AdvFem Meshfree 2016class

LS-DYNA Advanced FEM and Meshfree Methods
for Solid and Structural Analyses

Manufacturing Applications
LS-DYNA Training Class
Jun. 15th- 16th, 2016
Detroit
Wei Hu*
C. T. Wu
Yong Guo
Bo Ren
Youcai Wu
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
Download links of class material:

http://ftp.lstc.com/anonymous/outgoing/whu/Class/AdvFem_Meshfree_2016Class.pdf
http://ftp.lstc.com/anonymous/outgoing/whu/Class/AdvFem_Meshfree_2016Workshop.pdf
http://ftp.lstc.com/anonymous/outgoing/whu/Class/AdvFem_Meshfree_2016Workshop_Input.zip
Methods for Solid and Structural Analyses in LS-DYNA

Rubber Materials: FEM, EFG, MEFEM, SPG
Foam materials: FEM, SPH, EFG, SPG
Metal materials: FEM, SPH, EFG, MEFEM, Adaptive FEM/EFG, SPG
Ductile material failure: FEM, SPG
Brittle and quasi-brittle material fracture: FEM, SPH, EFG, SPG, Peridynamics
E.O.S. materials and high speed applications: ALE, SPH, SPG
Shells and fracture: FEM, EFG, XFEM
Soil: ALE, SPH, SPG
Discrete materials: Discrete element method (DEM)
Composites and Unit cell analysis: FEM, EFG, SPG
Advanced FEM/Meshfree methods in LS-DYNA for

Solids and Structures Analysis
Macromechanics
EFG for shells
(metal, fabric,.)
Adaptive EFG for shells
(metal)
XFEM for shells
(metal, brittle)
EFG for solids (foam)

Adaptive FEM/EFG for solids
(metal)
Cohesive EFG for solids
(brittle &semi-brittle)
Multi-scale (bridging)
ME-FEM for solids (rubber)

SAMG Multi-grid solver
SPG for solids
(metal, form, soil, composite ..)
H-XFEM for shells
(metal, composite)
SPG for shell/solid shell
(metal, brittle, composite)
Peridynamics for shells
(metal, brittle, composite )
Fusion Particle method
Bond-based Peridynamics
(glass, composite)
Micromechanics
Nanomechanics
Periodic Unit Cell method Poly-crystal DDD (single crystal)

Discrete Dislocation Dynamics
LS-DYNA Current Practice for Manufacturing Applications

Metal Forming: EFG & Adaptivity
o Sheet forming
EFG shell ELFORM=41,42 (*SECTION_SHELL_EFG)
ADPOPT=1,2,4 (*CONTROL_ADAPTIVITY)
o Bulk forming
EFG solid ELFORM=41,42 (*SECTION_SOLID_EFG)

ME-FEM solid ELFORM=43 (*SECTION_SOLID)
ADPOPT=7 (*CONTROL_ADAPTIVITY)
o Hot forming: Thermal-mechanical coupling
o Spring back analysis

o Explicit/Implicit analysis
Machining and Jointing: Adaptivity & SPG

o Cutting / Riveting / Drilling
Adaptivity: No material failure; Non-physical based Material separation

SPG solid ELFORM=47 (*SECTION_SOLID_SPG):
Brittle & Ductile material failure; Physical based material separation
o Welding (FSW)
Adaptivity: No material fusion

SPG: Fusion particle method (residual-based homogenization with numerical stabilization,
under development)
Course Outline
1. Introduction and Overview
2. Workshop I
3. EFG, SPG and MEFEM in Solids/Structures and Their Keywords
4. Workshop II
5. Standard 3D Adaptive FEM/EFG and Its Keywords
6. Workshop III
7. Advanced 3D Adaptive EFG and Its Keyword
8. Workshop IV
9. EFG, XFEM, SPG and Peri-Dynamics Failure Analyses and Keywords
10. Conclusions, Q&A
1. Introduction and Overview
What is the Meshfree/Meshless/Particle Method ?

No mesh is needed for the construction of shape functions;
shape functions are constructed from sets of particles
Meshfree shape function
Meshfree/Particle Application Range

Meshfree Solution looking for problems
F
E.B.C.
Material
Strength
Elastic Fluid
Solid
Equation of State
Fluid
Gas
Metal Forming
Incompressible fluid
SPG Extrusion Particle FEM
Conservative Element
MEFEMForging
Foam packing
Airbag
Peridynamics Adaptive Eulerian FEM
Compressible fluid Particle Airbag
EFG
Crashworthiness
ALE
Fracture
Sloshing
Discrete Element
Hydroplaning
Bird strike
Explosion
Penetration
Momentum
SPG
Splashing
SPH
Classification of Particle Methods
Continuum
Implicit Meshfree Galerkin

RKPM, EFG, Particle Finite Element (PFEM),
Explicit
Hydrocode
Particle
Method
Meshfree Collocation
SPH, Finite point,
Meshfree Galerkin
RKPM, EFG, PFEM, SPG,
Peridynamics,
Molecular Dynamics
Discrete
Lattice Boltzmann, DEM,
History and Research Trend

Meshfree Method
Meshfree Collocation Method
Smooth Particle Hydrodynamics (SPH) [Monaghan 1977]

Finite Point Method [Onate et al.1996]
Peridynamics [Silling 1998]
Meshfree Galerkin Method

Element Free Galerkin (EFG) [Belytschko et al. 1994]
Reproducing Kernel Particle Method (RKPM) [Liu et al. 1995]
Partition of Unity Method [Babuska and Melenk 1995]
HP-Clouds [Duarte and Oden 1996]
Free-Mesh Method [Yagawa et al. 1996]
Natural Element Method [Sukumar et al.1998]
Meshless Local Petrov-Galerkin Meshfree Method(MLPG) [Atluri et al.1998]
Local Boundary Integral Equation (LBIE) [Atluri et al. 1998]
Finite Sphere Method [Bathe 1998], Particle Finite Element Method [Idelsohn et al. 2004]
Smoothed Particle Galerkin (SPG) method [Wu et al. 2013]
Meshfree least square method,

(FEM, Control Volume, BEM ) + Meshfree Method
Coupled FEM/Meshfree Method [Wu et al. 2001]
Extended FEM Method [Belytschko et al. 1999]
Meshfree-enriched FEM (MEFEM) [Wu et al. 2011]
10
Computational Challenges
Advantages of Using Meshfree Method
Large material distortion, e.g., crashworthiness, hyper-velocity impact
Interfaces, moving boundaries, free surface, e.g., FSI
Adaptive procedure, e.g., forging and extrusion
Multiple-scale phenomenon, e.g., shear band
Moving discontinuities, e.g., material failure, crack propagation
Immersed Structures, e.g., reinforced composites
Disadvantages of Using Meshfree Method

High CPU and memory in implicit/explicit analysis (EFG, MEFEM, SPG)
Complicated in parallel (EFG)
Tensile instability and low-energy mode (SPH)
Difficult essential boundary condition treatment (SPH)
Does not pass Patch Test (most mesh-free methods);
11
Tensile Instability in SPH

FEM
SPH
12
Low Energy Mode in SPH
EFG
SPH
13
Computational Trade-off in Mesfree/Particle Methods
High CPU and Memory
Better computer capacity
Tensile instability
Accurate integration rule
Low-energy mode
Adapivity
Boundary effect
Implicit
Dispersed waves
Parallelization
Feasibility
Large deformation,
Material separation,
Trade-off
Accuracy, Convergence
Stability
14
Overview on Smooth Particle Hydrodynamics ( SPH )

Basic SPH Equation of Motion
Weak Form
Strong Form
d
v
=
dt
x
dv
1
=
dt
x
dE
v
=
dt
x
v(x) = Tu = wa (x s)u(s)ds
Kernel approximation
mj
d i
= i
vi v j Wij ,
dt
j j
i = m jWij
j
i j
dvi
= m j ( 2 + 2 )Wij ,
i
j
dt
j
mj
dvi
=
( i
j )Wij ,
dt
j i j
dEi i
= 2
dt
i
m (v v )W
j
ij ,
in LS-DYNA
15
SPH Applications
Spinning Test
Liquid Shifting Test
16
Overview on Element Free Galerkin Method (EFG)

NP
u (x) = wa[n] (x xI ) u( xI )xI

h
I =1
Moving Least-Squares approximation

or Reproducing Kernel approximation
6444444474444444
8
-1
wa[n] (x x I ) = H T[n](0)M [n] (x)H [n] (x - xI ) wa (x x I )
14444
4244444
3 14243
nth order completeness
weighting function
[n]
aI
w (x J ) IJ
AT MA1d&& + AT KA1d = AT R
[n]
a
( x xI ) = 1, x
More neighboring nodes

Accurate and convergent
Stable and no low-energy mode
More integration points
Adaptivity to handle severe deformation
CPU and memory demanding
17
Overview on Smoothed Particle Galerkin Method (SPG)

u ( X ) % (Y ; X ) u ( X ) d + % (Y ; X ) u ( X ) (Y - X ) d
1
(2)
% (Y ; X ) (2) u ( X ) ( 2 ) (Y - X ) d
2!
= u ( X ) % (Y ; X ) d + u ( X ) % (Y ; X )(Y ) d X
% (Y ; X ) d )
(2)
2 1
+ (2)u ( X ) ( ) % (Y ; X )(Y - X ) d
2!
(2)
2 1
= u ( X ) % (Y ; X ) d + (2)u ( X ) ( ) % (Y ; X )(Y - X ) d
2!
( 2)
(2 )
= u ( X ) + u ( X ) ( X )
A MA U&& = A-T f ext f int

-T
More neighboring nodes

Convergent
Stable and no low-energy mode
Particle integration
Can handle severe deformation
CPU demanding; Under intensive
development
18
Meshfree Kernels
EFG
SPH
EFG
SPH
SPG
EFG
SPG
Lagrangian Kernel:
1.Support is defined in the initial configuration
2.Support covers the same set of material points throughout time
3.Neighbor searching only once
Eulerian Kernel:
1.Support is defined in the current configuration
2.Support covers the different material points throughout time
3.Neighbor searching regularly
Updated Lagrangian or Semi-Lagrangian Kernel:
1.Support is defined in the current configuration
2.Support covers about the same number of material points throughout time
3.Neighbor searching regularly
19
SPH Lagrangian Kernel and Eulerian Kernel
Eulerian Kernel
Zero-energy mode
Tensile instability
Lagrangian Kernel
Zero-energy mode
Can not handle severe deformation
Time step drops in several orders
20
SPG Updated Lagrangian Kernel and Eulerian Kernel

Vz
Updated Lagrangian
No zero-energy mode; Capable to handle severe deformation; Time step drops slightly
Eulerian
t=0.2
0.4
0.6
0.8
1.0
No zero-energy mode; Time step does not drop; Tensile instability
21
Spatial Domain Integration

Integration Cell or Background Element
Valid within Lagrangain description
e
SPH and SPG nodal

integration
Usually in solid and structure !

Only needed in initialization phase !
Define the physical domain
and various contact conditions
Provide the mass computation
Impose boundary conditions
Create stress points
non-overlapping
mesh
EFG stress point

integration
22
Comparison of SPH, EFG and SPG

SPH
EFG
SPG
Explicit Lagrangian Collocative

method
Explicit/implicit
Full/SemiLagrangian Galerkin
method
Explicit
Impact/penetration
Manufacturing
Impact/penetration
compressible flow
Crashworthiness
Crashworthiness
2D and 3D
2D, 3D and shell
3D
Full/Semi-Lagrangian
Galerkin method
Efficient
Accurate
Accurate
Handle free surface fluid flow
Adaptive
Handle material failure
Difficult Boundary condition

Tension instability
Low energy mode
Slow
No adaptive (currently)
Under intensive development
23
When Should EFG be Considered in the Analysis ? (1)

Uniform Compression with Rigid Shell Impactor
Rigid shell
Result: EFG is same as FEM
Adding 20% viscous contact damping

(VDC=20.) to reduce oscillation in contact.
24
FEM
EFG
25

Non-uniform Compression with Half Rigid Solid Impactor
Half Rigid solid
FEM
Result: EFG advances further
EFG
26
What is the EFG Limitation in Large Deformation Analysis?

Negative Volume in EFG ?
Beyond Lagrangian Description !
x = ( X , t)
dx = F dX
J = det( F )
Lagrangian
Is continuously differentiable
Is one-to-one
J>0
F is invertible
No material overlapping
No gap
No negative volume
27
EFG Negative Volume due to Lagrangain kernel
Particle
Stress point
J>0
What will be the worst ?

As bad as FEM !
28
Beyond Finite Strain Problems
adaptivity
Global Refinement
FEM/Mesh-free adaptivity
Severe material deformation
SPG (Updated Lagrangian Kernel)
Introduce damage EFG, SPG (Updated

Material separation
Lagrangian Kernel or Eulerian Kernel)

Local Refinement
Strong discontinuity XFEM, SPG (Eulerian Kernel)

29
2. Workshop I
3. Standard EFG, SPG and MEFEM in

Solids/Structures and the Keywords
*SECTION_SOLID_EFG (1)
Card 1
Variable
SECID
ELFORM
Type
ELFORM EQ.41: EFG formulation

EQ.42: for 4-noded EFG or adaptive EFG
Card 2 and Card 3 define only for the EFG option

Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
1.01
1.01
1.01
-1
0.01
Variable
IPS
STIME
IKEN
SF
MID
IBR
DS
ECUT
Type
Default
I
F
I
I
F
F
ForF Fracture
Analysis
& Other
Features
1.e+20
0.0
1.01
0.1
*SECTION_SOLID_EFG
5, 41
1.1, 1.1, 1.1, , , 3, 1,
1
32
Normalized dilation parameters (normalized support size) in X, Y and Z directions
Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
1.01
1.01
1.01
-1
0.01
1.0 DX, DY, DZ 2.0 is recommended

CPU time increases with support size
Some Guidelines for DX, DY and DZ

Regular mesh
Irregular mesh
Foam
1.0~1.2
1.0~1.2
Metal
1.2~1.4
1.0~1.2
Fluid or E.Q.S.
1.4~1.6
1.2~1.4
33
EFG kernel function
Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
1.01
1.01
1.01
-1
0.01
ISPLINE EQ.0: Cubic spline function (default) (with base function 1, x, y, z)

EQ.1: Quadratic spline function
EQ.2: Cubic spline function (circular shape)
EQ.10 Cubic spline function with bilinear basses function (1, x, y, z, xy, yz, xz)
Cubic B-spline
ISPLINE: other values
ISPLINE=2
Only DX is active
*SECTION_SOLID_EFG
5, 41
1.1, , ,
34
Choice for the dilation parameter
Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
1.01
1.01
1.01
-1
0.01
IDILA EQ.0: Maximum distance based on background elements (default)

EQ.1: Maximum distance based on surrounding nodes
I
y
rxI = d
rxI = d / 2
if IDILA=0
if IDILA=1
d
X-support in computation = rxI DX
35
Essential boundary condition treatment
Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
1.01
1.01
1.01
-1
0.01
IEBT EQ. 1: Full transformation

EQ.1: (w/o transformation) (default)
EQ. 2: Mixed transformation
EQ. 3: Coupled FEM/EFG (most efficient)
= Smoothed Finite Element Method (SFEM)
Wu et. al. IJNME (2014); Comp. Mech. (2014)
EQ. 4: Fast transformation

EQ.4: (w/o transformation)
EQ. 5: Fluid particle (trial version)
EQ. 7: Maximum Entropy approximation
Mixed transformation is equivalent to Full transformation with improved efficiency.
36
MLS Approximation in EFG (1)
1.Define a local approximation of a given function u(x) in the neighborhood

of x=s by a set of basis functions
u( x )
uh(x,s)
wa ( x s )
x=s
x
Quadratic B-spline
Ls u( x ) u ( x, s ) = ( s )i bi[ n ] ( x ) H [n] ( s )b [n] ( x )

T
i =1
H [n] ( s ) = [1, s, L, ( s ) n ]
[n]: n-th order basis function
(in multi - dimensional H [n] ( x ) = [1, x, y , r cos , r sin , L])

2
2
n
Ls u( x ) u ( x, s ) = ( x-s )i bi[ n ] ( x ) H [n] ( x-s )b [n] ( x )

h
i =1
37

2. Minimization of weighted L2 norm
J ( u h (x)) = [u h (x, s) - u(s)] 2 wa (x - s)ds
u h ( x, s ) = bi[ n ] ( x )( x s )i
where
i =1
n
J ( x )
h
0 = [n]
= 2 [u (x, s) - u(s)]wa (x s) (x - s) j db [j n ] ( x )ds; i = 1,L n
bi ( x )
j =1
3. The corresponding stationary condition is
[n]
(x - s)[ H
[n] T
(x - s)b [n] (x) - u(s)]wa (x s)ds = 0
b [n] (x) = M [n] ( x ) H [n] (x - s)u(s)wa (x s)ds

-1
where M [n] ( x ) = H [n] (x - s)H [n] T (x - s)w (x s)ds

a
Moment matrix
38

4. The local approximation is
T
-1
) )
) )
u h (x, s) = H [ n ] (x - s)M [n] ( x) H [n] (x - s )u(s )wa (x s )ds
5. The global approximation u h (x) = u(s)wa ds can be obtained by setting
s = x in the local approximation
u ( x) = H
[n] T
( 0) M
[n] -1
) )
) )
( x) H [n] (x - s )u(s )wa (x s )ds u(s)wa[n] (x s)ds
6. Compare to the original approximation u h (x) = u(s)wa (x s)ds
-1
wa[n] (x s) = H [n] ( 0 ) M [n] ( x )H [n] (x - s)wa (x s)ds

1424
3 1424
3 14243
( 1n )
( nn )
Modified weighting function
(n1)
39

1-st order MLS Shape Functions
( x ) = 1, x
I
( x ) = 1, x
I
40
EFG Fast Transformation Method
Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
4, -4
Momentum equation
v& = x + b
Continuity equation
& = x v
r ( xI , t )
v v&d = v : d + v bd + v d
mI v& I = xI ( x s ) sVs
I
& s = s v I xI ( x s )
I
u h ( x ) = I[n] ( x ) J[m] ( x I ) x J I[m] ( x ) x I

I
x
I
x
Particle
Stress point
41
EFG Modified Maximum Entropy Method (1)
Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Local MAXENT (Ortiz and Arroyo, 2006, Wu et al, 2009; 2011))
(MAXENT )
maximize H ( p ) = ( x ) pi xi x + pi log pi
i =1
subject to
i =1
pi 0, i = 1,..., N
N
=1
( xi x ) = 0
i =1
N
p
i =1
for [0,+ ) , H(p) is continuous and strictly convex in solution (wellbehaved mass matrix, monotonicity, variation diminishing )
less dependent
difficult to decide
42
EFG Modified Maximum Entropy Method (2)
N
Define the partition function Z : Z ( x, ) i ( x )e ( x - xi ) / ri

i =1
where i ( x ) is the kernel function at node i

ri is the support size of kernel at node i
Non-negative approximation
Smoothness in irregular nodes
Less dependence
Kronecker-Delta at boundary
The unique solution of MAXENT is proven to be

pi ( x, ) =
i ( x )e f ( x, )
i
Z ( x, )
pi 0, i = 1,..., N
satisfying
=1
i =1
N
( xi x ) = 0
i =1
where f i ( x, ) = [( x xi ) / ri ]
Implicit solve (3~5 iterations)
43
Domain integration method
Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
1.01
1.01
1.01
-1
0.01
IDIM EQ. 1: Local boundary condition method

EQ. 2: Two-points Guass integration (default)
EQ.-1: Stabilized EFG method (apply to 8-noded, 6-noded and combination of them)
EQ.-2: Fractured EFG method (apply to 4-noded & smp only)
IDIM= 1 for compressible materials
IDIM= 2 for compressible and nearly incompressible materials
IDIM=-1 applicable for both compressible/incompressible materials (very fast in 8-noded cell)
l
l
xl
L
L
xL
44
Different EFG Integration Cell (Background Element)
Tetrahedron Element in FEM
1.
4-noded constant stress (#10)
2. 10-noded 5-stress points (#16)

3.
4-noded nodal pressure for bulk forming(#13)
Background Element in EFG
EFG solid
EFG shell
45
Stabilized EFG Method (1)
Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
-1
The Stabilized EFG Method (for 8-noded cells)

Is a one-point integration scheme + gradient type hourglass control.
Assumed strain method for nearly incompressible materials.
Designed especially for foam and rubber materials.
The speed is between FEM reduced integration element (#1) and full
integration element (#2)
A switch to full integration (rubber) or Semi-Lagrangian kernel (foam) is allowed
in large deformation range.
Available in SMP explicit and MPP explicit.
46
Deformation tolerance for the activation of Semi-Lagrangian kernel
Card 2
Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
1.01
1.01
1.01
0.01
TOLDEF
TODELF < 1.0
= 0.0 : Lagrangian kernel

> 0.0 : Semi-Lagrangian kernel
< 0.0 : Eulerian kernel
Time control for the activation of Full integration or Semi-Lagrangian kernel
Card 3
Variable
IGL
STIME
IKEN
Type
Default
1.e+20
47
Stabilized EFG Method (3): Rubber Bushing Analysis
Mooney-Rivlin Rubber
Poissons =0.4995
Stabilized EFG explicit analysis
Switched to full integration at t=100
Completion at t=150
CPU comparison at t=50
Methods
S-FEM(#1)
F-FEM(#2)
EFG
S-EFG
CPU
1.0
4.1
5.4~12.9
2.6
48
Deformation tolerance for the activation of Semi-Lagrangian kernel or Eulerian kernel
Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
1.01
1.01
1.01
-1
0.01
TOLDEF
TODELF < 1.0
= 0.0 : Lagrangian kernel

> 0.0 : Semi-Lagrangian kernel
< 0.0 : Eulerian kernel
The larger number of the TODELF , the earlier activation of Semi-Lagrangian
or Eulerian kenrel and more CPU time is expected.
Semi-Lagrangian kernel is suggested in foam materials.
Eulerian kernel (SMP only) is suggested in fluid and E.O.S. materials.
Mass scaling is NOT supported in Eulerian kernel.
49
Foam Compression Simulation
*SECTION_SOLID_EFG
5, 41
1.1, 1.1, 1.1, , , 4, ,0.01
FEM
EFG
EFG
Semi-Lagrangian Kernel
50
Dummy with Side Impact (1)
Hyper-elastic Jacket
Foam Ribs
Total 84583 nodes
15966 EFG nodes
Copyright by GM
FEM
51
Dummy with Side Impact (2)
FEM/Meshfree
Internal Energy (Ribs)
Internal Energy (Jacket)
52
Crashworthiness: ODB Model (1)

Aluminum Cover
Honeycomb Bumper
Copyright by GM
53
Crashworthiness: ODB Model (2)

Impact Force
FEM
FEM + high
viscosity
FEM/Meshfree(movie)
FEM/Meshfree
54
Brain Injury Simulation
Visco-hyperelastic Material
FEM/Meshfree
DOT/NHTSA
SIMon model
55
Variational Formulation (1)

The strong form of the initial-boundary-value problem is stated as follows:
Given bi ( x ), hi ( x ), g i ( x ), ui0 ( X ),vi0 ( X ),find ui ( X ,t) such that they satisfy

u&&i = ij , j + bi in
with boundary conditions
ij n = hi on xhi
ui = g i on xgi
and initial conditions
ui ( X ,0) = ui0 ( X )
u&i ( X ,0) = vi0 ( X )
56

The (Bubnov-) Galerkin weak form of the problem is stated as follows:
Given bi ( x ), hi ( x ), g i ( x ), ui0 ( X ),vi0 ( X ),

find ui ( X ,t) H g1 ( H g1 = { v : v H 1 ( ), vi = g i on xgi }
such that for all ui ( X ,t) H 01 ( H 01 = { v : v H 1 ( ),vi = 0 on xgi }
the follwoing equation is satisfied
u u&& d + u
i
ij d ui bi d
i,j
hi
x
ui hi d = 0
with
ui ( X ,0) = ui0 ( X )
u&i ( X ,0) = vi0 ( X )
H 1 : ( u ,x )2 d <
x
Sobolev space of degree one

57

Linearized updated Lagrangian equation
u u&& d + u
i
i,j
ij d ui bi d
x
hi
x
ui hi d = 0
u a d + C
i
ij
ijkl
kl d + ui , jTijkl uk ,l d ui bi d
ui 0 ai d
hi
x
ui hi d = u T R
Lumped mass is different from FEM
a lg
where Cijkl = Cijkl

Cijkl
Cijkl
= ij kl + ( il jk + jl ik + ik jl + jk il )
2
Tijkl = ik jl
NP
u (x) = wa[n] (x xI ) u ( xI )xI

I =1
58
Lagrangian and Eulerian Kernel (1)

Material velocity
Lagrangian Kernel :
NP
NP
I =1
I =1
NP
u( X ,t) = wa[n] ( X ; X X I )u I X I = H [ n ] (0 ) M [ n ] ( X ) H [ n ] ( X X I ) wa ( X X I )X I u I I ( X )u I
ui ( X ,t )
t
vi ( X ,t ) =
[X]
I =1
= I X ( X )d&iI ( t )
I
Eulerian Kernel :
NP
NP
u( x ,t) = w ( x ; x x I )u I xI = H
[n]
a
I =1
vi ( x, t ) =
[ n ]T
(0 ) M
[ n ]1
NP
( x ) H ( x x I ) wa ( x x I )xI u I I ( x )u I
[n]
I =1
ui ( x, t )
t
x
I
( x; h(t))
= [
t
I
[X]
ui ( x , t )
t
[ x]
I =1
ui ( x , t )
v j ( x, t )
x j
144
2443
Ix ( x , t )
x
&
iI (t )]
[ x ] iI (t ) + I ( x ) iI (t ) +v j ( x , t )
x j
&&
v = ( 1 A )1 ( b + c ) v& = u
59
Lagrangian and Eulerian Kernel (2)

Spatial derivatives of the Lagrangian kernel
Path-dependent materials: the spatial derivative of material shape functions is needed
I X ( X ) I X ( X ) 1
F ji ( x( X ,t ))
=
xi ( X ,t )
X i
f ( x )d = f ( x( X ,t ) )J ( X ,t )d ,
x
F 1
x
IX ( X )
=
= d I
+ I
X I
X
J = det( F )
60
*SECTION_SHELL_EFG (1)
Card 1
Variable
SECID
ELFORM
SHRF
NIP
PROPT
Type
Default
ELFORM EQ. 41: EFG shell (local projection)

EQ. 42: EFG shell (iso-parametric mapping)
EQ. 43: EFG 2D plane strain
EQ. 44: EFG 2D axisymmetric (y-axis of symmetry)
Card 3 define only for the EFG option

Variable
DX
DY
ISPLINE
IDILA
IEBT
IDIM
Type
Default
1.1
1.1
*SECTION_SHELL_EFG
6, 41
1.1, 1.1, , , 4, 1,
61
DX, DY, ISPLINE same as in *SECTION_SOLID_EFG
IDILA: not available
Essential boundary condition treatment

Variable
DX
DY
ISPLINE
IDILA
IEBT
IDIM
Type
Default
1.01
1.01
-1
IEBT EQ. 1: Full transformation (default)

EQ. 3: Coupled FEM/EFG = Smoothed Finite Element Method (SFEM)
Wu et. al. IJNME (2014); Comp. Mech. (2014)
62
Variable
DX
DY
ISPLINE
IDILA
IEBT
IDIM
Type
Default
1.01
1.01
-1
ELFORM = 41
IDIM EQ.1: first-kind Local boundary condition method
EQ.2: Gauss integration (default)
ELFORM = 42
IDIM EQ.1: first-kind Local boundary condition method (default)
EQ.2: second-kind Local boundary condition method
ELFORM = 41 is more suitable for crashworthiness analysis
ELFORM = 42 is more suitable for metal forming analysis
63
Meshfree Shell Surface (1)

ELFORM = 41: Global Approach
Meshfree Shell Surface Representation
E 0 := {X mid R 3 X mid (, ) = (, )}
Surface parameterization based on FE mesh + MLS [Krysl and Belystchko 1996]
Lagrange polynomials + MLS [Noguchi et al. 2000]
3D RKPM with extra constraints [Chen and Wu 2001]
Angle-based triangular flattening [Sheffer and Sturler, 2001] + MLS
N
minimize F ( ) = (ij ij )2 wij

i =1 j =1
subject to g i,(1)j ij 2 > 0, for i = 1...N,

j = 1...3, and some 2 > 0
Projection
Advantage: Handle complex manifold surface; Conforming shape functions
Disadvantage: Requires multiple parametric domains for spherical & cylindrical structures
64
Meshfree Shell Surface (2)

ELFORM = 42: Local Approach
zi
zI
yi
yI
M-plane
nM
I J xI
K
M-plane
xi
J K
nM +1
cos -1 ( nM , nM +1 ) Critical
Z,w
Y,v
Advantage: Handle complex geometry

Disadvantage: Non-conforming shape functions
I ( X J )
X,u
M plane
I ( X J )
N plane
Remedy: (Area-weighed) smoothing

% I ( X J ) =
NP
I ( X J ) ANIE
% I ( X )X iIN = X iN X E 0 / plate
IE =1
I =1
ANIE
NIE
where
NIE is the number of surrounding projected planes evaluated at X J
65
Constructed Meshfree Surface

Meshfree Global Approach
Meshfree Local Approach
Meshfree Global Approach

66
Meshfree Shell Formulation (1)

First-order shear deformable shell theory with 5/6 parameter approach
B := x R 3 x ( , , , t ) = ( , , t ) + V3 ( , , t ) with ,+
2 2
V3 = V2 + V1
NP
xI ( x ) I ( , )x
I =1
NP
ref
I
+ I ( , ) V3 I ; V3 I = x Itop x Ibottom
2
I =1
A co-rotational coordinate system is embedded at each in-plane integration

zone and defined by the convected coordinates
z y z y
s
V3 V2
Two approximations for local velocity
x s
V1
NP
NP
~
~ t
v i = I v iI + I I [ V2iI
2
I =1
I =1
&
xI
NP
NP
~
~ t I &
v i = I v iI + I yI
2
I =1
I =1
0

&
V1iI ] &I ; V3in +1 = Rij ( )V3in
I
y
x
with | v3 z |< 0.01 (Belytschkos element)

67

Lagrangian smoothed strains in co-rotational system [Chen and Wu 1998]
~
~
m = BIm d I
~
~
b = BIb d I
1
~
BIm ( xl ) =
Al
1
~
BIb ( xl ) =
Al
1
~
BIs ( x L ) =
AL
~ij1 ( X L ) =
h
1
=
2 AL
~
~
s = BIs d I
~ ~
I ( X )WJ = 0
BIm nd
X
l
J =1

X
l
b nd
B
I
where
~ ~
I ( X ) = 0
I =1
BIs nd
~ ~
I ( X ) X iI2 = X i
I =1
1
AL
l
ij d
h
NP
I =1 L
{[ I ( , )
(t I 1IV1 Ii + t I 2IV2 Ii ) ] n j + [ I ( , ) (t I 1IV1 Ij + t I 2IV2 Ij )] ni d

2
68

Internal nodal force
~ T
~T
~T
f Iint = BIm d + BIb d + BIs d
Equilibrium Equation
(K L + K N )d = f t + t t
NP
K L = ([ BI0 ( L ) + BI1 ( L ) ]TT CT [ BJ0 ( L ) + BJ1 ( L ) ] AL

T
L =1
NP
0
( L ) + BNI1 ( L )]TT ST [ BNJ0 ( L ) + BNJ1 ( L )] AL
K N = ([ BNI
L =1
NP
= ([ BI0 ( L ) + BI1 ( L ) ]TT AL

T
L =1
69
Crushing Tube using Meshfree Shell

FEM
Meshfree
70
Complex Channel Forming
Adaptivity
71
S-Rail Forming
8000 4-noded
Shell elements
72
Full Car Test (1)

Decomposition on 8 processors
Total nodes 285879

Total solid elements 2969
Total shell elements 269107
EFG solid: foam bumper
EFG shell: front hood
Normalized MPP CPU Time

No. of CPU
s
16
Full FEM
1.00
0.66
0.33
0.19
0.11
Coupled FE/EFG
1.08
0.72
0.37
0.22
0.15
Clusters of HP RX2600 (2 1.5 MHz CPUs per node)

OS: HP-UX 11.23
LS-DYNA: double precision MPP 971
73
Full Car Test (2)
74
Smoothed particle Galerkin (SPG) Method

Main Features
Is a pure particle integration method without integration cell.
Has explicit/implicit versions. Currently implemented by explicit method.
Has thermal-mechanics coupling (currently only in SMP).
Improves the low-energy modes due to rank deficiency in nodal.
Is related to residual-based Galerkin meshfree method.
Can be related to non-local or gradient types inelasticity.
NO stabilization control parameters.
Stability analysis via Variational Multi-scale analysis.
Frist-order convergence in energy norm.
Is capable to provide a physical-based failure analysis.
Released in R8.0.
75
*SECTION_SOLID_SPG (1)
Card 1
Variable
SECID
ELFORM
Type
ELFORM EQ.47: SPG formulation
Card 2 and Card 3 define only for the SPG option

Variable
DX
DY
DZ
ISPLINE
KERNEL
LSCALE
SMSTE
P
SUKTIME
Type
Default
1.5
1.5
1.5
Reserved
15
Reserved
Variable
Type
For Failure Analysis & Other Features
Default
*SECTION_SOLID_SPG
5, 47
1.6, 1.6, 1.6, , 1, , ,
76
Option of kernels in the analysis
Variable
DX
DY
DZ
ISPLINE
KERNEL
LSCALE
SMSTE
P
SUKTIME
Type
Default
1.5
1.5
1.5
Reserved
15
Reserved
KERNEL EQ. 0: Updated Lagrangian Kernel

EQ. 1: Eulerian Kernel
Some Guidelines KERNEL

Material
Note
Updated
Lagrangian
Solids
(Rubber-like, Foam, )
No failure
Eulerian
Solids
EOS, Solid fluid
Less shear deformation

Failure analysis
Extreme deformation
77
Interval of time step to conduct displacement regularization
Variable
DX
DY
DZ
ISPLINE
KERNEL
LSCALE
SMSTE
P
SUKTIME
Type
Default
1.5
1.5
1.5
Reserved
15
Reserved
Time to switch from updated Lagrangian kernel to Eulerian kernel

Variable
DX
DY
DZ
ISPLINE
KERNEL
LSCALE
SMSTE
P
SUKTIME
Type
Default
1.5
1.5
1.5
Reserved
15
Reserved
*SECTION_SOLID_SPG
5, 47
1.6, 1.6, 1.6, , 1, , 20, 0.15
78
Residual-based Meshfree Galerkin Principle

u ( X ) % (Y ; X ) u ( X ) d + % (Y ; X ) u ( X ) (Y - X ) d
1
(2)
% (Y ; X ) (2)u ( X ) ( 2 ) (Y - X ) d
2!
= u ( X ) % (Y ; X ) d + u ( X ) % (Y ; X )(Y ) d X
Displacement approximation
% (Y ; X ) d )
(2)
2 1
+ (2)u ( X ) ( ) % (Y ; X )(Y - X ) d
2!
(2)
1
= u ( X ) % (Y ; X ) d + (2)u ( X ) ( 2 ) % (Y ; X )(Y - X ) d
2!
= u ( X ) + (2 ) u ( X ) ( 2 ) ( X )
Gradient type nonlocal strain
a h (u , u ) = l (u ) u V h
( ) ( )
(2)
(2)
a h ( u , u ) = ( s u ) : C : ( s u ) d + u : C : u d
h
h
= astan
( u , u ) + astab
( u , u )
( ) ( )
(2)
(2)
h
astab
( u , u ) = u : C : u d
(2)
u =
1
: u (2) + (2)u : )
(
2
2
l ( u ) = u fd + u td ( ) u : fd
Wu et. al submitted to J.
Comput. Physics. (2014)
79
Well-defined Mathematical Property

Coercivity
2
(2)
s 2
u 1 c1 u c1 u + u
0
0
0
c1
h
h
astan
( u , u ) + astab
( u , u ) )
(
min ( C )
2
= c2 a h ( u , u ) , c1 , c2 > 0, u V h
Continuity
( ) ( )
( C ) ( ( u ) d ) + ( ( v ) d )
+ c ( h ( u ) d ) + ( h ( v ) d )
a h ( u , v )
( u ) : C : ( v ) d +
s
max
1/ 2
(2)
(2)
u : C : v d
1/ 2
1/ 2
1/ 2
max ( C ) c4 { u 1 v 1} c5 u 1 v 1 , c3 ,c4 ,c5 > 0
u , v V h
Unique solution !
80
Error Estimation via VMS (Variational Multi-scale Method)
(
(v
) (
) + a (v
) ( )
) = l (v ) v
a h v h , u h + a h v h , u b = l v h v h V h
coarse-scale equation
ah
fine-scale equation
, uh
, ub
Bh
B h ( ) := v b : v b H 1 , v b = 0 on
NP
global residual-free fine-scale space
~
u h ( x ) = J ( x )u J
J =1
NP
NP
~
~
= J ( x ) K ( x J )u
K
J =1
BP
BP
~ b ( x ) = b ( x )u
~ b ( x ), x Z b
ub ( I ( x ) I ( x ))u
I
I
I
I
I
I
I
I =1
I =1
K =1
NP NP
~
~
= K ( x J ) J ( x )u
K
K =1 J =1
NP
~
= K ( x )u
K
K =1
~
T ~
~
bT ~ 1 ~
u = u + u = K K K R + R + b K 1 R
h
a h u u h , u u h = a h u u h , u u h + astab
u uh , u uh
uu
h 2
e
+h uu
2
e
2
2
2
c( , )h 2 u 2, + c~( , )h 2 u 2, c( , )h 2 u 2,
Error estimation
in energy-norm
81
Nonlinear SPG Implementation

Implicit formulation
= ij Cijkl kl d + ui , jTijkl u k ,l d ui f i d ui ti d
x
( )
~
~
U T K nv+1 U
~
= U T Rnv+1
v +1
n +1
~
U = A -1U
NP
AIJ = J ( X I )I = K ( X I ) J ( X K )I
K =1
A-T K nv+1 A 1 (U )n+1 = A-T Rnv+1

v +1
Explicit dynamic formulation
A-T MA1U&& = A-T f ext f int
MU&& = A-T f ext f int

M
RS
I
NP
NP
Corresponding coordinate system

in SPG computation
T
1
= M IJ = AIK
M KM AML
NP
d I
~
~& ( x )
&
= I uI = I u
J
J ,x
I
dt
J =1
( )
Currently implemented in LS-DYNA
82
2D Prandtls Punch Problem

dy
Quasi static
Effective Plastic Strain
Elastic-plastic
Direct nodal integration method (DNI)
SCNI method (Chen et al. IJNME 2001)
SPG
83
2D Bushing problem
L

a1
dy
Core
a2
L = 1.5
a1 = 2.5
DNI
a2 = 1.0
Metal
Quasi static
Elastic-plastic
SCNI
Present
SPG
84
Taylor Bar Impact (1)
Bottom
view
Progressive deformation with EPS
t=0.004
t=1.7610-5
0.012
1.7210-5
0.020
1.6610-5
0.028
1.6610-5
85
Taylor Bar Impact (2)

FEM
SPG
EPS
EPS
Temperature
Temperature
86
Plate Impact with Updated Lagrangian Kernel (1)
Ball: rigid, R=5.0

Plate: R=20.0, thickness=5.0
Particles: 25721, updated
Lagrangian
Elastic perfectly-plastic material:
0=7.8510-3
E=6.9x104
v=0.3
y=200.0
Vz=-600.0
87
Plate Impact with Updated Lagrangian Kernel (2)

Bottom view
t=0.04
t=7.4410-6
t=0.02
t=7.4710-6
t=0.08
t=7.3910-6
88
Meshfree-enriched Finite Element Method (MEFEM)

Standard displacement-based low-order Lagrange elements
Attractive due to simplicity and economy in element formulation

Mathematically and numerically well-established
Volumetric locking in incompressible regime
Shear locking in bending-dominated problems when coarse mesh is adopted
Non-physical locking
modes in elastic
eigenvalue analysis
Q1 element
89
Existing Numerical Techniques

Find
u h V h H 01 ( )
such that A ( uh , v h ) = l (v h ) v h V h
A(u h ,v h ) = 2 (u h ) : (v h )d + ( u h )( v h )d
(P)
l (v h ) = f v h d + t v h d
N
In near-incompressible regime u h 0
as ( or v 0 .5 )
Divergence-free condition u h = 0
FEM
Meshfree
Reduced/selective integration
Reduced integration and hourglass control
Taylor expansion method
Mixed formulation
Assumed/Enhanced strain method
Non-conforming element
F-bar
Pressure projection method;
Average nodal pressure element
Pseudo-divergence-free interpolation
Pressure projection
B-bar
Mixed formulation
90
Meshfree-enriched Finite Element Method (MEFEM)

Has explicit/implicit versions.
Is a Hybrid element: meshfree bubble enriched 5-noded tetrahedral element.
Is a smoothed strain displacement-based Galerkin FEM method.
Is equivalent to a low-order inf-sup stable mixed formulation.
Surpasses the inf-sup deficiency through multi-scale analysis.
Suitable for rubber-like materials.
Currently needs a special mesh re-partitioning scheme for 4-noded elements.
91
*SECTION_SOLID
Card 1
Variable
SECID
ELFORM
Type
ELFORM EQ.43: Meshfree-enriched finite element formulation
*SECTION_SOLID
5, 43
92
Meshfree-enriched Bilinear Element (1)

Goal: uh ( gi )
0 as Qe M h , i = 1,2,3,4
Qe
Integration point
Fe : Qe Qe , x = Fe ( ) = F , F
[ ]
2
e
1
e
V h ( ) = u h : u h H 01 , u h
5
= xi i ( , ),
i =1
= u h o Fe , u h P1 (Qe ) for all Qe M h

1
Qe
(
)
y
for all Qe
i i
i =1
where P1 (Qe ) = span{i , i = 1,L5} denotes the space contains a set of basis functions in Q e
93
Meshfree-enriched Bilinear Element (2)

Sukuma IJNME (2004); Arroyo and Ortiz IJNME (2006)
Wu et al. IJNME (2011)
94
Area-weighted Strain Smoothing

% uh = 1
% h =
Am
1
~
( ( g1 , g1 )det (J1 ) + ( g 2 , g 2 )det (J 2 ))
h =
2 Am
u h d
= b mb
T
T
det ( J1 ) + det ( J 2 ) det ( GF G x )1 + det ( GF G x )2
Am =
=
2
2
h
% u = tr ( % ) = tr
u
d
Am m
1
=
tr ( u h ) d
Am m
h
Am =
det( J 2e1 ) + det( J 3e1 ) + det( J1e 2 ) + det( J 4e 2 )

2
e1
e1
e2
e2
det ( GF G xT ) + det ( GF G xT ) + det ( GF G xT ) + det ( GF G xT )

2
=
=
1
Am
uh d
uh ( g1 ) det ( J1 ) + uh ( g 2 ) det ( J 2 )
det ( J1 ) + det ( J 2 )
The resulting element formulation

Is divergence-free in incompressible limit
Contains no spurious zero energy mode
Passes patch test in compressible case
=0
95
Mathematical Properties
Find uh V h such that
Ah ( u h , v h ) = l ( v h ) v h V h
Ah ( uh , v h ) = C h ( uh ) : h ( v h ) d
Theorem 1
The Jacobian of the reference mapping for meshfree-enriched finite element is bijective .
Theorem 2
The smoothed gradients of the meshfree-enriched triangular elements satisfy the
integration constraint and the resultant element formulation passes the patch test.
Theorem 3
The modified bilinear form Ah ( , )
positive constant such that
is bounded on V h V h , i.e., there exists a
Ah u h , v h Cb u h
Theorem 4
The modified bilinear form Ah ( , )
positive constant such that
vh
u h , v h V h
is coercive on V h V h
Ah v , v
) C
h 2
1
, i.e., there exists a
v h V h
Wu and Hu CMAME (2011); Wu et al. IJNME (2012); Wu and Hu CM (2012)
96
Convergence in Incompressible Limit

Theorem 5: inf-sup condition
Find
( u , p ) V P
A ( u , v ) + B div v
h
Find uh V h such that
(R)
Ah ( uh , v h ) = C h ( uh ) : h ( v h ) d
p = tr h ( u
h
) ) =
div ( u
such that
, p h = l ( v h ) v h V h
_______
h
A ( uh , v h ) = 2 h ( uh ) : h ( v h ) d
Subject to a stability condition between the

displacement space and an implicit
pressure space induced by
__________
h
_______h h 1
B div u , q B ( p h , q h ) = 0 q h P h
(M)
Ah ( uh ,v h ) = l ( v h ) v h V h
B ( p, q ) = pqd
Saddle point
problem of penalized
Stokes equation
_______
h
in
div u = tr h ( uh )
P h = q h : q h L20 ( ) ,q h
L20 ( ) = q L2 ( ) ,
P0 ( m ) m M h
qd = 0
discrete inf-sup condition for bilinear function B div u h , q h
_______
infh
h
sup
q P \{0} v h V h \{0}
________
h
q div ( v ) d
h
% ( v h )
qh
1 h
~ (v h ) 0
q
97
Extension to Triangular and Tetrahedral Elements
Fe ( , )
98
2D Rubber Tube Inflation (1)
GMF-GI
NICE-T3
ME-Tri-AW
99
100
(a) semi-uniform discretization
(b) random discretization
Deformation plot using ME-Tri-AW in nearly-incompressible case: initial (dash lines);

analytical (thick red lines); numerical (thick blue lines)
101
2D RubberBushing (1)
102
2D RubberBushing (2)
103
2D RubberPunch (1)
104
2D RubberPunch (2)
105
Particle-reinforced Rubber Composite
Displacement-based
meshfree Galerkin method
Displacement-based
meshfree Galerkin method
with pressure smoothing
ME-FEM
106
4. Workshop II
5. Standard 3D Adaptive EFG and Its

Keywords
Adaptive Methods for Manufacturing Simulations

Reasons for Adaptivity
High accuracy requirement
Surface representation, High gradient
Residual stress effects the crash result
Multi-stage analysis involving moving-boundary
Typical applications:
Hot Forging/Extrusion, Cold forming
Machining, Riveting, Rolling, Tapping
Current Numerical Limitations

R-adaptivity for solids
H-adaptivity is limited to shell structures
Failure analysis is limited to metal cutting problems

Do not apply to rubber-like materials
109
Riveting (Self-Piercing) Analysis (1)
Tool
Rigid
Working Pieces
(to
be connected)
Rigid
EFG Implicit solver

Multiple-part adaptivity
110
111
IVT=2
IVT=1
112
Frictional Self-Riveting Simulation
Tool
Adaptive Parts
113
Complex Forming
MPP Implicit solver
Local adaptivity
114
Squeezing
MPP Implicit solver

Local adaptivity
Deformable tools
Work piece
Top tool
115
Metal Cutting Analysis
v=30m/s
MPP Implicit solver

Local adaptivity
116
Friction Stir Welding Analysis (1)

MPP Implicit solver
Local adaptivity
Temperature Distribution
117
Friction Stir Welding Analysis (2)
118
Forming with Sharp Tool
119
Piercing
Rigid
EFG Implicit solver

Local adaptivity
Forming_..._Mortar Contact
Working piece
Metal
No material failure (treated as solid fluid)
Piercing force
120
Main Features
An explicit/implicit solver coupled with thermal analysis.

Using 4-noded integration cell.
Automatic switch from 6/8-noded cell to 4-noded cell.
A second-order interpolation scheme for state variable transfer.
Pressure smoothing to improve accuracy.
Including global/local adaptive refinements.
Interactive adaptivity for efficiency as well as accuracy.
A special surface reconstruction algorithm for metal cutting analysis.
Available for multiple adaptive parts.
Manual remeshing and mesh editing.
Available in SMP and MPP.
121
Card 1
Variable
SECID
ELFORM
Type
I
42

EQ.42: for 4-noded adaptive EFG

Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
1.01
1.01
1.01
-1
0.01
Variable
IGL
STIME
IKEN
SF
MID
IBR
DS
ECUT
Type
Default
F
I
FAnalysis
I
I
F
F
For
Fracture
& Other
Features
1.e+20
0.0
1.01
0.1
*SECTION_SOLID_EFG
5, 41
1.1, 1.1, 1.1, , , 3, 1,
1
122
Essential Boundary Conditions
Card 2
Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
1.01
1.01
1.01
0.01
IEBT EQ. 1: Full transformation (default)

EQ. 2: Mixed transformation
EQ. 3: Coupled FEM/EFG
EQ. 4: Fast transformation
EQ. 5: Fluid particle (trial version)
EQ. 7: Modified Maximum Entropy approximation
*CONTROL_ADAPTIVE
.
*CONTROL_REMESHING
.
*SECTION_SOLID_EFG
5, 41
1.2, 1.2, 1.2, , , 3, 2,
1
*PART
Workpiece
100, 4, , , , ,2(ADPORT)

EQ.-1: Stabilized EFG method (apply to 8-noded, 6-noded and combination of them)
123
Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
Displacement approximation
KP
[m]
d
(
x
)d
;
x
L
iL
FEM
xL FEM
h
ui ( x ) ui ( x ) = NP
MP
[n]
wa ( x ; x x I )d iI + L[m] ( x )d iL ; x Meshfree R d
144
42444
3
14243
L
x I
x
L
Interface
I Meshfree
Completeness condition
NP
MP
[n]
a
i
1I
j
2I
( x ; x x I )x x +
[m]
J
( x ; x x J )x1i J x 2j J = x1i x 2j , i + j = 0 ,K , n
x I Meshfree
x J Interface
NP
or H
[n]
(0 ) =
[n]
a
MP
( x ; x x I )H
[n]
( x xI ) +
x I Meshfree
[m]
J
( x ; x x J )H [ n ] ( x x J )
x J Interface
Interface Constraint
~
~
I (x) = 0 {I : supp( I ) Interface 0}
x Interface
124
*CONTROL_ADAPTIVE (1)
Card 1
Variable
ADPFERQ
ADPTOL
ADPORT
MAXLVL
TBIRTH
TDEATH
LCADP
IOFLAG
Type
Default
none
1.e20
0.0
1.e20
*CONTROL_ADAPTIVE
0.1, ,7, , 0.0, 0.8
*PART
,,,
, , , , , , 2 (ADPORT),
125
*CONTROL_ADAPTIVE (2)
Card 2
Variable
ADPSIZE
ADPASS
IREFLAG
ADPENE
ADPTH
MEMORY
ORIENT
MAXEL
Type
0.0
inactive
inactive
inactive
Default
Adapene = 0 (Global refinement) :

Segment based contact (SOFT=2 in *CONTACT) is
recommended
Adapene >0 (Local refinement):
Support shell/solid rigid/deformable tool and multiple adaptive
parts.
Support surface_to_surface contact types (SMP & MPP)
FORMING_..._MORTAR (recommended)
Standard and AUTOMATIC
ADPENE represents a distance from the tooling surface within
which the adapted mesh refinement of the adaptive part is
influenced by the radius of curvature of the tooling surface.
This feature is currently unavailable in SOFT = 2 in *CONTACT.
Global Refinement
Local Refinement
126
Self-Riveting Simulation
Temperature Contour
127
*CONTROL_REMESHING
Variable
RMIN
RMAX
VF_LOSS
MFRAC
DT_MIN
ICURVE
CID
SEGANG
Type
1.0
0.0
0.0
Default
RMAX = 3~5 RMIN
*CONTROL_REMESHING
1.2, 4.8, , , , 4
128
*CONTROL_CONTACT
Card 1
Variable
SLSFAC
RWPNAL
ISLCHK
SHLTHK
PENOPT
THKCHG
ORIEN
ENMASS
Type
Default
.1
none
SLSFAC: Scale factor for sliding interface penalties (1.0~6.0 recommended in adaptivity)
SHLTHK: Shell thickness in contact (also in *CONTACT_)
= 1 thickness is considered but rigid bodies are excluded
(*CONTACT_AUTOMATIC_SURFACE_TO_SURFACE)
= 2 thickness is considered including the rigid bodies
(*CONTACT_SURFACE_TO_SURFACE) offset 0.5*shell thickness
PENOPT: Penalty stiffness value (4 or 5 are recommended for high gradients)
= 4: use slave node value, area or mass weighted.
= 5: proportional to shell thickness
129
*CONTACT_...
Optional
Card A
Variable
SOFT
SOFSCL
LCIDAB
MAXPAR
SBOPT
DEPTH
BSORT
FRCFRQ
Type
Default
0.1
1.025
0.0
10-100
SOFT: Soft constraint option:

EQ.0: penalty formulation
Supported in local adaptivity
EQ.1: soft constraint formulation
EQ.2: segment-based contact
Preferred in global adaptivity
EQ.4: constraint approach for FORMING contact option
BSORT: Number of cycles between bucket sorts.
Preferred contact types in adaptivity:
*CONTACT_(AUTOMATIC)_SURFACE_TO_SURFACE
*CONTACT_(FORMING/AUTOMATIC)_SURFACE_TO_SURFACE_MORTAR
130
*CONSTRAINED_GLOBAL or LOCAL
*SECTION_SOLID_EFG
5, 41
1.2, 1.2, 1.2, , , 7, 2,
*CONTROL_REMESHING
2.0, 5.0
*CONSTRAINED_LOCAL
10410 nodes
2769 nodes
EFG
Adaptive EFG
131
Implicit Time Step Size

*CONTROL_IMPLICIT_AUTO
Card 1
Variable
IAUTO
ITEOPT
ITEWIN
DTMIN
DTMAX
DTEXP
Type
Default
11
DT/1000.
DT*10.0
NONE
KFAIL
KCYCLE
IAUTO: Automatic time step control flag

EQ. 0: constant time step size
EQ. 1: automatically adjusted time step
DTMAX: Maxmium allowable time step size
*CONTROL_IMPLICIT_GENERAL
Card 1
Variable
IMFLAG
DT0
IMFORM
NSBS
IGS
CNSTN
FORM
ZERO_V
Type
Default
NONE
DT0: Initial time step size for implicit analysis
132
*CONTROL_IMPLICIT_SOLUTION
Card 1
Variable
NSOLVR
ILIMIT
MAXRE
F
DCTOL
ECTOL
RCTOL
LSTOL
ABSTOL
Type
Default
11
15
0.001
0.01
1.-E+10
0.90
1.e-10
IIMIT: Iteration limit between automatic stiffness reformations

MAXREF: Stiffness reformation limit per time step
133
*INTERFACE_SPRINGBACK_LSDYNA
Card 1
Variable
PSID
NSHV
Type
Default
none
none
NSHV: Number of history variables

*KEYWORD_ID
Tool_solid
*CONTROL_IMPLICIT_INERTIA_RELIEF
1, 1.0
*CONTROL_IMPLICIT_SOLUTION
2, 1 (ILIMIT: Stiffness reformation limit per time step)
*CONTROL_IMPLICIT_GENERAL
, , , , 2 (IGS: Geometric stiffness matrix)
*INCLUDE
Tool_solid.dynain (initial stress )
134
*CONTROL_THERMAL
Card 1
Variable
ATYPE
PTYPE
SOLVER
CGTOL
GPT
EQHEAT
FWORK
SBC
Type
Default
1.0e-4
1.
1.
0.
ATYPE: = 0 Steady state analysis

= 1 Transient analysis
CGTOL: = 1.0e-3~1.0e-4
*CONTROL_THERMAL
1, , , 1.0e-3
*CONTROL_THERMAL_TIMESTEP
*INITIAL_TEMPERATURE
*MAT_THERMAL_ISOTROPIC
135
The Upsetting Process with Thermal (1)
Initial Temperature 20 0C
Low carbon steel Ck105
No heat transfer to the environment
Internal Energy
Force
136
The Upsetting Process with Thermal (2)
Final stage
Initial Temperature 20 0C
Low carbon steel Ck105
Springback
Von Mises Stress Contour
137
Extrusion with Thermal Coupling (1)
15997 nodes
13969 nodes
15091 nodes
15003 nodes
Contact
Force
15086 nodes
138
Extrusion with Thermal Coupling (2)
Temperature
139
Extrusion with Thermal Coupling

Local refinement
*Contact_Forming_Surface_To_Surface_Mortar
Interactive adaptivity
140
Standard Practice of 3D Adaptivity
Adaptive frequency is controlled by an user-defined time dependent

parameter or curve.
Mass scaling is allowable in explicit dynamic analysis.

Support local refinement
Define local coordinates for constrained and symmetric planes.
Element erosion is not supported.
Implicit analysis is suggested for most manufacturing problems.
141
6. Workshop III
7. Advanced 3D Adaptive EFG and Its

Keywords
Advanced Adaptive Method

1.
2.
3.
4.
5.
6.
Meshfree Interactive Adaptive Method

Monotonic Mesh Resizing
Meshfree Pressure Smoothing Scheme
Internal Variables Transfer Methods
Meshfree for Orbital Forming
Manual Remeshing and Users Control File for Adaptivity
*CONTROL_REMESHING_EFG
0.0015, 0.0060
1, 3, 1, , 1
0.40,4.5,
*SECTION_SOLID_EFG
5, 42
1.1, 1.1, 1.1, , , 3, 2,
1
144
Meshfree Interactive Adaptive Method

Dynamically detect distortion and maintain the quality of EFG discretization.
Three indicators were implemented to detect distortion including shear
deformation, nodal distributions and volume change.
More efficient than traditional non-interactive adaptivity using constant interval.
More robust than non-interactive adaptivity in large deformation analysis.
Available in R5 version for SMP and MPP.
145
*CONTROL_REMESHING_EFG (1)
Card 1
Variable
RMIN
RMAX
VF_LOSS
MFRAC
DT_MIN
ICURVE
Type
1.0
0.0
0.0
Default
Options in interactive adaptive EFG method

Card 2
Variable
IVT
IAT
IAAT
MM
Type
Default
0.
0.
0.
Tolerance in interactive adaptive EFG method

Card 3
Variable
IAT1
IAT2
IAT3
Type
Default
1.0e+20
1.0e+20
1.0e+20
146
Card 2
Variable
IVT
IAT
IAAT
Type
Default
0.
0.
IAT: Interactive adaptivity type

= 0: No interactive adaptivity
= 1: Interactive adaptivity combined with predefined adaptivity
Extra adaptivity is triggered interactively within every period defined by ADPFREQ.
= 2: More interactive adaptivity
The time interval between two successive adaptive steps is bounded by ADPFREQ.
= 3: Purely interactive adaptivity
IAAT: Interactive adaptivity adjustable tolerance
= 0: The tolerance to trigger interactive adaptivity is not going to be adjusted in run-time
= 1: The tolerance to trigger interactive adaptivity is going to be adjusted in run-time
This is designed to avoid over-activation of interactive adaptivity.
147
Tolerance in interactive adaptive EFG method
Card 3
Variable
IAT1
IAT2
IAT3
Type
Default
1.0e+20
1.0e+20
1.0e+20
IAT1: The tolerance for shear deformation. (0.1~0.5)

IAT2: The tolerance for nodal distribution. (~RMAX/RMIN for no local refinement)
IAT3: The tolerance for volumetric change. (~0.5).
The tolerance is automatically adjustable at runtime when IAAT=1.

The tolerance adjustment is affected by:
(1) How frequently interactive adaptivity is triggered
(2) Material deformation
(3) ADPFREQ
The rate of change in three different indicators over ONE time step is also considered.
If the rate is over 50%, interactive adaptivity is also triggered.
148
Forging Simulation (1)

*CONTROL_ADAPTIVE
0.0002
0.0015
0.0060
3
0.40
4.5
Resultant force
Mixed adaptivity
* SMP with 1 CPU

IAT
Normalized CPU time
1.00
1.60
1.14
1.31
# of adaptive steps
22
73
29
32
Traditional adaptivity
Purely interactive adaptivity
149
Forging Simulation (2)
150
Wheel Forging Simulation (1)

*CONTROL_ADAPTIVE
2.5
0.15
0.30
3
1
0.20
3.5
0.80
Resultant force
* SMP with 6 CPUs
Traditional adaptivity
IAT
Normalized CPU time
1.0
0.72
# of adaptive steps
50
22
Purely interactive adaptivity
151
Wheel Forging Simulation (2)

Purely Interactive Adaptivity (IAAT=1: adjusted tolerance)
Interactive adaptivity triggered by rate of indicator change
Tolerance
Shear
deformation
Indicator
t (sec)
Unbalanced
nodal
distribution
Volumetric
change
Standard adaptivity triggers remeshing in predefined timetable regardless of mesh distortion

Interactive adaptivity is triggered only when distortion is detected by indicators.
152
Metal Cutting Simulation (1)

w/ interactive
w/o interactive
Stop due to local distortion
Shell
rigid
MAT_003
*CONTROL_ADAPTIVE
0.010
0.20
1.2
3
1
0.15
3.5
Resultant force
153
Metal Cutting Simulation (2)

Purely Interactive Adaptivity (IAAT=1: adjusted tolerance)
Tolerance
Shear
deformation
Indicator
Unbalanced
nodal
distribution
Volumetric
change
Interactive adaptivity is able to detect distortion that occurs frequently and irregularly
in metal cutting analysis, which is hard to be handled by traditional adaptivity.
Automatic adjustment of the user defined tolerance is able to avoid over-activation of
interactive adaptivity, which results in an improved efficiency.
154
Gear Forging Simulation
155
Monotonic Mesh Resizing (1)

Card 2
Variable
IVT
IAT
IAAT
IER
MM
Type
Default
0.
0.
0.
0.
MM: Monotonic mesh resizing

= 0: Off
= 1: On
Capture high gradient
Maintain mesh density for not losing accuracy through adaptivity
Especially useful when local refinement is not available
Local refinement is currently driven ONLY by curvature of contact surface
156

Contour Effective Plastic Strain (0~1.8)
Adaptive EFG
Norm.
CPU
Adaptive EFG
Interactive
Adaptive EFG
Interactive
Monotonic resizing
Adaptive EFG
Monotonic
resizing
Adaptive EFG
Adaptive EFG
Interactive
Adaptive EFG
Monotonic resizing
Adaptive EFG
Interactive
Monotonic resizing
1.0
0.82
3.1
2.0
157

*CONTROL_REMESH
0.004
0.05
0.2
-2
2,,,1
1.2
0.9
*SECTION_SOLID_EFG
1
41
1.1, 1.1, 1.1 , , , 3, 2
1
Off
On
158
Meshfree Pressure Smoothing Scheme (1)

*SECTION_SOLID_EFG
Card 1
Variable
SECID
ELFORM
Type
*SECTION_SOLID_EFG
5, 42
1.1, 1.1, 1.1, , , 3, 2,
1

EQ.42: for 4-noded adaptive EFG

Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
1.01
1.01
1.01
-1
0.01
Variable
IPS
STIME
IKEN
SF
MID
IBR
DS
ECUT
Type
Default
F
I
FAnalysis
I
I
F
F
For
Fracture
& Other
Features
1.e+20
0.0
1.01
0.1
IPS EQ. 0: (default) No pressure smoothing

EQ. 1: Moving-least squared pressure recovery
159

*SECTION_SOLID_EFG
Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
1.01
1.01
1.01
-1
0.01
Variable
IPS
STIME
IKEN
SF
MID
IBR
DS
ECUT
Type
Default
F
I
FAnalysis
I
I
F
F
For
Fracture
& Other
Features
1.e+20
0.0
1.01
0.1
Pressure smoothing is recommended for IEBT=3 or IDIM=1 (including

other IEBT approximations) in adaptive EFG method
Pressure smoothing is NOT recommended when there is hard contact
condition or implicit analysis has convergence problem
(Hard contact condition means the contact causes the numerical difficulty to
obtain the solution, e.g. local refinement with severe deformation and locally
high gradients)
160

Incompressibility is satisfied in a weak sense
2
( P ) =
S ij,i I , j PI
j =1
from flow formulation

L2 ( )
~
W
( P ) = Qp
J
L2 ( )
Minimization
~
~
1
T W
P = QM Q
where M = Q T Qd
J
Linearization
2 ~
~
1
T W
1 u k
P = QM Q
JF
d
lk
J2
X l
K nv+1d = f ext f int
v
n +1
~ ~
v
K nv+1 = B T ( D + T )n +1 Bd + k B T Bd
~
B = QM 1 Q T gBd
161
*SECTION_SOLID_EFG
5, 42
1.1, 1.1, 1.1, , , 3, 2,
0
Smeared gradient
Adaptive EFG w/
Pressure smoothing
Adaptive EFG w/o

Pressure smoothing
162
Multi-stage Hot Forming Analysis

*MAT_ELASTIC_VISCOPLASTIC_THERMAL
Curling is due to unbalanced stress distribution through thickness
Pressure smoothing helps to improve stress calculation
w/o pressure
smoothing
w/ pressure
smoothing
163
Internal Variables Transfer Methods (1)

Internal variable Transfer in Adaptive EFG
Card 2
Variable
IVT
IAT
IAAT
Type
Default
0.
0.
IVT: = 1 Moving Least square approximation with Krocknet-delta property

= -1 Moving Least square approximation without Krocknet-delta property
= 2 Parition of Unity approximation with Krocknet-delta property
= -2 Parition of Unity approximation without Krocknet-delta property
IVT=1 is recommended in general case
IVT=2 is recommended in local refinement with significant change in mesh
density
164
Internal Variables Transfer Methods (2)

0.0015, 0.0060
-2,
*SECTION_SOLID_EFG
5, 42
1.1, 1.1, 1.1, , , 3, 2,
1
Smeared gradient
Adaptive EFG
IVT=1
Adaptive EFG
IVT=2
(Low CPU cost)
165
Meshfree for Orbital Forming (1)

*PART
Card 2
Variable
PID
SECID
MID
EOSID
HGID
GRAV
ADPORT
TMID
Type
A8
A8
A8
A8
A8
Default
none
none
none
ADPORT EQ. 2: R-adaptive for 3D.

EQ. 3: for 6-noded/8-noded orbital forming
*SECTION_SOLID_EFG
Card 2
Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
1.01
1.01
1.01
-1
0.01
IDIM EQ.-1: Stabilized EFG method (apply

to 8-noded, 6-noded and
combination of them)
*PART
2, 1, 1, , , , 3,
*SECTION_SOLID_EFG
5, 41
1.1, 1.1, 1.1, , , 3, -1,
166

*CONTROL_REMESHING
Card 1
Variable
RMIN
RMAX
VF_LOSS
MFRAC
DT_MIN
ICURVE
CID
SEGANG
Type
1.0
0.0
0.0
Default
CID: Define the orbital axis

The orbital axis has to be in parallel to the global z-axis in current pratice
SEGANG: Define angular mesh size
Orbital
axis
*CONTROL_REMESHING
, , , , , , 5, 2.0
*DEFINE_COORDINATE_SYSTEM
5, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0
1.0, 1.0, 0.0
167

EPS
168
Users Adaptivity Control Files (1)

Runtime control on adaptivity (Beta development version)
Manually trigger additional adaptive remeshing
Manually remesh and edit mesh
Define adaptive parameters for individual adaptive part
IADPFC=1 (7th flag in card 4, *CONTROL_ADAPTIVE)

Two control files in the job folder
adapt.fc1
1. Trigger additional adaptive steps either immediately or any time later
2. Manually remesh/edit the mesh instead of using LS-Dyna remesher
adapt.fc2
Set specific Rmin/Rmax, adaptive birth/death time for individual adaptive part
169

adapt.fc1: Three parameters C1, C2, C3
Examples:
(1) 1,0.0,0: trigger additional normal adaptivity immediately
(2) 1, 2.5e-3,0: trigger additional normal adaptivity at t=2.5e-3
(3) 1, 2.5e-3,3: trigger additional special adaptivity with manual
remeshing on adaptive part 3 at t=2.5e-3.
Manual remeshing and mesh editing:
(1) LS-DYNA will hang up waiting for new mesh from user
(2) "user.mesh" in the job folder containing the mesh for users to remesh/edit
(3) Change C1 to -1 to continue LS-DYNA with new "user.mesh"
170

adapt.fc2: Parameter list for adaptive parts
Example:
Line 1: 2 (number of adaptive parts)
Line 2: 2, 0.0,0.1,1,4 (for adaptive part 2, adaptive birth time is 0.0,
adaptive death time is 0.1, RMIN=1, RMAX=4)
Line 3: 3, 0.01,0.2,2,4 (for adaptive part 3, adaptive birth time is 0.01,
adaptive death time is 0.2, RMIN=2, RMAX=4)
If new mesh quality of some adaptive parts using manual remeshing is NOT good
enough for the remesher in LS-Dyna, it is recommended to stop adaptivity by
setting corresponding adaptive death time to be just slightly larger than C2 in
adapt.fc1 to avoid error termination due to failure of automatic remeshing.
171

Manually edit the mesh
Erode a thin layer of mesh right before material separation
Stop adaptivity right after manually editing the mesh
172

Set different control parameters on different adaptive parts
173

Manually erode a thin layer of elements
Manually cut the mesh using Hypermesh
Stop adaptivity of top sheet after erosion
Adaptivity of top sheet continues on
174
8. Workshop IV
9. Failure Analysis and Its Keywords
Modeling Material Separation

1.
Weak Discontinuities : discontinuous deformation gradients

Continuum damage constitutive equation + Nonlocal strain smoothing + Material erosion
Implicit Cracks: Crack is an assumed width
Polynomial basis is inadequate to represent the fine scale.
Time step tents to be very small in explicit analysis with fine mesh
2.
Strong Discontinuities : discontinuous displacement

Cohesive model + (Interface element, or elemental enrichment EFEM, or
nodal enrichment XFEM, or EFG)
Phenomenological failure + nonlocal strain (SPG, bond/state-based peridynamics)
Explicit Cracks : remove the influence of mesh size and orientation
No direct correlation between the strain softening and critical energy release rate.
Time step:
2
2k
3.
max
; max =
Weak + Strong Discontinuities

Loss of uniqueness as a criterion for changing from continuum damage mechanics to discete
177
Material Fracture v.s. Numerical Fracture
Material separation due to

numerical fracture in SPH
Physical material fracture before numerical fracture
Enlarge numerical support !
178
SPG for Ductile Failure

Can be related to non-local or gradient types material failure model.
Currently provide three failure mechanism:
Continuum damage mechanics based
Phenomenological strain based
Max principle stress based
More failure criterions can be added upon request.
Availabel in R8.0
Will be extended to brittle materials.
179
Variable
DX
DY
DZ
ISPLINE
KERNEL
LSCALE
SMSTE
P
SUKTIME
Type
Default
1.5
1.5
1.5
Reserved
15
Reserved
Card 2
KERNEL EQ. 0: Updated Lagrangian Kernel

EQ. 1: Eulerian Kernel
Some Guidelines KERNEL

Material
Note
Updated
Lagrangian
Solids
(Rubber-like, Foam, )
No failure
Eulerian
Solids
EOS, Solid fluid
Less shear deformation

Failure analysis
Extreme deformation
180
Failure Control
Card 3
Variable
IDAM
SF
Type
Default
IDAM EQ. 0: Continuum damage mechanics (default).

EQ. 1: Phenomenological strain-based failure criteria.
EQ. 2: Principle stress based failure criteria.
SF: Failure EPS / Max-Principle-Stress.
*SECTION_SOLID_SPG
5, 47
1.1, 1.1, 1.1, , 1, , ,
1, 0.5
181
Plate Impact (1)
Ball: rigid, R=5.0

Plate: R=20.0, thickness=5.0
Particles: 25721, Eulerian
Elastic perfectly-plastic material:
0=7.8510-3
E=6.9x104
v=0.3
y=200.0
Vz=-600.0
Phenomenological strain-based failure criteria

IDAM=1
182
Plate Impact (2)
v=300
t=0.12
v=400
t=0.09
v=600
t=0.06
Front view
Top view
Bottom view
183
Plate Impact (3)

Bottom view
v=400
t=0.04
t=7.4410-6
t=0.06
t=7.4710-6
t=0.02
t=7.4810-6
t=0.08
t=7.4810-6
t=0.09
t=7.4410-6
184
Plate Impact (4)
v=300
v=400
v=600
185
A Low-speed Ball Penetrating through the Metal Plate(1)
EPS
Contact Force
Von Mises stress
186
A Low-speed Ball Penetrating through the Metal Plate(2)
Using existing FEM mesh.

No element/particle erosion or manual cut of the model.
Time step size does not drop.
Top view
Bottom view
187
Metal Cutting Analysis (1)
Aluminum
0=2.710-6 kg/mm3
E=78.2GPa
v=0.3
y=0.29(1+125ep)0.1
Strain-based failure criteria fail = 0.5
Cutting Speed = 10 m/s
Fixed t=3.010-5
188

Different cutting angles
Fixed t=3.010-5
189

Different depth
Fixed t=3.010-5
190
191
Metal Shearing Analysis (1)
fail = 0.5
Fixed t=1.010-5
Effective plastic strain is monotonically increased w/o diffusion !
192
Metal shearing analysis (2)
Part
Burr Geometry
Scrap
SPG simulation
Experiment
Major applications in blanking, bolt/rivet shearing, AHSS trimming
193
Hole Punching in Metal (1)

Time-Punch Force
Effective Stress Contour (full model)
Pressure Contour (Cross-section view)
194
Effective Stress Contour (Cross-section view)
t=7.69~8.16 10-8
195
196
Self-pierce Riveting Analysis (1)
Rigid
Constan
tv
Deformable tool
(FEM)
Solid plates
(SPG)
Rivet model
EPS
197
EPS
Upper plate
Lower plate
198
Von Mises stress
Upper plate
Lower plate
199
Flow Drill Screw (FDS) Analysis (1)

stage 1
Head with internal of
external drive system
Thread forming zone

Flow drilling zone
stage 2
Stage 1
Courtesy of Ford Motor Co.
EPS
200

Stage 1
EPS
von Mises stress
201

Stage 2
EPS
202
Pull Out Test in FDS

Constant v
Stage II
Stage I
Rigid
Solid plates
(SPG)
Constant v
v decreases by x10
Rotational speed increases by x4
Rotation creates material vertical transportation in the direction opposite to

the prescribed v
Start unscrewing
203
FDS with Thermal Effect (1)
Temperature
204
EPS
Max: 7.0
With thermal
Max: 26.1
Without thermal
205
Contact force
Temperature
at final stage
206
Metal Tearing (1)
Multiple cracks
Sharing nodes with FEM
207
Metal Tearing (2)
von Mises stress
EPS
208
SPG for Composites (1)

Works for several composites
Does not erode materials
MAT_COMPOSITE_MSC with AOPT=2 and a 45 material angle
FEM
SPG
209
SPG for Composites (2)
Under development features (Beta version)
MAT_COMPOSITE_FAILURE
Courtesy of DynaS+
MAT_CONCRETE_DAMAGE
210
SPG for Water Sloshing

Supports several EOS materials
Uses regular NODE_TO_SURFACE contact
211
XFEM and Meshfree for Brittle Failure
Tmax
Initially-elastic
Initially-rigid
T = T ( )
GIC =
max
T ( )d
max
: displacement jump
The potential crack propagation plane is
idealized as a cohesive zone and is
assumed to support a traction field T.
The mechanical response of the
cohesive interface is described through a
constitutive law relating the traction field T
with a separation parameter.
212
Initially-Elastic and Initially-rigid Cohesive Law
Initially-elastic
Cohesive Interface
Element
Initially-rigid
EFG or XFEM
Size: 5.0 X 10.0

Displacement-control (1.2 in 0.3 sec)
Tvergaard-Law I cohesive law
Tmax = 5.0
= 1.0
n = 1.0
t = 1.0
cr = 1. / 3.
= 1.0e 6; E = 167.0; = 0.3
213
Interface element, EFG and XFEM
Cohesive zones
failed up to here
EFG
XFEM
CFEM
Initial Rigid
CFEM
Initial Elastic
214
Overview on Meshfree Method and XFEM

Extended FEM : Level set + Local PU [Belytschko et al. 2000]
( )uI + I ( X )qI
uh ( X ) = FEM
I
I =1
I w
FEM
( )(H ( f ( X )) H ( f ( X I ))) fully cut element
I
I ( X ) = FEM
*
I ( )(H ( f ( X )) H ( f ( X I ))) contain crack tip
f =0
Meshfree Method: MLS + Visibility [Belytschko et al. 1996]

uh ( X ) = I ( X )uI
0+
I =1
I = P ( X ) A( X ) P ( X I )W ( X X I , h )
T
A( X ) = P ( X J )P T ( X J )W ( X X I , h )
X ( )
Extended Meshfree Method: MLS + Local PU + Visibility [Rabczul and Zi, 2006]
uh ( X ) = I ( X )uI +
I =1
( X )H ( f ( X ))a
J
J f
( X ) Q ( X )b
K
J t
KJ

Q = [ r cos , r sin , r cos sin( ), r sin sin( )]
2
2
2
2
215
EFG Brittle Failure Analysis in Solids

Meshfree Brittle Failure Analysis and Its Assumptions
Is a discrete approach.
Crack initiation and propagation are governed by cohesive law.
Crack currently is cell-by-cell propagation and is defined by visibility.
Minimized mesh sensitivity and orientation effects.
No material fusion.
Applied mainly to quasi-brittle materials and some ductile materials.
Current Pratice
Apply mainly to brittle and semi-brittle materials (Mode-I).

Only for 4-noded background cell.
Currently SMP only.
216
Discrete Cracks
Crack in Meshfree: Visibility Criterion [Belytschko et al.1996]
Intrinsic (Implicit crack) : no additional unknowns
(
)
(
(
)
)
(
(
)
)
x ( ) = FEM
X
X
u
X
u
+
I
I
J
J
J
J
2 J 0+
I =1
J 0
0+
X ( )
( ) + 1 u J ( X ) + u J ( X ) X ( )
x ( ) 2
FEM
I
= XI
J
J
2 J 0+
X
X
I =1
J 0
Initially-rigid Cohesive Law: Redefined Displacement Jump (Sam, Papoulia and Vavasis 2005)
2
= un ( + ) + 2 ut ( + )
0n
n
0t
t

Tefs T + Tt 2 = Tmax

1 un Tmax
Tn =
and
n 1 cr
2
n
Tt =
1 ut Tmax
t 1 cr
= cr = 0.005 = cr = 0.01
217
Computation Procedures
W kin = W int W ext + W coh u ( X ) u0
1. Representation of Cracks
W kin = u 0u&&d0
0
u
: Pd0
0 X
W int =
2. Cohesive Law
Crack initiation/propagation
W ext = u 0bd0 + u t 0dt0

t0
W coh = [[u]] c dc
c
kin
= f
int
ext
+f
3. Branching/Multiple cracks
coh
&&
f ekin = e 0 N T NH ((1) e f ( X ))d e0 u
0
int
e
= e B H ((1) f ( X ))d
T
4. State Variables Transfer
e
0
f eext = e 0 N T bH ((1) e f ( X ))d e0 + e N T tH ((1) e f ( X ))d0e,t

0
0 ,t
5. Numerical Integration
f ecoh = (1) e e N T c n0 d0e,t

0 ,t
218
*MAT_COHESIVE_TH (1)
Tvergaard and Hutchinson [1992] cohesive model
Card 1
Variable
MID
RO
ROFLAG
INTFAIL
SIGMAX
NLS
TLS
Type
INTFAIL: Number of integration points required for the cohesive element to be deleted
= 0 : no deletion
= 1 : deletion
219
*MAT_COHESIVE_TH (2)
Card 1
Variable
MID
RO
ROFLAG
INTFAIL
SIGMAX
NLS
TLS
Type
SIGMAX: (maximum traction)

NLS: Critical displacement jump in normal direction
*MAT_COHESIVE_TH
100, , ,1, 330.0, 0.0001
= cr = 0.005 = cr = 0.01
220
Card 2
Variable
DX
DY
DZ
ISPLINE
IDILA
IEBT
IDIM
TOLDEF
Type
Default
1.01
1.01
1.01
-1
0.01

EQ.-1: Stabilized EFG method (apply to 8-noded, 6-noded and combination)
221
Material identification for EFG fracture analysis
Card 3
Variable
IGL
STIME
IKEN
SF
MID
IBR
DS
ECUT
Type
Default
1.e+20
0.0
1.01
0.1
MID Material identification used for EFG fracture analysis

*SECTION_SOLID_EFG
5, 41
1.1, 1.1, 1.1, , ,4,-2,
,,,, 100, 1, 2.0, 0.2
*MAT_COHESIVE_TH
100,1.0e-07, ,1, 330.0, 0.0001
Currently, only mode-I is considered and MAT_COHESIVE _TH is available.

Only 4-noded TET integration is implemented.
222
Strain filter in facture analysis
Card 3
Variable
IGL
STIME
IKEN
SF
MID
IBR
DS
ECUT
Type
Default
1.e+20
0.0
1.01
0.1
SF : For the stress-based cohesive law, a strain filter is recommended as an extra

condition for the crack initiation under slow loading. Only when the strain reaches to this
value, the crack is allowed to initiate. However, under high dynamic loading, this value
should be null or very small to allow the appearance of Spall fracture.
*SECTION_SOLID_EFG
*SECTION_SOLID_EFG
5, 41
5, 41
1.1, 1.1, 1.1, , ,4,-2,
1.1, 1.1, 1.1, , ,4,-2,
,,, 0.0, 100, 1, 2.0, 0.01
,,, 1.5, 100, 1, 2.0, 0.01
*MAT_ELASTIC
*MAT_PLASTIC_KINEMATIC
Brittle
Ductile
223
Card 3
Variable
IGL
STIME
IKEN
SF
MID
IBR
DS
ECUT
Type
Default
1.e+20
0.0
1.01
0.1
IBR = 1: No branching is allowed.

= 2: Branching is allowed.
DS : Normalized support defined for computing the displacement jump in fracture analysis
ECUT : 0.0~0.5
Define the minimum edge cut in an integration cell in fracture analysis.
0.1
crack
224
Minimization of Mesh Size Effect in Mode-I Failure Test

Failure is limited
in this area
Tn
D = cr = 0.005
Coarse elements
D = cr = 0.01
Fine elements
D=0.01
D=0.005
225
Kalthoff Plate Crack Propagation (1)
100mm
75mm
100mm
y
x
100mm
v0
25mm
50mm
= 8000 kg/m 3 , E = 190 GPa , = 0 .30

G Ic = 1.213 10 4 N/m, max = 5.245 10 5 m
v 0 = 16.5 m/s
226
Kalthoff Plate Crack Propagation (2)
Maximum Principle Stress
Total Displacement
227
EFG 3D Edge-cracked Plate Under Loading

d
101 x 31 x 6 nodes
Elastic
EFG Fracture
Linear Cohesive Law
Explicit analysis
Front
Resultant Displacement Contour
Back
228
Rigid Ball Impact on Concrete Plate

Time-velocity of the rigid ball
Elastic
EFG Fracture
Linear Cohesive Law
Explicit analysis
Progressive Crack Propagation
229
Windshield Under Impact (1)

Front
101 x 101 x 4 nodes
Elastic + Rubber
EFG Fracture
Linear Cohesive Law
Explicit analysis
Back
Failure Contour
230
Windshield Under Impact (2)
Rigid ball
Metal ball
231
XFEM Failure Analysis in 2D and Shells (1)
XFEM Failure Analysis and Its Assumptions
Is a discrete approach.
Brittle and ductile fractures are treated differently.
Crack initiation and propagation are governed by cohesive law.
Crack currently is cell-by-cell propagation and is defined by level set.
Preferred in a pre-cracked domain.
Minimized mesh sensitivity and orientation effects.
No material fusion.
Applied mainly to quasi-brittle materials and ductile materials.
Newly added strain regularization scheme to minimize mesh sensitivity.
232
XFEM Failure Analysis in 2D and Shells (2)

Extended FEM : Level set + Local PU [Belytschko et al. 2000]
( )uI + I ( X )qI
uh ( X ) = FEM
I
I =1
I w
FEM
( )(H ( f ( X )) H ( f ( X I ))) fully cut element
I
I ( X ) = FEM
*
I ( )(H ( f ( X )) H ( f ( X I ))) contain crack tip
f =0
Discontinuity defined by two implicit functions: f (X) and g(X)

Signed distance function
f ( X ) = min X X sign[n (X X )]
X
Discontinuity
X 0 if f (X ) = 0 and g (X, t ) > 0

Define implicit functions locally
f (X ) = f N (X )
I
g ( X, t ) replaced by index for elementwise crack propagation
233
*SECTION_SHELL_XFEM (1)
Card 1
Variable
SECID
ELFORM
Type
ELFORM EQ.52: Plane stain, using FEM #13 as base element

EQ.54: Shell, using FEM#16 (default) as base element
Card 2 define for the XFEM option

Variable
MCID
BASELM
DOMINT
FAILCR
PROPCR
LPRINT
Type
*SECTION_SHELL_XFEM
5, 52
100, 13, 0, 1,
*MAT_COHESIVE_TH
100,
234
*SECTION_SHELL_XFEM (2)
Card 3 define for the XFEM option
Variable
MCID
BASELM
DOMINT
FAOLCR
PROPCR
FS
LS
NC
Type
BASELM: Base element for shell

EQ. 2: Belytschko-Lin-Tsay shell.
EQ.16: Assumed stress, assumed strain shell.
DOMINT: Option for domain integration in XFEM:
EQ.0: Phantom element integration.
EQ.1: Subdomain integration with triangular local boundary integration
(mainly for plane strain).
FAILCR: Option for different failure criteria:
EQ. 1: Maximum tensile stress.
EQ. 2: Maximum shear stress.
EQ.-1: Effective plastic strain.
FS: Failure value.
LS: Length scale for strain regularization
235
*BOUNDARY_PRECRACK
Card 1
Variable
PID
CTYPE
NP
Type
PID: Part ID to define a pre-crack in.

CTYPE: Pre-crack type
EQ.1: straight line.
NP: Number of points defining the pre-crack.
(10, 10, 0)
(0, 5, 0)
Card 2
Variable
Type
X,Y,Z: Coordinates of the points defining the pre-crack.
*BOUNDARY_PRECRACK
100, , 2
0, 5, 0
10, 10, 0
236
Strain Regularization
NP
p = ia ip
i =1
NP is number of integration points

within regularization zone
ia
is the meshfree shape function

with kernel size a
a is the length scale of the
regularization zone, a material
constant
237
Mode I Test in Ductile Fracture (1)
Regular mesh 1
(1840 elements)
FEM Erosion
Regular mesh 2
(7360 elements)
Regular mesh 3
(29440 elements)
XFEM
238
Shell element type 2
Standard effective plastic strain criterion (FS=0.15)
239
Regularized effective plastic strain criterion (FS=0.15, LS=0.5)
240
Edge-cracked Plate under Impulsive Loading (1)
100mm
75mm
100mm
25mm
50mm
y
x
1
v0
2
100mm
v0
{
symmetry
= 8000 kg/m3, E = 190 GPa, = 0.30

4
50x50 elements
GIc = 2.213 10 N/m, max = 5.245 10 m

v 0 = 16.5 m/s
241
Edge-cracked Plate under Impulsive Loading (2)
242
XFEM Thin Cylinder Shell Pulling
Pre-crack
Clamped edge
Rigid diaphragms
1860 elements
Failure Contour
Elastic-plastic Shell
XFEM Fracture
Linear Cohesive Law
Explicit analysis
243
XFEM Thin Cylinder Shell Twisting
Constant w
Failure Indicator
Pre-crack
Effective Plastic
Strain
Fixed
*MAT_POWER_LAW_PLASTICITY
FS=0.15
244
XFEM Fracture of Container Under Pressure
245
XFEM Three-point Bending
Damage Indicator
246
XFEM Three-point Bending with a Hole
Damage Indicator
247
A Summary of Current LS-DYNA Peridynamics

Current version is mainly for brittle fracture analysis.
Is a discontinuous Galerkin (DG) approach with bond-based peridynamics theory.
Failure is based on critical energy released rate.
No element deletion is needed to advance the cracks.
Branching of the cracks is an outcome of the DG approach.
Self-contact between cracks is possible but CPU time consuming.
Accommodates for non-uniform mesh and allow the direct enforcement of
boundary conditions and constraints.
Available in R9.0.
Is mainly for windshield or plastic panels damage analysis in crashworthiness.
248
Peridynamics Theory (1)
Courtesy of Dr. Steward Silling at Sandia Nat. Lab.
249

Strong Form for Explicit Dynamics Analysis
Bond
Horizon
f = cs
+
+
, s=
250

Peridynamics horizon
Similar to nodal support size or domain of

influence in meshfree methods.
Box function is used in most bond-based
peridynamics.
Bond
Conservation of linear momentum

Horizon
Newtons third law

Like MD !
Conservation of angular momentum
Suggested in solid mechanics.

251

Autonomous and multiple cracks growth
A bond breaks when

its stretch exceeding a
critical value.
When a bond breaks, its load is

shifted to its neighbors, leading to
progress failure in brittle material.
252
Discontinuous Galerkin Weak Form-Explicit Dynamics (1)
The weak form of the governing equation:
253

The domain is discretized by elements.
The kinematic quantities are interpolated in each element as FEM:
The Galerkin weak form can be derived as:
This two integrations can be discretized by the summation of

Gaussian points in each element:
254

All elements do not share nodes. It implies that the discretization is piecewise continuous, i.e., the crack is allowed to go along the edges of elements.
255
SECTION_SOLID_PERI
Card 1
Variable
SECID
ELFORM
Type
ELFORM EQ.48: Peridynamic formulation for 4, 6, or 8 -noded elements
Card 2
Variable
DR
PTYPE
Type
Default
1.01
DR
.
. is recommend to determine the horizon size
based on the characteristic length of element
PTYPE
EQ.1: bond based formula (currently implemented)

EQ.2: state based formula
256
MAT_ELASTIC_PERI
Card 1
Variable
MID
RO
Type
Default
1.0E28
MID: User defined material ID

RO: mass density
E: Classic elastic modulus E.
G: Fracture energy release rate G
257
Horizon (1)
The regular horizon is defined as
Concave shape
Horizon Searching in Concave shape algorithm
258
Horizon (2)
The mesh based horizon searching
Horizon can not access the detached element
Advantage
The concave shape is represented according the
grid connectivity.
Disadvantage
The adjustment of the horizon size is limited for
deficient candidates: the surrounding gaussion
points.
Characteristic length: Diagonal length
259
Horizon (3)
Recall the Galerkin solution space:
In a FEM solution, the inner force is calculated:

r=dx
Here, A loops in a boundary element.

In the case of Peridynamics, the inner force is calculated:
Here, A loops in a boundary element, P loops through the boundary element and
adjacent element.
260
Horizon (4)
Boundary treatment-like FEM!

Continuous problem:
Continuous Galerkin
Discontinuous Galerkin
261
Horizon (5)
Discontinuous problem:
Continuous Galerkin
Discontinuous Galerkin
262
Material Constants (1)

Elastic modulus E
The classic elastic energy density under small deformation condition:
A stretched bond has the micro energy:
Solve this linear equations, each bond has its own micro modulus based
on the material constant E and its horizon.
263

Elastic modulus E
With calibrated c
With constant c
1D line with displacement loading

on one side and fixed another side
264

Energy Release Rate G
To form a crack surface, all bonds crossing that surface must be broken.
The energy required to break these

bonds is related to fracture mechanics
concept: energy release rate, G:
The critical bond stretch of each bond is calculated

from Energy Release Rate G and its horizon.
265

Energy Release Rate G
266
Wave Propagation 3D Bar with Initial Velocity
Displacement history at middle point
267
Cantilever Beam (1)
Thickness: 1
Thickness: 0.3
Thickness: 0.2
268
Cantilever Beam (2)
Thickness: 0.1
Thickness: 0.05
269
Mode I Crack
38400
elements
600
elements
Crack velocity
4800
elements
270
Kalthoff-Winkler Problem
30800 elements
52272 elements
Element: 97608
271
Three Point Bending of Concrete Beam (1)
272
Three Point Bending of Concrete Beam (2)
Experimental
results
=0
=00
=0.5
=210
=0.7
=290
273
3D Glass-PC-Glass System (1)
Essential boundary
Fixed in vertical direction
Rigid body
E=211Gpa,
=2g/cm2
V=30 m/s
Polycarbonate (FE)
Elastic material
E=2Gpa
=0.25
Soda-Lime glass
Peridynamic model
E=72Gpa
G=8J/m2
274
Back View
Top View
Damage patterns
275
Top View
Back View
Damage evolution
276
The cone shape main damage zone in glass indentation

277
Contact force
Ball velocity
278
Windshield Impact (1)

Glass layers (Peridynamic Model, MAT_ELASTIC_PERI)
Vinyl layer, FEM Model,

MAT_PIECEWISE_LINEAR_PLASTICITY
Interface of vinyl and glasses:
CONTACT_TIED_SURFACE_TO_SURFACE_OFFSET
279

Damage pattern (top)
Damage pattern (bottom)
EPS
Vynl layer
280

von Mises stress on vinyl interlayer
Top view with ball
Bottom view
281
Conclusions
Make sure the numerical method is physical-based and

numerical-convergent.
Attack the Error Not the Feature !
-Thomas JR Hughes
Thank you!
282

AdvFem Meshfree 2016class

Uploaded by

Copyright:

Available Formats

AdvFem Meshfree 2016class

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

AdvFem Meshfree 2016class

Uploaded by

Copyright:

Available Formats

What are some of the different methods discussed for solid and structural analysis?

What are some of the different methods discussed for solid and structural analysis?

What are some applications of these methods discussed in manufacturing?

What are some applications of these methods discussed in manufacturing?

LS-DYNA Advanced FEM and Meshfree Methods

for Solid and Structural Analyses

Download links of class material:

Methods for Solid and Structural Analyses in LS-DYNA

Advanced FEM/Meshfree methods in LS-DYNA for

EFG for solids (foam)

ME-FEM for solids (rubber)

Periodic Unit Cell method Poly-crystal DDD (single crystal)

LS-DYNA Current Practice for Manufacturing Applications

EFG solid ELFORM=41,42 (*SECTION_SOLID_EFG)

o Spring back analysis

Machining and Jointing: Adaptivity & SPG

Adaptivity: No material failure; Non-physical based Material separation

Adaptivity: No material fusion

1. Introduction and Overview

What is the Meshfree/Meshless/Particle Method ?

Meshfree shape function

Meshfree/Particle Application Range

Classification of Particle Methods

Implicit Meshfree Galerkin

History and Research Trend

Smooth Particle Hydrodynamics (SPH) [Monaghan 1977]

Meshfree Galerkin Method

Meshfree least square method,

Disadvantages of Using Meshfree Method

Tensile Instability in SPH

Low Energy Mode in SPH

Computational Trade-off in Mesfree/Particle Methods

High CPU and Memory

Better computer capacity

Accurate integration rule

Overview on Smooth Particle Hydrodynamics ( SPH )

Liquid Shifting Test

Overview on Element Free Galerkin Method (EFG)

u (x) = wa[n] (x xI ) u( xI )xI

Moving Least-Squares approximation

More neighboring nodes

Overview on Smoothed Particle Galerkin Method (SPG)

A MA U&& = A-T f ext f int

More neighboring nodes

SPH Lagrangian Kernel and Eulerian Kernel

SPG Updated Lagrangian Kernel and Eulerian Kernel

No zero-energy mode; Time step does not drop; Tensile instability

Spatial Domain Integration

SPH and SPG nodal

Usually in solid and structure !

EFG stress point

Comparison of SPH, EFG and SPG

Explicit Lagrangian Collocative

2D, 3D and shell

Handle free surface fluid flow

Handle material failure

Difficult Boundary condition

When Should EFG be Considered in the Analysis ? (1)

Result: EFG is same as FEM