Beruflich Dokumente
Kultur Dokumente
Nam-Ho Kim
Introduction
• Linear systems
– Infinitesimal deformation: no significant difference between the
deformed and undeformed shapes
– Stress and strain are defined in the undeformed shape
– The weak form is integrated over the undeformed shape
• Large deformation problem
– The difference between the deformed and undeformed shapes is
large enough that they cannot be treated the same
– The definitions of stress and strain should be modified from the
assumption of small deformation
– The relation between stress and strain becomes nonlinear as
deformation increases
• This chapter will focus on how to calculate the residual
and tangent stiffness for a nonlinear elasticity model
2
Introduction
• Frame of Reference
– The weak form must be expressed based on a frame of reference
– Often initial (undeformed) geometry or current (deformed)
geometry are used for the frame of reference
– proper definitions of stress and strain must be used according to
the frame of reference
• Total Lagrangian Formulation: initial (undeformed)
geometry as a reference
• Updated Lagrangian Formulation: current (deformed)
geometry
• Two formulations are theoretically identical to express
the structural equilibrium, but numerically different
because different stress and strain definitions are used
Table of Contents
• 3.2. Stress and Strain Measures in Large Deformation
• 3.3. Nonlinear Elastic Analysis
• 3.4. Critical Load Analysis
• 3.5. Hyperelastic Materials
• 3.6. Finite Element Formulation for Nonlinear Elasticity
• 3.7. MATLAB Code for Hyperelastic Material Model
• 3.8. Nonlinear Elastic Analysis Using Commercial Finite
Element Programs
• 3.9. Fitting Hyperelastic Material Parameters from Test
Data
• 3.9. Summary
• 3.10.Exercises
4
3.2
Stress and Strain Measures
6
Mild vs. Rough Nonlinearity
x Xu x )( X, t) X u( X, t)
displacement )
:x
:0 u
Q
P
X x
), ) 1 : One-to-one mapping
Continuously differentiable
10
Deformation Gradient
• Infinitesimal length dX in :0 deforms to dx in :x
• Remember that the mapping is continuously differentiable
wx :x Q'
dx dX dx Fd X P'
wX :0
dX
u dx
Q
P
• Deformation gradient:
wxi wu 1 [Gij ],
Fij F 1 1 0 u
wXj wX w w
0 , x
– gradient of mapping ) wX wx
– Second-order tensor, Depend on both :0 and :x
– Due to one-to-one mapping: det F { J ! 0. dX F 1dx
– F includes both deformation and rigid-body rotation
11
12
Green-Lagrange Strain
• Why different strains?
2 2
• Length change: dx dX dxT dx dXT dX
dXT FT FdX dXT dX
dXT (FT F 1)dX
1 § wu wuT wuT wu ·
E ¨ ¸
2 © wX wX wX wX ¹
Displacement gradient
2
1 u uT uT u
0 0 0 0
Higher-order term
– When 0 u 1 , E|
1
2
0 u 0 uT
14
Example – Rigid-Body Rotation
• Rigid-body rotation
x1 X1 cos D X2 sin D
x2 X1 sin D X2 cos D D
x3 X3
E 1 (FT F
2
1) 0
15
ª 2(1 cos D) 0 0º
0 uT 0 u « 0 2(1 cos D ) 0 »
« »
«¬ 0 0 0 »¼
E 1 ( u 0 uT 0 uT 0 u) 0
2 0
16
Example – Rigid-Body Rotation cont.
• What happens to engineering strain?
u1 x1 X1 X1 (cos D 1) X2 sin D
u2 x2 X2 X1 sin D X2 (cos D 1)
u3 x3 X3 0
ª cos D 1 0 0º
H « 0 cos D 1 0 »
« »
«¬ 0 0 0 »¼
17
• Length change: dx 2 dX 2
d x T d x d XT d X
d x T d x d x T F T F 1 d x
dxT (1 F T F 1 )dx
dxT (1 b1 )dx
18
Eulerian Strain Tensor cont.
• Properties
– Symmetric
wu
– Approach engineering strain when 1
wx
– In terms of displacement gradient
1 § wu wuT wuT wu · w
e ¨ ¸ x
2 © wx wx wx wx ¹ wx
1
2
x u x uT x uT x u Spatial gradient
E FT eF
19
4
° x(s, t)
°
¦ NI (s, t)xI 0.35(1 t)
I 1
® 4
°
° y(s, t) ¦ NI (s, t)yI s 1
¯ I 1
20
Example – Lagrangian Strain cont.
• Deformation gradient
:x Q'
wx wx ws P'
F :0 u dx
wX ws wX dX Q
P
ª 0 .35 º ª 4 / 3 0 º
«1 0 »¼ «¬ 0 2 »¼
¬ x(s ,t )
ª 0 0.7 º X(s ,t )
«4 / 3 0 »¼
¬ Reference
domain (s, t)
• Green-Lagrange Strain
1 T ª 0.389 0 º Tension in X1 dir.
E (F F 1) « 0
2 ¬ 0.255 »¼ Compression in X2 dir.
21
ª 0.52 0 º
e 1
2 1 b1 « 0 0.22 »¼
Compression in x1 dir.
Tension in x2 dir.
¬
• Engineering Strain
ª 1 0.7 º
0 u F 1 « »
¬ 1.33 1 ¼
ª 1 0.32 º
H 1
2 0 u 0 uT « 0.32 1 »
Artificial shear deform.
Inconsistent normal deform.
¬ ¼
Which strain is consistent with actual deformation?
22
Example – Uniaxial Tension
• Uniaxial tension of incompressible material (O1 = O!)
• From incompressibility x1 O1X1
O1O2O3 1 O2 O3 O 1/2 x2 O2X2
x3 O3X3
• Deformation gradient and deformation tensor
ªO 0 0 º ª O2 0 0 º
« » « 1
»
F «0 O 1/2
0 » C « 0 O 0 »
«0 1/2 » «0 0 O 1 »
¬ 0 O ¼ ¬ ¼
• G-L Strain
ª O2 1 0 0 º
1« 1
»
E « 0 O 1 0 »
2« 1 »
¬ 0 0 O 1 ¼
23
ª O 2 0 0º ª 1 O 2 0 0 º
« » 1« »
b1 « 0 O 0» e 0 1 O 0
2« »
« 0 0 O »¼ « 0 0 1 O »¼
¬ ¬
• Engineering Strain
ªO 1 0 0 º
« »
H « 0 O 1/2 1 0 »
10%
strain
« 0 0 O 1/2 1 »¼
¬
• Difference
1 2 1
E11 (O 1) e11 (1 O 2 ) H11 O 1
2 2 24
Polar Decomposition
25
dx Q U dX F VQ
V Q dX
e1 e3
Q e2 V
3 e3
E3 1 e1
E1 E2 3E3 2 e2
U
2E2 Q
1E1
F QU
• Eigenvectors of U: E1, E2, E3
• Eigenvectors of V: e1, e2, e3
• Eigenvalues of U and V:O1, O2, O3 26
Polar Decomposition cont.
• Relation between U and C
U2 C U C
– U and C have the same eigenvectors.
– Eigenvalue of U is the square root of that of C
• How to calculate U from C?
• Let eigenvectors of C be ) [E1 E2 E3 ]
• Then, / )T C) where
ª O12 0 0º
« » Deformation tensor in
/ «0 O22 0» principal directions
«0 0 O23 »¼
¬
27
/ «0
« O2 0 »» U ¦ OiEi
Ei
i 1
«¬ 0 0 O3 »¼ 3
Q ¦ ei
Ei
i 1
3
• General Deformation b ¦ Oi2ei
ei
dx Fd X b QUdX b i 1
3
1. Stretch in principal directions V ¦ Oiei
ei
2. Rigid-body rotation i 1
3
3. Rigid-body translation F ¦ Oiei
Ei
i 1 28
Generalized Lagrangian Strain
• G-L strain is a special case of general form of Lagrangian
strain tensors (Hill, 1968)
Em
1
2m
U2m 1
29
• Rotational Tensor
30o X 1, x 1
1
ª 32 12 º
Q FU « »
«¬ 1 2 3 2 »¼
X2, x2
3 2½ 1 ½ 12 ½ 1.15 ½
Q® ¾ ® ¾ , Q ® ¾ ® ¾
¯ 1 2 ¿ ¯0 ¿ ¯ 5 2 3 ¿ ¯ 1 ¿
– 30o clockwise rotation
30o
T
ª5 3 6 12 º X 1, x 1
V FQ « »
¬« 1 2 3 2 ¼»
32
Example – Polar Decomposition cont.
• A straight line X2 X1 tan T will deform to
X1 x1 kx2 , X2 x2
X2, x2
x2 (x1 kx2 ) tan T
x1 tan1 T k x2
• Consider a diagonal line: T = 45o
25o
x2 1 X1, x1
tan D D 24.9q
x1 1 k
X2, x2
• Consider a circle
X12 X22 r2 X1, x1
(x1 kx2 )2 x22 r2
x12 2kx1x2 (1 k2 )x22 r2 Equation of ellipse
33
Deformation of a Volume
• Infinitesimal volume by three vectors
– Undeformed: dV0 dX1 (dX2 u dX3 ) erstdXr1dXs2dXt3
– Deformed: dVx dx1 (dx2 u dx3 ) eijk dxi1dxj2dxk3
• Volumetric Strain
dVx dV0
J 1
dV0
• Incompressible condition: J = 1
• Transformation of integral domain
³³³: x
f d: ³³³: 0
fJ d:
35
x1 O1X1 O1 L / L0 L0
x2 O2X2 O2 h / h0
x3 O3X3 O3 h / h0
h0
• Deformation gradient h0
ª O1 0 0º J det F O1O2O3
F «0 O2 0 »» 2 L
« L § h · LA
h
«¬ 0 0 O3 »¼ L0 ¨© h0 ¸¹ L0A0
h
• Constant volume
L0 L0
J 1 h h0 A A0
L L
36
Deformation of an Area
• Relationship between dS0 and dSx
NdS0 dX1 u dX2 NdS
i 0 eijk dXj1dXk2
ndSx dx1 u dx 2 nrdSx erstdxs1dxt2
wXj wXk 1 2
NdS
i 0 eijk dxsdxt
wxs wxt
N n
dX2 dx2
X dS0 x dS
x
S0 dx1 Sx
dX1
Undeformed Deformed
37
nrdSx
• Use the second relation:
ndSx JF T NdS0 T F T N
n&F N n
F T N
dSx J F( x) T N( X) dS0
38
Stress Measures
• Stress and strain (tensor) depend on the configuration
• Cauchy (True) Stress: Force acts on the deformed config.
'f
– Stress vector at :x: t lim Vn
'Sx o 0 'Sx
Cauchy Stress, sym
f
N
S0
P Sx
P n
39
'f
T lim PT N
'S0 o 0 'S0 First Piola-Kirchhoff Stress
Not symmetric
40
Stress Measures cont.
• Unsymmetric property of P makes it difficult to use
– Remember we used the symmetric property of stress & strain
several times in linear problems
• Make P symmetric by multiplying with F-T
1
S P F T J F 1 V F T V F S FT
J
Second Piola-Kirchhoff Stress, symmetric
– Just convenient mathematical quantities
• Further simplification is possible by handling J differently
W JV F S FT
41
• Observation
– For linear problems (small deformation): H | E | e
– For linear problems (small deformation): V | W | P | S
– S and E are conjugate in energy
– S and E are invariant in rigid-body motion
42
Example – Uniaxial Tension
F
• Cauchy (true) stress: V11 , V22 = V33 = V12 = V23 = V13 = 0
A
• Deformation gradient:
L0
ª O11 0 0 º
« »
F 1 « 0 O21 0 », J 1
« 0 0 O31 »¼ h0
¬
h0
• First P-K stress
F 1 F A F L
P11 (JF 1 V11
A O1 A A0 A0 h
h
• Second P-K stress F
F 1 F A2 FA F
S11 (JF 1 V F T )11 No clear physical
A O12 A A02 A02 A0 O1 meaning
43
Summary
• Nonlinear elastic problems use different measures of
stress and strain due to changes in the reference frame
• Lagrangian strain is independent of rigid-body rotation,
but engineering strain is not
• Any deformation can be uniquely decomposed into rigid-
body rotation and stretch
• The determinant of deformation gradient is related to the
volume change, while the deformation gradient and
surface normal are related to the area change
• Four different stress measures are defined based on the
reference frame.
• All stress and strain measures are identical when the
deformation is infinitesimal
44
3.3
Nonlinear Elastic Analysis
45
Goals
• Understanding the principle of minimum potential energy
– Understand the concept of variation
• Understanding St. Venant-Kirchhoff material
• How to obtain the governing equation for nonlinear elastic
problem
• What is the total Lagrangian formulation?
• What is the updated Lagrangian formulation?
• Understanding the linearization process
46
Numerical Methods for Nonlinear Elastic Problem
• We will obtain the variational equation using the principle
of minimum potential energy
– Only possible for elastic materials (potential exists)
• The N-R method will be used (need Jacobian matrix)
• Total Lagrangian (material) formulation uses the
undeformed configuration as a reference, while the
updated Lagrangian (spatial) uses the current
configuration as a reference
• The total and updated Lagrangian formulations are
mathematically equivalent but have different aspects in
computation
47
X x
Iteration
48
Total Lagrangian Formulation cont.
• In TL, the undeformed configuration is the reference
• 2nd P-K stress (S) and G-L strain (E) are the natural choice
• In elastic material, strain energy density W exists, such
that
wW
stress
wstrain
49
50
St. Venant-Kirchhoff Material
• Strain energy density for St. Venant-Kirchhoff material
W(E) 1E :D:E
2 Contraction operator: a : b aijbij
E 1 (FT F
2
1) 1 ( UT Q T Qu
2
1) 1 ( U2
2
1)
52
Example Y
QE E
O 17,308 P 11, 538
(1 Q)(1 2Q) 2(1 Q)
1 ª 1, 872 0 º
V FSFT « 0
J ¬ 21, 516 »¼ 53
ª1 kº 1 T 1 ª0 k º
F «0 1 » E (F F 1) « »
¬ ¼ 2 2 ¬ k k2 ¼
X1, x1
• Material properties
QE E
O 40MPa P 40MPa
(1 Q)(1 2Q) 2(1 Q)
20
• 2nd P-K stress Cauchy
10
ª k2 2k º
Shear stress
55
Variational Formulation
• We want to minimize the potential energy (equilibrium)
3int: stored internal energy
3ext: potential energy of applied loads
3( u) 3int ( u) 3 ext ( u)
³³: 0
W(E) d: ³³
:0
uT fb d: ³ s uT t d*
*o
56
Variational Formulation cont.
• Variation of Potential Energy (Directional Derivative)
d
3( u, u ) 3( u Wu ) We will use “over-bar” for variation
dW W 0
– 3 depends on u only, but 3 depends on both u and
– Minimum potential energy happens when its variation becomes
zero for every possible
– One-dimensional example
3(u)
57
k u
f
³³: 0
S : E d: ³³: 0
u T fb d : ³ s u T t d *
*o
59
2
1 FT u u T F
0 0
E ( u, u ) sym( 0 u T F)
60
Variational Formulation cont.
• Variational Equation
³³: 0
S : E d: ³³: 0
u T fb d : ³ s u T t d *
*o
for all '
a( u, u ) A( u )
a( u, u ) A( u ), u '
Linearization (Increment)
• Linearization process is similar to variation and/or
differentiation
– First-order Taylor series expansion
– Essential part of Newton-Raphson method
• Let f(xk+1) = f(xk + 'uk), where we know xk and want to
calculate 'uk
df( x)
f( xk 1 ) f( xk )
'uk H.O.T.
dx
• The first-order derivative is indeed linearization of f(x)
d wf
L[f] { f( x Z'u) 'u Linearization
dZ Z 0 wx
d wf
Gf f { f( x Wu ) u Variation
dW W 0 wx
62
Linearization of Residual
• We are still in continuum domain (not discretized yet)
• Residual R( u) a( u, u ) A( u )
• We want to linearize R(u) in the direction of 'u
– First, assume that u is perturbed in the direction of 'u using a
variable W. Then linearization becomes
T
wR( u W'u) ª wR º
L[R( u)]
wW « wu » 'u
W 0 ¬ ¼
f(xk+1)
k 1 'uk
• Update state u u 'u
k k
xk+1 xk x
k 1 k 1
x Xu
ª wR( uk ) º
• We know how to calculate R(uk), but how about « » ?
¬ wu ¼
w w
[R( u)] [a( u, u ) A( u )]
wu wu
– Only linearization of energy form will be required
– We will address displacement-dependent load later
64
Linearization cont.
• Linearization of energy form
§ wx · § w( X u) · w'u
'F '¨ ¸ '¨ ¸ 0 'u
© wX ¹ © wX ¹ wX
65
Linearization cont.
• Strain increment 'E 1 ( 'FT F FT 'F)
2
1 ( 'uT F FT 'u)
2 0 0
sym( 0 'uT F)
!!! Linear w.r.t. 'u
• Inc. strain variation 'E '[sym( 0 u T F)]
sym( 0 u T 'F)
sym( 0 u T 0 'u) !!! Linear w.r.t. 'u
a * ( n u k ; 'u k , u ) A( u ) a( n u k , u ), u '
• In discrete form
{ d }T [ nKTk ]{ 'dk } { d }T { nRk }
n k n k
• What are [ KT ] and { R } ?
67
• Strain variation
du du du
E11 u2 (1 u2 )
dX dX dX
• Strain energy density and stress
1 E (E )2 wW § 1 ·
W(E11 ) 11 S11 E E11 E ¨ u2 (u2 )2 ¸
2 wE11 © 2 ¹
• Energy and load forms
L0
a(u,u) ³0 S11E11AdX S11AL0 (1 u2 )u2 A(u) u2F
• N-R iteration
[E(1 u2k )2 S11
k
]AL0 'u2k F S11
k
(1 u2k )AL0
69
70
Updated Lagrangian Formulation
• The current configuration is the reference frame
– Remember it is unknown until we solve the problem
– How are we going to integrate if we don’t know integral domain?
• What stress and strain should be used?
– For stress, we can use Cauchy stress (V)
– For strain, engineering strain is a pair of Cauchy stress
– But, it must be defined in the current configuration
1 § wu wu ·
T
H ¨ ¸ sym( x u)
2 ¨ wx wx ¸
© ¹
71
Variational Equation in UL
• Instead of deriving a new variational equation, we will
convert from TL equation
1 § wu ·
T
1 T wu
V F S FT E ¨ FF ¸
J 2 ¨ wX wX ¸
© ¹
S JF 1 V F T
1 T § T wu wu 1 ·
T
F ¨F F ¸F
2 ¨ wX wX ¸
© ¹
1 T § wX wu wu wX ·
T T
F ¨ ¸F
2 ¨ wx wX wX wx ¸
Similarly © ¹
1 T § wu wu ·
T
'E FT 'H F F ¨ ¸F
2 ¨ wx wx ¸
1 § w'u w'u ·
T
© ¹
'H ¨ ¸
2 ¨ wx wx ¸ FT H F
© ¹
72
Variational Equation in UL cont.
• Energy Form
a( u, u ) ³³: 0
S : E d: ³³: 0
(JF 1VF T ) : (FT HF) d:
³³: 0
S : E d: ³³: 0
V : H Jd: ³³: x
V : H d:
• Variational Equation
Linearization of UL
• Linearization of ax ( u, u ) will be challenging because we
don’t know the current configuration (it is function of u)
• Similar to the energy form, we can convert the linearized
energy form of TL
• Remember a * ( u; 'u, u ) ³³ : [E : D : 'E S : 'E ] d :
0
0
75
76
Linearization of UL cont.
• From equivalence, the energy form is linearized in TL and
converted to UL
L[a( u, u )] ³³: 0
[ H : c : 'H V : K]J d:
a * ( u; 'u, u ) ³³: x
[ H : c : 'H V : K] d:
• N-R Iteration
a * ( n uk ; 'uk , u ) A( u ) a( n uk , u ), u '
• Observations
– Two formulations are theoretically identical with different
expression
– Numerical implementation will be different
– Different constitutive relation
77
1 1 2
• Cauchy stress: V11 FS F E(u2 u )(1 u2 )
J 11 11 11 2 2
u2
• Strain variation: H11 (u) F11T E11F111
1 u2
L
• Energy & load forms: a(u,u) ³0 V11H11 (u)Adx V11Au2 A(u) u2F
78
Example – Uniaxial Bar
1
• Spatial constitutive relation: c1111 FFFFE (1 u2 )3 E
J 11 11 11 11
L
• Linearization: ³0 H11 (u)c1111H11 ('u)Adx EA(1 u2 )2 u2'u2
L V11A
³0 V11K11 ( 'u, u)Adx u 'u
1 u2 2 2
L
a * (u; 'u,u) ³0 H11 (u)c1111H11 ('u) V11K('u,u) Adx
V11
EA(1 u2 )2 u2 'u2 Au 'u
1 u2 2 2
79
Section 3.5
Hyperelastic Material Model
80
Goals
• Understand the definition of hyperelastic material
• Understand strain energy density function and how to use
it to obtain stress
• Understand the role of invariants in hyperelasticity
• Understand how to impose incompressibility
• Understand mixed formulation and perturbed Lagrangian
formulation
• Understand linearization process when strain energy
density is written in terms of invariants
81
What Is Hyperelasticity?
• Hyperelastic material - stress-strain relationship derives
from a strain energy density function
– Stress is a function of total strain (independent of history)
– Depending on strain energy density, different names are used,
such as Mooney-Rivlin, Ogden, Yeoh, or polynomial model
• Generally comes with incompressibility (J = 1)
– The volume preserves during large deformation
– Mixed formulation – completely incompressible hyperelasticity
– Penalty formulation - nearly incompressible hyperelasticity
• Example: rubber, biological tissues
– nonlinear elastic, isotropic, incompressible and generally
independent of strain rate
• Hypoelastic material: relation is given in terms of stress
and strain rates 82
Strain Energy Density
• We are interested in isotropic materials
– Material frame indifference: no matter what coordinate system is
chosen, the response of the material is identical
– The components of a deformation tensor depends on coord. system
– Three invariants of C are independent of coord. system
• Invariants of C
I1 tr(C ) C11 C22 C33 O12 O22 O32 No deformation
I1 = 3
I2 1 ª (tr C )2 tr(C2 ) º O12O22 O22O32 O32O12 I2 = 3
2 ¬ ¼
I3 = 1
I3 det C O12O22O32
84
Strain Energy Density cont.
• Strain Energy Density Function
– Yeoh model
85
-50
Nominal stress
-100
Neo-Hookean
-150
-200
-250
-0.8 -0.4 0 0.4 0.8
Nominal strain
86
Example – St. Venant Kirchhoff Material
• Show that St. Venant-Kirchhoff material has the following
strain energy density
O 2
W(E) ª¬ tr(E) º¼ Ptr(E2 )
2
wW(E) wtr(E) wtr(E2 )
S Otr(E) P
wE wE wE
• First term
wtr(E)
tr(E) 1: E 1
wE
wtr(E)
Otr(E) O1(1 : E) O(1
1) : E
wE
• Second term
wEijEji
Gik G jlEji EijG jk Gil Elk Elk 2Elk
wEkl
87
88
Nearly Incompressible Hyperelasticity
• Incompressible material
– Cannot calculate stress from strain. Why?
• Nearly incompressible material
– Many material show nearly incompressible behavior
– We can use the bulk modulus to model it
• Using I1 and I2 enough for incompressibility?
– No, I1 and I2 actually vary under hydrostatic deformation
– We will use reduced invariants: J1, J2, and J3
89
Locking
• What is locking
– Elements do not want to deform even if forces are applied
– Locking is one of the most common modes of failure in NL analysis
– It is very difficult to find and solutions show strange behaviors
• Types of locking
– Shear locking: shell or beam elements under transverse loading
– Volumetric locking: large elastic and plastic deformation
• Why does locking occur?
– Incompressible sphere under hydrostatic pressure
No unique pressure
Pressure
Volumetric strain
90
How to solve locking problems?
• Mixed formulation (incompressibility)
– Can’t interpolate pressure from displacements
– Pressure should be considered as an independent variable
– Becomes the Lagrange multiplier method
– The stiffness matrix becomes positive semi-definite
Displacement
Pressure
4x1 formulation
91
Penalty Method
• Instead of incompressibility, the material is assumed to be nearly
incompressible
• This is closer to actual observation
• Use a large bulk modulus (penalty parameter) so that a small volume
change causes a large pressure change
• Large penalty term makes the stiffness matrix ill-conditioned
• Ill-conditioned matrix often yields excessive deformation
• Temporarily reduce the penalty term in the stiffness calculation
• Stress calculation use the penalty term as it is
Unique pressure
Pressure
x1 DX1 ªD 0 0 º ª D2 0 0 º
° «0 D 0» « »
® x2 DX2 F « » C «0 D2 0 »
°x «0
¯ 3 DX3 «¬ 0 0 D »¼
¬ 0 D2 »¼
• Invariants
I1 3D2 I2 3D 4 I3 D6 I1 and I2 are not constant
• Reduced invariants
93
J3,E 1 (I 1/2 )I
2 3 3,E
I1,E 21 I1,E 21
I2,E 4(1 trE)1 4E I2,E 2(I1 1 C )
I3,E (2 4trE)1 4E [ 94 eimn ejrsEmrEns ] I3,E 2I3C 1
96
Example
• Show I1,E 21, I2,E 2(I1 1 C ), I3,E 2I3C 1
• Let I1 tr(C ), I2 1 tr( CC ),
2
I3 1 tr( CCC )
3
• Then I1 I1 , I2 1 I2
2 1
I2 , I3 I3 61 I13 I1I2
• Derivatives
wI1 wI2 wI3
Gij, Cji, Cjk Cki
wCij wCij wCij
and
w w
2
wC wE
97
Mixed Formulation
• Using bulk modulus often causes instability
– Selectively reduced integration (Full integration for deviatoric
part, reduced integration for dilatation part)
• Mixed formulation: Independent treatment of pressure
WH (J3, p) p(J3 1)
– Pressure p is additional unknown (pure incompressible material)
– Advantage: No numerical instability
– Disadvantage: system matrix is not positive definite
• Perturbed Lagrangian formulation
1 2
WH (J3, p) p(J3 1) p
2K
a(r, r ) ³³: 0
ª¬ S : E pH º¼ d:
p
H J3 1 Volumetric strain
K
99
X1, x1
I1 4, I2 4, I3 1
I1,E 21
ª 6 2 0 º
I2,E 2(I1 1 C ) «« 2 4 0 »»
«¬ 0 0 6 »¼
ª 4 2 0 º
I3,E 2I3C 1 «« 2 2 0 »»
«¬ 0 0 2 »¼
100
Example – Simple Shear cont.
ª 5 4 0 º
J1 I1I31/3 4 4 2« »
J1,E I1,E I3,E 4 1 0
3 3« »
«¬ 0 0 1 »¼
J2 I2I32/3 4
ª 7 5 0 º
2« »
J2,E I2,E 83 I3,E 5 2 0
J3 I31/2 1 3« »
«¬ 0 0 1 »¼
101
I1 C1 C2 C3
I2 C1C2 C1C3 C2C3 C4C4 C5C5 C6C6
I3 (C1C2 C4C4 )C3 (C4C6 C1C5 )C5 (C4 C5 C2C6 )C6
{I1,E } 2{1 1 1 0}
{I2,E } 2{C2 C3 C3 C1 C1 C2 C4 C5 C6 }
{I3,E } 2{C2C3 C52 C3C1 C62 C1C2 C42
C5C6 C3C4 C6C4 C1C5 C4C5 C2C6 }
• Material stiffness
wS
D A10J1,EE A01J2,EE K(J3 1)J3,EE KJ3,E
J3,E
wE
a * ( u; 'u, u ) ³³: 0
ª¬ E : D : 'E S : 'E º¼ d:
103
Linearization cont.
• Second-order derivatives of reduced invariants
1 1 4 4 7 1 4
J1,EE I1,EE I3 3 I3 3 (I1,E
I3,E I3,E
I1,E ) I1I3 3 I3,E
I3,E I1I3 3 I3,EE
3 9 3
23 2 53 10 38 2 5
J2,EE I2,EE I3 I3 (I2,E
I3,E I3,E
I2,E ) I2I3 I3,E
I3,E I2I3 3 I3,EE
3 9 3
1 3
1 1
J3,EE I3 2 I3,E
I3,E I3 2 I3,EE
4 2
I1,EE 0
I2,EE 41
1 I
I3,EE 4I3C 1
C 1 I3C 1 IC 1
104
MATLAB Function Mooney
• Calculates S and D for a given deformation gradient
%
% 2nd PK stress and material stiffness for Mooney-Rivlin material
%
function [Stress D] = Mooney(F, A10, A01, K, ltan)
% Inputs:
% F = Deformation gradient [3x3]
% A10, A01, K = Material constants
% ltan = 0 Calculate stress alone;
% 1 Calculate stress and material stiffness
% Outputs:
% Stress = 2nd PK stress [S11, S22, S33, S12, S23, S13];
% D = Material stiffness [6x6]
%
105
Summary
• Hyperelastic material: strain energy density exists with
incompressible constraint
• In order to be material frame indifferent, material
properties must be expressed using invariants
• Numerical instability (volumetric locking) can occur when
large bulk modulus is used for incompressibility
• Mixed formulation is used for purely incompressibility
(additional pressure variable, non-PD tangent stiffness)
• Perturbed Lagrangian formulation for nearly
incompressibility (reduced integration for pressure term)
106
Section 3.6
Finite Element Formulation for
Nonlinear Elasticity
107
Voigt Notation
• We will use the Voigt notation because the tensor
notation is not convenient for implementation
– 2nd-order tensor º vector
– 4th-order tensor º matrix
108
4-Node Quadrilateral Element in TL
• We will use plane-strain, 4-node quadrilateral element to
discuss implementation of nonlinear elastic FEA
• We will use TL formulation
• UL formulation will be discussed in Chapter 4
X2 4 t
3 (–1,1) (1,1)
1 2
X1 (–1,–1) (1,–1)
FiniteElementat ReferenceElement
undeformeddomain
109
N1 1 (1 s)(1 t)
4
1 (1 s)(1 t) Interpolation (shape) function
N2 4
1 (1 s)(1 t) • Same for all elements
N3 4
1 (1 s)(1 t)
• Mapping depends of geometry
N4 4
110
Displacement and Deformation Gradients
• Displacement gradient
Ne
wu wN ( s) Ne
wX
¦ wIX uI ui,j ¦ NI,j (s)uIi
I 1 I 1
wNI (s)
– How to calculate ?
wX
• Deformation gradient
111
Green-Lagrange Strain
• Green-Lagrange strain
ª O 2P O 0º
[D] « O O 2P 0 »
« »
«¬ 0 0 P »¼
112
Variation of G-R Strain
• Although E(u) is nonlinear, E ( u, u ) is linear
E ( u, u ) sym( 0 u T F) {E } [BN ]{ d }
ª º
« F11N1,1 F21N1,1 F11N2,1 F21N2,1 " F11N4,1 F21N4,1 »
« »
« »
[BN ] « F12N1,2 F22N1,2 F12N2,2 F22N2,2 " F12N4,2 F22N4,2 »
« »
« »
« F11N1,2 F21N1,2 F11N2,2 F21N2,2
"
F11N4,2 F21N4,2 »
« F N F N F N F N F12N4,1 F22N4,1 »¼
¬ 12 1,1 22 1,1 12 2,1 22 2,1
Function of u
Different from linear strain-displacement matrix
113
Variational Equation
• Energy form
a( u, u ) ³³: 0
S : E d:
| { d }T ³³ [BN ]T { S } d:
:0
{ { d }T {Fint }
• Load form
A( u ) ³³: 0
u T fb d : ³* S
0
u T t d*
¦ uIT ^ ³³: `
Ne
| NI (s)fb d: ³ S NI (s)t d*
0 *0
I 1
{ { d }T {Fext }
• Residual
{ d }T {Fint ( d)} { d }T { F ext }, { d } 'h
114
Linearization – Tangent Stiffness
• Incremental strain { 'E } [BN ]{ 'd }
• Linearization
ª S11 S12 0 0 º
«S 0 »»
[6] « 12 S22 0
« 0 0 S11 S12 »
« »
¬ 0 0 S12 S22 ¼
[KT ] ³³: 0
ª [BN ]T [D][BN ] [BG ]T [6][BG ] º d:0
¬ ¼
116
Summary
• For elastic material, the variational equation can be
obtained from the principle of minimum potential energy
• St. Venant-Kirchhoff material has linear relationship
between 2nd P-K stress and G-L strain
• In TL, nonlinearity comes from nonlinear strain-
displacement relation
• In UL, nonlinearity comes from constitutive relation and
unknown current domain (Jacobian of deformation
gradient)
• TL and UL are mathematically equivalent, but have
different reference frames
• TL and UL have different interpretation of constitutive
relation.
117
Section 3.7
MATLAB Code for
Hyperelastic Material Model
118
HYPER3D.m
• Building the tangent stiffness matrix, [K], and the residual
force vector, {R}, for hyperelastic material
• Input variables for HYPER3D.m
Variable Array size Meaning
MID Integer Material Identification No. (3) (Not used)
PROP (3,1) Material properties (A10, A01, K)
UPDATE Logical variable If true, save stress values
LTAN Logical variable If true, calculate the global stiffness matrix
NE Integer Total number of elements
NDOF Integer Dimension of problem (3)
XYZ (3,NNODE) Coordinates of all nodes
LE (8,NE) Element connectivity
119
%
% Residual forces
FORCE(IDOF) = FORCE(IDOF) - FAC*BM'*STRESS;
%
% Tangent stiffness
if LTAN
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*BM + BG'*SHEAD*BG;
GKF(IDOF,IDOF)=GKF(IDOF,IDOF)+FAC*EKF;
end
end; end; end;
end
end
122
Example Extension of a Unit Cube
• Face 4 is extended with a stretch ratio O = 6.0
• BC: u1 = 0 at Face 6, u2 = 0 at Face 3, and u3 = 0 at Face 1
• Mooney-Rivlin: A10 = 80MPa, A01 = 20MPa, and K = 107
% Nodal coordinates X2
XYZ=[0 0 0;1 0 0;1 1 0;0 1 0;0 0 1;1 0 1;1 1 1;0 1 1]; Face 1
% 4 3
% Element connectivity
LE=[1 2 3 4 5 6 7 8];
% 8 7 Face 4
% No external force
EXTFORCE=[]; Face 6 2
% X1
% Prescribed displacements [Node, DOF, Value] 1
SDISPT=[1 1 0;4 1 0;5 1 0;8 1 0; % u1=0 for Face 6
1 2 0;2 2 0;5 2 0;6 2 0; % u2=0 for Face 3 5 6
1 3 0;2 3 0;3 3 0;4 3 0; % u3=0 for Face 1 X3 Face 3
2 1 5;3 1 5;6 1 5;7 1 5]; % u1=5 for Face 4
%
% Load increments [Start End Increment InitialFactor FinalFactor]
TIMS=[0.0 1.0 0.05 0.0 1.0]';
%
% Material properties
MID=-1;
PROP=[80 20 1E7];
123
z 125
S A10J1,E A01J2,E
128
Section 3.9
Fitting Hyperelastic Material
Parameters from Test Data
129
70
uni-axial
bi-axial
60
pure shear
50
Nominal stress
40
30
20
10
0
0 1 2 3 4 5 6 7
Nominal strain
130
Elastomer Tests
• Data type: Nominal stress vs. principal stretch
F L F
L
F
L
Data Preparation
• Need enough number of independent experimental data
– No rank deficiency for curve fitting algorithm
• All tests measure principal stress and principle stretch
Pure shear test Stretch ratio O= L/L0 Nominal stress TE = F/A0
132
Data Preparation cont.
• Uni-axial test O1 O, O2 O3 1/ O
wU
T 2(1 O 3 )(A10 O A01 )
wO
A ½
T(A10 ,A01 , O) { x }T { b} ª 2(O O 2 ) 2(1 O 3 ) º ® 10 ¾
¬ ¼ A
¯ 01 ¿
• Equi-biaxial test O1 O2 O, O3 1 / O2
1 wU
T 2( O O 5 )(A10 O2A01 )
2 wO
• Pure shear test O1 O, O2 1, O3 1 / O
wU
T 2(O O 3 )(A10 A01 )
wO
133
Type 1 1 1 ! 4 4 ! 4
O O1 O2 O3 ! Oi Oi1 ! ONDT
T E T1E T2E T3E ! Ti E TiE1 ! TNDT
E
134
Curve Fitting for Mooney-Rivlin Material
• Need to determine A10 and A01 by minimizing error
between test data and model
NDT
¦
2
minimize TkE T(A10 ,A01 , Ok )
A10 ,A01 k 1
{ e }T { e } { T E T }T { T E T }
{ T E Xb}T { T E Xb}
{ T E }T { T E } 2{b}T [ X]T { T E } { b}T [ X]T [ X]{ b}
136
Stability of Constitutive Model
• Stable material: the slope in the stress-strain curve is
always positive (Drucker stability)
• Stability requirement (Mooney-Rivlin material)
dH : D : dH ! 0
ªD D12 º dH1 ½
^ dH1dH2 ` « 11
D D » ® dH ¾ ! 0
¬ 21 22 ¼ ¯ 2¿
• In order to be P.D.
D11 D22 ! 0
D11D22 D12D21 ! 0
137