduce explicit fluctuations in URANS 17
Shock reflection case: expe. TU Delft 18
19
URANS Turbulence model: k L Mach number at two times and fluctuations at point in shock region 20
Unphysical solution! Huge detachment and unsteadiness due to poorly performing turbulence model. Reason: turbulence is strongly damped in incoming boundary layer. 21
URANS Turbulence model: Spalart-Allmaras More reasonable shock detachment but steady flow 22
Perspective: introduce explicit fluctuations at inlet BL fluctuations are sufficient to trigger SWBLI unsteadiness Synthetic fluctuations injection 23
Perspectives of UFAST project Control of experimental flows Simulations with control Large eddy simulations and hybrid RANS/LES methods Comparison experiment/simulation results and synthesis, with industry partners Database available on web: 2009 24