Professional Documents
Culture Documents
Chap 4 FEA For Elastoplastic Problems: Nam-Ho Kim
Chap 4 FEA For Elastoplastic Problems: Nam-Ho Kim
Nam-Ho Kim
Introduction
• Elastic material: a strain energy is differentiated by
strain to obtain stress
– History-independent, potential exists, reversible, no permanent
deformation
• Elatoplastic material:
– Permanent deformation for a force larger than elastic limit
– No one-to-one relationship between stress and strain
– Constitutive relation is given in terms of the rates of stress and
strain (Hypo-elasticity)
– Stress can only be calculated by integrating the stress rate over
the past load history (History-dependent)
• Important to separate elastic and plastic strain
– Only elastic strain generates stress
Table of Contents
• 4.2. 1D Elastoplasticity
• 4.3. Multi-dimensional Elastoplasticity
• 4.4. Finite Rotation with Objective Integration
• 4.5. Finite Deformation Elastoplasticity with
Hyperelastcity
• 4.6. Mathematical Formulation from Finite Elasticity
• 4.7. MATLAB Code for Elastoplastic Material Model
• 4.8. Elastoplasticity Analysis Using Commercial Programs
• 4.9. Summary
• 4.10. Exercises
4.2
1D Elastoplasticity
Goals
• Understand difference between elasticity and plasticity
• Learn basic elastoplastic model
• Learn different hardening models
• Understand different moduli used in 1D elastoplasticity
• Learn how to calculate plastic strain when total strain
increment is given
• Learn state determination for elastoplastic material
Plasticity
• Elasticity – A material deforms under stress, but then
returns to its original shape when the stress is removed
• Plasticity - deformation of a material undergoing non-
reversible changes of shape in response to applied forces
– Plasticity in metals is usually a consequence of dislocations
– Rough nonlinearity
• Found in most metals, and in general is a good description
for a large class of materials
• Perfect plasticity – a property of materials to undergo
irreversible deformation without any increase in stresses
or loads
• Hardening - need increasingly higher stresses to result in
further plastic deformation
Behavior of a Ductile Material
Terms Explanation
Proportional limit The greatest stress for which the stress is still proportional to
the strain
Elastic limit The greatest stress without resulting in any permanent strain on
release of stress
Young’s Modulus Slope of the linear portion of the stress-strain curve
Yield stress The stress required to produce 0.2% plastic strain
Strain hardening A region where more stress is required to deform the material
Ultimate stress The maximum stress the material can resist
Necking Cross section of the specimen reduces during deformation
V
Ultimate
stress Fracture
Yield stress
Proportional
limit Young’s
modulus
H
Strain Necking
hardening
Elastoplasticity
• Most metals have both elastic and plastic properties
– Initially, the material shows elastic behavior
– After yielding, the material becomes plastic
– By removing loading, the material becomes elastic again
• We will assume small (infinitesimal) deformation case
– Elastic and plastic strain can be additively decomposed by
H H e Hp
– Strain energy density exists in terms of elastic strain
U0 1 E( H )2
2 e
ʍ ʍ
^ƚƌĂŝŶŚĂƌĚĞŶŝŶŐƐůŽƉĞ͕ƚ
/ŶŝƚŝĂůůŽĂĚŝŶŐ
ȴʍ
ůĂƐƚŝĐ ʍz
ZĞůŽĂĚŝŶŐ
ƐůŽƉĞ͕
ȴɸĞ ȴɸƉ
hŶůŽĂĚŝŶŐ
ɸz ɸ
ȴɸ ɸ
Ž ɸ
Plastic Modulus
• Strain increment 'H 'He 'Hp ʍ
^ƚƌĂŝŶŚĂƌĚĞŶŝŶŐƐůŽƉĞ͕ƚ
• Stress increment 'V E'He
'V
• Plastic modulus H ȴʍ
'Hp ʍz
• Relation between moduli ȴɸĞ ȴɸƉ
Analysis Procedure
• Analysis is performed with a given incremental strain
– N-R iteration will provide 'u º 'HH
– But, we don’t know 'HHe or 'HHp
• When the material is in the initial elastic range, regular
elastic analysis procedure can be used
• When the material is in the plastic range, we have to
determine incremental plastic strain
ʍ
'V H'Hp
'H 'He 'Hp 'Hp 'Hp
E E
§H · ȴʍ
'Hp ¨ 1 ¸ ʍz
©E ¹
'H ȴɸĞ ȴɸƉ
'Hp
1H/E
ɸz ɸ
Only when the material is on the plastic curve!! ȴɸ
1D Finite Element Formulation
• Load increment
– applied load is divided by N increments: [t1, t2, …, tN]
– analysis procedure has been completed up to load increment tn
– a new solution at tn+1 is sought using the Newton-Raphson method
– iteration k has been finished and the current iteration is k+1
• Displacement increments
n 1 k
– From last increment tn: 'dk d nd u1 ½
n 1 k 1
d ® ¾
– From previous iteration: Gdk d n 1dk ¯ u2 ¿
X X
3 3
[ [
/
1D FE Formulation cont.
• Interpolation
'u ½
'u(x) [N1 N2 ] ® 1 ¾ N 'd
¯ 'u2 ¿ Gu N Gd
d ª 1 1 º 'u1 ½ GH B Gd
'H
dx
'u «L ® ¾
L »¼ ¯ 'u2 ¿
B 'd
¬
u Nd
H Bd
• Weak form (1 element)
– Internal force = external force
L u1 ½
d T ³ BT n 1 Vk 1Adx d T n 1F, d R2 d ® ¾
0 ¯u2 ¿
1D FE Formulation cont.
• Stress-strain relationship (Incremental)
n 1 wV
Vk 1 | n 1
Vk GH n 1
V k D ep GH
wH
– Elastoplastic tangent modulus
E if elastic
Dep ®
¯ Et if plastic
d T ª ³ BTDepBAdx º Gd
L L
d T n 1F d T ³ BT n 1 VkAdx
«¬ 0 »¼ 0
1D FE Formulation cont.
• Tangent Stiffness
ADep ª 1 1 º
kT
L «¬ 1 1 »¼
• Residual
L ° n 1F1 n 1 VkA ½°
n 1 k
R F³ B
n 1 T n 1 k
V Adx ® n 1 ¾
n 1 k
0
¯° F2 V A ¿°
• State Determination: n 1 Vk f( n V, n Hp , 'Hk , !)
Will talk about next slides
ɸƉ ɸĞ ɸ ɸƉ ɸ
V Yn
V Yn
ftr V tr Vny ɸ
ȴɸ
ftr (1 R)E'H
R: Fraction of 'Vtr to the yield stress
State Determination (Isotropic Hardening) cont.
• If ftr d 0 , material is elastic
Vn 1 Vtr
Either initial elastic region or unloading
1. If ftr d 0 (elastic)
n 1
– Remain elastic: V V tr , Hnp 1 Hnp ; exit
2. If ftr ! 0 (plastic)
ftr
a. Calculate plastic strain: 'Hp
EH
b. Update stress and plastic strain (store them for next increment)
Vn 1 Vtr sgn(Vtr )E'Hp Hnp1 Hnp 'Hp
ʍ
ʍ Vn
Vn V Y
V Y
V Y D n V Y
Dn ϮɸƉ ɸ
ɸƉ ɸĞ ɸ
• Elastic predictor
Vtr Vn E'H, D tr Dn , Ktr Vtr D tr ʍ
ʍƚƌ Đ
• Check yield status
Trial yield function ʍŶнϭ
Ě Ğ
f tr
K tr
V 0y
Ktr
• If ftr d 0, material is elastic ʍŶ Ă ď
Vn 1 Vtr ɲŶнϭ Ő
ɲŶ
Ĩ ɸ
Either initial elastic region or unloading ȴɸ
D tr Dn
Ktr Vtr D tr
ftr Ktr V0Y
2. If ftr d 0 (elastic)
n 1
– Remain elastic: V V tr , Dn 1 Dn , Hnp 1 Hnp ; exit
3. If ftr ! 0 (plastic)
ftr
a. Calculate plastic strain: 'Hp
EH
b. Update stress and plastic strain (store them for next increment)
Vn 1 Vtr sgn(Ktr )E'Hp Dn 1 Dn sgn(Ktr )H'Hp
• Since trf tr
K 0 VY ! 0, material yields in compression
• Plastic strain f tr
'Hp 0.444 u 10 3
EH
• State update n 1
V tr
V sgn( tr K)E'Hp 161.1MPa
n 1
D tr
D sgn( tr K)H'Hp 38.9MPa
Combined Hardening Model
• Baushinger effect
– conditions where the yield strength of a metal decreases when the
direction of strain is changed
– Common for most polycrystalline metals
– Related to the dislocation structure in the cold worked metal. As
deformation occurs, the dislocations will accumulate at barriers
and produce dislocation pile-ups and tangles.
• Numerical modeling of Baushinger effect
– Modeled as a combined kinematic and isotropic hardening
0dEd1
E = 0: isotropic hardening
E = 1: kinematic hardening
ftr
'Hp
EH
MATLAB Program combHard1D
%
% 1D Linear combined isotropic/kinematic hardening model
%
function [stress, alpha, ep]=combHard1D(mp, deps, stressN, alphaN, epN)
% Inputs:
% mp = [E, beta, H, Y0];
% deps = strain increment
% stressN = stress at load step N
% alphaN = back stress at load step N
% epN = plastic strain at load step N
%
E=mp(1); beta=mp(2); H=mp(3); Y0=mp(4); %material properties
ftol = Y0*1E-6; %tolerance for yield
stresstr = stressN + E*deps; %trial stress
etatr = stresstr - alphaN; %trial shifted stress
fyld = abs(etatr) - (Y0+(1-beta)*H*epN); %trial yield function
if fyld < ftol %yield test
stress = stresstr; alpha = alphaN; ep = epN;%trial states are final
return;
else
dep = fyld/(E+H); %plastic strain increment
end
stress = stresstr - sign(etatr)*E*dep; %updated stress
alpha = alphaN + sign(etatr)*beta*H*dep; %updated back stress
ep = epN + dep; %updated plastic strain
return;
Ex) Two bars in parallel cont.
while conv > tol && iter < 20
delu = Res / (Dep1*A1/L1 + Dep2*A2/L2);
u = u + delu;
delE = delu / L1;
[Snew1, Anew1, epnew1]=combHard1D(mp1,delE,nS1,nA1,nep1);
[Snew2, Anew2, epnew2]=combHard1D(mp2,delE,nS2,nA2,nep2);
Res = P - Snew1*A1 - Snew2*A2;
conv = Res^2/(1+P^2);
iter = iter + 1;
Dep1 = E1; if epnew1 > nep1; Dep1 = Et1; end
Dep2 = E2; if epnew2 > nep2; Dep2 = Et2; end
nS1 = Snew1; nA1 = Anew1; nep1 = epnew1;
nS2 = Snew2; nA2 = Anew2; nep2 = epnew2;
fprintf('\n %3d %7.4f %7.3f %7.3f %7.3f %7.3f %8.6f %8.6f %10.3e',...
iter,u,nS1,nS2,nA1,nA2,nep1,nep2,Res);
end
Summary
• Plastic deformation depends on load-history and its
information is stored in plastic strain
• Stress only depends on elastic strain
• Isotropic hardening increases the elastic domain, while
kinematic hardening maintains the size of elastic domain
but moves the center of it
• Major issue in elastoplastic analysis is to decompose the
strain into elastic and plastic parts
• Algorithmic tangent stiffness is consistent with the state
determination algorithm
• State determination is composed of (a) elastic trial and (b)
plastic return mapping
1D Elastoplastic Analysis Using ABAQUS
• Material Card
*MATERIAL,NAME=ALLE
*ELASTIC
200.E3,.3
*PLASTIC
200.,0.
220.,.0009
220.,.0029
Plastic strain
Yield stress
4.3
Multi-Dimensional
Elastoplastic Analysis
Goals
• Understand failure criteria, equivalent stress, and
effective strain
• Understand how 1D tension test data can be used for
determining failure of 3D stress state
• Understand deviatoric stress and strain
• Understand the concept of elastic domain and yield
surface
• Understand hardening models
• Understand evolution of plastic variables along with that
of the yield surface
Multi-Dimensional Elastoplasticity
• How can we generalize 1D stress state (V11) to 3D state (6
components)?
– Need scalar measures of stress and strain to compare with 1D test
– Equivalent stress & effective strain
– Key ingredients: yield criteria, hardening model, stress-strain
relation
• We will assume small (infinitesimal) strains
• Rate independent elastoplasticity- independent of strain
rate
• Von Mises yield criterion with associated hardening model
is the most popular
Failure Criteria
• Material yields due to relative sliding in lattice structures
ʹVz
Ud d Ud (tension test)
Hm 1 tr( H )
3
1H
3 v
1 (H
3 11
H22 H33 )
e H Hm 1 Idev : H Idev I 31 1 1
Failure Criteria cont.
• Example: Linear elastic material
V [O1
1 2PI] : H { D : H
volumetric deviatoric
Bulk modulus
3O 2P
• Distortion energy density K
3
U 1V :H 1 (V 1 s) : ( H m 1 e ) 3V H 21 s : e
2 2 m 2 m m
Ud 1s :e 1 s :s
2 4P
Ud 1 s :s 1 2 V2 1 V2
1D 4P 4P 3 6P
Ud 1 s :s 1 V2 1V e Effective strain
4P 6P e 2 e e
ee 1 V 1 3s :s 1 3 2Pe : 2Pe 2e :e
3P e 3P 2 3P 2 3
ª2 0 0 º 2
(1 Q)H « » § (1 Q)H ·
e 0 1 0 e:e 6¨ ¸
3 « » © 3 ¹
¬« 0 0 1 ¼»
2(1 Q)
ee 2e
3
:e H Effective strain for 1D tension
3
ɸ ɸ
Von Mises Criterion
• Material yields when Ve = VY J2 1s :s
2
Ve 3s
2
:s 3J2 2nd invariant of s
1
J2 ª (Vx V y )2 (V y Vz )2 ( Vz Vx )2 º W2xy W2yz W2zx
6¬ ¼
1
J2 ª (V V2 )2 (V2 V3 )2 (V3 V1 )2 º In terms of principal stresses
6¬ 1 ¼
• Yield criterion ʍϮ
J2 1 [s
2
: s tr(s)2 ] 1s
2
:s
Impossible
• Von Mises yield function s2
state
Elastic
state
3J2 V2Y 0 Material
yields
3s
2
: s V2Y 0 s1
s:s 2V
3 Y
0
s 2V
s 2V
3 Y
0 3 Y
radius
s s:s W2 W2 W 2
– Yield surface:
1 V 6DIHUHJLRQ
2W 2V
3 Y
W VY
3
Example
• Uniaxial tensile test
ªV 0 0º ª2V 0 0 º
«0 0 0» «3 »
V « » s « 0 3V
1 0 »
«¬ 0 0 0 »¼ « 0 0 31 V »¼
¬
s 4 V2 91 V2 91 V2 2V
9 3
– Yield surface
2V 2V 0 V VY Consistent with uniaxial
3 3 Y
tension test
V V
Hardening Model
• For many materials, the yield surface increases
proportional to plastic deformation º strain hardening
• Isotropic hardening: Change in radius
• Kinematic hardening: Change in center
/ƐŽƚƌŽƉŝĐŚĂƌĚĞŶŝŶŐ <ŝŶĞŵĂƚŝĐŚĂƌĚĞŶŝŶŐ
ʍϮ ʍϮ
ʍϭ ʍϭ
/ŶŝƚŝĂůLJŝĞůĚƐƵƌĨĂĐĞ /ŶŝƚŝĂůLJŝĞůĚƐƵƌĨĂĐĞ
– H = 0: elasto-perfectly-plastic material
s2
• Kinematic hardening model (linear)
s
– The center of yield surface : back stress D
D
– Shifted stress: K = s – D s1
K 2V
3 Y
0
Radial
– D moves proportional to ep K
D p
2He direction
3 K
Hardening Model cont.
• Combined Hardening
– Many materials show both isotropic and kinematic hardenings
– Introduce a parameter E [0, 1] to consider this effect
– Baushinger effect: The yield stress increases in one directional
loading. But it decreases in the opposite directional load.
– This is caused by dislocation pileups and tangles (back stress).
When strain direction is changed, this makes the dislocations easy
to move
K 2 [V 0
3 Y (1 E)Hep ] 0
K
D p
2E He
3 K
– Isotropic hardening: E = 0
– Kinematic hardening: E = 1
a) Isotropic hardening
s 2 ( V0
3 Y Hep ) 2V
3
2 (400
3
200 u 0.1) 0
V 420 MPa
b) Kinematic hardening
sD 2 V0
3 Y
0
s D 2 V0
3 Y
2V
3
2He
3 p 2V0
3 Y
0
V 420 MPa
Ex) Uniaxial Bar with Hardening
c) Combined hardening
Rate-Independent Elastoplasticity
• Additive decomposition
H H e Hp H H e H p From small deformation assumption
D (O 23 P)1
1 2PIdev
Rate-Independent Elastoplasticity cont.
• Stress cont.
– Volumetric stress: V m 3
)
1 tr( V (O 23 P)tr(H ) (3O 2P)H m
f( K, ep ) K 2 N(e )
3 p d0 Vtr
g3
g2
g1
Rate-Independent Elastoplasticity cont.
• Associative flow rule
– Flow potential = yield function
wf( K, [)
H p J Unit deviatoric tensor
wK normal to the yield surface
wf w K w K:K K K
N
wK wK wK K:K K
wf K
H p J J J N
wK K
e p 2e
3
p : e p 2
3
e p
– Note: plastic deformation only occurs in deviatoric components
H p e p H p J N J
[ (D, ep )
e p 2
3
J [ (V, [)
Jh
Rate-Independent Elastoplasticity cont.
• Kuhn-Tucker conditions
– The plastic consistency parameter must satisfy
J t 0 J f 0 fd0
b. Neutral loading f 0 J 0 J f 0
Classical Elastoplasticity
• Elastoplasticity boils down to how to calculate plasticity
consistency parameter
• Classical plasticity uses the rate form of evolution
relations to calculate it
• Plastic consistency condition J f 0
– J is only non-zero when continues plastic deformation
J ! 0 V, [)
f( 0
V, [) wf wf
f( : V [ 0
wV w[
wf wf
: D : (H H p ) J h 0
wV w[
wf wf wf
: D : H : D : J r J h 0 Solve for plastic
wV wV w[ consistency parameter
Classical Elastoplasticity cont.
• Plastic consistency parameter
wf Assume the denominator is positive
: D : H
wV x if x ! 0
J x ®
wf wf ¯ 0 if x d 0
:D:r h
wV w[
wf
J ! 0 : D : H ! 0
wV
wf
: D : H
cos T wV wf
E normal
wf T wV
V
D : H
wV
T < 90o : plastic loading D : H trial stress rate
T = 90o : neutral loading
T > 90o : elastic unloading
V D : (H H p )
wf
wV
: D : H
V D : H D : J r D : H D : r wf wf
wV
:D:r w[
h
ª D : r
wwVf : D º
V «D » : H
wf wf
«¬ : D : r w[ h »
wV ¼
D : r
wwVf : D
Dep D wf wf
wV
:D:r w[
h Elastoplastic tangent operator
Nonlinear Hardening Models
• Nonlinear kinematic hardening model
§ ep ·
D H(ep )e p H(ep ) H0 exp ¨ f ¸ Saturated hardening
¨ e ¸
© p ¹
N(ep ) V 0Y (1 E)Hep D p
2 E He
3
wf wf wf
f : s : D ep N : s N : D 2 (1 E)Hep 0
ws wD wep 3
s
2P e e p 2Pe 2PJN
D p
2 E He 2 E HJN
3 3
ep 2J
3
Example: Linear hardening model cont.
• Plastic consistency parameter cont.
f 2PN : H 2PJN : N 23 EHJN : N 23 (1 E)HJ 0
N:N 1
2PN : H N : e N : H
J
2P 23 H
– No iteration is required
• Elastoplastic tangent stiffness
V D : H D : H p D : H JD : N D (O 23 P)1
1 2PIdev
D:N 2PN
2PN : H ª 4P2 º
V D : H 2PN «D N
N » : H
2P 23 H «¬ 2P 23 H »¼
Dep
ª 0.1 0 0 º ª 0.08 0 0 º
'H « 0 0.02 0 », 'e « 0 0.04 0 »
« » « »
«¬ 0 0 0.02 »¼ «¬ 0 0 0.04 »¼
Ex) Plastic Deformation of a Bar
• Purely elastic at tn: nD = 0, nep = 0. nK = ns – nD = ns
• Trial states:
ª 360 0 0 º
tr
K tr
s n
s 2P'e « 0 180 0 » MPa
« »
«¬ 0 0 180 ¼»
tr
K 3602 1802 1802 180 6MPa
ª2 0 0 º
tr
K 1 « »
N 0 1 0
tr
K 6« »
«¬ 0 0 1 »¼
• Yield function
f( tr K, tr ep ) tr
K 2 N( n e )
3 p 180 6 300 2
3
80 6 ! 0
Plastic state!
ª 1.54 0 0 º
n 1
D n
D 23 EHJN « 0 0.77 0 » MPa No
« » equilibrium!!
«¬ 0 0 0.77 ¼»
n 1
ep n
ep 2J
3
0.0774
Numerical Integration
• Plastic evolution is given in the rate form
• We will use backward Euler method to integrate it
yn 1 yn
y f(t, y) f(tn 1 yn 1 )
't
Numerical Integration cont.
2. Plastic corrector cont.
2. If f > 0 (return mapping to yield surface)
unknown 'Hp 'JN
sn 1 str 2P 'H p
sn 1 str 2P'JN
Dn 1 D tr HD 'JN
Kn 1 Ktr
N Known from trial state
Kn 1 Ktr So, everything boils down to 'J
f( Kn 1 , epn 1 ) 0
f( Kn 1 , epn 1 ) Kn 1 2 N(en 1 )
3 p 0
Numerical Integration cont.
2. Plastic corrector cont.
– Plastic consistency condition
Ktr 2P HD (epn 1 ) 'J 2 N(en 1 )
3 p 0
f
Ktr 2P HD (epn 1 ) 'J 2 N(en 1 )
3 p
df dHD dN
(2P HD ) 2 'J 23
dJ 3 dep dep
f
'J 'J
df / dJ
epn 1 epn 1 2 'J
3
– Stop when f ~ 0
Vn 1 Vn D : 'H 2P'JN
Difference from the Rate Form
• Rate form (linear hardening)
2PN : H f(nI,nep) = 0 ns
2N%e
J trs
2P 23 H f(trI,trep)
• Incremental form nB
2NN%e
Ktr 2 N(en )
3 p
'J
2P 2H
3
w'V w'J wN
D 2PN
2P'J
w'H w'H w'H
(1) (2)
Consistent Tangent Operator cont
wf
• Term (1) f( Kn 1 , epn 1 ) f( Kn , epn ) 0 0
w'HH
wf w ª tr 2 N(en 1 ) º
K (2P HD (epn 1 ))'J 0
w'H w'H ¬ 3 p ¼
w'J
w'J 2
2PN 2P HD (epn 1 ) 2H 'J 23 N,e 0
3 D,ep p
w'HH
3 w'HH
w'J 1
2P HD (epn 1 )
2PAN A
2H
3 D,ep
'J 23 N,e
w'H
p
wN 1 2P
ª¬ I N
N º¼ : 2PIdev ª¬ Idev N
N º¼
w'H Ktr Ktr
HD (epn 1 ) 2 EH
3
HD,e 0
p
• Consistency condition
f Ktr 2P 23 EH 'J 2
3 N(e )
n
p
2 (1
3
E)H'J 0
Ktr 2 N(en )
3 p
'J No iteration is required
2P 2H
3
1 2
2P H
A 3
Variational Equation
• Variational equation
a( n [ ; n 1 u, u ) A( u ), u ]
a( n [ ; n 1 u, u ) { ³³: H( u ) :
n 1
V d:
a * ( n [, n 1 u ; G u, u ) { ³³: H ( u ) : D
alg
: H ( Gu) d : .
• Update displacement
n1 k 1 n1 k
u u Guk
Implementation of Elastoplasticity
• We will explain for a 3D solid element at a Gauss point
• Voigt notation
{V } [ V11 V 22 V 33 V12 V 23 V13 ]T
{ 'H } [ 'H11 'H22 'H33 2 'H12 2 'H23 2 'H13 ]T
^ 'dI1 'dI3 `
T
• Inputs 'dI 'dI2
^ Vn11 `
T
Vn Vn22 Vn33 Vn12 Vn23 Vn13
^ Dn11 `
T
[n Dn22 Dn33 Dn12 Dn23 Dn13 epn
]
[ [ ±± ±
±
[
[
K
[
[ ±±
[
±±
x
±
[ [
x
x D)LQLWH(OHPHQW E5HIHUHQFH(OHPHQW
T
– Unit tensor 1 ª¬ 1 1 1 0 0 0 º¼
Vn 1 V tr D alg D
– Exit
Implementation of Elastoplasticity cont.
ª 2 1 1 0 0 0º
• Consistent tangent matrix « 31 3 3
»
«3 31 0 0
2 0»
3
4P2 4P2 'J « 1 1 2 0 0 0»
c1 c2 « 3 »
2P 23 H Ktr I dev 3 3
« 0 0 0 21 0 0»
« »
« 0 0 0 0 21 0»
D alg D (c1 c2 )NNT c2 Idev « 1»
¬ 0 0 0 0 0 2¼
Program combHard.m
%
% Linear combined isotropic/kinematic hardening model
%
function [stress, alpha, ep]=combHard(mp,D,deps,stressN,alphaN,epN)
% Inputs:
% mp = [lambda, mu, beta, H, Y0];
% D = elastic stiffness matrix
% stressN = [s11, s22, s33, t12, t23, t13];
% alphaN = [a11, a22, a33, a12, a23, a13];
%
Iden = [1 1 1 0 0 0]';
two3 = 2/3; stwo3=sqrt(two3); %constants
mu=mp(2); beta=mp(3); H=mp(4); Y0=mp(5); %material properties
ftol = Y0*1E-6; %tolerance for yield
%
stresstr = stressN + D*deps; %trial stress
I1 = sum(stresstr(1:3)); %trace(stresstr)
str = stresstr - I1*Iden/3; %deviatoric stress
eta = str - alphaN; %shifted stress
etat = sqrt(eta(1)^2 + eta(2)^2 + eta(3)^2 ...
+ 2*(eta(4)^2 + eta(5)^2 + eta(6)^2));%norm of eta
fyld = etat - stwo3*(Y0+(1-beta)*H*epN); %trial yield function
if fyld < ftol %yield test
stress = stresstr; alpha = alphaN; ep = epN;%trial states are final
return;
else
gamma = fyld/(2*mu + two3*H); %plastic consistency param
ep = epN + gamma*stwo3; %updated eff. plastic strain
end
N = eta/etat; %unit vector normal to f
stress = stresstr - 2*mu*gamma*N; %updated stress
alpha = alphaN + two3*beta*H*gamma*N; %updated back stress
Program combHardTan.m
function [Dtan]=combHardTan(mp,D,deps,stressN,alphaN,epN)
% Inputs:
% mp = [lambda, mu, beta, H, Y0];
% D = elastic stiffness matrix
% stressN = [s11, s22, s33, t12, t23, t13];
% alphaN = [a11, a22, a33, a12, a23, a13];
%
Iden = [1 1 1 0 0 0]';
two3 = 2/3; stwo3=sqrt(two3); %constants
mu=mp(2); beta=mp(3); H=mp(4); Y0=mp(5); %material properties
ftol = Y0*1E-6; %tolerance for yield
stresstr = stressN + D*deps; %trial stress
I1 = sum(stresstr(1:3)); %trace(stresstr)
str = stresstr - I1*Iden/3; %deviatoric stress
eta = str - alphaN; %shifted stress
etat = sqrt(eta(1)^2 + eta(2)^2 + eta(3)^2 ...
+ 2*(eta(4)^2 + eta(5)^2 + eta(6)^2));%norm of eta
fyld = etat - stwo3*(Y0+(1-beta)*H*epN); %trial yield function
if fyld < ftol %yield test
Dtan = D; return; %elastic
end
gamma = fyld/(2*mu + two3*H); %plastic consistency param
N = eta/etat; %unit vector normal to f
var1 = 4*mu^2/(2*mu+two3*H);
var2 = 4*mu^2*gamma/etat; %coefficients
Dtan = D - (var1-var2)*N*N' + var2*Iden*Iden'/3;%tangent stiffness
Dtan(1,1) = Dtan(1,1) - var2; %contr. from 4th-order I
Dtan(2,2) = Dtan(2,2) - var2;
Dtan(3,3) = Dtan(3,3) - var2;
Dtan(4,4) = Dtan(4,4) - .5*var2;
Dtan(5,5) = Dtan(5,5) - .5*var2;
Dtan(6,6) = Dtan(6,6) - .5*var2;
Program PLAST3D.m
function PLAST3D(MID, PROP, ETAN, UPDATE, LTAN, NE, NDOF, XYZ, LE)
%***********************************************************************
% MAIN PROGRAM COMPUTING GLOBAL STIFFNESS MATRIX RESIDUAL FORCE FOR
% PLASTIC MATERIAL MODELS
%***********************************************************************
%%
....
%LOOP OVER ELEMENTS, THIS IS MAIN LOOP TO COMPUTE K AND F
for IE=1:NE
DSP=DISPTD(IDOF);
DSPD=DISPDD(IDOF);
%....
% LOOP OVER INTEGRATION POINTS
for LX=1:2, for LY=1:2, for LZ=1:2
%
% Previous converged history variables
NALPHA=6;
STRESSN=SIGMA(1:6,INTN);
ALPHAN=XQ(1:NALPHA,INTN);
EPN=XQ(NALPHA+1,INTN);
....
% Computer stress, back stress & effective plastic strain
if MID == 1
% Infinitesimal plasticity
[STRESS, ALPHA, EP]=combHard(PROP,ETAN,DDEPS,STRESSN,ALPHAN,EPN);
....
%
% Tangent stiffness
if LTAN
if MID == 1
DTAN=combHardTan(PROP,ETAN,DDEPS,STRESSN,ALPHAN,EPN);
EKF = BM'*DTAN*BM;
....
Summary
• 1D tension test data are used for 2D or 3D stress state
using failure theories
– All failure criteria are independent of coordinate system (must
defined using invariants)
• Yielding of a ductile material is related to shear stress or
deviatoric stress
• Kinematic hardening shift the center of elastic domain,
while isotropic hardening increase the radius of it
• For rate-independent J2 plasticity, elastic predictor and
plastic correct algorithm is used
• Return mapping occurs in the radial direction of deviatoric
stress
• During return mapping, the yield surface also changes
4.4
Elastoplasticity with Finite
Rotation
Goals
• Understand the concept of objective rate and frame-
indifference (why do we need objectivity?)
• Learn how to make a non-objective rate to objective one
• Learn different objective stress rates
• Learn how to maintain objectivity at finite rotation
• Understand midpoint configuration
• Understand how to linearize the energy form in the
updated Lagrangian formulation
• Understand how to implement update Lagrangian frame
u sym(u) skew(u)
strain rotation
• Objective tensor T Q T QT
Velocity Gradient
• In two different frames
Velocity gradient is related to incremental
wv wv displacement gradient in finite time step
L , L
wx wx w'u
L't |
wx
• Time differentiate of x Qx
x x
Qv Q
v Tx
Q v QQ Velocity is not objective
• Spatial differentiation of v
wv
L
wx
wv wx wx
Q Q Q T
wx wx wx
Q Velocity gradient is not objective
QLQ Q
T T
Rate of Deformation and Spin Tensor
• Rate of Deformation
d sym(L ) d sym(L )
d sym(L ) 1
2 Q L QT Q LT QT Q QT Q Q T
Q QT 1 Q T Q Q T
Q 0
d Q 21 (L LT ) Q T Q d QT Objective
This is incremental stain
• Spin tensor
W 1 (L
2
LT ) W 1 (L
2
LT )
W 1 (L LT ) 1 (Q QT Q Q
L Q T Q LT Q T Q T)
2 2
W Q W Q T 21 ( Q Q T Q Q
T)
E E
– Coordinate transformation is opposite to rotation x
E
[V]xyz [ Q][V]xyz [ Q]T
z
zc
Objective Rate
• If T is an objective tensor, will its rate be objective, too?
– This is important because in plasticity the constitutive relation is
given in terms of stress rate
• Differentiate an objective tensor T Q T QT
T QT Q T QT Q T Q
T
T Q
and Q
– Not objective due to Q T
Q L Q QL T
Q Q T LT LT Q T
T QT( Q T LT LT Q T )
T (LQ QL ) TQ T QTQ
LQTQ T QLTQ T QTQ T QTQ T LT QTLT Q T
T TLT QTLT Q T
LT QLTQ T QTQ
T LT TLT Q( T LT TLT )QT
T W T T W
• Convected rate
T LT T T L
• These objective rates are different, but perform equally
• When T is stress, they are objective stress rate
Finite Rotation and Objective Rate
• Since constitutive relation should be independent of the
reference frame, it has to be given in terms of
objective rate
• Cauchy stress is an objective tensor, but Cauchy stress
rate is not objective rate
• Instead of rate, we will use increment (from previous
converged load step to the current iteration)
• Consider a unit vector ej in spatial Cartesian coordinates
under rigid body rotation from material vector Ej
1 § w'u w'u ·
T
ej Q Ej W ¨ ¸ 'Q Q T
2 ¨ wx wx ¸
© ¹
W: spin tensor
'ej 'Q Ej W Q Ej W ej Q: rotation tensor
• Constitutive relation
Vn R Vn RT
This takes care of rigid body rotation
Dn R Dn R T
Linearization
• Increment of deformation gradient
w ª w( x Z'u) º w'u
'F 0 'u
wZ «¬ wX »
¼Z 0 wX
• Increment of Jacobian
Fmn 1e e F F F
'J 'F J div( 'u) 6 ijk rst ir js kt
Linearization cont.
• Linearization of energy form
' ³³ ( 0 uF 1 ) : VJ d:
:0
³³: 0
ª ( 0 u'F 1 ) : VJ ( 0 uF 1 ) : 'VJ ( 0 uF 1 ) : V'J º d:
¬ ¼
³³: 0
ª 0 uF 1n 1 'u : V 0 uF 1 : 'V 0 uF 1 : Vdiv( 'u) º Jd:
¬ ¼
³³: n 1
ª¬ n 1 un 1 'u : V n 1 u : 'V n 1 u : Vdiv( 'u) º¼ d:
³³: n 1
n 1 u : ª¬ 'V Vdiv( 'u) V(n 1 'u)T º¼ d:
Linearization cont.
• Linearization of energy form cont.
L[a([ n ; u n 1 , u )] ³³: n 1 u : ª¬ 'V Vdiv( 'u) V(n 1 'u)T º¼ d:
n 1
³³: n 1
n 1 u : ª¬ 'V J WV VW Vdiv( 'u) V(n 1 'u)T º¼ d:
– Spin term
1 ( w'ui w'um w'uk
WimVmj 2 wxm
wxi
)Vmj 1 V (G G
2 mj ik ml
Gmk Gil ) wxl
1 (V G VkjGil )[n 1 'u]kl
2 lj ik
VimWmj 1 (V G
2 il jk Vik G jl )[n 1 'u]kl
w'uk
Vij wxk
VijGkl[n 1 'u]kl
w'uj
Vim wxm
VilG jk [n 1 'u]kl
Linearization cont.
• Linearization of energy form cont.
L[a( u, u )] ³³: n 1
n 1 u : ª¬ 'V J WV VW Vdiv( 'u) V(n 1 'u)T º¼ d:
– Define
Dijkl
*
VijGkl 21 ( VilG jk V jlGik Vik G jl V jk Gil )
Linearization cont.
• Linearization of energy form cont.
L[a([ n ; u n 1 , u )] ³³: ª n 1 u : (D alg D * ) : n 1 'u V : K 'u u ) º d:
n 1
¬ ¼
{ a * ([ n , u n 1 ; 'u, u )
• N-R iteration
History-dependent Bilinear
(implicit)
Implementation
• We will explain using a 3D solid element at a Gauss point
using updated Lagrangian form
• The return mapping and consistent tangent operator will
be the same with infinitesimal plasticity
• Voigt Notation
{V } [ V11 V 22 V33 V12 V23 V13 ]T
{ 'H } [ 'H11 'H22 'H33 2 'H12 2 'H23 2 'H13 ]T
^ 'dI1 'dI3 `
T
• Inputs 'dI 'dI2
^ `
T
Vn Vn11 Vn22 Vn33 Vn12 Vn23 Vn13
^ Dn11 `
T
[n Dn22 Dn33 Dn12 Dn23 Dn13 epn
Implementation cont.
• In the updated Lagrangian, the derivative is evaluated at
the current configuration (unknown yet)
• Let the current load step is n+1 (unknown) and k+1 N-R
iteration
• Then, we use the configuration at the previous iteration
(n+1, k) as a reference
• This is not ‘true’ updated Lagrangian, but when the N-R
iteration converges, k is almost identical to k+1
• Caution: we only update stresses at the converged load
step, not individual iteration
• All derivatives and integration in updated Lagrangian must
be evaluated at (n+1, k) configuration
• Displacement increment 'u is from (n+1,0) to (n+1,k)
Implementation cont.
• Stress-displacement matrix (Two approaches)
1. Mapping between current (n+1, k) and reference configurations
ª wx1 wx2 wx3 º ª w º ª w º
« w[ w[ w[ » « wx » « w[ »
« » « 1» « »
« wx1 wx2 wx3 » « w » 1 « w »
J « wK J
wK wK » « wx » « wK »
« » « 2» « »
« wx1 wx2 wx3 » « w » « w »
« w] w] w] »¼ « wx » «¬ w] »¼
¬ ¬ 3¼
Implementation cont.
1. Obtain midpoint configuration (between k and k+1)
1
wxn ª 1 w'u º w'u w'u wxn
«1 2 n »
wxn 1 2 ¬ wx ¼ wxn 1 2 wxn wxn 1 2
2. Rotation matrix: R 1 (1 21 W) 1 W
3. Rotate stresses: Vn R Vn RT Dn R Dn RT
4. Return mapping with Vn , Dn
– This part is identical to the classical return mapping
– Calculate stresses: Vnk 11 , Dnk 11
– Calculate consistent tangent operator Dalg
Implementation cont.
• Internal force
4 NG This summation is similar to
fint
¦¦ (BIT Vnk 11 J )K ZK assembly (must be added to the
I 1K 1
corresponding DOFs)
2 V23
1
2 (V22 V33 )
1
2 V12
1
2 V12
1 »
»
2 (V11 V33 ) ¼»
1
I 1 J 1K 1
Implementation cont.
• Solve for incremental displacement
[KT KS ]{ Gdk 1 } { f ext } { fint }
• Update displacements
dkn11 dkn 1 Gdk 1
'dk 1 'dk 1 Gdk 1
Program PLAST3D.m
function PLAST3D(MID, PROP, ETAN, UPDATE, LTAN, NE, NDOF, XYZ, LE)
%***********************************************************************
% MAIN PROGRAM COMPUTING GLOBAL STIFFNESS MATRIX RESIDUAL FORCE FOR
% PLASTIC MATERIAL MODELS
%***********************************************************************
....
% Computer stress, back stress & effective plastic strain
elseif MID == 2
% Plasticity with finite rotation
FAC=FAC*det(F);
[STRESSN, ALPHAN] = rotatedStress(DEPS, STRESSN, ALPHAN);
[STRESS, ALPHA, EP]=combHard(PROP,ETAN,DDEPS,STRESSN,ALPHAN,EPN);
....
%
% Tangent stiffness
if LTAN
elseif MID == 2
DTAN=combHardTan(PROP,ETAN,DDEPS,STRESSN,ALPHAN,EPN);
CTAN=[-STRESS(1) STRESS(1) STRESS(1) -STRESS(4) 0 -STRESS(6);
STRESS(2) -STRESS(2) STRESS(2) -STRESS(4) -STRESS(5) 0;
STRESS(3) STRESS(3) -STRESS(3) 0 -STRESS(5) -STRESS(6);
-STRESS(4) -STRESS(4) 0 -0.5*(STRESS(1)+STRESS(2)) -0.5*STRESS(6) -0.5*STRESS(5);
0 -STRESS(5) -STRESS(5) -0.5*STRESS(6) -0.5*(STRESS(2)+STRESS(3)) -0.5*STRESS(4);
-STRESS(6) 0 -STRESS(6) -0.5*STRESS(5) -0.5*STRESS(4) -0.5*(STRESS(1)+STRESS(3))];
SIG=[STRESS(1) STRESS(4) STRESS(6);
STRESS(4) STRESS(2) STRESS(5);
STRESS(6) STRESS(5) STRESS(3)];
SHEAD=zeros(9);
SHEAD(1:3,1:3)=SIG;
SHEAD(4:6,4:6)=SIG;
SHEAD(7:9,7:9)=SIG;
EKF = BM'*(DTAN+CTAN)*BM + BG'*SHEAD*BG;
....
6KHDUVWUHVV03D
6PDOO VWUDLQ
)LQLWH
URWDWLRQ
6KHDUVWUDLQ
Summary
• Finite rotation elastoplasticity is formulated using
updated Lagrangian (reference frame moves with body)
• Finite rotation elastoplasticity is fundamentally identical
to the classical plasticity. Only rigid-body rotation is taken
into account using objective stress rate and integration
• We must use an objective stress rate to define the
constitutive relation because the material response should
be independent of coordinate system
• Objectivity only applies for spatial vectors and tensors
• In the finite rotation, the midpoint configuration is used
to reduce errors involved in non-uniform rotation and spin
• Linearization is performed after transforming to the
undeformed configuration
4.5
Finite Deformation
Elastoplasticity with
Hyperelasticity
Goals
• Understand the difference between hypoelasticity and
hyperelasticity
• Learn the concept of multiplicative decomposition and
intermediate configuration
• Understand the principle of maximum dissipation
• Understand the plastic evolution in strain space and stress
space
• Learn J2 plasticity in principal stress space
Finite Deformation Plasticity
• So far, we used small strain elastoplasticity theory
• Finite rotation has been taken care of using the deformed
configuration with an objective rate
• However, still, the strain should be small enough so that
the elastic and plastic strains are decomposed additively
• This is fundamental limitation of hypoelasticity
• How can we handle large strain problem?
• On the other hand, hyperelasticity can handle large strain
• However, it is not easy to describe plastic evolution in 2nd
P-K stress. It is given in the current configuration (Cauchy
stress)
• How can we handle it? Transformation between references
Intermediate Configuration
• Let’s take one step back and discuss different references
• Lee (1967) proposed that the deformation gradient can
be multiplicatively decomposed
F( X) Fe ( X)Fp ( X)
Additive decomposition: H H e Hp
Intermediate Configuration cont.
• Fp(X): deformation through the intermediate configuration
(related to the internal plastic variables)
• Fe-1(X): local, stress-free, unloaded process
• Decomposition of F(X) into the intermediate configuration
followed by elastic deformation
F
hŶĚĞĨŽƌŵĞĚ ƵƌƌĞŶƚ
ŽŶĨŝŐƵƌĂƚŝŽŶ ŽŶĨŝŐƵƌĂƚŝŽŶ
Fp Fe
ůĂƐƚŝĐ
ĞĨŽƌŵĂƚŝŽŶ
/ŶƚĞƌŵĞĚŝĂƚĞŽŶĨŝŐƵƌĂƚŝŽŶ
Elastic Domain and Free Energy
• Elastic domain
E { ^ ( W, q) f( W, q) d 0 `
w\
– strain-like internal variables vector: q
w[
Dissipation Function
• Dissipation function (ignoring thermal part)
d
D {W:d \(be , [) t 0
dt
• Thus, we have
be Lbe be LT Lv be
Elastic Plastic
d
D W:d \(be , [)
dt
w\ w\
W : d e : be [
wb w[
w\
W : d e : Lbe be LT Lv be q [
wb
w\ § w\ ·
W :d2 be : L ¨ 2 be ¸ : ¬ª 21 (Lv be )be1 ¼º q [
wbe © wbe ¹
§ w\ · § w\ · 1
¨ W 2 wb be ¸ : d ¨ 2 wb be ¸ : ª¬ 2 (Lv be )be º¼ q [ t 0
1
© e ¹ © e ¹
)RUDV\PPHWULFPDWULFHV$%& $&%
For a symmetric S and general L, S:L = S:sym(L)
Principle of Maximum Dissipation
• Principle of Maximum Dissipation
– For all admissible stresses and internal variables, the inequality
must satisfy
§ w\ · § w\ · 1
D W ¸ : ª¬ 2 (Lv be )be º¼ q [ t 0
2 b : d 1
¨ wbe e ¸ ¨2 be
© ¹ © wbe ¹
– If we consider the material is elastic, then no plastic variable will
change Lv be [ 0
– In order to satisfy the inequality for any d (especially d1 = - d2)
w\
W 2 b Constitutive relation
wbe e
D V : H p q [ t 0
wf wf
Hp & H p J
wV wV
Principle of Maximum Dissipation cont.
• Principle of Maximum Dissipation cont.
– For given rates {Lv be , [ }, state variables {W, q} maximize the
dissipation function D
D ( W W * ) : ª¬ 21 (Lv be )be1 º¼ ( q q* ) [ t 0, ^ `
W * , q* E
wf( W, q) w f( W, q)
21 Lv be J be [ J
wW wq
J t 0, f( W, q) d 0, Jf( W, q) 0.
– Plastic evolution is still in a rate form
– Stress is hyperelastic (total form)
– Plastic evolution is given in terms of strain (be and [)
– We need to integrate these equations
Time Integration
• Given: {Fn ben [n } and 'u
f
• Relative deformation gradient Fn
wxn 1 :Q
f( x) 1 n 'u :Q
wx n
:
Fn 1 f Fn
wu wu wxn 1
f Lf
wxn wxn 1 wxn
Time Integration cont.
• Elastic predictor cont.
betr Fetr FetrT f Fen FenT fT fben fT
f tr f betr fben fT [ tr [n
ben 1 betr [n 1 [ tr
Wn 1 W tr qn 1 qtr
Elastic Plastic
ª wf( W e , q) º
ben 1 betr exp « 2'J » 'J J't
¬ wW ¼
wf( W e , q)
[n 1 [ tr 'J
wq
'J t 0, f( W, q) d 0, 'J f( W, q) 0
Spectral Decomposition
• Objective: want to get a similar return mapping algorithm
with classical plasticity
• Return-mapping algorithm for principal Kirchhoff stress
• For isotropic material, the principal direction of W is
parallel to that of be
• Spectral decomposition
3 3
be ¦ Oi2 nˆi
nˆi W ¦ Wpi nˆi
nˆi
i 1 i 1
ben 1 betr exp ª¬ " º¼
\( e, [) 1 O[e ˆ [)
e2 e3 ]2 P[e12 e22 e32 ] K(
2 1
ª wf( W e , q) º
ben 1 betr exp « 2'J »
¬ wW ¼
ª wf( W, q) º
log(ben 1 ) log(betr ) log exp « 2'J
¬ w W »¼
wf
3
wfˆ
f( W, q) f( Wp , q)
wW
¦ wW nˆi
nˆi
i 1 pi
wf̂( Wp , q)
2en 1 2etr 2'J
wWp
wf̂( Wp , q)
ce en 1 ce etr 'Jce
wWp
wf̂( Wp , q)
Wp Wptr 'J ce W ptr c e e tr
wWp
wf̂( Wp , q)
[n 1 [n 'J
wq
'J t 0, ˆ W , q) d 0,
f( ˆ W , q)
'J f( 0
p p
Return Mapping Algorithm
• Deviatoric principal stress
s Wp 31 ( Wp 1 ˆ
ˆ)1 1dev Wp
• Yield function
K sD
f( K, ep ) K 2 N(e )
3 p d0
t
• Return mapping
ep ³0 2
3
ep (t) dt
W ptr c e e tr
D tr Dn
Wnp1 W ptr 2P'JN eptr epn
Dn 1 D tr 'JHDN Kn 1 Ktr
N
Kn 1 Ktr
epn 1 eptr 2 'J
3
– Solve for 'J using N-R iteration, or directly for linear hardening
– Derivative wf 1
(2P HD 2H
3 D,ep
'J 23 N,e )
w'J p
A
• Recovery
– Once return mapping converged, recover stress and strain
3 3
W n 1
¦ Wnpi1 n
n
ˆi ˆi ¦ Wnpi1 mi mi nˆi
nˆi
i 1 i 1
3
ben 1 ¦ exp(2ein 1 )mi en 1 e tr 'JN
i 1
Ex) Incompressible Elastic Cube
• Elastic deformation: x1 DX1 , x2 EX2, x3 EX3
• Deformation gradient
ªD 0 0º ª D2 0 0º
« 0 E 0 », « »
F « » b FFT « 0 E2 0»
« 0 E2 ¼»
¬« 0 0 E ¼» ¬ 0
• Incompressibility: det(F)=1 E 1/ D
• Eigenvalues and eigenvectors:
O1 D2 , n1 [1 0 0]T
O2 D 1 , n2 [0 1 0]T
O3 D 1 , n3 [0 0 1]T
• Logarithmic stretches:
^ 2log D log D `
T
e log D
• Kirchhoff stress
3 ª2 0 0 º
W ¦ Wipni
ni 2P log D «« 0 1 0 »»
i 1
«¬ 0 0 1 »¼
Consistent Tangent Operator
• Relation b/w material and spatial tangent operators
wS
D E : D : 'E [FT H( u )F] : D : [FT H( 'u)F] H( u ) : c : H( 'u)
wE
wSrs
cijkl FirFjsFkmFlnDrsmn FirFjsFkmFln
wEmn
– FirFjs: transform stress to material frame W = FSFT
– FkmFln: differentiate w.r.t. E and then transform to spatial frame
wWij
• But, cijkl z ww ww T
F F
wHkl wH wE
¦i 1 Wnpi1 mi
3
• Let Wn 1
wW w §
3 · 3 ª wWnpi1 iº
n 1 wm
c ¨¨ ¦ Wnpi1 mi ¸¸ ¦ « wH
m Wpi wH »
i
wH wH © i 1 ¹ i 1« »¼
¬
3 ª 3 wWnpi1 wejtr iº
n 1 wm
c ¦ «« ¦ wetr wH
m Wpi wH »»
i
i 1 ¬j 1 j ¼
Consistent Tangent Operator cont.
wWp consistent tangent operator in principal stress
(1) c alg Same as classical return mapping (3×3)
wetr
wejtr wejtr
(2) 2Fe FeT mj
wH wCe
These are elastic
wm i
wm Ti
(3) 2Fe F 2cˆi
wH wCe e
3 3 3
wW
c
wH
¦¦ cijalgmi
m ¦ 2Wpicˆi
j
i 1j 1 i 1
³³: 0
W : H( u ) d :
• Linearization
a * ( n [, n b e , u ; ' u, u ) { ³³ 0
:
ª¬ H ( u ) c : H ( ' u ) W K( ' u, u ) º¼ d :
• N-R iteration
a * ( n [, n 1 uk 1 ; ' uk 1 , u ) A( u ) a( n [ ; n 1 uk , u ), u ]
MATLAB Code MULPLAST
function [stress, b, alpha, ep]=mulPlast(mp,D,L,b,alpha,ep)
%mp = [lambda, mu, beta, H, Y0];
%D = elasticity matrix b/w prin stress & log prin stretch (3x3)
%L = [dui/dxj] velocity gradient
%b = elastic left C-G deformation vector (6x1)
%alpha = principal back stress (3x1)
%ep = effective plastic strain
%
EPS=1E-12;
Iden = [1 1 1]'; two3 = 2/3; stwo3=sqrt(two3); %constants
mu=mp(2); beta=mp(3); H=mp(4); Y0=mp(5); %material properties
ftol = Y0*1E-6; %tolerance for yield
R = inv(eye(3)-L); %inc. deformation gradient
bm=[b(1) b(4) b(6);b(4) b(2) b(5);b(6) b(5) b(3)];
bm = R*bm*R'; %trial elastic left C-G
b=[bm(1,1) bm(2,2) bm(3,3) bm(1,2) bm(2,3) bm(1,3)]';
[~,P]=eig(bm); %eigenvalues
eigen=sort(real([P(1,1) P(2,2) P(3,3)]))'; %principal stretch
%
% Duplicated eigenvalues
TMP=-1;
for I=1:2
if abs(eigen(1)-eigen(3)) < EPS
eigen(I)=eigen(I)+TMP*EPS;
TMP=-TMP;
end
end
if abs(eigen(1)-eigen(2)) < EPS; eigen(2) = eigen(2) + EPS; end;
if abs(eigen(2)-eigen(3)) < EPS; eigen(2) = eigen(2) + EPS; end;
%
% EIGENVECTOR MATRIX N*N' = M(6,*)
M=zeros(6,3); %eigenvector matrices
for K=1:3
KB=1+mod(K,3);
KC=1+mod(KB,3);
EA=eigen(K);
EB=eigen(KB);
EC=eigen(KC);
D1=EB-EA;
D2=EC-EA;
DA = 1 / (D1 * D2);
M(1,K)=((b(1)-EB)*(b(1)-EC)+b(4)*b(4)+b(6)*b(6))*DA;
M(2,K)=((b(2)-EB)*(b(2)-EC)+b(4)*b(4)+b(5)*b(5))*DA;
M(3,K)=((b(3)-EB)*(b(3)-EC)+b(5)*b(5)+b(6)*b(6))*DA;
M(4,K)=(b(4)*(b(1)-EB+b(2)-EC)+b(5)*b(6))*DA;
M(5,K)=(b(5)*(b(2)-EB+b(3)-EC)+b(4)*b(6))*DA;
M(6,K)=(b(6)*(b(3)-EB+b(1)-EC)+b(4)*b(5))*DA;
end
%
eigen=sort(real([P(1,1) P(2,2) P(3,3)]))'; %principal stretch
deps = 0.5*log(eigen); %logarithmic
sigtr = D*deps; %trial principal stress
eta = sigtr - alpha - sum(sigtr)*Iden/3; %shifted stress
etat = norm(eta); %norm of eta
fyld = etat - stwo3*(Y0+(1-beta)*H*ep); %trial yield function
if fyld < ftol %yield test
sig = sigtr; %trial states are final
stress = M*sig; %stress (6x1)
else
gamma = fyld/(2*mu + two3*H); %plastic consistency param
ep = ep + gamma*stwo3; %updated eff. plastic strain
N = eta/etat; %unit vector normal to f
deps = deps - gamma*N; %updated elastic strain
sig = sigtr - 2*mu*gamma*N; %updated stress
alpha = alpha + two3*beta*H*gamma*N; %updated back stress
stress = M*sig; %stress (6x1)
b = M*exp(2*deps); %updated elastic left C-G
end
Ex) Shear Deformation of a Square
Young = 24000; nu=0.2; mu=Young/2/(1+nu); lambda=nu*Young/((1+nu)*(1-2*nu));
beta = 0; H = 1000; sY = 200*sqrt(3);
mp = [lambda mu beta H sY];
Iden=[1 1 1 0 0 0]';
D=2*mu*eye(6) + lambda*Iden*Iden';
D(4,4) = mu; D(5,5) = mu; D(6,6) = mu;
6KHDUVWUHVV03D
Iden=[1 1 1]';
DM=2*mu*eye(3) + lambda*Iden*Iden';
L = zeros(3,3);
stressN=[0 0 0 0 0 0]';
deps=[0 0 0 0 0 0]';
alphaN = [0 0 0 0 0 0]'; 6PDOO VWUDLQ
epN=0; )LQLWH URWDWLRQ
stressRN=stressN; alphaRN=alphaN;epRN=epN; /DUJH VWUDLQ
bMN=[1 1 1 0 0 0]';
alphaMN = [0 0 0]';
epMN=0; 6KHDUVWUDLQ
for i=1:15
deps(4) = 0.004; L(1,2) = 0.024; L(2,1) = -0.02;
[stressRN, alphaRN] = rotatedStress(L, stressRN, alphaRN);
[stressR, alphaR, epR]=combHard(mp,D,deps,stressRN,alphaRN,epRN);
[stress, alpha, ep]=combHard(mp,D,deps,stressN,alphaN,epN);
[stressM, bM, alphaM, epM]=mulPlast(mp,DM,L,bMN,alphaMN,epMN);
X(i)=i*deps(4);Y1(i)=stress(4);Y2(i)=stressR(4);Y3(i)=stressM(4);
stressN = stress; alphaN = alpha; epN = ep;
stressRN = stressR; alphaRN = alphaR; epRN = epR;
bMN=bM; alphaMN = alphaM; epMN = epM;
end
X = [0 X]; Y1=[0 Y1]; Y2=[0 Y2]; Y3 = [0 Y3]; plot(X,Y1,X,Y2,X,Y3);
Summary
• In multiplicative decomposition, the effect of plasticity is
modeled by intermediate configuration
• The total form stress-strain relation is given by
hyperelasticity between intermediate and current config.
• We studied principle of max dissipation to derive
constitutive relation and plastic evolution
• Similar to classical plasticity, the return mapping
algorithm is used in principal Kirchhoff stress and
principal logarithmic elastic strain
• It is assumed that the principal direction is fixed during
plastic return mapping