Using HPC for Understanding fluid flows

Transcription

1 Using HPC for Understanding fluid flows Sanjay Mittal Department of Aerospace Engineering IIT Kanpur Acknowledgement: Students, Colleagues, Various funding agencies

2 Incompressible flow equations Unsteady, Non linear, Coupled, PDE's

3 Evolution of computing power

4 First simulation for an aircraft: First mesh: 1978 First computation: 1986

5

6 Finite Element Formulation (DSD/ST):

7 Parallel Computing: 32 node Linux Cluster. Each node: Dual processor 3.06 Ghz Xeon, 512 K, 4 GB RAM, 72 GB HDD Gigabit Switch Domain partitioning The non linear equations resulting from the finite element discretization are solved using GMRES method with diagonal preconditioner Using MPI Libraries

8 Parallel computing: Algorithm Data structure at three levels Node: data, coordinates Element: formation of matrices and vectors Equation: assembly of element level data and solution to the equation system. Behara & Mittal, Parallel Computing (2009)

9 Parallel computing: Mesh Statistics: in millions Mesh nn M1 M2 M3 M4 ne neq Super-linear Speed-up Behara & Mittal, Parallel Computing (2009)

10 Parallel computing: percentage time spent in various activities 1. Communications 2. Formation of RHS & preconditioner 3. Preconditioning 4. Matrix-free GMRES 5. Misc. operations Total

11 Flow past a circular cylinder Cylinder: bluff body with simple geometry has all the flow complexities Reynolds number=ud/ Williamson (1996)

12 Flow past a circular cylinder First convective wake instability: Re ~ 5; Monkewitz (1988) Re ~ 4; (present study) (Phy. of Fluids, 2008) Onset of flow separation: Re=6.29 (JFM 2009) First wake instability (self sustained) : Re ~ 47: leads to von Karman shedding (CMAME, IJNMF 2006) Shear layer instability (convective) : Wide scatter Re ~ 1300; Bloor (1964) Re ~ 350; Gerrard (1978) Re ~ 1900; Unal & Rockwell (1988) Re ~ 1200; Prasad & Williamson (1997) Re ~ 740; Rajagoplan & Antonia (2005) Re ~ 54; (present study) (Phy. Of Fluids, 2008; JFM 2009)

13 Drag Crisis: Role of shear layer instability Drag Crisis: Sudden loss in the drag coefficient. Transition of boundary layer from laminar to turbulent Question: What is the mechanism of this transition

14 Flow past a stationary cylinder: shear layer instability Singh & Mittal, IJNMF (2005)

15 Flow past a stationary cylinder: shear layer instability Singh & Mittal, IJNMF (2005)

16 Flow past a stationary cylinder: shear layer instability Mean Drag Coefficient Base Pressure Coeff. Singh & Mittal, IJNMF (2005)

17 Flow past a stationary cylinder: shear layer instability the onset of shear layer instability moves upstream with Re At the critical Re, the shear layer vortices cause mixing of flow in the boundary layer Singh & Mittal, IJNMF (2005)

18 3D Flow past a cylinder: understanding instabilities Cellular shedding & vortex dislocations

19 3D Flow past a cylinder: understanding instabilities Re=100 Effect of aspect ratio: End cell remains same

20 Mixed Compression Air intake Fundamental difference in the actual working of an intake, experiments and numerics in terms of end conditions!

21 Types of Air Intakes External Compression Internal Compression Mixed Compression

22 Operation of Mixed compression Air Intake Super critical Critical Sub critical

23 Mixed Compression Air intake

24 Mixed Compression Air intake: Euler Mach number distribution for various values of back pressure. Jain & Mittal, IJNMF (2003)

25 Mixed Compression Air intake: Euler unstarting of the air intake for back pressure larger than a critical value; p_b/p_i=32.42

26 Viscous flow: the finite element mesh

27 Viscous flow: M=3.0, Re=106 14% increase in throat area, no bleed Mach number red : 3.0 blue: 0.0

28 Viscous flow: M=3.0, Re=106 14% increase in throat area, no bleed

29 Viscous flow: bleed no bleed unstarts 9 % bleed lower bleed 3.8% on cowl 5.2% on ramp starts M = 3.0, Re = 106, 14% increase in throat area

30 Viscous flow: 6% bleed, M=3.0, Re=106 14% increase in throat area mach number vorticity

31 Viscous flow:m=3.0, Re=106 Mass flow rate at throat for various cases

32 Viscous flow: M=3.0, Re=106 14% increase in throat area, 27% bleed pb /pi = 21.0 Mach number red : 3.0 blue: 0.0 Vivek & Mittal, Jour. Of Propulsion & Power (2009)

33 Viscous flow: M=3.0, Re=106 14% increase in throat area, 27% bleed pb /pi = 21.0

34 Viscous flow:m=3.0, Re=106 Two kinds of buzz are possible: Little buzz: Ferri Nucci type (shear layer instability) Big buzz: Dailey type (pressure/acoustic waves) Both are driven by superharmonics of the closed organ pipe modes of the intake

35 Viscous flow:m=3.0, Re=106 Little buzz: 25% increase in throat area, 6% bleed, pb / pi = 10.92

36 Viscous flow:m=3.0, Re=106 Big buzz: 14% increase in throat area, 9% bleed, pb / pi = 11.2

37 Viscous flow:m=3.0, Re=106 Big and Little buzz: 14% increase in throat area, 9% bleed, pb / pi = 10.5

38 Flow past a multi element airfoil High lift devices: Flaps Slats

39 Flow past a multi element airfoil 3 element airfoil: schematic

40 Flow past a multi element airfoil the mesh

41 Flow past a multi element airfoil α = 19 deg.

42 Flow past a multi element airfoil α = 19 deg.

43 Flow past a multi element airfoil locations for reporting velocity profiles

44 Flow past a multi element airfoil α = 19 deg.

45 Flow past a multi element airfoil α = 19 deg.

46 Thank You