A COMPUTATIONAL FRAMEWORK FOR MULTI ... - Virginia Tech
Transcript of A COMPUTATIONAL FRAMEWORK FOR MULTI ... - Virginia Tech
A COMPUTATIONAL FRAMEWORK FOR
MULTI-MATERIAL FLUID-STRUCTURE
INTERACTION WITH DYNAMIC FRACTURE
Kevin G. Wang 1, Patrick Lea 2, Charbel Farhat 2,3,4
1 Department of Aerospace and Ocean Engineering,
Virginia Tech
2 Department of Aeronautics and Astronautics, 3 Department of Mechanical Engineering,
4 Institute for Computational and Mathematical Engineering,
Stanford University
experiment
FSI simulation
FSI problems with dynamic fracture
shock lithotripsy
pipeline explosion underwater explosion &
implosion
• shock waves at/near material interfaces
• large structural deformation
• fluid-induced fracture and/or fragmentation
• flow seepage through cracks
• multi-phase flow with large density/pressure
jump
• multiple time scales
computational challenges
MOTIVATION
before fracture after fracture
fluid 2
fluid 1
fluid 2
f-s interface
f-f interface
fluid 1
MATHEMATICAL MODELS
- Euler equations
where
- immiscible fluids / free surface
and
(level-set equation for interface tracking)
- equations of state (EOS)
Multi-material compressible fluids in Eulerian formulation
sss pep )1( (stiffened gas for liquids)
eeR
BeR
Ap
RR
0201
)2
1()1(001
left fluid right fluid
nvR
(JWL for explosion products)
MATHEMATICAL MODELS
- equation of motion
Elastic-plastic shells in Lagrangian formulation
- J2 plasticity, piecewise linear hardening
- EPS-based failure criterion before fracture after fracture
fluid 2
fluid 1
fluid 2
f-s interface
f-f interface
fluid 1
- 2nd-order Green’s strain tensor
Fluid-structure interaction
- Impermeable interface fluid-structure transmission conditions
(for inviscid fluid flows)
sss nunv
sss npn
“no interpenetration”
“equilibrium”
The standard finite volume semi-discretization
i Ci
- integrate over a control volume (Ci )
0)(
WF
t
W
0)()(
dnWFdt
W
ineij C
ijh
C
h
iji
- evaluate one numerical flux for each “facet” ( )ijC
j
ijC
ij
j
h
i
hijh EOSnWWRoednWF ),,,()(
ijC
- special treatment is required near material (i.e. fluid-fluid or fluid-structure)
interfaces
FIVER: FINITE VOLUME METHOD WITH MULTI-
MATERIAL RIEMANN SOLVERS
- Euler equations
1D fluid-structure Riemann problem
i j
fluid 1fluid 2
SF/S
* Wang, Rallu, Gerbeau, Farhat 2011
FIVER AT FLUID-STRUCTURE INTERFACE
0)(
WFW
ιjWW )0,(
ss nuV ),0(
*
ijW
0)(
WFW
jiWW )0,(
ss nuV ),0(
*
jiW
f-s transmission
condition
(embedded surface)
Fluid-fluid Riemann problem
i j
fluid 1fluid 2
SF/F
FIVER AT FLUID-STRUCTURE INTERFACE
(0 level-set)
0)(
WFW
)0,(W
**, jiij WWijW if
jiW
0
0if
- 2nd order accuracy at interfaces can be obtained by linear extrapolation
(Zeng, Main, Farhat, in press)
VALIDATION EXAMPLES
Underwater implosion
Flapping wings
experiment
simulation
experiment
simulation
-100-50050
100150200250300
-1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
Sensor A
Time (ms)
pre
ss
ure
(p
si)
(Farhat, Wang, et. al. IJSS, 2013)
(Farhat, Lakshminarayan, JCP, 2014)
FSI WITH DYNAMIC FRACTURE
FEM-based computational methods for fracture
element deletion
XFEM
- element deletion
- extended finite element method (XFEM)
- cohesive elements, remeshing, etc.
• delete element if failure criterion (e.g. EPS)
is met
robust and easy to implement
no defined crack path
loss of mass, momentum, energy
explicitly tracks crack propagation
difficult to track multiple interacting cracks
- “generic” FSI methods are desirable
• enrich the FE basis locally with step functions
I
III
h qXfHtuXNtXu }))(()(){(),(
• enrichment (qI) is injected when a criterion for
crack nucleation or growth is met
INTERFACE REPRESENTATION
A generic representation of fractured interfaces
- phantom elements
• for element deletion: deleted elements
• for XFEM: duplicate elements near crack*
* originally proposed by Song and Belytschko (2006) to simplify theimplementation of XFEM in legacy FE codes
• can be easily extended to many other
FE-based fracture methods
- augmented fluid-structure interface (S*F/S)
• real elements + phantom elements
XFEM with phantom nodes/elements- level-set function (f) defined in S*F/S
• gives signed distance to the closest crack
implicit representation of crack: 0)(|*
/ S XX SF f
implicit representation of fractured interface: 0)(|*
/ S XX SF f
• in practice, the precise value of f is only needed near cracks
INTERFACE TRACKING
Tracking a fractured fluid-structure interface
S*F/S
D h
real
phantom
- an intersection based algorithm* is used to track the embedded
fluid-structure interface with respect to the non-body-fitted CFD grid
* Wang, Gretarsson, Main, Farhat, IJNMF, 2012
- an intersection point x* in S*F/S is accepted if and only if
0)( * xf
pre-flawed aluminum
pipe (specimen)
clamp (torque applied)
notch
457 mm
916 mm
detonation wave
Mylar diaphragm0.89 mm
41.28 mm
detonation tube
I beam
aluminum pipe
(specimen)
detonation tube
* Performed by J. Shepherd et al. at California Institute of Technology.
pressure gauge
PIPE FRACTURE DUE TO DETONATION
Fracture of aluminum pipe due to internal detonation
EXPERIMENTAL RESULT
Crack propagation patterns
0
20
40
60
80
100
120
140
160
180
0 20 40 60 80 100
0
20
40
60
80
100
120
140
160
180
0 20 40 60 80 100
initial notch length (mm)
pe
ak b
last p
ressu
re (
kP
a)
EXPERIMENTAL RESULT
Blast pressure: the notch length series
- detonation modeled using the Chapman-Jouguet theory
- three fluid materials:
explosive gas (C2H2 + O2), detonation product, and water
- CFD grid: 1.4M nodes, 8.5M elements(on 100~200 proc. cores)
- initial notch: 25.4mm / 38.1mm / 50.8mm / 63.5mm / 76.2mm
HIGH-FIDELITY SIMULATION MODEL
Simulation setup
XFEM element deletion
notch is modeled as
an initial crack
notch is modeled by shells
with reduced thickness
HIGH-FIDELITY SIMULATION MODEL
Modeling the initial notch
SIMULATION RESULT
3D Visualization of simulation result
simulation with XFEM
simulation with XFEM
SIMULATION RESULT
Structural deformation, stress, and strain
simulation with XFEM
SIMULATION RESULT
2D cut-view of fluid pressure field
- XFEM: curving in the same direction or opposite directions
- element deletion: branching
- all these propagation patterns are observed in experiment
XFEM, L= 25.4 mm
XFEM, L= 38.1 mm
ED, L= 38.1 mm
VALIDATION
Fracture response
Blast pressure
VALIDATION
- summary
• simulations captured the crack
propagation modes
• linear trend in peak pressure vs. notch
length is captured
• discrepancy in peak blast pressure: 5~20%