Document
FUN3D Solutions for Nose Landing Gear
Veer N. Vatsa, David P. Lockard And Mehdi R. Khorrami NASA Langley Research Center, Hampton, VA
Outline
• Objectives
• Numerical Method
• Configuration and Flow Conditions
• Grids
• Results
• Computational Resources
• Observations
Objectives
• Assess the applicability of an unstructured grid flow solver
FUN3D for Nose Landing Gear configuration
• Examine grid and turbulence modeling sensitivity
Numerical Method
• Equations solved
Unsteady Reynolds-averaged Navier-Stokes (URANS) equations Fully unstructured node-based flow solver (FUN3D) Turbulence models – Hybrid RANS/LES model (Ref. Lynch et al. AIAA Paper 2008-3854) – Modified Delayed Detached Eddy Simulation (MDDES) model (Ref. Vatsa and Lockard AIAA Paper 2010-4001)
• Spatial and temporal discretizations
Roe’s flux -difference splitting scheme without flux limiter Optimized second-order backward difference (BDF2OPT) scheme for temporal discretization: Dual-time stepping with 15 subiterations
• Boundary Conditions
Constant temperature, no-slip floor & gear Inviscid side walls & ceiling subsonic inflow/outflow for inlet and exit planes – Outlet pressure specified – Inlet total pressure and temperature specified
Configuration and Flow Conditions
• Re = 73,000 based on post diameter
- Flow code run in fully turbulent mode
• M = 0.166
Computational grids
• Unstructured, mixed-element grids using VGRID • Sequence of 3 successively refined grids: 9, 25 and 71 million nodes • Locally enriched 47 million node grid Inviscid Inflow Tunnel ceiling plane Viscous Outflow Tunnel floor plane
Results
• Time step
-6 4.92x10 seconds
• Number of time steps run
Total : minimum of 80,000 time steps Sampling : Minimum of 50,000 time steps
• Convergence information
Cp and Cp checked after every 10,000 time steps rms
Surface Pressure comparisons
(starboard wheel)
Turbulence modeling Grid sensitivity sensitivity
Surface Pressure comparisons
o
(port wheel transverse cut at 237 )
Turbulence modeling Grid sensitivity sensitivity
Surface Pressure comparisons at door
(Rows 2-4) Row 2 Row 4 FUN3D-25M-HRLES FUN3D-71M-HRLES
Surface Pressure comparisons at door
(Rows-5-8) Row 5 Row 8 FUN3D-25M-HRLES FUN3D-71M-HRLES
Power Spectral Density Comparisons
Power Spectral Density Comparisons … (2)
Partial view of grid near torque-arm
25 M node grid 47 M node grid
2-D Turbulence Kinetic Energy
at wheel wake centerline
FUN3D-25M-HRLES FUN3D-71M-HRLES Exp. PIV data
Spanwise vorticity
at wheel wake centerline
FUN3D-25M-HRLES FUN3D-71M-HRLES Exp. PIV data
Spanwise vorticity at torque arm wake
FUN3D-25M-HRLES FUN3D-71M-HRLES Exp. PIV data
Iso-surfaces of Q-criterion
• Colored with perturbation pressure FUN3D-9M-HRLES FUN3D-25M-HRLES
Computational Resources
• Computer hardware
CPU: NAS Pleiades, 2 quad-core Xeon E5472 Harpertown cpu’s /node, 1GB memory/core Interconnect:Infiniband
• Resources (for 25 M nodes, HRLES case)
CPU (or wall clock) Time / time step : 33.8 secs. using 960 cores – Minimum of 80,000 time steps in simulation – Minimum of 50,000 time-steps for data sampling
Observations
• What did you learn?
Computational challenges – Significant computational effort for statistically meaningful results – Constructing suitable grids very challenging New insights into the physics – Complex flow physics, difficult to simulate with fixed (non-adapting)grids Manual, local refinement effective but tedious – Tunnel inflow/outflow b.c.’s could influence computations – Transition difficult to simulate, could impact flow on smaller components Assessment of state-of-the-art based on your simulation for the problem category of interest – Encouraging results, solutions capture salient flow features – Uncertainty due to grids, transition and turbulence modeling Recommendations for follow-on efforts – Need test data to quantify Reynolds number sensitivity – Need systematic grid refinement/adaptation studies, better turbulence/transition modeling