This is the final public presentation for my PhD defense.
It covers a variety of topics such as object oriented design for scientific computing, high performance computing, parallelization, physical modeling and numerical algorithms for simulating complex aerothermodynamics phenomena, involving chemically reacting hypersonic flows in equilibrium and nonequilibrium
Automating Google Workspace (GWS) & more with Apps Script
An Object Oriented and High Performance Platform for Aerothermodynamics Simulation
1. COOLFluiD Framework
Aerothermodynamics
Conclusions
An Object Oriented and High Performance
Platform for Aerothermodynamics Simulation
Candidate: Andrea Lani
Promoter: Prof. Herman Deconinck
PhD presentation @ULB, 4th December 2008
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
2. COOLFluiD Framework
Aerothermodynamics
Conclusions
Presentation Overview
COOLFluiD Framework
Introduction
Object Oriented Design
High Performance Techniques
Aerothermodynamics
Physical Modeling
Numerical Methods
Numerical Results
Conclusions
COOLFluiD Gallery
Conclusion and Future Work
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
3. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
Outline
1 COOLFluiD Framework
Introduction
Object Oriented Design
High Performance Techniques
Validation of the COOLFluiD Framework
2 Aerothermodynamics
Physical Modeling
Numerical Methods
Numerical Results
3 Conclusions
Conclusions
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
4. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
COOLFluiD Platform
Co-developed together with T. Quintino, T. Wuilbaut and D. Kimpe
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
5. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
Component-based Software Architecture
Plug-in policy for a modular integration of new developments
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
6. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
What is a COOLFluiD Simulation?
From user-defined inputs to engineering solutions
Physics
COOLFluiD
Numerics
Mesh Data
Input Mesh CFD Simulation Flowfield
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
7. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
Object Oriented Design
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
8. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
MeshData: Topological Region Sets (TRS)
The domain is subdivided in topologically different regions
GEOMETRIC
Mesh Data ENTITY BUILDER
SHAPE
CELL
FUNCTION
TRS TR GEOMETRIC
ENTITY
FACE SHAPE
FUNCTION
NODE STATE
Boundary TRSs Inner TRS
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
9. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
MeshData: Geometric Entities (GE)
GE’s are algorithm-dependent agglomerations of degrees of freedom
GEOMETRIC
Mesh Data ENTITY BUILDER
SHAPE
CELL
FUNCTION
TRS TR GEOMETRIC
ENTITY
FACE SHAPE
FUNCTION
NODE STATE
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
10. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
MeshData: Data Storage
Facade managing serial/distributed data creation and access
GEOMETRIC
Mesh Data ENTITY BUILDER
SHAPE
CELL
FUNCTION
TRS TR GEOMETRIC
ENTITY
SHAPE
FACE FUNCTION
DATA STORAGE
NODE STATE
quot;nodesquot; NODE
quot;statesquot; STATE
quot;normalsquot; NORMAL
... ...
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
11. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
Physics: Perspective pattern
Multiple interfaces offering multiple views of the physics
CONVECTIVE DIFFUSIVE REACTIVE
PHYSICAL
VARSET VARSET VARSET
MODEL
Concrete Concrete Concrete
Concrete
Convective Diffusive Reaction
Convective
VarSet VarSet VarSet
Term
CONVECTION Concrete
DIFFUSION Diffusive
Term
REACTION
Concrete
Reaction
Term
Concrete
Physical
Model
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
12. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
Numerics: Method Command Strategy (MCS) pattern
Flexible and uniform way to implement numerical algorithms
BaseMethod
action1()
action2() ConcreteMethodData
getStrategyA()
getStrategyB()
ConcreteMethod getStrategyC()
action1() ...
Action1−>execute() Concrete
... StrategyA
action2() StrategyA
StrategyB Concrete
Command StrategyB
Action1
execute() execute() Concrete
StrategyC
StrategyC
Command Action2
execute() execute()
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
13. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
Example: MCS combined with Perspective pattern
Flexible and uniform way to implement numerical algorithms
SpaceMethod
Physical Model
applyBC()
computeRHS() FVM_MethodData
getVarSet()
Concrete
getExtrapolator()
Physical Model
FVM_Method
getFluxSplitter()
applyBC() ...
WallBC−>execute() Concrete
... VarSet
computeRHS() VarSet
Extrapol Concrete
Command Extrapol
WallBC
execute() execute() Concrete
FluxSplit
FluxSplit
Command ComputeRHS
execute() execute()
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
14. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
High Performance Techniques
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
15. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
High Performance Computing
The growing complexity of scientific simulations requires parallelization
Remote access from workstations to Modern HPC clusters include 1000’s
multi-processor supercomputers CPUs (SGI ICE Altix in photo)
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
16. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
Parallel Functionalities
Parallel IO: reading and writing
User and developer-friendly layer
(numerics independent!)
Scalability up to 1024 CPUs
Parallel mesh partitioning with ParMetis
Robust algorithm for arbitrarily complex
unstructured (hybrid) meshes
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
17. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
N-layer Overlap Region
N-layer overlap region for tunable inter-process data exchange
Schematic of overlap region Example of 1- and 2-layer overlap
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
18. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
Parallel speedup and efficiency
Cylinder mesh (3,426,300 hexa’s) Cylinder mesh (20,557,753 tetra’s)
FV on SGI Altix ICE and ICE+ RD on SGI Altix ICE
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
19. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
Validation of the COOLFluiD Framework
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
20. Introduction
COOLFluiD Framework
Object Oriented Design
Aerothermodynamics
High Performance Techniques
Conclusions
Validation of the COOLFluiD Framework
Gallery of COOLFluiD results
An overview of some CF applications partially reusing our work
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
21. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
Outline
1 COOLFluiD Framework
Introduction
Object Oriented Design
High Performance Techniques
Validation of the COOLFluiD Framework
2 Aerothermodynamics
Physical Modeling
Numerical Methods
Numerical Results
3 Conclusions
Conclusions
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
22. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
Physical Modeling
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
23. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
What is Aerothermodynamics?
Thermo-chemical regimes (Da = τf /τc )
1 Frozen flows (Da ≈ 0)
2 Equilibrium flows (Da 1)
3 Nonequilibrium flows (Da ≈ 1)
Rotation
Translation
Vibration
Truly multi-physical science Electronic
gasdynamics
Different systems of PDE’s
statistical thermodynamics
Non-reacting Navier-Stokes
chemical kinetics
LTE-FEF or LTE-VEF
quantum mechanics
TCNEQ (multi-temperature)
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
24. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
Chemical Equilibrium and Nonequilibrium
Flow is modeled as a mixture of Ns perfect gases
R
p= ps , ps = ρs T, ρs = ρys
s
Ms
Example of gas mixtures used in this work
Nitrogen-2: N, N2
+ +
Air-11: e − , N, O, N2 , NO, O2 , N + , O + , N2 , NO + , O2
Chemical models
Equilibrium (LTE): ys = ys (p, T , Ye )
LTE-FEF: Ye = const
LTE-VEF: ∂ρYe + · (ρYe u) = − · Je
∂t
∂ρys
Nonequilibrium: ∂t + · (ρs u) = − · (ρs ud ) + ωs
s ˙
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
25. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
LTE vs. NEQ: Temperature field
LTE-FEF (top) vs. LTE-VEF (bottom) CNEQ (top) vs. TCNEQ (bottom)
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
26. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
Thermal Nonequilibrium
Disequilibration of energy amongst different modes
e = et (Tt ) + ee (Te ) + ef atoms
e = et (Tt ) + er (Tr ) + ev (Tv ,m ) + ee (Te ) molecules
e = et (Te ) free electrons
Examples of multi-temperature models
3-T model (ionized mixtures): Tt = Tr = T , Tv ,m = Tv , Te
2-T model (ionized mixtures): Tt = Tr = T , Tv ,m = Te = TV
Multi-T (neutral mixtures): Tt = Tr = T , Tv ,m
Prototype electron-electronic or vibrational energy conservation equation
∂ρe∗
+ · (ρe∗ u) = − · q∗ + Ω ∗
∂t
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
27. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
Governing equations for TCNEQ
Advection-diffusion-reaction PDE’s
∂U ∂P
+ · Fc = · Fd + S
∂P ∂t
Conservative and natural variables for Multi-T model
U = [ρs ρu ρE ρm ev ,m ]T , P = [ρs u T Tv ,m ]T
Fluxes and Source Terms for Multi-T model
ρs u −ρs us ωs
˙
ρuu + pˆ
I ¯
τ 0
Fc = d
, F = (τ · u)T − q
, S=
ρuH ¯ ˜ 0
ρm uev ,m −qv ,m Ωv ,m
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
28. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
Diffusive Fluxes and Source Terms
Viscous stresses Heat fluxes
∂uj ∂ui 2
τij = µ + − · u δij q = −λ T +
˜ ym qv ,m + ρs us hs
∂xi ∂xj 3 m s
Mass production term
qv ,m = −λv ,m Tv ,m − ρm um hv ,m
Nr
ω s = Ms
˙ b f
(αs,r − αs,r )(Rf ,r − Rb,r ) Energy relaxation (Landau-Teller)
r =1
Ns α∗ ∗
ρs s,r
(ev ,s − ev ,s )
R∗,r = k∗,r (T , Tv ,m ) Ωv ,m = ρm ˜ ˙
+ Dm ω m
s=1
Ms τm
The Mutation library (T. Magin, M. Panesi) has been used to compute
all thermodynamics, transport, chemistry, energy relaxation
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
29. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
Numerical Methods
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
30. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
Implicit Time Stepping
˜ ∂U ∂P
R(P) = + RSM (P) = 0
∂P ∂t
Newton method
˜
∂R “ n ”
n+1 n n
“ ” “ ”
˜
R P ˜
= R P + P ∆P = 0
∂P
8 » –
˜ “ ”
∂ R Pk
>
> ∆Pk = −R(Pk )
˜
< ∂P
>
Pk+1 Pk + ∆Pk
>
:
=
n+1 k last +1
P = P ⇒
(Steady case = k = 0)
Implicit time integration schemes
U(P) − U(Pn )
˜
R(P) = Ω + R(P) Backward Euler
∆t
U(P) − U(Pn ) 1 n
˜
R(P) = Ω+ [R(P) + R(P )] Crank-Nicholson
∆t 2
3U(P) − 4U(Pn ) + U(Pn−1 )
˜
R(P) = Ω + R(P) 3-Point Backward
2∆t
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
31. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
Finite Volume Method (FV)
Integral form of the PDE’s
d
U dΩi + Fc · n d∂Ωi = Fd · n d∂Ωi + S dΩi
dt Ωi ∂Ωi ∂Ωi Ωi
Cell-centered discretization
∂U dPi
(Pi ) Ωi + RFV (Pi ) = 0
∂P dt
Nf Nf
RFV (Pi ) = Fc Σ f −
f Fd Σ f − Si Ω i
f
f =1 f =1
Linear Reconstruction + Flux Limiter Φ
˜
P(xq ) = Pi + Φi Pi · (xq − xi )
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
32. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
Finite Volume Method (FV)
Upwind schemes for convective flux
¯
1
2 (Fc + Fc ) − | A | (UR − UL )
R L Roe
Fc =
f F+ + F− = A+ UL + A− UR S-W
m1/2 ΨL/R + p1/2
˙ AUSM
Central discretization for diffusive flux
Fd = Fd (Pf , Pf , nf )
f
Nl
1 1 ¯
Pf = P n dΣv = Pl nl Σv
l
Ωv Σv Ωv s=1
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
33. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
Residual Distribution Method (RD)
Conservation law
∂U ∂P
+ · Fc = · Fd + S
∂P ∂t
Vertex-centered discretization
∂U dPl
(Pl ) Vl + RRD (Pl ) = 0
∂P dt
FE linear interpolation
d
h X
P (x, t) = Pj (t)Nj (x), Nj (xk ) = δjk
j=1 RRD (Pl ) = Φc − Φd − Φs
l l l
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
34. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
Residual Distribution Method (RD)
Convective term discretization Flux contour integral (CRD)
Φc = BΩ (K± ) Φc,Ω ∂Fc ∂U
l l Φc,Ω = i
dΩ = F · next d∂Ω
Ω∈Ξl Ω ∂U ∂xi ∂Ω
Galerkin discretization of diffusive term
1 ˜
Φd = −
l Fd (P, P) · nl dΩ
Ωd Ω
Ω∈Ξl
Petrov-Galerkin discretization of source term
1-point
Φs =
l wlΩ S dΩ =⇒ B Ω Sc Ω
l
Ω∈Ξl Ω Ω∈Ξl
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
35. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
Numerical Results
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
36. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
RTO RTG 43: Double Cone Run 42
Double cone geometry definition
Computational mesh (131,584 nodes)
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
37. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
RTO RTG 43: Double Cone Run 42
v
Nitrogen-2, TCNEQ 2T (T , TN2 ), M∞ = 11.5
Schematics of the flowfield
(from Nompelis’ PhD thesis)
Mach number (CRD-Bx)
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
38. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
RTO RTG 43: Double Cone Run 42
v
Nitrogen-2, TCNEQ 2T (T , TN2 ), M∞ = 11.5
Roto-translational temperature (CRD-Bxc) Vibrational temperature of N2 and
mass fraction of atomic nitrogen
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
39. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
RTO RTG 43: Double Cone Run 42
v
Nitrogen-2, TCNEQ 2T (T , TN2 ), M∞ = 11.5
Surface pressure: COOLFluiD (CRD-Bxc) vs. FV solvers
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
40. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
RTO RTG 43: Double Cone Run 42
v
Nitrogen-2, TCNEQ 2T (T , TN2 ), M∞ = 11.5
Surface heat flux: COOLFluiD (CRD-Bxc) vs. FV solvers
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
41. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
RTO RTG 43: Cylinder Case III
Cylinder mounted in HEG facility (DLR)
Computational mesh (3,426,300 hexa)
Thanks to Janos Molnar
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
42. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
RTO RTG 43: Cylinder Case III
v v
Air-5, TCNEQ 3T (T , TN2 ,TO2 ), M∞ = 8.8
Mach number
Roto-translational temperature
AUSM+, LS 2nd order
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
43. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
RTO RTG 43: Cylinder Case III
v v
Air-5, TCNEQ 3T (T , TN2 ,TO2 ), M∞ = 8.8
Surface pressure: blind comparison vs. experiments
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
44. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
RTO RTG 43: Cylinder Case III
v v
Air-5, TCNEQ 3T (T , TN2 ,TO2 ), M∞ = 8.8
Surface heat flux: blind comparison vs. experiments
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
45. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
Stardust Sample Return Capsule
Stardust capsule after landing
Computational mesh (68300 quads)
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
46. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
Stardust Sample Return Capsule
Air-11, TCNEQ 2T (T , Tve ), M∞ = 42
Mach number Stagnation temperatures profiles
AUSM+, LS 2nd order COOLFluiD vs. NASA DPLR
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
47. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
EXPErimental Re-entry Test-bed (EXPERT) Vehicle
Model of the EXPERT vehicle
mounted in VKI wind tunnel
Computational mesh (2,872,584 hexa)
Thanks to Fabio Pinna
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
48. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
EXPERT Vehicle
v v
Air-5, TCNEQ 3T (T , TN2 ,TO2 ), M∞ = 18.4
Mach number
Liou-Steffen AUSM, LS 2nd order Roto-translational temperature
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
49. COOLFluiD Framework Physical Modeling
Aerothermodynamics Numerical Methods
Conclusions Numerical Results
EXPERT Vehicle
v v
Air-5, TCNEQ 3T (T , TN2 ,TO2 ), M∞ = 18.4
Vibrational temperature of N2 Vibrational temperature of O2
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
50. COOLFluiD Framework
Aerothermodynamics Conclusions
Conclusions
Outline
1 COOLFluiD Framework
Introduction
Object Oriented Design
High Performance Techniques
Validation of the COOLFluiD Framework
2 Aerothermodynamics
Physical Modeling
Numerical Methods
Numerical Results
3 Conclusions
Conclusions
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
51. COOLFluiD Framework
Aerothermodynamics Conclusions
Conclusions
Contributions of this thesis
Co-development of a multi-purpose computational framework
Co-development of OO design techniques for scientific computing
Parallel algorithms for HPC simulation
Integration of multiple systems of PDE’s for Aerothermodynamics:
=⇒ N-S, LTE, TCNEQ, Collisional Radiative
Parallel implicit multi-physics FV solver
Parallel implicit multi-physics RD solver
=⇒ Application of CRD to handle TCNEQ flows
Validation of the solvers on challenging testcases
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami
52. COOLFluiD Framework
Aerothermodynamics Conclusions
Conclusions
Thank you all for the attention!
Any questions? Remarks?
Candidate: Andrea Lani Promoter: Prof. Herman Deconinck An Object Oriented and High Performance Platform for Aerothermodynami