TD4 : Numerical elastoplasticity

Local aspects

- Nonlinear problems: elastoplasticity (local aspects)

== > Local integration of the elastoplastic constitutive law

== > Radial return algorithm

- A simple example

- A Von Mises type model with nonlinear isotropic hardening

 TD4: Numerical elastoplasticity
Local aspects
Equilibrium at instant t (weak form)

  :  (v)d   T .vd   f .vd  v  C (0)

d d

 ST 
 t  0, T 

Kinematic conditions: u  C (u )
d
u  u on Su
d

Constitutive equations


 t   A :  t    p t   Elasticity
f
 p t    Yield function

f  ,...  0 ;   0 ; f  ,...   0 normality

To be precised :
Yield function - plastic multiplier - complementarity - Elasticity law (A)
- Yield function (f)
- Hardening (isotropic, kinematic)
Initial condition:  p (t  0)  0
TD4 : Von Mises – Isotropic hardening
Assumptions:
- linear isotropic elasticity
- Von Mises criterion; plastic rate respects normality law
- Isotropic hardening

Questions:
f 3 s
- For f ( , p)   eq  R( p) prove 
 2  eq
- Prove   p tr ( )
s   1
- Deduce the constitutive equations 3
tr ( )
e  1
Elasticity 3

2
Yield function p   : 
3 p p
normality

- Linear isotropic hardening: R( p)  h p   0

 TD4 : Numerical elastoplasticity
Equilibrium at instant t (weak form)

         .vd  v  C (0)
d d
: ( v ) d T .vd f
 ST 
 t  0, T 


Kinematic conditions: u  C (u d ) u  u d on Su 
Constitutive equations

Elasticity

Yield function

normality

Algorithm:

== > Incremental approach (step by step):

 TD4 : Radial return algorithm
Knowing: (t=tn) ; and (t=tn+1)

Equilibrium at instant t=tn+1 (weak form)

Knowing: and find Local treatment of the

constitutive equations
----
TD4 Radial return algorithm

== > Find

TD5
 TD4 : Radial return algorithm
Knowing: and find Local treatment of the
constitutive equations
----
Radial return algorithm

Constitutive equations

Local Local treatment of the constitutive equations:

+ Temporal discretization: tn+1= tn+Dt

+ Numerical integration

== > Discretization of the constitutive equations

+ Radial return algorithm:
For a given Dn (n+1 = n +Dn), compute Sn+1 at tn+1
 TD4 : Constitutive equations
Discrete version

Numerical integration : tn== > tn+1= tn+Dt

1- Deduce the discretization of the constitutive equations at instant tn+1(implicit scheme)


 n 1   n  Tr (D n ) 1  2 De n  D np  Constitutive equations
(discrete version)
(P)
  
f(  n 1 , pn  Dpn )   n 1   0  h pn  Dpn  0
eq

3S n 1 3
D np  Dpn  Dpn N n 1 ; Dpn  0 ; Dpn f(  n 1 , pn  Dpn )  0
2 n 1
eq
2

2- Radial return algorithm ?

Solve problem (P): for a given Dn (n+1 = n +Dn), compute Sn+1 at instant t=tn+1
 TD4 : Radial return algorithm
Elastic prediction:
 elas
n 1
  n  Tr( D n )1  2 D en
S n 1  S n  2 D en
elas

f nelas
1  f( elas
n 1
, pn )   n 1  ( 0  hpn )
eq

1  f ( n 1 , pn )  f ( n 1 , pn  Dpn )
elas
Convexity of f == > f nelas


 n 1   elas
n 1
if f nelas
1  0   
 D
 np  0 ; D  p
n
0

Otherwise, solve
 p 3
   elas  2 D p D
 n  Dp N n 1

n
n 1 n 1 2
if f nelas
1  0   
n
with 
S n 1  S n 1  2 D n
elas p
S  2  eq N
 n 1 3
n 1 n 1

f ( n1 , pn  Dpn )  0
 TD4 : Radial return algorithm
Plastic correction:
 p 3
   elas  2D p D
 n  Dp N n 1

n
n 1 n 1 2
1  0
n
if f nelas  with 
S n 1  S n 1  2D n
elas p
S  2 eq N
 n 1 3
n 1 n 1

f ( n1 , pn  Dpn )  0

 2 2 
S n1  S n 1  2D n S n 1    neq1  2  Dpn  N n 1
elas p elas
We have We obtain
 3 3 

N n1  N n1
elas
and, as a consequence

It gives
3

2 eq

 n1  3Dpn N n1  S elas
n 1 and finally  eq
n 1  eq elas
n 1  3Dpn

We now write the consistency condition: f ( n1, pn  Dpn )   neq1  R( pn  Dpn )  0

== > solve with respect to Dpn For linear isotropic hardening:
f nelas
 neq1  3Dpn  R( pn  Dpn )  0
elas
Dpn  1
h  3
TD4 : Radial return algorithm - RR_VonMises.m
Matlab code: all-elements/RR_VonMises.m

Input

increment
parameters

parameters
State
Output

Strain
at

Elastic

Plastic
t=tn

Dp ,   RR _ VonMises E, , h, ,

n n 1 0 n
, pn , D n 

Deps  Δε11; Δε 22 ; 2ΔΔ12  Plane strain D 33  0

Tr(Δε)  Δε11  Δε 22  0

De  De11; De22 ; De33; De12   Δε11; Δε 22 ; 0; Δε12   Tr(Δε)1; 1; 1; 0

1
3

- Linear isotropic hardening:

f nelas
R( p )  h p   0 Dpn  1
h  3
 TD4 : Example (see strip_plast.m)

Exact solution available:

+ plane strain
+ homogeneous solution Material: homogeneous, elastic,
- xx, yy (other 0) perfectly plastic (R’(p)=0)
- xx, zz (other 0)

Compute yy, xx, zz such that

Kinematic admissibility
== > xx= q/2L
Constitutive relation
=== > radial return algorithm
Equilibrium
== > yy = 0
 TD4 : A simple example
Algorithm (strip_plast.m)

 n ,  n , pn known Plane strain

1   
Temporal discretization (loading increment): Dq E 
0

A   1  0 
(1  )(1  2 )
find  n1 ,  n1   n  D n , pn1  pn  Dpn  0 0 (1  2 ) / 2

Initialization  xxn,0  Dq / 2 L (imposed)

D  
n,0 n,0
 yyn,0   xxn,0 A(1,2) / A(2,2) such as ( 0 ) yy  ( A :   n,0 ) yy  0 (choice )

Radial return algo == > but 0 yy may not be zero !

while  yyn1,k  0  xxn,k  0 (imposed)

n , k 1  yyn,k  ( yyn 1,k  A(1,2) xxn,k ) / A(2,2)    yyn 1,k / A(2,2)
D  D 
n,k n,k

n 1, k
such as (  A:  ) yy  0 (choice )
n,k

(k) (k)
Radial return algo == > Strip_plast.m
Iterate until convergence
 TD4 : A simple example

Example : Strip_plast.m
Ex: A Von Mises type model with nonlinear isotropic hardening
Geometry : L = 0,05m
Material parameters
(Acier 316L) :
1- Repeat the analysis with the yield function


f ( , p)   eq   0  Hp1/ M  E= 200 000 Mpa
 = 0.3
2- Include modifications in RR_VonMisesNL.m 0 = 138 Mpa
H = 430 Mpa
3- Simulate load path q(t): from 0 up to qfin = L/100 M = 4.55

Other Yield functions …

Drucker-Prager
f   eq  3Tr( )  H   0

Kinematic hardening
f ( , X )    X    0
eq

X  2C 
p

use   S  X
DM: A Von Mises type model with nonlinear isotropic hardening
Geometry : L = 5 cm
Material parameters
(Acier 316L) :
1- Repeat the analysis with the yield function
E= 200 000 Mpa

f ( , p)   eq   0  Hp1/ M   = 0.3
0 = 138 Mpa
2- Include modifications in RR_VonMisesNL.m H = 430 Mpa
M = 4.55
3- Prove the initial yield limit q0 is
Send by mail :
4- Simulate load path q(t) (cycle) - Modified matlab code
1 : q(t) = 0 up to q(t)= 2q0 - RR_VonMisesNL.m
2 : q(t) = 2q0 up to q(t)= - 4q0 - Strip_plast.m
- 2 pages (max) of comments
3: q(t) = - 4q0 up to q(t)= 0
and analysis + results
(figures)
5- Simulate several cycles

6- Plot xx in function of xx