NASCART-GT - Georgia Tech
Transcription
NASCART-GT - Georgia Tech
Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin NASCART-GT: A Viscous SolutionNASCARTSolutionAdaptive Cartesian Grid Flow Solver Stephen M. Ruffin [email protected] Associate Professor, School of AE Georgia Institute of Technology Atlanta, GA Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Overview of Cartesian Grid Approach Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Introduction Overview of Cartesian Grid Approach • • • • Unstructured method with cell faces aligned with coordinate directions Control volume approach used to solve governing equations Internal cells created by subdividing parent cells Boundary cells “cut” by surface Overlayed Cartesian Cell Cut Cell Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Introduction Overview of Cartesian Grid Approach • Benefits – Automated grid generation and geometry definition separated from grid resolution selection – Less truncation error for cells since have orthogonal control volumes – Well-suited for high-order schemes due to orthogonal, regularlyspaced cells – Fewer terms in equations (example: Navier-Stokes momentum equation – 14 terms vs. 94 terms for generalized structured grid, Meakin 1997) • Drawbacks – Solid surfaces required more computational work due to the generalized finite volume formulations – Like other unstructured grid approaches, more bookkeeping needed to preserve grid topology than for structured grids – Novel boundary condition treatments needed to handle viscous derivatives near surface of cut cells. Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Viscous Flux Stencils Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Introduction Prior Viscous Cartesian Related Efforts • Navier-Stokes on Pure Cartesian Grids – Coirier and Powell (1993, 1996) demonstrated that traditional viscous flux formulations failed due to the nonsmoothness of the grids – The tradeoff between accuracy and positivity in the viscous flux stencil indicates that a new treatment for surface cells is needed – Non-physical fluctuation and instability due to non-body-fitted boundary cells • Existence of cell centers inside of the wall boundary • Sharp fluctuation of the distance from cell centers to the wall Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Viscous Wall BC • Example of initial attempts . . -1.5 -1.5 Direct implementation Smoothing treatment -1 -1 -0.5 Cp Cp -0.5 0 . . 0 0.5 AGARD Experiment Smoothing treatment 0.5 1 1 1.5 0 0.2 0.4 0.6 0.8 1 x/c Laminar: Cp over NACA0012 airfoil, Rec = 2.0E5, M∞ = 0.3, α = 3.59° 0 0.2 0.4 0.6 0.8 1 x/c Turbulent: Cp over NACA0012 airfoil, Rec = 1.86E6, M∞ = 0.3, α = 3.59° Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Boundary Condition Development Embedded Boundary Method • • Basic Idea: Interpolate to find state vector at reference cells (e.g. locations C3 and C5 in schematic) along surface normals to set state vector at “ghost cells” (e.g. cells 3 and 5) and perform finite volume integration all cells except “ghost cells” – Utilize flat wall and curve wall extrapolations similar to Extrapolation Method to set state vector at ghost cells – In finite volume integration over surface cells (e.g. cell 1), are treated as if uncut Benefits: – Eliminates viscous flux stencil positivity problem due to perfectly uniform cell centroids near surface (Extrapolation Method was slightly non-uniform) – Eliminates time integration limitation associated with the cut cells – All cells (surface and flow cells) utilize C3 4 2 1 3 C5 5 Ghost Cells Body Surface Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin NACA0012: Laminar • Fine grid solutions – Casalini : structured grid – NASCART : Embedded Boundary Interpolation Method -1.5 0.6 Casalini 1999 NASCART -1 0.4 -0.5 Cf Cp 0.2 0 0 0.5 -0.2 Casalini 1999 NASCART 1 1.5 -0.4 0 0.2 0.4 0.6 0.8 x/c < Pressure coefficient > 1 0 0.2 0.4 0.6 0.8 x/c < Skin friction coefficient > 1 Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Inviscid & Laminar Wall BC • Embedded boundary ghost cell approach using reference point 1 2 – Primitive variables at reference point (B, D) interpolated from its 3 closest neighbor cells using Linear Least Square Interpolation 5 6 B 3 4 7 8 δr D – Pressure p g = p ref =0 : δg V N,ref δr 12 w ∂T =0 ∂η ) κ : curvature p g = p ref ∂p − ∂η (δ g δr + Tw +δg ) 13 14 15 w Tref = Tw = Tg )δ r δg ( ≈ κ ρV δg V N, g = − A C 2 T ref Isothermal wall : Tg = (Tw − Tref – Normal velocity 11 w – Temperature Adiabatic wall : 10 9 δr ∂p ∂η Curved wall : ∂p ∂η Flat wall δr = cell diagonal length δg = distance from wall – Tangential velocity Inviscid : VT ,g = VT ,ref Laminar : VT ,g = − δg VT ,ref δr 16 Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Turbulent Wall BC • Wall function with standard k-ε model – Introduced by Launder and Spalding (1974) – Very grid efficient method for solving the RANS equations – Widely used in structured and tetrahedral unstructured grid solvers – Smooth variation of grid cell distance from the wall required – Conventional methods are not compatible with non-bodyfitted immersed Cartesian cells • Unphysical fluctuation and separation / unreal flow acceleration in transient solution • Complicated coordinate transformation required to integrate viscous flux → Increased truncation error & attenuate the advantage of Cartesian grid topology Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin New Turbulent Wall BC • The law of the wall – Spalding’s formulation : + 2 + u κ κu + + + −κB κu + y = u + e e − 1 − κu − − 2 6 τw δρ u u uτ ≡ y+ = w τ u+ = t ρw uτ µw ( ) ( ) 3 κ : von Karman constant, 0.41 B : related to roughness parameter – Valid for log-law layer, buffer layer and viscous sublayer – Excellent agreement with various experimental data, even for y+>300 – Wall temperature from Crocco-Busemann equation : Tw = Tref 2 r VT ,ref + 2 cp – Use of iterative method, i.e. Newton’s method to find wall shear VT ,ref stress from Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin New Turbulent Wall BC • Application of wall shear stress – Numerical approximation : τ xy ,1+ 1 2 ≈ 1 [(µ l + µ t )1 + (µ l + µ t )2 ] u 2 − u1 2 y 2 − y1 – Computed wall shear stress ≠ Actual wall shear stress – Computed wall shear stress should be corrected to the actual value obtained from the law of the wall Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Viscous Wall BC • Results of initial attempts of shear stress transformation in embedded Cartesian solver – Slip wall condition • Observed non-physical flow acceleration near wall in transient solution • Large CPU time required to converge – No-slip wall condition • Observed good solution in body-fitted Cartesian grid system, i.e. flat plate • Found non-physical fluctuation and instability in nonbody-fitted boundary cells • Fluctuations could not be removed in many attempts Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin New Turbulent Wall BC • Requirements – Remove non-physical fluctuation – Yield realistic transient solution – Eliminate complicated coordinate transformation – Give stable solution with relatively coarse grid at y+>300 • Development of new approach – When tangential velocity of ghost cell corrected to satisfy actual wall shear stress, this would eliminate the need for the coordinate transformation – Cells near wall would remain in numerically linear region Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin New Turbulent Wall BC Sample results from new model for NACA0012, Turbulent • -1.5 Flow conditions – M∞ = 0.3, α = 3.59°, Rec = 1.86 ×106 – • B = 5.0 (smooth wall) Computational grid – 5 chords apart from airfoil – 22×20 root cells – Refinement level : 8, 9, 10 for coarse, medium and fine grids Pressure distribution – Very good correlation observed for fine & medium grids – Coarse grid solution nicely correlated except slight under-prediction at suction peak -0.5 Cp • -1 0 0.5 Experiment NASCART, Fine (y+ = 81.1) NASCART, Medium (y+ = 154.2) NASCART, Coarse (y+ = 287.2) 1 1.5 0 0.2 0.4 0.6 x/c 0.8 1 Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Hemispheroid Turbulence Validation • Validation of k-epsilon turbulence model in NASCART-GT conducted for flat plates and axisymmetric bodies Flow conditions for hemispheroid tests 0.01 – M∞ = 0.063 in experiment 0.008 = 0.3 in computation 0.006 – ReL = 2.0×106 – B = 5.5 (from 0.004 experiment) Experiment, theta=0 (deg) Experiment, theta=180 (deg) NASCART Cf • Ramaprian’s (1981) hemispheroid Computed vs Experimental Skin friction • Good skin friction correlation observed 0.002 – Discrepancy at first 0 station caused by transitional flow (not-0.002 0 modeled) at that location 0.2 0.4 0.6 x/c 0.8 1 Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin NASCART-GT Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin NASCART-GT: Flow Modeling Overview www.ae.gatech.edu/~sruffin/nascart • Governing Equations: – Navier-Stokes • Laminar • Turbulent, k-Epsilon model – Euler + Integral Boundary Layer • Laminar, transition, & turbulent • Uncoupled viscous/inviscid • Coupled viscous/inviscid – Euler Thermal Models – Ideal gas, Equilibrium & NonEquilibrium Reacting flow Time Integration: – 2nd order, explicit – LU-SGS, implicit Solution-adaption parameters: – Divergence, vorticity, entropy magnitude, turbulent k gradient, turbulent epsilon gradient Inviscid Fluxes: – 2nd, 3rd or 5th order – Roe FVS – AUSM/PW+ (for hypersonic) • Boundary Conditions – Isothermal, Adiabatic, radiative equilibrium thermal BC – Moving & flexible body BC’s – Actuator Disk model – Riemann invariants at Inflow/Outflow for External Flows – Slip and no-slip surface conditions – Supersonic exhaust, Characteristic based for subsonic internal flows* *Future capability Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Sample Results Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin NASCART-GT Sample Perfect Gas Results Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Current Capability Results – Cessna 172 Mach=0.2, AoA=2.5 [deg], Alt=SL Pressure Contours 9.900e+4 [N/m2] 1.025e+5 [N/m2] Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Current Capability Results – Transonic Inviscid NACA 0012 Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Current Capability Results – Transonic Inviscid NACA 0012 Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Boundary Condition Development Extrapolation Method Circular Cylinder Drag Coefficient Circular Cylinder Analysis 100 Measurements by C. Wieselsberger Numerical results by Hamielec Asymptotic formula for Re near zero 10 Cd NASCART-GT Solution-adapted grid & entropy contours [NASCART-GT], Re=1.0e4 Computed: Sr=0.21 [NASCART-GT] Experimental: Sr=0.21 [Schlicting] 1 0.1 1.0E-1 1.0E+0 1.0E+1 1.0E+2 1.0E+3 1.0E+4 1.0E+5 Re • • NASCART-GT viscous solutions developed using solution adaption technique for vortex shedding Viscous boundary treatment of cut cells yields good agreement with experimental drag and Strouhal Number (Sr) 1.0E+6 Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Boundary Condition Development Extrapolation Method Circular Cylinder Analysis Predicted entropy contours [NASCART], Re=1.0e4 Predicted entropy contours [NASCART-GT], Re=140 Experimental Karman vortex street [Van Dyke], Re=140 Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Moving Geometry Capability Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Moving Geometry Capability • Three routines installed for analysis of moving and deforming geometries – Geometry motion module for each grid surface node and rigid body movement – Computational grid adaptation module – Flow boundary condition module • Developed and tested interface to LS-DYNA and performed aeroelastic simulation using modified Newtonian option of NASCART-GT Computed solution around NACA0012 Airfoil with active trailing-edge flap Computed Mach number & adapted grid for wing/pylon/store configuration at M=0.95 Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Rotorcraft Applications Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin GT Rotor Model • 3 test cases – Euler, no flap – RANS, no flap – RANS, flap • Flow conditions – M∞ = 0.3, α = 0°, ReL = 9.196 ×105 – CT = 0.009045 h/R = 0.3 • Computational grid – 5 body length apart from airfoil – 26×22×20 root cells, 8 level refinement – y+max = 169.9 (no flap) = 209.9 (flap) Rotor blade : NACA 0015 < GT rotor configuration > • Actuator disk model – Trimming on collective pitch angle only (no cyclic pitch in experimental model) Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin GT Rotor Model • Final grid configuration & Entropy contours – Cells refined around fuselage and disk – Entropy is a measure of vorticity – Vortices are well captured and cells adapted around vortex core – Stronger tip vortex and faster descending of vortex core in advancing side than retreating side due to BET Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin GT Rotor Model • Upper centerline Cp Experiment Outline Euler, no flap RANS, no flap RANS, flap 2 – Euler & RANS results are close except reduced pressure peak due to viscous dissipation and wakes – Peak pressure and its location estimated better by adding flap effect – Inconsistency with experiment near nose and local peak & drop at x/R=0.3 – Limitation of actuator disk model : not analyzing unsteady wake effect Cp 1 0 -1 0 1 2 x/R 3 Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin GT Rotor Model • Iso-surface of entropy – Vortex rollup at fore disk propagates as a vortex sheet & dissipates rapidly – Tip vortex rollup at lateral sides of disk merges to strong line vortex & propagates further – In reality, individual blade generates strong vortex filaments by tip vortex rollup to form helical shape line vortex (illustrated by O’Brien and Smith from overset method, 2005) → Unsteady effect – Actuator disk model based on time-averaged loading cannot produce these unsteady effect → Model limitation Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Current Capability Results – Robin Rotorcraft Test Case 1 • Fuselage alone Realistic Configuration • Vinf = 81.7 knots Described by Super-Ellipse Equations • AOA = 0.0o • Re/L = 2.88x106/m Extensively Tested Model Features: Test Case 2 • Fuselage with rotor • Vinf = 81.7 knots • AOA = 2.68o • Advance ratio = 0.15 NASCART-GT Grid and Computed Pressure Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin ROBIN Fuselage/Rotor Systems • Successful 3-D tests of vortex refinement with k-epsilon turbulence model in NASCART-GT conducted • Flow conditions for ROBIN test – M∞ = 0.3, α = 0°, 1.2° nose left, Re2L = 1.312 ×106 – Rotor: µ = 0.051, CT = 0.00636 • Effective mesh refinement demonstrate d in rotor wake for 3-D configuration s Entropy Contours, Cartesian Grid and Actuator Disk for ROBIN tests Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin ROBIN Model • Cp on upper centerline 6 4 2 Cp – Offset of pressure tap location makes two distinct value at same x-location – Both solutions well correlated with experiment – Two solutions very similar except pylon region where flow separation occurs 0 Experiment, averaged unsteady Experiment, steady Euler RANS -2 -4 0 0.4 0.8 1.2 x/L 1.6 2 Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin ROBIN Model • Sectional Cp 0.15 0.1 – Well correlated with experiment – Discrepancy on upper surface at x/L=1.17 & on lower surface at x/L=1.354 seems to be due to strut & rotor shaft in experiment z/L 0 x/L=1.170 x/L=0.353 -0.05 -0.1 0.15 -0.15 x/L=1.540 0.1 0.05 x/L=1.354 z/L – Difference between two solutions increased in x-dir due to enhanced viscous effect downstream 0.05 0 Exp. RHS Exp. LHS Euler RANS -0.05 -0.1 -0.15 -12 -8 -4 0 Cp 4 -12 8 -8 -4 0 Cp 4 8 Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin NASCART-GT Summary Integral Boundary Layer Option • • Rapid determination of viscous effects for laminar, transitional or turbulent flow Integration conducted along surface streamlines from 3-D solution to include pressure gradient effects (for attached flow) Robin Fuselage turbulent • Rexe – – 22,400 0.46 Re xe ReθTr = 1.1741 + Re xe – • laminar Laminar: Twaites Method Transition: Michel’s Criterion, Turbulent: Head’s Method Separation predicted by shape factor criteria. Cf Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Parallelization Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Intro: Space Filling Curve • Idea: - Fills every nodes of multidimensional space while preserving locality - Reduce the multidimensional space into 1D - Domain decomposition: “cutting” this 1D line Hilbert: Morton: •Why use a Space Filling Curve (SFC) for parallelization? •Communication Overhead: SFC maintains cells in contiguous blocks thus lowering inter-CPU communication time. •Load Balancing: SFC allows the number of cells on each processor to be equal providing efficient use of each computer node. •Simplicity: Does not take an iterative method to decompose domains Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Intro: Space Filling Curve • History of SFC - First described by Italian mathematician, Giuseppe Peano, 1890 - A year later, Hilbert curve was described by David Hilbert - Then several more versions are described by Jordan, Morton, Moore - Recently adopted by wide range of disciplines: image compression, vision sensing, ground mapping for GPS, motion pictures, CAD, CFD • Hilbert curve outperforms the others in preserving locality - Always connects the closest two nodes • Why use a Space Filling Curve (SFC) for parallelization? - Communication Overhead: SFC maintains cells in contiguous blocks thus lowering inter-CPU communication time. - Load Balancing: SFC allows the number of cells on each processor to be equal providing efficient use of each computer node. - Simplicity: Does not take an iterative method to decompose domains Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin ONERA M6 Test Case 14 CPUs 1.2 Million Cells 0.9 Million Surface Panel Mach Number = 0.84 Angle of attack = 3.06 Velocity Divergence Solution Adaption Inviscid mode Pressure Contours Sectional view of domain decomposition Smooth communication across CPU boundaries Hub Mid-section Tip Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Parallelization Results Decomposed domain distribution to 8 CPUs while performing solution adaption Initial Grids over 3D Sphere Grids after 39th Solution Adaption over 3D Sphere M = 1.2, A.O.A. = 3 degree Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Hypersonic and Reacting Flows Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Thermo--Chemical Nonequilibrium Thermo Development • Governing Equations - Euler or Euler + Integral boundary layer method - Navier Stokes with two equation (k-epsilon) turbulence model - Species conservation equations - Vibrational energy equation (Two Temperature Model), L-T vibrational relaxation model • Thermodynamic Model - Calorically perfect gas - Chemically reacting thermally perfect gas: (NASA Glenn Thermodynamic Curve Fit) • Chemistry Model - Air: Dunn/Kang, Gupta, Park85, Park87, Park91 - Titan N2-CH4-Ar: Nelson, Gokcen • Flux Calculation - Inviscid flux: AUSMPW+ scheme with variable ratio of specific heat for reacting gas MUSCL data reconstruction -Species flux + Chemistry Source Term: Point Implicit • Solution Adaption - Velocity Divergence, Vorticity, Gradient of K-epsilon, Species Gradient Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Equations for Thermochemical N.E. Total Energy: Governing Equations: Chemical Source Term: Point Implicit Formulations: Vibrational Source Term: Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin AUSMPW+ Scheme Idea : Fluxes splitting based on intuitively defined cell interface Mach number PW+ : Pressure Weighted, Critical Speed of Sound Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin 2D Cylinder Verification Results for Thermochemical N.E. (Air5 (Air5--Park93) Freestream Conditions : M ∞ = 12.7, ρ ∞ = 1.60 × 10−3 kg / m 3 , T∞ = 196K , P∞ = 90 pa Present Max: 5.96E+3 Min: 1.96E+2 Present Max: 1.54E-2 Min: 1.60E-3 Present Max: 1.88E+4 Min: 9.04E+1 Present Max: 4.77E+3 Min: 1.93E+2 DPLR Max: 6.11E+3 Min: 1.96E+2 DPLR Max: 1.55E-2 Min: 1.60-3 DPLR Max: 1.90E+4 Min: 9.04E+1 DPLR Max: 4.84E+3 Min: 1.96E+2 Temperature Density Pressure Vibrational T Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin 2D Cylinder Verification Results for Thermochemical N.E. (Air5 (Air5--Park93) Freestream Conditions : M ∞ = 12.7, ρ ∞ = 1.60 × 10−3 kg / m 3 , T∞ = 196K , P∞ = 90 pa Present Present Present Present Present DPLR DPLR DPLR DPLR DPLR N2 O2 NO O N Present Max: 7.67E-1 Min: 7.26E-1 Present Max: 2.33E-1 Min: 1.81E-2 Present Max: 8.01E-1 Min: 0.00 Present Max: 1.95E-2 Min: 0.00 Present Max: 1.49E-1 Min: 0.00 DPLR Max: 7.67E-1 Min: 7.24E-1 DPLR Max: 2.33E-1 Min: 1.93E-2 DPLR Max: 8.71E-1 Min: 0.00 DPLR Max: 1.94E-2 Min: 0.00 DPLR Max: 1.43E-1 Min: 0.00 Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin 2D Cylinder Verification Results for Thermochemical N.E. (Air5 (Air5--Park93) Freestream Conditions : M ∞ = 12.7, ρ ∞ = 1.60 × 10−3 kg / m 3 , T∞ = 196K , P∞ = 90 pa Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Huygen Probe Verification Results for Thermochemical N.E. (N2(N2-CH4 CH4--Ar 13) Freestream Conditions : M ∞ = 18.86, ρ ∞ = 2.96 × 10−4 kg / m 3 , P∞ = 15.621 pa Present Cartesian Grids 60 x 40 Structured Grids Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Huygen Probe Verification Results for Thermochemical N.E. (N2(N2-CH4 CH4--Ar) Freestream Conditions : M ∞ = 18.86, ρ ∞ = 2.96 × 10−4 kg / m 3 , P∞ = 15.621 pa Present DPLR Temperature Present DPLR Vibrational Temperature Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Huygen Probe Verification Results for Thermochemical N.E. (N2(N2-CH4 CH4--Ar) Freestream Conditions : M ∞ = 18.86, ρ ∞ = 2.96 × 10−4 kg / m 3 , P∞ = 15.621 pa DPLR Present Density Present Pressure DPLR Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Huygen Probe Verification Results for Thermochemical N.E. (N2(N2-CH4 CH4--Ar) Freestream Conditions : M ∞ = 18.86, ρ ∞ = 2.96 × 10−4 kg / m 3 , P∞ = 15.621 pa DPLR Present N2 DPLR Present CN Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Shock Structure Comparison for Trailing Toroidal Ballute • Calorically Perfect gas comparison with LAURA Navier-Stokes Solver – Good Agreement in Flow Structure: shock locations and shock interaction – Agreement in computed CD value within 5% – NASCART run in Euler mode so differences exist in Ballute trailing edge, separated flow region relative to viscous Laura result shown. Temperature Distribution in plane of symmetry at M = 26.3 calculated from LAURA Navier-Stokes solver, Cd = 1.39 (by P. Gnoffo, NASA LRC) Temperature Distribution in plane of symmetry at M = 26.3 calculated from NASCART-GT Euler solver, Cd = 1.318 Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Validation to Wind Tunnel Test • Coupled Solution obtained from NASCART-GT & LS-DYNA • Coupled Solution matches well with the wind tunnel data in axial displacement and shock shape Shock structure over deformed clamped ballute at freestream of Mach = 5.73, Density = 1.46x10e-3 kg/m^3, Temp = 254k by Reuben R. Rohrschneider Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Titan Clamped Ballute (Results) Temperature • Total of 13 species N2-CH4-Ar chemistry with two-temperature model • 187,043 active flow cells • Solution adaption based on velocity divergence & species gradient • CD = 1.393 • No shock interactions are observed • Max. Temp outside of T.B.L. = 7.87E+3 K • Max. Pressure at Surface = 1.09E+2 pa Pressure Vibrational Temperature Modified Newtonian Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin Titan Clamped Ballute (Results) N2 and N Mole Fraction Contours, X-plane: N2 Y-plane and Surface: N CN, H2 and H Mole Fraction Contours, X-plane: CN, Y-plane: H2 Surface: H CH4, CH and C Mole Fraction Contours, X-plane: CH4, Y-plane: CH, surface: C • Max. N2 Dissociation = 8.25 % (by volume) • Max. CN(Cyano Radical) = 0.875% (by volume) • All the hydrocarbon species disappear right after the shock wave • Only C, N2, N, H exist near the surface Georgia Institute of Technology School of Aerospace Engineering Computational Capabilities Stephen M. Ruffin, http://www.ae.gatech.edu/~sruffin NASCART-GT: Next Steps • Continue reacting flow CFD development and validation for hypersonic flows. • Develop validations of coupled viscous approach for skin friction and heat transfer including separated flows – Demonstrate efficiency & accuracy of approach. • Enhance efficiency of parallelized code to utilize multi-processor computational clusters. • Develop adjoint methodology for coupling with vehicle optimization tools. • Develop 6 DOF integration for store separation and moving body problems. • Develop flexible geometry integration with structural analysis solvers.