1 Aerospace & Mechanical Engineering Department Continuum Mechanics & Thermomechanics Tel:...
-
Upload
georgia-simon -
Category
Documents
-
view
219 -
download
0
Transcript of 1 Aerospace & Mechanical Engineering Department Continuum Mechanics & Thermomechanics Tel:...
1
Aerospace & Mechanical Engineering DepartmentContinuum Mechanics & Thermomechanics
Tel: +32-(0)4-366-48-26 Chemin des chevreuils 1 Fax: +32-(0)4-366-91-41 B-4000 Liège - Belgiume-mail: [email protected] www.ltas-mct.ulg.ac.be/who/noels/
A new energy-dissipative momentum-Conserving time integration scheme using the variational
updates framework
L. Noels, L. Stainier, J.-P. PonthotUniversity of Liège, LTAS-MCT, Belgium
Computational Methods in Structural Dynamics and Earthquake Engineering 2007
13 – 16 June 2007, Rethymno, Crete, Greece
2
Scope of the presentation
1. Scientific motivations
2. Conserving scheme in the non-linear range
3. New formulation using the variational updates approach
4. Numerical examples
5. Conclusions
3
1. Scientific motivationsDynamic simulations
Spatial discretization into finite elements Temporal integration of the balance equations 2 methods
– Explicit method
• Non iterative
• Small memory requirement
• Conditionally stable (small time step)
– Implicit method
• Iterative
• Larger memory requirement
• Unconditionally stable (large time step)
11intext1
1 ,deductionionapproximat
nnnnn
n
n
n
xxFFMx
x
x
x
),,,,(
),,,,(iterationsionextrapolat
111
111
extint
1
1
1
nnnnnn
nnnnnn
n
n
n
n
n
n
xxxxxfx
xxxxxfx
FFxM
x
x
x
x
x
x
Very fast dynamics
Slower dynamics
extint FFxM
4
1. Scientific motivationsNon-linear numerical example
Traditional methods can be unstable in the non-linear range Example (Armero & Romero [CMAME, 1999])
– Mass spring system with geometrical non-linearity– 1 revolution ~ 4s
Conservation laws are not satisfied leading to divergence, and motivating the development of conserving algorithms
Newmark implicit scheme (no numerical damping)
HHT implicit scheme
(numerical damping ∞= 0.9)
t=1s t=1.5s t=1s t=1.5s
-20 0 20-20
-10
0
10
20
X [m]
Y [
m]
Newmark; t = 1 s
-20 0 20-20
-10
0
10
20
X [m]
Y [
m]
Newmark; t = 1.5 s
-20 0 20-20
-10
0
10
20
X [m]
Y [
m]
HHT; t = 1 s
-20 0 20-20
-10
0
10
20
X [m]
Y [
m]
HHT; t = 1.5 s
5
2. Conserving scheme in the non-linear range Conservation laws
Conservation of linear momentum (Newton’s law )
– Time discretization: &
Conservation of angular momentum ( )
– Time discretization: &
Conservation of energy ( )
– Time discretization:
&
nodes
ext2/12/1
nodes nodes11 nnnnnn FxtxMxxMx
nodes
1ext
2/1intintint
1nodes
11 ][2
1
2
1nnnnnnnnn xxFDWWxxMxxM
nodes
ext2/1
nodes1 nnn FtxMxM
Wint: internal energy; Wext: external energy; Dint: dissipation (plasticity …)
extFt
xM
extFxt
xMx
intextint DWt
Wt
Kt
0nodes
int2/1 nF
0nodes
int2/12/1 nn Fx
nodes
1int
2/1intintint
1 ][ nnnnn xxFDWW
6
First Energy-Momentum Conserving Algorithm (EMCA) (Simo and Tarnow [ZAMP, 1992])
– Based on the mid point scheme:
– With and designed to verify conserving equations
– Example of the mass-spring system
2. Conserving scheme in the non-linear range EMCA
int2/1
ext2/1
1
nnnn FF
t
xxM
t
xxxx nnnn
11
2
nnnn
nnn xx
ll
lVlVF
122
1
1int2/1
)()(
ext2/1nF
EMCA, t=1s EMCA, t=1.5s
int2/1nF
-20 0 20-20
-10
0
10
20
X [m]
Y [
m]
EMCA; t = 1 s
-20 0 20-20
-10
0
10
20
X [m]
Y [
m]
EMCA; t = 1.5 s
0 100 200 300 400 500-250
-200
-150
-100
-50
0
Time [s]
Ang
ular
mom
entu
m [
kg m
2 /s]
EMCA; t = 1 sNewmark; t = 1 s
7
Introduction of numerical dissipation: Energy-Dissipative Momentum Conserving scheme (EDMC) (Armero & Romero [CMAME, 2001])
– Introduction of new terms in the EMCA:
– Same internal and external forces as in the EMCA
– With and designed to achieve positive numerical dissipation without
spectral bifurcation
– Example of the mass-spring system for first order accurate EDMC
2. Conserving scheme in the non-linear range EDMC
int2/1nF
ext
2/1nF
diss2/1nF
diss2/1
int2/1
ext2/1
1
nnnnn FFF
t
xxM
t
xxx
xx nnn
nn
1diss
11
2
diss1nx
nodes
1ext
2/1numintintint
1nodes
11 ][2
1
2
1nnnnnnnnn xxFDDWWxxMxxM
2
1
1
1diss1
nn
nn
nn
n
xx
xx
xxx
2
1
1
21''
diss2/1
1 nnll
nn
nlnl
n
xx
ll
VF nn
0 100 200 300 400 500-250
-200
-150
-100
-50
0
Time [s]
Ang
ular
mom
entu
m [
kg m
2 /s]
EDMC; = 0.9; t = 1 s
HHT; = 0.9; t = 1 s
0 100 200 300 400 5005
10
15
Time [s]
Spr
ing
leng
th [
m]
EDMC; = 0.9; t = 1 s
HHT; = 0.9; t = 1 s
9
2. Conserving scheme in the non-linear rangeFormulations in the literature
Elastic formulation– Saint Venant-Kirchhoff hyperelastic model (Simo et al. [ZAMP, 1992])
– General formulation for hyperelasticity (stress derived from a potential V)
(Gonzalez [CMAME, 2000])
with
– Classical formulation
Penalty contact formulation (Armero & Petöcz [CMAME, 1998-1999])
0xD
2
1
2
11
12/1
:2
22
2
nn
nnnn
nnn
O
VVV
V
CC
CC
CCC
CCC
CC
CS
F: deformation gradient; C: right Cauchy-Green strain; V: potential; : shape functions;
0
2int
V
dVDV
F
CF
)()( 2/12/1112/1dd
1 nnnnnnnnn uyuyxxngg
2/1dd
1
dd1cont
2/1
n
nn
nnn n
gg
gVgVF
0
2/11int
2/1 2V
nnn
n dVDF
SFF
10
2. Conserving scheme in the non-linear rangeFormulations in the literature
Elasto-plastic materials– Hyperelasticity with elasto-plastic behavior (Meng & Laursen [CMAME, 2001])
• energy dissipation of the algorithm corresponds to the internal dissipation of the material
• Isotropic hardening only
– Hyperelasticity with elasto-plastic behavior (Armero [CMAME, 2006])
• Energy dissipation from the internal forces corresponds to the plastic dissipation
• Modification of the radial return mapping
• Yield criterion satisfied at the end of the time-step
– Hypoelastic formulation• Stress obtained incrementally from a hardening law
• No possible definition of an internal potential!
• Idea: the internal forces are established to be consistent on a loading/unloading cycle
• Assumption made on the Hooke tensor
• Energy dissipation from the internal forces corresponds to the plastic dissipation
12
3. Variational updates approachPurpose of the work
Development of a general approach leading to conserving/dissipating algorithms for any material behavior!
What we want– No assumption on the material behavior
– Material model unchanged compared to the standard approach:• From a given strain tensor, the outputs of the model are the same
• Use of the same material libraries
– Expression of the internal forces for the conserving algorithm remains the same as in the elastic case
– Yield criterion satisfied at the end of the time step
Solution derives from the variational formulation of visco-plastic updates [Ortiz & Stainier, CMAME 1999] – Definition of an energy, even for complex material behaviors
Introduction of numerical dissipation via Deff
nnn
n
DCC
CS ,2 1
1
eff
1
C: right Cauchy-Green strain; S: second Piola-Kirchhoff stress; Deff: incremental potential
13
3. Variational updates approachUse of an incremental potential
Description of the variational updates for elasto-plasticity– Multiplicative plasticity
– Plastic flow
– Functional increment
– Effective potential:• Minimization with respect to pl satisfies yield criterion
• Minimization with respect to N satisfies radial return mapping
– Usual stress derivation:
Introduction of numerical dissipation
– Modification of potential
– Dissipation function
plelFFF
plplpl1 )exp( nn FNF
ttWW
WWD
nnnn
nnnnnnnnnn
plpl1*plplpl
1pl
pl1plelpl1
1pl11
elplpl11 ),(),(),,,,(
NFFNFFNFF
),,,,(min pl1
pl11
,1
eff
pl1
NFFFN
nnnnn DDn
1
1eff
1 2
n
nn
D
C
FS
Wel: reversible potential; Wpl: dissipation by plasticity; : dissipation by viscosity
N: plastic flow direction
ttDD nn
nnnnnn
elel1disspl
1pl
11,
1eff ),,,,(min
pl1
CCNFFF
N
elelel
el2el
elel1diss ::
2C
CCC
CC
V
ttnn
14
3. Variational updates approachUse of an incremental potential
Internal forces directly obtained from Gonzalez elastic formulation [CMAME 2000]
with
Dissipation forces are part of
C
C
CCC
CCC
CC
CS
2
1eff
1eff
1eff
2/1
:2
,2
22
nnnn
nnn
DD
D
2
1
2
11
12/1
:2
22
2
nn
nnnn
nnn
O
VVV
V
CC
CC
CCC
CCC
CC
CS
0
2/11int
2/1 2V
nnn
n dVDF
SFF
int2/1nF
diss
2/1nF
15
3. Variational updates approachUse of an incremental potential
Modification to quasi-incompressible form– Definition of a mean volumetric deformation on the element e.
– Definition of distortion gradient
– Split of the stress tensor
– Volume part
– Deviatoric part
with , and resulting from the minimization of
dGSS 2/1dev
2/12/1 2 nnn p
CC
CC
CdG
2
2/1
2/1:n
n
JJJ
e
e
ee
en
ene
nen
e
en
en
n
DD
Dp
2
1eff
1eff
1eff
2/1
22,
22
C
C
CCC
CC
C
CS
2
el2/1
effelel
1eff
el2/1
effdev
2/1
ˆˆ,ˆ
2ˆ
2e
DD
D
nnn
e
nn
FF 3/1ˆ J
T
nenn
nne
n DJ
D pl2/1
elel2/1
eff1pl
2/13/22/1
el2/1
eff
ˆ
ˆ,ˆDEV
ˆF
C
CCF
C
C
n
n
nnnD pl2/1pl1 ,,,,2
CNCCel
2/1ˆ
nCpl2/1nF2/1nJ
16
3. Variational updates approachUse of an incremental potential
Properties:– 2 material configurations computed
• Mid configuration trough
• Final configuration trough
– Material model unchanged
– Yield criterion verified at configuration n+1
– Conservation of linear and angular momentum
– Numerical dissipation of energy (first order accurate)
nnD CC ,1eff
21
effnnD CC
C
dVttWWWWxxFV
nnnnnodes
nnn
0
diss*plpl1
elel11
int2/1 ][
dVV
xxmxxxMdVtDV
nnnnn
V
00
elelel
el2el2
1diss
11dissnum ::
22C
CCC
nodes
1ext
2/1numintintint
1nodes
11 ][2
1
2
1nnnnnnnnn xxFDDWWxxMxxM
dVWWV
0
elint
dVtWWDV
nn
0
*plpl1
int
17
4. Numerical ExamplesSimulation of a tumbling beam
Tumbling beam– Initial symmetrical loads (t < 10s)– Elasto-perfectly-plastic hyperelastic material
EMCA, t=0.5 s EDMC, t=0.5s, ∞= 0.7
18
4. Numerical ExamplesSimulation of a tumbling beam
Time evolution of the results
0 50 100 150 2000
5
10
15
20
25
Time [s]
Ene
rgy
[J]
Dint
W int
Dint + W int
Fint . x
0 50 100 150 2000
20
40
60
Time [s]
Ene
rgy
[J]
K Dint + W int
Dint + W int + K Fext . x
EMCA, t=0.5 s EDMC, t=0.5s, ∞= 0.7
0 50 100 150 2000
5
10
15
20
25
Time [s]
Ene
rgy
[J]
Dint
W int
Dint + W int
(Fint. x-V t diss)
0 50 100 150 2000
20
40
60
Time [s]
Ene
rgy
[J]
K
Dint + W int
Dint + W int + K Fext . x
).( 0V
dV txF dissint
19
4. Numerical ExamplesSimulation of a tumbling beam
Influence of time step size (∞= 0.7)
Influence of ∞ (t = 0.5)
0 50 100 150 2000
20
40
60
Time [s]
Din
t + W
int +
K [
J]
EDMC; t = 0.125 sEDMC; t = 0.25 sEDMC; t = 0.5 sEDMC; t = 0.75 sNewmark; t = 0.75 s
0 50 100 150 2000
5
10
15
20
25
Time [s]
Din
t [J]
EDMC; t = 0.125 sEDMC; t = 0.25 sEDMC; t = 0.5 sEDMC; t = 0.75 sNewmark; t = 0.75 s
0 50 100 150 2000
20
40
60
Time [s]
Din
t + W
int +
K [
J]
EDMC; = 0.95
EDMC; = 0.9
EDMC; = 0.8
EDMC; = 0.7
Newmark; = 0.7
0 50 100 150 2000
5
10
15
20
25
Time [s]
Din
t [J]
EDMC; = 0.95
EDMC; = 0.9
EDMC; = 0.8
EDMC; = 0.7
Newmark; = 0.7
21
4. Numerical ExamplesImpact of two 2D-cylinders
Two cylinders (Meng & Laursen)
– Left one has an initial velocity (initial kinetic energy 14J)
– Elasto-perfectly-plastic hyperelastic material
22
4. Numerical ExamplesImpact of two 2D-cylinders
Results comparison at the end of the simulation
0 0.099 0.197 0.296 0.394Equivalent plastic strain
EDMC, t=1.5 ms, ∞= 0.7
0 0.098 0.195 0.293 0.390Equivalent plastic strain
EDMC, t=15 ms, ∞= 0.7
0 0.098 0.196 0.293 0.391Equivalent plastic strain
CH, t=1.5 ms, ∞= 0.7
0 0.189 0.379 0.568 0.758
CH, t=15 ms, ∞= 0.7
Equivalent plastic strain
23
4. Numerical ExamplesImpact of two 2D-cylinders
Results evolution comparison
0 1 2 3 40
5
10
15
Time [s]
Din
t [J]
EDMC; t = 1.5 msEDMC; t = 15 msCH; t = 1.5 msCH; t = 15 ms
0 1 2 3 40
5
10
15
Time [s]
K [
J]
EDMC; t = 1.5 msEDMC; t = 15 msCH; t = 1.5 msCH; t = 15 ms
0 1 2 3 410
12
14
16
18
20
Time [s]
Din
t +W
int +
K [
J]
EDMC; t = 1.5 msEDMC; t = 15 msCH; t = 1.5 msCH; t = 15 ms
0 1 2 3 4
0
1
2
3
4
Time [s]
F
ext .
x
[J]
EDMC; t = 1.5 msEDMC; t = 15 msCH; t = 1.5 msCH; t = 15 ms
24
4. Numerical ExamplesImpact of a square tube
Impact (50 m/s) of a square tubes on a rigid punch– Saturated hardening law (aluminum)– Masses added at tube extremities (to simulate structure inertia)
EDMC, ∞= 0.8, no point mass EDMC, ∞= 0.8, point masses = 100% of tube mass
25
4. Numerical ExamplesImpact of a square tube
Time evolution of the results
0 0.5 1 1.5 20
5
10
15
20
Time [ms]
Ene
rgy
[J]
Dint
W int
K
0 0.5 1 1.5 2-0.2
-0.15
-0.1
-0.05
0
Time [ms]
F
ext .
x
[J]
0 0.5 1-0.2
-0.15
-0.1
-0.05
0
Time [ms]
F
ext .
x
[J]
EDMC, ∞= 0.8, no point mass EDMC, ∞= 0.8, point masses = 100% of tube mass
0 0.5 10
2
4
6
8
10
12
Time [ms]
Ene
rgy
[J]
Dint
W int
K
26
Developed a visco-elastic formulation leading to a conserving time integration scheme
Use of the variational updates formulation:– The formulation derives from a energy potential
– The formulation is general for any material behavior
– The formulation includes numerical dissipation
The internal forces expression remains the same as for elasticity The momenta are conserved Numerical dissipation of energy The yield criterion is satisfied at the end of the time step Numerical examples demonstrate the robustness
5. Conclusions