international conference on fracture 11, torino, 2005
TRANSCRIPT
Simulation of Impact andFragmentation with theMaterial Point Method
Biswajit BanerjeeJ. Guilkey, T. Harman, J. Schmidt, P. McMurtry
Center for the Simulation of Accidental Fires and ExplosionsUniversity of Utah
March 21, 2005
C−SAFEE
UNIVRSIT OF UTAH
Y
Outline
�The Problem.
�The Tool: Material Point Method.
�Failure Simulation: The Approach.
�Simulations and Results.
�Conclusions and Future Work.
C−SAFEE
UNIVRSIT OF UTAH
Y
2
Problem: The Goal
Predict fragment velocities and
fragment size distributions
in various fire scenarios
C−SAFEE
UNIVRSIT OF UTAH
Y
5
Tools: Material Point Method
Sulsky et al.,1995, Computer Physics Communications, 87, 236-252
C−SAFEE
UNIVRSIT OF UTAH
Y
6
Tools: Material Models - I� Additive decomposition of the rate of deformation.
d = de + dth + dp (1)
� Mie-Gruneisen equation of state.
p =ρ0C
20ζ[1 +
(1− Γ0
2
)ζ]
[1− (Sα − 1)ζ ]2 + Γ0E(2)
where p = pressure, C0= bulk speed of sound, ζ = (ρ/ρ0 − 1), E = internal energy,
Γ0 = Gruneisen’s gamma at reference state, Sα = linear Hugoniot slope coefficient.
(Zocher et al., 2000, ECOMAS, Barcelona.)
� Hypoelastic law for deviatoric stress.
dev(σ) = 2 µ(p, T ) dev(εe) (3)
7
Tools: Material Models - II� Huber-von Mises yield condition.
f (σ, εp, εp, T ) = ‖dev(σ)‖ −√
2
3σy(ε
p, εp, T ) ≤ 0
(4)
� Associative rate-independent plasticity.
εp = λ∂f (σ, εp, εp, T )
∂σ(5)
� Johnson-Cook plasticity model.
σy(εp, εp, T ) = [A + Bεn
p ][1 + C ln(ε∗p)][1− TmH ] (6)
where εp =√
23‖d
p‖, ε∗p = εp/εp0, εp =∫ t
0 εp(τ ) dτ , TH = T−TrTm−Tr
, and A, B, C, n,m
are material constants. (Johnson and Cook, 1983, Proc. 7th Intl. Symp. Ballistics, The Hague.)
8
Tools: Material Models - III� Temperature- and pressure-dependent shear modulus.
µ(p, T ) =1
J
[(1− T )[µ0 +
∂µ
∂p(p
η)] +
ρkT
Cm
](7)
where T = TTm
, η = ( ρρ0
)1/3, J = 1+exp[ T−1
ζ(1− T1+ζ )
], and ζ, C,m are material constants.
(Nadal and Le Poac, 2003, J. Appl. Phys., 93(5), 2472-2480).
� Pressure-dependent melt temperature.
Tm(p) = Tm0 exp
[2a
(1− 1
η
)]η2(Γ0−a−1/3) (8)
where Tm0 is the melt temperature at ρ = ρ0 and a is a correction to Gruneisen’s
gamma Γ0. (Steinberg et al., 1980, J. Appl. Phys., 51(3), 1498-1504).
9
Tools: Heating� Isotropic thermal expansion rate.
dth = αT1 (9)
where α is the thermal expansion coefficient.
� Plastic work converted into a plastic heating rate.
T p =χ
ρCpσ : dp (10)
where χ is the Taylor-Quinney coefficient, and Cp(T ) is the specific heat at constant
pressure.
� Heat conduction (summed over grid points).
Tg = T sg + T p
g −κ
ρCv∇g ·∇pT (11)
κ = 0 for adiabatic heating.
10
Failure Simulation: Approach�Determine failed material point.
•Evolve porosity and a scalar damage variable andcheck TEPLA-F criterion.
•Check loss of hyperbolicity of the incrementalgoverning equations.
•Check for melting.
� If failed.• Incrementally lower the material point stress tozero.
•Assign a separate velocity field to failed materialpoints.
•Allow “failed” and unfailed material points to in-teract via contact.
11
Failure Simulation: Evolution Rules� Porosity evolution.
f = fnucl + fgrow (12)
where fgrow = (1 − f )Tr(dp), fnucl =fn
(sn
√2π)
exp
[−1
2
(εp − εn)2
s2n
]εp, fn is the
volume fraction of void nucleating particles, εn is the mean of the distribution of
nucleation strains, and sn is the standard deviation of the distribution.
(Chu and Needleman, 1980, ASME J. Engg. Mater. Tech., 102, 249-256).
� Scalar damage evolution (Johnson-Cook model).
D = εp/εfp (13)
where εfp =
[D1 + D2 exp
(D3
3σ∗
)][1 + D4 ln(εp
∗)] [1 + D5TH ], σ∗ =Tr(σ)
σeq, D
is the scalar damage variable, εfp is the fracture strain, and D1, D2, D3, D4, D5 are
constants. (Johnson and Cook, 1985, Int. J. Eng. Fract. Mech., 21, 31-48).
12
Failure Simulation: Failed ?� TEPLA-F failure criterion satisfied.
(f/fc)2 +(εp/ε
fp
)2= 1 (14)
(Johnson and Addessio, 1988, J. Appl. Phys., 64(12), 6699-6712).
� Drucker stability postulate violated.
σ : dp ≤ 0 (15)
(Drucker, 1959, J. Appl. Mech., 26, 101-106.)
� Loss of hyperbolicity of the incremental equations.
det(A) ≤ 0 (16)
where A = n · M · n + n · σ · (n1), M is the incremental tangent modulus tensor,
and n is the normal to the localization band.
(Perzyna, 1998, Localization and Fracture Phenomena in Inelastic Solids, 99-241).
13
Simulations: Impact
Chhabildas et al, 1998, Int. J. Impact Engrg., 23, 101-112.
C−SAFEE
UNIVRSIT OF UTAH
Y
15
Simulations: Impact L1
Z Z
Aluminum SphereVelocity = 1480 m/s
9.52
90
VISAR Reading ofAxial Velocity
31.8
Aluminum Plate
S5 S6
S1 S2 S3
4819
Axial Strain Gages
78 Aluminum Plate
Hollow Aluminum Cylinder
13.614
28.6
2
X
S4
Y
(All dimensions are in mm. Not to scale)
(0.12,2.5)
Aluminum Sphere
C−SAFEE
UNIVRSIT OF UTAH
Y
16
Results: Impact L1
0 20 40 60 80 1000
500
1000
1500
Time (µ sec)
Ene
rgy
(J)
Kinetic EnergyStrain EnergyTotal Energy
Energy.
0 20 40 60 80 1000
0.5
1
1.5
2
Time (µ sec)
Mom
entu
m (k
g m
/s)
Momentum (mag)X−MomentumY−MomentumZ−Momentum
Momentum.
0 10 20 30 40 50 60 70 80 90 100
−50
0
50
100
150
200
250
300
Time (µ sec)
Axi
al V
eloc
ity (m
/s)
Expt.Inner Circle2nd Circle3rd Circle4th Circle
Velocity.
0 10 20 30 40 50 60 70 80 90 100−8
−6
−4
−2
0
2
4
6
8x 10−3
Time (µ sec)
Axi
al S
trai
n
Expt.Simulation
Strain at S4.18
Simulations: Impact L3
Z Z
9.52
90
VISAR Reading ofAxial Velocity
31.8
Aluminum Plate
S5 S6
S1 S2 S3
4819
Axial Strain Gages
78 Aluminum Plate
Hollow Aluminum Cylinder
X
S4
Y
(All dimensions are in mm. Not to scale)
Aluminum Sphere
13.919
28.6
Aluminum SphereVelocity = 1470 m/s
(11.4,−3.8)
C−SAFEE
UNIVRSIT OF UTAH
Y
19
Results: Impact L3
0 20 40 60 80 1000
500
1000
1500
Time (µ sec)
Ene
rgy
(J) Kinetic Energy
Strain EnergyTotal Energy
Energy.
0 20 40 60 80 1000
0.5
1
1.5
2
Time (µ sec)
Mom
entu
m (k
g m
/s)
Momentum (mag)X−MomentumY−MomentumZ−Momentum
Momentum.
0 10 20 30 40 50 60 70 80 90 100−10
0
10
20
30
40
50
Time (µ sec)
Axi
al V
eloc
ity (m
/s)
Expt.Inner Circle2nd Circle3rd Circle4th Circle
Velocity.
0 10 20 30 40 50 60 70 80 90 100−0.012
−0.01
−0.008
−0.006
−0.004
−0.002
0
Time (µ sec)
Axi
al S
trai
n
Expt.Simulation
Strain at S4.
20
Results: Fragmentation: 2D
Coarse Grid - no tensile stresses in failed particles.
Fine Grid - no tensile stresses in failed particles.
C−SAFEE
UNIVRSIT OF UTAH
Y
24
Results: Fragmentation: 2D
Coarse Grid - mass of failed particles removed.
Fine Grid - mass of failed particles removed.
C−SAFEE
UNIVRSIT OF UTAH
Y
25
Results: Fragmentation: 2D
Coarse Grid - zero deviatoric stress in failed particles.
Fine Grid - zero deviatoric stress in failed particles.
C−SAFEE
UNIVRSIT OF UTAH
Y
26
Results: Fragmentation: 2D
Coarse Grid - Equation of State for PBX9501.
Fine Grid - Equation of State for PBX9501.
C−SAFEE
UNIVRSIT OF UTAH
Y
27
Conclusions/Future Work
� Wave arrival times and velocity peaks captured.
� Reasonable fragment distribution in 2D.
� Mesh dependence of results.
� Viscoplastic regularization.
� Nonlocal plasticity/continuum damage.
� Strong discontinuity approach to fracture.
C−SAFEE
UNIVRSIT OF UTAH
Y
28