Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Optimal-Transportation MeshfreeApproximation Schemes
M OrtizM Ortiz
Approximation Schemes for Fluid and Plastic Flows
M. OrtizM. OrtizCalifornia Institute of Technology
In collaboration with: In collaboration with: Bo Li, Feras Habbal (Caltech), Bo Li, Feras Habbal (Caltech),
B. Schmidt (TUM), A. Pandolfi (Milano), F. Fraternali (Salerno)
UCLA, February 23, 2010
Michael OrtizUCLA 02/22/10
UCLA, February 23, 2010
ASC/PSAAP Centers
Michael Ortiz UCLA 02/22/10
Caltech PSAAP CenterObjective: Predict hypervelocity impact phenomena with quantified margins and uncertainties
NASA Ames Research Center E fl h f h l it Energy flash from hypervelocity
test at 7.9 Km/s
Michael Ortiz UCLA 02/22/10
Hypervelocity impact test bumper shield(Ernst-Mach Institut, Freiburg Germany)
QMU – Center’s assets Experimental Science Simulation codes
SPHIRPhysics models UQ tools
VTF OTMHSRTPhysics models UQ tools
Michael Ortiz UCLA 02/22/10
Plasma/EoS Probability/CoM UQ pipelineStrength/Fracture
Simulation requirements• Hypervelocity impact: Grand challenge in
scientific computingp g• Main simulation requirements:
– Hypersonic dynamics, high-energy density (HED)– Multiphase flows (solid, fluid, gas, plasma)– Free boundaries + contact– Fracture fragmentation perforationFracture, fragmentation, perforation– Complex material phenomena:
• HED/extreme conditions • Ionization, excited states, plasma• Multiphase equation of state, transport• Viscoplasticity, thermomechanical coupling
Michael Ortiz UCLA 02/22/10
p y, p g• Brittle/ductile fracture, fragmentation...
Optimal-Transportation Meshfree (OTM)
• Time integration (OT):– Optimal transportation methods:Optimal transportation methods:
• Geometrically exact, discrete Lagrangians
– Discrete mechanics, variational time integrators:• Symplecticity, exact conservation properties
– Variational material updates, inelasticity:• Incremental variational structurea a a o a u u
• Spatial discretization (M):– Max-ent meshfree nodal interpolation:
• Kronecker-delta property at boundary
– Material-point sampling:• Numerical quadrature material history
Michael Ortiz UCLA 02/22/10
• Numerical quadrature, material history– Dynamic reconnection, ‘on-the-fly’ adaptivity
Optimal transportation theory
Gaspard MongeBeaune (1746), Paris (1818)
"Sur la théorie des déblais et des
Leonid V. KantorovichSaint Petersbourg (1912)
Moscow (1986)
Michael Ortiz UCLA 02/22/10
Sur la théorie des déblais et des remblais" (Mém. de l’acad.
de Paris, 1781)
Moscow (1986)Nobel Prize in
Economics (1975)
Mass flows ─ Optimal transportation• Flow of non-interacting particles in
• Initial and final conditions:
Michael Ortiz UCLA 02/22/10
Mass flows ─ Optimal transportation• Benamou & Brenier minimum principle:
• Reformulation as optimal transportation problem:
• McCann’s interpolation:
Michael Ortiz UCLA 02/22/10
Euler flows ─ Optimal transportation• Semidiscrete action:
inertia internal energy
Discrete Euler Lagrange equations:• Discrete Euler-Lagrange equations:
Michael Ortiz UCLA 02/22/10
geometrically exact mass conservation!
Optimal-Transportation Meshfree (OTM)
• Optimal transportation theory is a useful tool for generating geometrically-exact tool for generating geometrically exact discrete Lagrangians for flow problems
• Inertial part of discrete Lagrangian measures distance between consecutive mass densities (in sense of Wasserstein)Di t H ilt i i l f t ti • Discrete Hamilton principle of stationary action: Variational time integration scheme:– Symplectic time reversibleSymplectic, time reversible– Exact conservation properties (linear and angular
momenta, energy)S i i l i h f Γ
Michael Ortiz UCLA 02/22/10
– Strong variational convergence in the sense of Γ-convergence (B. Schmidt)
OTM ─ Spatial discretization nodal points:
materialipoints
Michael Ortiz UCLA 02/22/10
OTM ─ Spatial discretization
Michael Ortiz UCLA 02/22/10
Steel projectile/aluminum plate: Nodal set
OTM ─ Spatial discretization
Michael Ortiz UCLA 02/22/10
Steel projectile/aluminum plate: Material point set
OTM ─ Spatial discretization nodal points:
materialipoints
Question: How can we
Michael Ortiz UCLA 02/22/10
Qreconstructfrom nodal coordinates?
OTM ─ Max-ent interpolation• Problem: Reconstruct function from nodal
sample so that:– Reconstruction is least biased– Reconstruction is most local
• Optimal shape functions (Arroyo & MO, IJNME, 2006):
shape function width information entropy
Michael Ortiz UCLA 02/22/10
OTM ─ Max-ent interpolation
Michael Ortiz UCLA 02/22/10
OTM ─ Max-ent interpolation
• Max-ent interpolation is smooth, meshfreemeshfree
• Finite-element interpolation is recovered in the limit of β→∞β
• Rapid decay, short range• Monotonicity, maximum principle• Good mass lumping properties• Kronecker-delta property at the
boundary:boundary:– Displacement boundary conditions – Compatibility with finite elements
Michael Ortiz UCLA 02/22/10
p y
OTM ─ Spatial discretization nodal points:
materialipoints
Michael Ortiz UCLA 02/22/10
OTM ─ Spatial discretization nodal points:
materialipoints
Michael Ortiz UCLA 02/22/10
OTM ─ Spatial discretization
Michael Ortiz UCLA 02/22/10
Np = local neighborhoodof material point p
OTM ─ Spatial discretization nodal points:
materiali
• Max-ent interpolation at material point p determined points material point p determined by nodes in its local environment NpL l i t • Local environments determined ‘on-the-fly’ by range searches
• Local environments evolve continuously during flow (dynamic reconnection)( y )
• Dynamic reconnection requires no remapping of history variables!
Michael Ortiz UCLA 02/22/10
history variables!
OTM ─ Flow chart(i) Explicit nodal coordinate update:
(ii) Material point update:
position:
deformation:
ol mevolume:
density:
(iii) Constitutive update at material points
(iv) Reconnect nodal and material points (range
Michael Ortiz UCLA 02/22/10
( ) p ( gsearches), recompute max-ext shape functions
OTM ─ Tensile stability(m
/s)
(m/s
)
velo
city
(
velo
city
(
tip
tip
time (s) time (s)
Finite elements OTM
Michael Ortiz UCLA 02/22/10
OTM is free from tensile instabilities!
OTM ─ Riemann problem ) o
rm
(Kg/m
3)
1 e
rror
no
den
sity
den
sity
Lconvergencerate ~ 1
position (m) mesh size (h)
d rate ~ 1
Michael Ortiz UCLA 02/22/10
computed vs. exactwave structure
density convergence (L1 norm)
OTM ─ Shock tube problem
Michael Ortiz UCLA 02/22/10
Shock tube problem – velocity snapshots
OTM ─ Shock tube problemm m
erro
r nor
erro
r norm
oci
ty L
2 e
nsi
ty L
1 e
convergenceconvergence
mesh size (h) mesh size (h)
vel
den
grate ~ 1
grate ~ 1
mesh size (h) mesh size (h)
density convergence (L1 norm)
velocity convergence (L2 norm)
Michael Ortiz UCLA 02/22/10
Shock tube problem – convergence plots
(L norm)(L norm)
OTM ─ Taylor anvil test
t=0 t 7 5 t 0 t=7.5 µscopper rod@ 750 m/s
Michael Ortiz UCLA 02/22/10t=15 µs t=28 µs
OTM ─ Taylor anvil test
t=0 t 7 5 t 0 t=7.5 µscopper rod@ 750 m/s
Michael Ortiz UCLA 02/22/10t=15 µs t=28 µs
OTM ─ Bouncing balloons
OTM fluid(water air)
FE membrane(rubber Kapton) (water, air)(rubber, Kapton)
Michael Ortiz UCLA 02/22/10
OTM ─ Bouncing balloons
OTM fluid(water air)
FE membrane(rubber Kapton) (water, air)(rubber, Kapton)
Michael Ortiz UCLA 02/22/10
OTM ─ Bouncing balloons
OTM fluid(water air)
FE membrane(rubber Kapton) (water, air)(rubber, Kapton)
Michael Ortiz UCLA 02/22/10
OTM ─ Bouncing balloons
OTM fluid(water air)
FE membrane(rubber Kapton) (water, air)(rubber, Kapton)
Michael Ortiz UCLA 02/22/10
OTM ─ Terminal ballistics
1500 m/s
steel projectile
aluminum plate
Michael Ortiz UCLA 02/22/10
aluminum plate
OTM ─ Seizing contact
body 1 body 2linearmomentum
ll ti !cancellation!
nodesnodesmaterial points
Seizing contact (infinite friction)
Michael Ortiz UCLA 02/22/10
Seizing contact (infinite friction) is obtained for free in OTM!
(as in other material point methods)
OTM ─ Seizing contact
body 1 body 2linearmomentum
ll ti !cancellation!
nodesnodesmaterial points
Seizing contact (infinite friction)
Michael Ortiz UCLA 02/22/10
Seizing contact (infinite friction) is obtained for free in OTM!
(as in other material point methods)
Variational Fracture & fragmentation
Michael Ortiz UCLA 02/22/10
M. Ortiz and A.E. Giannakopoulos, Int. J. Fracture, 44 (1990) 233-258.
Variational Fracture & fragmentation
rate
(G
)-r
elea
se r
Ener
gy-
crack extension (Δa)
Michael Ortiz UCLA 02/22/10
M. Ortiz and A.E. Giannakopoulos, Int. J. Fracture, 44 (1990) 233-258.
OTM – Fracture & fragmentation
M. Ortiz and A.E. Giannakopoulos, Int J Fracture 44 (1990) 233 258
Crack growth in mixed mode
Int. J. Fracture, 44 (1990) 233-258.
• Fracture energy over-estimated as h → 0!Non con e gence fo gene al paths meshes!
Michael Ortiz UCLA 02/22/10
• Non-convergence for general paths, meshes!
OTM – Fracture & fragmentation
ε-neighborhood gconstruction
Michael Ortiz UCLA 02/22/10
Variational Fracture & fragmentation
• Proof of convergence of i ti l l t i t variational element erosion to
Griffith fracture:– Schmidt, B., Fraternali, F. and , , ,
Ortiz, M. “Eigenfracture: An eigendeformation approach to variational fracture,” SIAM J. Multiscale Model. Simul., 7(3) (2009) 1237-1366.
• OTM implentation: Variational
crack
perosion of material points (by ε-neighborhood construction)
Michael Ortiz UCLA 02/22/10
crack
OTM ─ Back to terminal ballistics
1500 m/s
steel projectile
aluminum plate
Michael Ortiz UCLA 02/22/10
aluminum plate
QMU – Simulation codes – OTM
Michael Ortiz UCLA 02/22/10
QMU – Simulation codes – OTM
Michael Ortiz UCLA 02/22/10
QMU – Simulation codes – OTM
Michael Ortiz UCLA 02/22/10
QMU – Simulation codes – OTM
erim
ent
expe
nm
ula
tion
Michael Ortiz UCLA 02/22/10
si
OTM ─ Summary and outlook• Optimum-Transportation-Meshfree method:
– OT is a useful tool for generating geometrically-OT is a useful tool for generating geometricallyexact discrete Lagrangians for flow problems
– Max-ent approach supplies an efficient meshfree, continuously adaptive remapping free FEcontinuously adaptive, remapping-free, FE-compatible, interpolation scheme
– Material-point sampling effectively addresses the i f i l d hi i blissues of numerical quadrature, history variables
• Extensions include:Contact (seizing contact for free!)– Contact (seizing contact for free!)
– Fracture and fragmentation (provably convergent)
• Outlook: Parallel implementation, UQ…
Michael Ortiz UCLA 02/22/10
p , Q
OTM ─ Summary and outlook
Thank you!y
Michael Ortiz UCLA 02/22/10