Vous êtes sur la page 1sur 74

Crashworthiness engineering

with LS-DYNA

Part 4 :
Material modeling

4.1

Content :

4.1 : classification of materials for numerical


simulation
4.2 : general notions :
4.2.1 : stress in solids
4.2.2 : strain and strain rate in solids
4.2.3 : hypo-elasticity
4.2.4 : elasto-plasticity and radial return
4.3 : material models in LS-DYNA for 2D elements
4.3.1 : car body panels : thin metal sheets
4.3.2 : castings
4.3.3 : trimpanels : Woodstock and Lignotok
4.3.4 : trimpanels : polymers
4.4 : material models in LS-DYNA for 3D elements
4.4.1 : visco-elastic material models
( rubber and plastic )
4.4.2 : non-linear elastic material models
( reversible foams )
4.4.3 : elasto-plastic material models
( irreversible foams and honeycomb )
4.4.4 : development of user-subroutines

4.2

4.1 : Classification of materials for numerical simulation :

Main decisions :
is the structure 2D or 3D ?
is the material reversible or irreversible ?

Reversible

Elastic ML
Visco-elastic ML
2D (Shell)

3D (Solid)

Irreversible
(permanent
deformations)
Elasto-plastic ML

Tissue (Bag-Belt) Mild steel


Plastic (Polymers) High strength steel
Aluminium
Magnesium
Plastic (Polymers)
Rubber
Honeycomb
(Elastomers)
Aluminium foam
Soft foams
Crushable PU
Confor foam
EPF (EPP,EPS...)
EA-PU

Sometimes the choice is hard since part of the deformation is


reversible and part is permanent. (Plastics !)
4.3

Part 4 :
Material Modeling :

Part 4.2 :
General notions :

4.4

4.2.1 : Measures of stress in solids :

We consider Cauchy stresses (force over current area)


We will write the stress tensor as :

xx xy xz
T

_ = yx yy yz =

zy
zz
zx
Since the stress tensor is symmetric, we can always find a
proper orthogonal matrix (=rigid body rotation) that
diagonalizes it :

1 0
T 
R R = 0 2

0 0

The diagonal components are the principal stresses and alow


a 3D representation of the state of stress in a point of a
continuum
4.5

Invariants :

The state of stress in a point is determined by the invariants


of the stress tensor , the three invariants are :
_p=

xx + yy + zz
3

2
2
2
3
2
2
2
_ vm =
p
p
p
2
2
2

+
+

+
+

+
+

(
)
(
)
xx
yy
zz
xy
yz
xz

2

_

The first and second invariant are called pressure and


von Mises stress resp.

By invariance we mean invariance under a proper


orthogonal transformation :

T  
_ = R 0 R
_ p = p0
_ vm = vm0


4.6

Objectivity of the stress tensor :

Example of a bar in compression with rotation of the


reference system :


= 0
0

0 0
0


0 0 = 0
0
0 0

=
vm = vm

p = p =

3
4.7

0
0

Invariant Space :

For an isotropic material, the plasticity law will depend upon


the stress invariants only, if we further restrict to the first 2
invariants, we can represent the material law in a 2D plane :

_ vm
3
1
p

Here :
- the horizontal axis corresponds to a hydrostatic loading
path
- the vertical axis is pure shear
- the line with tangent 1/3 is uniaxial compression
- the line with tangent -1/3 is uniaxial tension
- a line of constant von Mises stress is horizontal
- etc...

4.8

Uniaxial state-of-stress :

xx

= 0
0

0 0

0 0
0 0
p=

vm =

xx
3

2
2
3 2 xx
xx

+ 2
= xx
2 3
3

4.9

vm = 3 p _(compression)

= 3 p _(tension)
0 vm

Deviatoric Stresses :

The pressure or first invariant is related to the change in


volume of the solid, the deviation from a hydrostatic state
of stress is linked to the change in shape.

The stress deviator is defined as :


 
_ s = + pI
s
xx
_ syx
s
zx

sxy
syy
szy

sxz xx xy xz p


syz = yx yy yz + 0

szz zx zy zz 0

0
p
0

The second invariant in terms of stress deviators becomes :

_ vm =

3 2
2
sxx + syy
+ szz2 + 2 sxy2 + 2 sxz2 + 2 syz2
2

A surface of constant von Mises stress in deviatoric space


(9D) or in principal deviatoric space (3D) is thus a sphere
4.10

Representation in the invariant plane :

_ vm


= ( p, vm )

p

pI = ( p,0)

4.11

Remarks :

A hydrostatic state of stress is the kind that can exist in a


liquid or a gas. It has such a high degree of symmetry that
the corresponding deformation can only be a change of size,
without change of shape (cube remains cube, sphere remains
sphere etc...), as long as the material is isotropic.
Similarly the deviatoric stresses can produce only shear
deformation changing the angle between 2 faces of a block or
between 2 of its diagonal planes. Indeed the deviatoric
stresses have a zero trace (volumetric component) and can
thus be represented as the superposition of 5 simple shearstress systems :
sxx + syy + szz 0.
thus _:
0

s = sxy

0
sxx

_ 0

sxy
0
0
0
sxx
0

0 0 0

0 + 0 0

0 0 syz

0 0

syz + 0

0 sxz

0 0 0

0 + 0 szz

0 0 0

szz

4.12

0 sxz
0 0 +

0 0

xy = yx 0

yy = xx < 0
4.13

Concept of material law in hydrocodes :

Material law

Constitutive law
Shear
Change in shape

Equation of state
pressure
change in volume

4.14

4.2.2 : Measures of Strain :

All measures of strain are a function of the velocity


gradient and/or the deformation gradient :

 x
L= 
x

x

x
y
=
x

z
x

 x
F= 
x0

x0
y
=
x0
z
x
0

x
y
y
y
z
y
x
y0
y
y0
z
y0

x

z
y
z
z
z
x

z 0
y

z 0
z
z0
4.15

Evaluate the first component


of the velocity gradient :

Evaluation is in an integration point P


of a finite element :
z

P 1
2
y

x
The velocity field must be evaluated in two points on an
axis parallel to x through P

x x 2 x1

x x 2 x1

4.16

The most familiar strain measure is the small strain tensor :


u
  
u = x x0 = v
w

x 0

 1 u
v

E =
+
2 y 0 x 0

1 u + w

2 y
0 z 0

1 u
v

+
2 y 0 x 0
v
y 0
1 u w

+
2 y 0 z 0

1 u w

+
2 y 0 z 0
1 w v

+
2 y 0 z 0

z 0

The small strain tensor is the symmetric part of the


displacement gradient :

T
 1 ( x x0 ) ( x x0 )

+
E=


2 x 0
x 0

 1  T

E = F + F 2I
2

4.17

The small strain tensor is the classical total strain measure. It


is however useful only for linear small displacement analysis
since it does not vanish under a rigid body rotation.
Indeed :

_ if _:
 
_ x = Rx 0

 x

_F =  = R
x 0
 1  T

 
1  T
_ E = ( F + F 2I ) = ( R + R 2I ) 0
2
2

If this strain measure would be used in a large rotation


context, a rigid body motion would cause strains, thus stresses
and the body would start to expand or compress.

Consequently the small strain tensor is not used at all in LSDYNA which is designed to treat large displacement
problems.

4.18

small strain for a 2D rotation :

x
1

u1 = 0
u2 = l cos l
x01 = 0
x02 = l
Exx =

u2 u1
= cos 1
x0 2 x01

4.19

Most formulations in LS-DYNA are incremental,


rather then total and are based upon the strain rate or rate of
deformation tensor that is defined as the symmetric part of
the velocity gradient :

 1  T
= L+L
2

x

 1 x
y
=
+
2 y x

1 x
z
+

x
2 z

1 x
y
+
2 y x
y
y
z
1 y
+
2 z
y

1 x
z
+
2 z
x

1 z
y
+
2 y z

z

The strain rate components have a dimension ( 1/s)

4.20

Strain rates clearly vanish in the case of rigid body rotation :


_ if _ :
  
_ x = x
_ then _ :
 x 
_L=  =
x

 1  
1  
1  
_  = L + LT = + T = 0
2
2
2

The strain rate is thus an objective, incremental


measure of strain.

4.21

strain rate for a 2D rotation :

x
1

x1 = 0
x 2 = 0
x1 = 0
x2 = l
E xx =

x 2 x1
=0
x2 x1

4.22

It is important to realize that the time derivative of the small


strain tensor and the strain rate are not equivalent. Actually
they are the same only if both rotations and deformations
remain small. (e.g. civil engineering problems).

The latter can be illustrated for the uniaxial case :

u x
x


E xx =
xx =
=
t x 0 x 0
x
x x E 
0

xx

xx

4.23

The physical analog for the different strain measures can be


obtained from the change in length of an elementary truss :

2
l0

E xx =

xx

u x 2 x1 x 2 0 + x10 l l0

=
x 0
x 2 0 x10
l0

l
x 2 x1
l
dl
= xx dt
dt = dt = = ln
x 2 x1
l
l
l0

The logarithmic strain tends towards minus infinity


if the length of the truss approaches zero :

l0

4.24

Strain rate Invariants :

We define the first invariant as the volumetric strain rate :

v = xx + yy + zz


x y z
v = + +
x y z
For a small material element this can be written as follows :
lx ly lz
v = + +
lx l y lz

v =
v =

lxl y l z + l xly l z + l xl y lz


l xl y l z
V
V

The volumetric strain rate thus equals the rate of change of


(material) volume over the actual
volume

4.25

Deviatoric strain rates :

Deviatoric strain rates are defined based upon


the volumetric strain :


 1 
d = v I
3

d xy d xz xx
d xx





d yx d yy d yz = yx





zx
dzx
d
d

zy
zz

1 0 0
xy xz
1



yy yz v 0 1 0
3

zy zz
0 0 1

Clearly the trace (sum of diagonal components)


is zero for the deviatoric strain rates as it was
for the stress :


tr(  ) = xx + yy + zz = v

tr( d ) = d xx + d yy + d zz = 0
The deviatoric strain rate thus has no volumetric
component.

4.26

Second invariant of strain rate :

The effective (or equivalent) strain rate is defined as :

eff =

2  
d :d
3

Or :

eff =

2 2
2
2
2
2
2
+ dzz
+ 2dxy
+ 2dyz
+ 2dzx
dxx + dyy
_
3

4.27

Note that under uniaxial loading : (in x-direction) :


 xx

 = 0
0

0
 xx

0
 xx
0

v = xx (1 2 )
2 0 0


xx
(1 + ) 0 1 0
d =
3
0 0 1

eff = xx

2
(1 + )
3

4.28

Effective strain rate under uniaxial loading :

Consider an incompressible elastic material :

= 0.5 eff =  xx
The consider a cyclic uniaxial loadcase :

eff ,  xx

strain rate and effective strain rate are state variables :


they can be determined from the current configuration

4.29

Effective strain under uniaxial loading :

Consider the effective strain in the previous case :


t

eff =  eff dt =  xx dt
Then consider a cyclic uniaxial loadcase :

eff , xx

The effective strain is a history variable : computation


requires knowledge about the entire loading history

4.30

In elasto-plasticity, strain rates will be split in


elastic and plastic parts. Each part can then
again be split in volumetric and deviatoric
components :


=


ed +


e =
ev I

pd +


p =
pv I

Metals will usually exhibit a zero volumetric


plastic strain component : plastic deformation
in metals occurs at constant volume, this is
not the case however in foams.

4.31

The second and third strain rate invariants will not often be
used, more commonly the second invariant of the plastic
strain rate tensor is defined as follows :

p =  e
p =

2 2
2
2
2
2
2
pdxx + pdyy
+ pdzz
+ 2pdxy
+ 2pdyz
+ 2pdzx
_
3

p = p dt

The last two equations are the definitions for the equivalent
(or effective) plastic strain rate and equivalent plastic strain
respectively.

Note that this definition is valid , also if the plastic


strains are not deviatoric.
Clearly the equivalent plastic strain must always
be a monotonically increasing function of time.

4.32

Elastic and plastic Poisson coefficient :

The Poisson coefficient relates to elastic deformations only


and gives the relationship between longitudinal and
transversal deformation under uniaxial loading :

 yye
 yy  yyp
=
=
xxe
 xx  xxp
Similarly we can introduce a plastic Poisson coefficient based
upon plastic deformations only :

yyp
p =
xxp
And a total Poisson coefficient based on total deformations
(the only one we can measure ) :

 yy
t =
xx

4.33

Values of effective strain rate and equivalent plastic strain


rate under uniaxial loading conditons :

For elastic metals we have that :

eff = 0.866 xx
= t = 0.3
 p = 0.
For plastic metals we have that :

= 0 .3
p = 0 .5 t

0.866 xx eff  xx

 p =  p , xx

For plastic metals we also have that :

= 0 .3
p = 0.5 t

eff xx

p = p , xx xx

sin ce
xx p , xx >> e , xx
So whereas in the plastic regime effective strain and
equivalent plastic strain are always nearly equal, the same is
not true for effective strain rate and equivalent plastic strain
rate due to the oscillatory nature of the elastic strain rates.

4.34

For foams we have that :

2
3
= p =t 0
2
 p =  p , xx
3

eff =  xx

And after time integration :

= p =t 0

2
3
xx
2
= xx

3
E

eff = xx
2
3

p = p , xx

4.35

Remarks :

There is some arbitrareness in the definition of


the invariants :

- the first stress invariant is positive in compression


rather then in tension
- the second stress invariant (von Mises stress) was
chosen since it corresponds to the plasticity limit
in uniaxial tension for steel and most metals
- the second invariant of the plastic strain rate is
defined so that its product with the von Mises
stress yields a rate of plastic deformation energy

e p = vm  p

4.36

4.2.3. Isotropic Hypoelastic material :

Most explicit codes implement a hypoelastic law as a linear


relationship between stress rate and strain rate, expressed in
some corotational reference system.

The deviatoric stress rate will depend upon deviatoric


deformation rates only and similarly, volumetric strain rate
determines the pressure :



s = 2Gd
V
p = Kv = K
V

V
p = K ln = K ln
V0
0

Hypoelasticity is equivalent to linear elasticity if both the


rotations AND the strains are small

4.37

Hypoelasticity in terms of Youngs modulus and Poissons


ratio becomes :


_ =

E 
+
v I

(1 + ) (1 2 )

With :
E
2( 1 + )
E
K=
3(1 2 )

G=

If the Poisson ratio equals zero we obtain :



_ = E
and every stress component is proportional to the
corresponding strain component : the individual stress
components are uncoupled.

Foams usually exhibit very small or zero Poisson


effects in compression.

4.38

Material law 1 in LS-DYNA (MAT_ELASTIC) corresponds


to this hypoelastic formulation

Numerically the algorithm is trivial and central thus second


order accurate :

t(n)

t(n+1/2)

t(n+1)



n +1 / 2

n +1

New stresses must be calculated from old stresses and strain


rates at the half timepoint

4.39

Numerical integration of hypoelastic law :

first split in deviatoric and volumetric components :

(
xx , n + yy , n + zz , n )
pn =
n
3



sn = n + pn I
v , n +1 / 2 =  xx , n +1/ 2 +  yy , n +1 / 2 +  zz ,n +1 / 2



n +1/ 2


1
d ,n +1/ 2 = n +1/ 2 v , n +1/ 2 I

3


then integrate the material law :

p = Kv
p n +1 p n
= Kv ,n +1 / 2
t
p n +1 = p n K v

4.40

and for the deviatoric part :



s = 2G d



sn +1 sn
= 2G d ,n +1 / 2
t



sn +1 = sn + 2G d

and assemble the new stresses :

n +1



= s n +1 p n +1 I

4.41

Note that :
3 
s :s
2

2  
Eeff =
Ed : Ed
3
vm = 3GEeff


s = 2GEd

vm =

Under uniaxial loading :

Svm

Svm
3G

Eeff

Exx

4.42

4.2.4 : Elasto-plasticity for metals :

We will treat the case of perfect plasticity : the von Mises


condition then is :

vm y 0
Under uniaxial loading :

Such a material law will be sensitive to strain localisation

4.43

A full description of isotropic elasto-plastic


behaviour is given by :

1: Material law based on introduction of plastic


strain rates and hypoelasticity :

) (



 
+ = 2G d dp + K v vp I
taking symmetrie into account, this is a system
of 6 equations with 12 unknowns : 6 stresses and
6 plastic strain rates.

4.44

2 : A flow rule that makes an assumption about


the plastic strain rates : a good description
of metals is obtained if we assume the plastic
strain rate to be proportional to the deviatoric
stress :

3 
s

p =  p

2 vm

 p =

2  
dp : dp
3

 p =

2 2
2
2
2
2
2
+ dpzz
+ 2dpxy
+ 2dpyzx
+ 2dpzx
dpxx + dpyyx
3

This reduces the number of unknowns in the plastic strains


from 6 to 1, knowledge of the equivalent plastic strain rate
will now be sufficient to solve the problem.
Since metallic plastic strains are deviatoric, the equivalent
plastic strain rate is best interpreted as the magnitude of the
plastic strain rate vector

4.45

Consequences of the flow rule choice :

The principal directions of plastic strain rate, deviatoric


stresses and stresses all coincide.
Since plastic strain rate and deviatoric stress are
proportional, the trace of the plastic strain rate is zero and
the plastic strain rate is equal to the deviatoric plastic strain
rate :



p s

 pxx +  pyy +  pzz s xx + s yy + szz = 0




p pd
Consequently the equivalent plastic strain rate can be based
upon the total plastic strains and the volumetric plastic
strain rate is zero :

vp 0.
 p =

2  
p : p
3

Plastic deformation does occurs at constant volume


(confirmed experimentally by P.`Bridgman).
4.46

Reformulation of the material law :

The flow rule allows to rewrite the material


law as follows :

(
p = K ( 

)
)






s = 2G d dp = 2G d p
v

vp

becomes_:


s = 2Gd 2G p

2 
s

3 vm

p = Kv
Thus :
The pressure depends upon the volumetric strain (density)
only and is not affected by the plasticity. Thus a nonlinear
EOS relating pressure and density can be used without
adding complexity.
The deviatoric stresses are obtained from an elastic
increment minus a stress reduction that is applied parallel to
the deviatoric stress vector. To determine the stress
reduction we need to know the equivalent plastic strain.

4.47

3 : Loading and unloading conditions :

Define a yield surface described by a function f of the stress


components and some material parameters, in the case of a
von Mises yield condition :


f ( , y ) = vm y 0.

_ vm
vonMises

The state of stress is elastic iff :


The state of stress is below the yield surface
The equivalent plastic strain remains constant
Thus the equivalent plastic strain rate is zero
The state of stress is plastic iff :
The state of stress is on the yield surface
The equivalent plastic strain increases
Equivalent plastic strain rate is > zero and can be
determined from the equation f=0
4.48

Other yield surfaces :

The Drucker-Prager criterion can correspond to a parabola


or an ellips in the invariant plane :

f ( , Ai ) =

_ vm

2
vm

A0 A1 p A2 p 2 0.

Drucker-Prager

These formulations correspond to material laws 5 and 75 in


LS-DYNA

4.49

Solving the elasto-plastic problem numerically :

Numerically, the elasto-plastic problem can be


stated as follows : find the stresses at timepoint
(n+1) from the known stresses at timepoint (n)
and the known total strain rates at (n+1/2).
Thus :

 
_ n ,

1
n+
2

n +1

In order to achieve this we will also have to calculate the new


values of the hardening parameters (yield stress) and the
equivalent plastic strain from the old values :

yn , pn y ( n +1) , p ( n +1)
The usual way to solve this problem in explicit codes is by
using the radial return algorithm.

This extremely simple and accurate algorithm (Wilkins,


1964) is one of the main reasons for the efficiency and succes
of ecplicit codes.
4.50

The radial return algorithm :

STEP 1 : Calculate elastic trial stresses :

Define the incremental strain :

 
d = 

1
d ,n +
2

v = 

v ,n +

1
2

v ,n +1 = v ,n + v
Assume the total strain increment to be elastic :




se ,n +1 = sn + 2G d
pe ,n+1 = pn K v
or _:

pe ,n+1 = pe ,n +1 v ,n +1
(EOS)

4.51

STEP 2 : check if the elastic trial stress fulfills


the yield condition :

f ( e ,n +1 , yn ) 0.



3
_ vm ,e ,n +1 =
se ,n+1 : se ,n+1
2
_ vm ,e ,n +1 y ,n 0 ?
If this condition is fulfilled, the process is elastic
and the numerical treatment ends here.
If not, the stress must be returned to the yield
surface

4.52

STEP 3 : if the process is plastic, return stress


to the yield surface :

The radial return algorithm consists in discretizing


the material law for the deviatoric stresses and
using the trial elastic stress to estimate the
direction of the plastic strain rate :



s = 2Gd 2G p

3 
s

2 vm






sn +1 sn se , n +1 sn
3
=
2G
 1 se, n +1
t
2 e, vm, n +1 p , n + 2
t



3

sn +1 = se , n +1 1 2G p

2 vm, e, n +1

We thus obtain that the new elasto-plastic stress is obtained


by multiplying the elastic trial stress with a scalar function of
the equivalent plastic strain increment.

4.53

The radial return algorithm :

Since we do not alter the pressure we obtain :

_ pn+1 = pe ,n+1


_ sn +1 = kse ,n+1
_ vm,n+1 = kvm,e ,n+1
_ k 1.
We determine k directly from the yield condition
at time n+1 if the material is perfectly plastic :

_ vm,n+1 y ,n+1 = 0
_ k vm,e ,n+1 y ,n = 0
_k =

y ,n
vm,e ,n+1

4.54

The deviatoric stresses are scaled by the yield stress over the
new trial elastic von Mises stress.
The radial return is radial in deviatoric space :

R = vm,e ,n +1

s3
Se,n+1

R = y ,n+1

Sn+1
s2
s1

The resulting stress return path is parallel to the deviatoric


stress vector as suggested by the constitutive law. This
intuitively explains the high accuracy of the radial return
algorithm.

4.55

A representation in the invariant plane clearly shows how


the pressure remains unaffected by the plasticity :

_ vm

vm,e ,n +1

e ,n+1


n+1

vm,n +1

p n + 1 = p e , n +1

The stress projection depends upon the flow rule only and is
independent of the choice of yield surface. Material laws 10,
3 , 12, 24 and many others are all based on the same
metallic flow rule.

4.56

Algorithmic Setup :

The algorithmic setup illustrates the simplicity of radial


return :




se ,n+1 = sn+1 + 2Gd ,n+1/ 2 t
pn+1 = pn+1 ( v ,n+1 )

?

3
vm,e,n+1 =
se ,n+1 : se ,n+1 y
2
IF _ YES _ :



n+1 = se,n+1 pn+1 I

IF _ NO _ :


n+1


= se,n+1

y
vm,e ,n+1

4.57


pn+1 I

4.2.5 : elasto-plasticity with strain hardening :

The von Mises condition becomes :

vm y ( p ) 0
Under uniaxial loading :

Consider linear hardening first


Such a material law will be much less sensitive to strain
localisation
4.58

To describe the material behaviour we need additionally :

4 : A hardening rule allowing to calculate the


evolution in time of the hardening parameters,
Perfect plasticity is obtained if the hardening
parameters do not depend upon the plastic
strain.
For metals, with a von Mises yield condition
however the yield stress is a function of the
equivalent plastic strain and a hardening rule
is necessary :

_ y = y ( p )

y
_  y =
 p = H p
p
_ y ,n+1 = y ,n + H p ,n+1/ 2 t
If the hardening rule is assumed linear in the equivalent
plastic strain rate, all numerical algorithms simplify
considerably.
4.59

Simple elasto-plastic material laws with linear hardening are


available in LS-DYNA as :
MAT_PLASTIC_KINEMATIC (law 3)
For brick and shell elements, kinematic hardening
(Bauschinger effect) is optional
MAT_ISOTROPIC_ELASTIC_PLASTIC (law 12)
For brick elements only, no plane stress condition is enforced
for shells

4.60

The radial return algorithm applies as before.


The same scaling as for the individual deviatoric stress
components must obviously apply for the von Mises stress :

vm ,n +1 =


3
s n +1 : s n +1
2

vm ,n +1 =



3
3

1
2
G
s
:
s

p
e , n +1
e , n +1
2
2 vm ,e ,n +1

vm ,n +1 = vm ,e ,n +1 1 2G p

vm ,n +1 = vm ,e ,n +1 3G p

4.61

2 vm,e, n +1
3

If hardening is present in the metal, the stress scale factor


must be calculated from the increment in equivalent plastic
strain.
The equivalent plastic strain increment is solved from the
yield condition at time n+1, for linear hardening this can be
done in a single step :

_ vm , n +1 y , n +1 = 0

3G

_ vm ,e, n +1 1
p y , n H p = 0
vm ,e, n*1

_ vm ,e, n +1 3G p y , n H p = 0

_ p =

vm,e , n +1 y , n
3G + H

The last formula allows to determine the incremental plastic


strain, which in turn allows to calculate the stress scale
factor k from the yield stress value at t(n+1)
4.62

Clearly nothing changes in principle, the stress scale factor


must be based on the updated value of the yields stress :

y ,n+1
_k =
vm,e,n+1

4.63

Algorithmic Setup :

The algorithmic setup illustrates the simplicity of radial


return :




se , n +1 = sn +1 + 2G d , n +1 / 2 t
pn +1 = pn +1 ( v , n +1 )

3

se, n +1 : se , n +1 ? y , n
2
IF _ YES _ :



n +1 = se, n +1 pn +1I

vm,e , n +1 =

IF _ NO _ :
p =

vm,e , n +1 y , n

3G + H
y , n +1 = y , n + H p

y , n +1

n +1 = se, n +1
pn +1I


vm,e , n +1

4.64

Plastic flow in shells and thickness change :

The plastic flow corresponds to a discontinuity in the Poisson


effect of the material (0.3 in the elastic region and 0.5 in the
plastic region).
Plastic deformation occurs at constant volume.
Plane stress plasticity must be solved for iteratively since the
through-the-thickness strain component is not known from
the shell kinematics :

A first elastic estimate for incremental strains is made :


xx = xx t
yy =  yy t
xy = xy t
yz =  yz t
zx = zx t

zz =
( xx + yy )
1

4.65

The apply radial return and check the stress component in


the through-the-thickness direction :





s e = 2G d = s e ,n +1 s n
p = K v = p n +1 p n

vm ,e,n +1 =


3
s e, n +1 : s e,n +1
2

y ,n +1


s n +1 = s e,n +1

vm,e,n +1

zz ,n +1 = zz

y ,n +1
= 2G d , zz
+ K v = 0 ?
vm,e,n +1

If the state-of-stress is plastic, the plane stress condition will


not be fulfilled and we calculate a second point with :

zz = ( xx + yy )
(Assume a fully plastic increment).

4.66

The true z-strain increment is then determined by secant


iteration :

i +1
zz

i 1
zz

i 1
zz

i-1

zzi zzi 1
zzi zzi 1

i+1

3 options are possible :


radial return
3 iterations
full iterative plasticity

(set on CONTROL_SHELL)
4.67

This allows correct thickness updates of the shell if required


on the CONTROL_SHELL card. (ISTUPD=1)

t/2

t =

zz

dt

t / 2

Influence on element membrane and bending stiffness is


usually negligeable in crashworthiness applications. If this
option is activated, bulk viscosity should be added to shell
elements in order to ensure numerical stability.

4.68

visco-plasticity

The von Mises yield condition is replaced by a constitutive


law introducing a viscous stress :
vm y ( p ) 0
vm y ( p ) v ( p ) = 0

in a tabulated approach this can be rewritten as :


vm A0 ( p , 0 ) [ A0 ( p , p ) A0 ( p , 0 ) ] = 0
vm A0 ( p , p ) = 0

this condition must be fulfilled iff the state of stress is


viscoplastic, in general we use the plastic consistency
parameter
= p
= p
vm ,n +1 A0 ( n +1 , n +1 ) = 0

4.69

stability of the numerical algorithm will require a


backward Euler approach :

n +1 =
t

vm ,n +1 A0 n + ,
=0

introducing the elastic trial stress and the associated flow


rule completes the equation to tbe solved :
e
vm ,n +1 = vm
,n +1 3G

e
n + ,
vm
=0
,n +1 3G A0

in the general case this equation is solved by Ridders


method
finally update deviatoric stresses as usual :

s n+1

3G p

= 1
se,n+1

vm ,e ,n +1

4.70

for linear hardening and viscosity the solution requires no


iteration :

vm
=0
,n +1 3G A0 n + ,

e
vm

3
G

=0
,n +1
y ,n
t
e
vm
,n +1 y ,n
=

3G + H +
t

power laws are often used to describe hardening and rate


effects

4.71

closest point algorithm

in the closest point algorithm an implicit approach is used


to estimate the direction of the plastic strain rate :



3 
s = 2G d 2G p
s
2vm




se ,n +1 sn

sn +1 sn
3

=
2G
p,n + 1sn +1
t
t
2vm ,n +1
2

3 
sn +1 1 + G p
= se ,n +1

vm ,n +1

4.72

giving for the second invariant :

3
2
2

vm ,n +1 1 + G p

vm ,e ,n +1

vm ,n +1
2

2
( vm ,n +1 + 3G p ) = vm
,e ,n +1
2
2
2
2
vm
+
6
G

+
9
G

,n +1
p vm ,n +1
p
vm ,e ,n +1

and for the plastic increment (as before) :


e
vm ,n +1 = vm
,n +1 3G

3
G

A
+

,
=0
vm

,n +1
0 n

the stress update :


sn + 1


se ,n +1
=

3G p
1 +

vm ,n +1

3G p

= se ,n +1 1

vm ,e ,n +1

4.73

4.74

Vous aimerez peut-être aussi