Lecture 4 (Dirk Mohr, ETH, Integration Algorithms in Plasticity)

151-0735: Dynamic behavior of materials and structures
Lecture #4:
Integration Algorithms for
Rate-independent Plasticity (1D)
by Dirk Mohr
ETH Zurich,
Department of Mechanical and Process Engineering,
Chair of Computational Modeling of Materials in Manufacturing
2015
2/15/2016
D. Mohr
Lecture #4 Fall 2015
1 1
Recall: Important difference

ELASTO-PLASTIC
NON-LINEAR ELASTIC
(e.g. metals, concrete, thermoplastics )
(e.g. rubbers, foams)
4.00E+02
3.50E+02
Elasto-plastic
loading
Elastic
loading
3.00E+02
2.50E+02
Elastic
unloading
2.00E+02
1.50E+02
Elastic
loading
1.00E+02
Elastic
unloading
5.00E+01
0.00E+00
0.00E+00
5.00E-02
1.00E-01
1.50E-01
2.00E-01
2.50E-01
2/15/2016
D. Mohr
2 2
Rate-independent perfect plasticity

Simplified rheological model:
frictional device
linear spring
INITIAL
CONFIGURATION
DEFORMED (CURRENT)
CONFIGURATION
p
The strain is split into an elastic and a plastic part
e p
i.e. the elastic strain is
e p
2/15/2016
D. Mohr
3 3
Rate-independent perfect plasticity - Summary

i. Constitutive equation for stress
E ( p )
ii. Yield function
iii. Flow rule
f [ , k ] k
p sign [ ]
iv. Loading/unloading conditions
0 if f 0
0 if f 0 and
0 if f 0 and
f 0
f 0
Material model parameters: (1) Youngs modulus E, and (2) flow stress k.
2/15/2016
D. Mohr
4 4
Rate-independent perfect plasticity - Application
k/E
k
time
Total strain
Plastic strain
k/E
time
Stress
k
time
2/15/2016
D. Mohr
5 5
Rate-independent isotropic hardening plasticity
Elasto-plastic
loading
Elastic
loading
Elasto-plastic
loading
Elastic
Elastic
re-loading
unloading
E
E
The magnitude of the stress increases due to strain hardening

when the material is deformed in the elasto-plastic range. For
isotropic hardening materials, it is described through an evolution
equation for the flow stress k.
2/15/2016
D. Mohr
6 6

Firstly, we introduce a scalar valued non-negative function
p dt
to measure the amount of plastic flow (slip). This measure is often
called equivalent plastic strain. Unlike the plastic strain, the
magnitude of the equivalent plastic strain can only increase!
It is then assumed that the flow stress is a monotonically increasing
smooth differentiable function of the equivalent plastic strain
k k [ p ]
This equation describes the isotropic hardening law.
2/15/2016
D. Mohr
7 7

Frequently used parametric forms of the function k k [ p ] are the
Swift and Voce laws:
E+02
4.00E+02
E+02
3.50E+02
E+02
3.00E+02
3.00E+02
E+02
2.50E+02
2.50E+02
E+02
2.00E+02
2.00E+02
E+02
1.50E+02
1.50E+02
E+02
1.00E+02
1.00E+02
Swift
Hardening saturation
dk
0, k k0 Q
d p
5.00E+01
E+00
0.00E+00
3.50E+02
Swift-Voce
5.00E+01
2/15/2016
D. Mohr
Voce
E+01
k S A( p 0 ) n
4.00E+02
0.00E+00
kV k0 Q 1 exp[ p ]
k (1 )kV k S
8 8

In engineering practice, the isotropic hardening function is often
represented by a piece-wise linear function
k [ MPa ]
PEEQ
0.000
0.020
0.050
0.100
0.200
400
350
300
250
200
k
199.1
246.3
283.9
321.0
365.6
150
100
50
p []
2/15/2016
D. Mohr
0.05
0.1
0.15
0.2
9 9
Isotropic hardening plasticity - Summary

i. Constitutive equation for stress
E ( p )
ii. Yield function
iii. Flow rule
f [ , p ] k [ p ]
p sign[ ]
iv. Loading/unloading conditions
0 if f 0
0 if f 0 and f 0
0 if f 0 and f 0
v. Isotropic hardening law

k k [ p ] with
2/15/2016
D. Mohr
p dt
10
10 10
Differential equation to be solved

First-order ordinary differential equation
E ( p )
p sign[ ]
Initial condition
p [ t 0] 0
p sign [t ] p [t ]E
Prescribed loading
[t ]
Multiplier to satisfy the constraints
0 if f 0
0 if f 0 and f 0
0 if f 0 and f 0
with f [ , p ] k [ p ] , E ( p ) and p dt
2/15/2016
D. Mohr
11
11 11
Numerical solution of Differential Equations

Without the loading and unloading conditions, the plasticity problem
reduces to solving an ordinary first-order differential equation for
the plastic strain, considering time as the only independent variable:
dy
p sign [t ] p [t ] E
g[ y ]
D.E.
dt
p [ t 0] 0
I.C.
y [ t 0] y 0
Such equations are solved numerically using integration algorithms.
Instead of the calculating the exact analytical solution, we limit our
attention to calculating the approximated solution
y n y [ tn ]
at equally-spaced instants tn, n=1,,N with the time step Dt,

Dt tn 1 tn
2/15/2016
D. Mohr
12
12 12

A first popular method is the so-called forward (explicit) Euler
algorithm:
y0
y1 y0 Dtg[ y0 ]
y2 y1 Dtg[ y1 ]
yn 1 yn Dtg[ yn ]
Starting with the initial condition, the approximations can be

progressively calculated.
2/15/2016
D. Mohr
13
13 13

Recall that y ' g[ y ] and thus the forward (explicit) Euler algorithm
may also be written as
y
exact derivative
y0
approximation
y1 y0 Dty 0
y2 y1 Dty1
yn 1 yn Dty n
tn tn+1
In other words, the time derivative at time tn is given by the

approximation
y yn
y[t n ] n 1
Dt
2/15/2016
D. Mohr
14
14 14

A second popular method is the so-called backward (implicit) Euler
algorithm:
y0
y1
which is obtained from solving the implicit equation
y1 y0 Dtg[ y1 ]
y2 which is obtained from solving the implicit equation y2 y1 Dtg[ y2 ]
yn 1
which is obtained from solving the implicit equation
yn 1 yn Dtg[ yn 1 ]
Starting with the initial condition, the approximations can be

progressively calculated. However, at each time step ti, an often
implicit equation needs to be solved.
2/15/2016
D. Mohr
15
15 15

According to the backward (implicit) Euler algorithm, the time
derivative at time tn is given by the approximation
yn yn 1 Dtg[ yn ]
y [t n ]
y g[ y ]
yn yn 1
Dt
approximation
exact derivative
tn-1 tn
2/15/2016
D. Mohr
16
16 16
Example:
Illustration
dy
y
dt
(differential equation)
y [ t 0] 1
(initial condition)
y exp[t ]
(exact solution)
The approximate solution with forward (explicit) Euler algorithm for

a time step of Dt=1 reads (we have g[y]=y and y0=1):
y0 1
160
y1 y0 Dtg[ y0 ] 1 1 1 2
y2 y1 Dtg[ y1 ] 2 1 2 4
y3 y2 Dtg[ y2 ] 4 1 4 8
y4 y3 Dtg[ y3 ] 8 1 8 16
y5 y4 Dtg[ y4 ] 16 1 16 32
Dt 0.01
exact
140
Dt 0.1
120
100
80
60
40
Dt 1
20
0
0
Observe from the graph that the method converges for Dt

2/15/2016
D. Mohr
17
17 17
Comparison implicit vs. explicit

The approximate solutions with forward (explicit) Euler and
backward (implicit) Euler algorithms for a time step of Dt=0.1
implicit
200
180
160
140
120
100
80
60
40
20
0
exact
explicit
2/15/2016
D. Mohr
18
18 18
back to the plasticity problem

p sign [t ] p [t ]E
Differential equation:
np1 np Dtg[tn 1 ]
np Dt sign n 1 np1 E
D
np D sign n 1
Initial condition:
0p 0
State variable:
np1 np D
0p 0
plus discrete evolution constraints:
Dependent variables:
n 1 E ( n 1 np1 )
k n 1 k [ np1 ]
f n 1 n 1 kn 1
f n 1 0
D 0
( D ) f n 1 0
2/15/2016
D. Mohr
if D 0 then f n 1 0
if f n 1 0 then D 0
19
19 19
back to the plasticity problem

p sign [t ] p [t ]E
Differential equation:
np1 np Dtg[tn 1 ]
np Dt sign n 1 np1 E
D
np D sign n 1
Initial condition:
0p 0
State variable:
np1 np D
0p 0
plus discrete evolution constraints:
Dependent variables:
n 1 E ( n 1 np1 )
k n 1 k [ np1 ]
f n 1 n 1 kn 1
f n 1 0
D 0
( D ) f n 1 0
2/15/2016
D. Mohr
Main unknown:
if D 0 then f n 1 0
if f n 1 0 then D 0
Plastic multiplier
which makes our problem more complex

than solving an ordinary first order
differential equation!
20
20 20
Return Mapping Algorithm

We solve the plasticity problem assuming a strain-driven process, i.e.
for a given increment in the applied total strain,
D n 1 n
we determine numerical approximations of the corresponding stress

and state variables at time tn+1 based on their values at time tn.
Applied total strain
increment
OUTPUT:
State variables at time tn
np , np
State variables at time tn+1

RETURN MAPPING
ALGORITHM
np1 , np1
Stress at time tn+1
n 1 E ( n 1 np1 )
2/15/2016
D. Mohr
21
21 21
Return mapping procedure

When computing the solution at time tn+1, we first compute the trial
elastic state by assuming that the material response is purely elastic
(no plastic evolution) when applying D :
p
p
ntrial
E
(
E
(
D
1
n 1
n
n
n ) n E ( D )
p
np,trial
1
n
np,1trial np
trial
p
f ntrial
k
[
1
n 1
n ]
2/15/2016
D. Mohr
then elastic loading step

if f ntrial
1 0
then plastic loading step
if f ntrial
1 0
22
22 22

ntrial
1 n E ( D )
ntrial
1
trial
n 1
n
f ntrial
1 0
elastic
step
2/15/2016
D. Mohr
ntrial
1
ntrial
1
n
ntrial
1
f ntrial
1 0
elastic
step
f ntrial
1 0
plastic
step
f ntrial
1 0
f ntrial
1 0
plastic
step
elastic
step
23
23 23

Elastic loading step:
D 0
f n 1 f ntrial
1 0
increment
D 0
np , np
2/15/2016
D. Mohr
Calculate
Trial State
trial
ntrial
,
f
1
n 1
OUTPUT:
f ntrial
1 0
np1 np
np1 np
Stress at time tn+1
n 1 E ( n 1 np1 )
24
24 24

Plastic loading step: f ntrial
D 0
1 0
In a plastic loading step, the plastic multiplier
D>0 must be determined such that the yield
condition at time tn+1 is full filled.
f n1 n1 kn1 (1)
Firstly, we express the absolute value of the stress n+1 as a function

of the unknown plastic multiplier:
D
p
n 1 E ( n 1 np1 ) E n 1 np ( np1 np ) ntrial

1 ED p
while
D p np1 np D sign n 1
And hence
n 1 E ( n 1 np1 ) ntrial
1 E ( D )sign[ n 1 ]
2/15/2016
D. Mohr
25
25 25

n 1 sign[ n 1] n 1
sign[ n 1] ntrial
1 E ( D )sign[ n 1 ]
sign[ n 1] ntrial
1 E ( D )
observe that
sign[ n 1] sign[ ntrial
1 ]
n 1 ntrial
1 ED (2)
Secondly, we express the flow stress kn+1 as a function of the

unknown plastic multiplier:
np1 np D
kn 1 k [ np D ] (3)
Then, using the results (2) and (3) in (1), we obtain the so-called
discrete consistency condition:
p
f n 1 n 1 kn 1 ntrial
E
D
k
[
1
n D ] 0
2/15/2016
D. Mohr
26
26 26

p
f n 1 ntrial
E
D
k
[
1
n D ] 0
ntrial
1
E ( D )
kn 1
n 1
kn
k [ np D ]
D D e
D
n
2/15/2016
D. Mohr
n 1
27
27 27
Solving the discrete consistency condition

Example #1: Linear hardening law
k [ p ] k 0 H p
with constant hardening modulus H
The discrete consistency condition then reads

p
f n 1 ntrial
E
D
H
(
1
0
n D )
p
ntrial
(
k
1
0
n ) H D E D
f ntrial
1 ( H E ) D 0
f ntrial
1
from which we determine the plastic multiplier

f ntrial
1
D
H E
2/15/2016
D. Mohr
y f ntrial
1 ( H E ) x
28
28 28

Example #2: General non-linear concave hardening law k [ p ]
k
The discrete consistency condition then reads
p
f n 1 ntrial
E
D
k
[
1
n D ]
kn
ntrial
E
D
k
[
1
n
n D ] k n
f ntrial
E
D
k
[
1
n D ] k n 0
Which corresponds to seeking the root of the

convex function
p
n
f ntrial
1
y[x] f ntrial
Ex
k
[
1
n x ] kn
D
2/15/2016
D. Mohr
29
29 29
x0
x1 x2
Seeking the root of a C1-continuous function is a standard problem

in applied mathematics. For example, it can be found using a
Newton-Raphson scheme:
y[ x0 ]
y[ x1 ]
y [ xn ]
x2 x1
xn 1 xn
x0 0 x1 x0
y ' [ x0 ]
y ' [ x1 ]
y ' [ xn ]
iterate until y[ xn 1 ] TOL then D xn 1
2/15/2016
D. Mohr
30
30 30
Elasto-plastic Tangent Modulus

The derivative d/d is called elasto-plastic tangent modulus.
During plastic tensile loading ( d 0, d 0, d p 0), we have
dk
d 0
df d dk E ( d d )
d
p
and thus
dk
E
d p
d E ( d d )
d
dk
p
2/15/2016
D. Mohr
E
d
d
dk
dk
E
d p
d
d
dk
p
31
31 31
Elasto-plastic Tangent Modulus

Formally, we note the incremental stress-strain response as
d Eep ( d )
with
Eep
2/15/2016
D. Mohr
Eep
E
if
dk
dk
if
32
32 32
Summary: Return Mapping Algorithm

D 0

increment
OUTPUT:
Calculate
Trial State
trial
ntrial
,
f
1
n 1
np , np
f ntrial
1 0
Stress at time tn+1
f ntrial
1 0
Solve:
n 1 E ( n 1 np1 )
p
ntrial
E
D
k
[
1
n D ] 0
D 0
OUTPUT:
np1 np ( D )sign[ ntrial

1 ]
np1 np D
2/15/2016
D. Mohr
np1 np
np1 np
Stress at time tn+1
n 1 E ( n 1 np1 )
33
33 33
Reading Materials for Lecture #4

J.C. Simo and T.J.R. Hughes, Computational Inelasticity (first chapter):
http://link.springer.com/book/10.1007%2Fb98904
M.E. Gurtin, E. Fried, L. Anand, The Mechanics and Thermodynamics of
Continua, Cambridge University Press, 2010.
2/15/2016
D. Mohr
34
34 34

Lecture 4 (Dirk Mohr, ETH, Integration Algorithms in Plasticity)

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Lecture 4 (Dirk Mohr, ETH, Integration Algorithms in Plasticity)

Transféré par

Droits d'auteur :

Formats disponibles

151-0735: Dynamic behavior of materials and structures

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

Recall: Important difference

(e.g. metals, concrete, thermoplastics )

(e.g. rubbers, foams)

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

Rate-independent perfect plasticity

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

Rate-independent perfect plasticity - Summary

iv. Loading/unloading conditions

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

Rate-independent perfect plasticity - Application

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

Rate-independent isotropic hardening plasticity

The magnitude of the stress increases due to strain hardening

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

Rate-independent isotropic hardening plasticity

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

Rate-independent isotropic hardening plasticity

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

Rate-independent isotropic hardening plasticity

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

Isotropic hardening plasticity - Summary

iv. Loading/unloading conditions

v. Isotropic hardening law

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

Differential equation to be solved

Multiplier to satisfy the constraints

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

Numerical solution of Differential Equations

at equally-spaced instants tn, n=1,,N with the time step Dt,

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

Numerical solution of Differential Equations

Starting with the initial condition, the approximations can be

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

Numerical solution of Differential Equations

In other words, the time derivative at time tn is given by the

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

Numerical solution of Differential Equations

which is obtained from solving the implicit equation

y2 which is obtained from solving the implicit equation y2 y1 Dtg[ y2 ]

which is obtained from solving the implicit equation

Starting with the initial condition, the approximations can be

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

Numerical solution of Differential Equations

Lecture #4 Fall 2015

151-0735: Dynamic behavior of materials and structures

The approximate solution with forward (explicit) Euler algorithm for

Observe from the graph that the method converges for Dt