CompContactMech Yastrebov

NONLINEAR COMPUTATIONAL MECHANICS
Computational
Contact
Mechanics
th
19 March, 2014 Vladislav A. Yastrebov
Ecole des Mines de Paris
Centre des Matériaux
France
MINES ParisTech, CNRS
N O N L I N E A R C O M P U T A T I O N A L M E C H A N I C S :: C O M P U T A T I O N A L C O N T A C T M E C H A N I C S
Outline
Introduction
Governing equations
Optimization methods
Examples
2/34
1. Introduction
Industrial and natural contact problems

1 Assembled parts, e.g. engines
Aircraft’s engine GP 7200

www.safran-group.com
[1] M. W. R. Savage
J. Eng. Gas Turb. Power, 134:012501 (2012)
4/34

2 Railroad contacts
High speed train TGV www.sncf.com
Wilde/ANSYS wildeanalysis.co.uk
4/34

2 Railroad contacts
3 Gears and bearings
Bearings
www.skf.com
[1] F. Massi, J. Rocchi, A. Culla, Y. Berthier

Mech. Syst. Signal Pr., 24:1068-1080 (2010)
4/34

2 Railroad contacts
Helical gear www.tpg.com.tw
www.mscsoftware.com
4/34

2 Railroad contacts
4 Breaking systems
Assembled breaking system
www.brembo.com
www.mechanicalengineeringblog.com
4/34

2 Railroad contacts
4 Breaking systems
5 Tire-road contact
Tire-road contact www.michelin.com
[1] M. Brinkmeier, U. Nackenhorst, S. Petersen,

O. von Estorff, J. Sound Vib., 309:20-39 (2008)
4/34

2 Railroad contacts
4 Breaking systems
5 Tire-road contact
6 Metal forming
Deep drawing www.thomasnet.com
[1] G. Rousselier, F. Barlat, J. W. Yoon

Int. J. Plasticity, 25:2383-2409 (2009)
4/34

2 Railroad contacts
4 Breaking systems
5 Tire-road contact
6 Metal forming Crash-test www.porsche.com
7 Crash tests
[1] O. Klyavin, A. Michailov, A. Borovkov

www.fea.ru
4/34

2 Railroad contacts
4 Breaking systems
5 Tire-road contact
6 Metal forming
7 Crash tests Human articulations
www.sportssupplements.net
8 Biomechanics
J. A. Weiss, University of Utah

Musculoskeletal Research Laboratories
4/34

2 Railroad contacts
4 Breaking systems
Sand dunes www.en.wikipedia.org
5 Tire-road contact
6 Metal forming
7 Crash tests
8 Biomechanics
9 Granular materials
E. Azema et al, LMGC90
4/34

2 Railroad contacts
4 Breaking systems
5 Tire-road contact
6 Metal forming Damage at electric contact zone
www.taicaan.com
7 Crash tests
8 Biomechanics
10 Electric contacts
Simulation of electric current

www.comsol.com
4/34

2 Railroad contacts
4 Breaking systems
5 Tire-road contact
6 Metal forming
7 Crash tests San-Andreas fault, by M. Rightmire
www.sciencedude.ocregister.com
8 Biomechanics
11 Tectonic motions
[1] J.D. Garaud, L. Fleitout, G. Cailletaud

Colloque CSMA (2009)
4/34

2 Railroad contacts
4 Breaking systems
5 Tire-road contact
6 Metal forming
Drill Bit tool Robit Rocktools;
extraction of geothermal energy (SINTEF,NTNU)
7 Crash tests
8 Biomechanics
11 Tectonic motions
12 Deep drilling [1] T. Saksala, Int. J. Numer. Anal. Meth.
Geomech. (2012)
4/34

2 Railroad contacts
4 Breaking systems
5 Tire-road contact
Impact crater, Arizona
6 Metal forming www.MrEclipse.com et maps.google.com
7 Crash tests
8 Biomechanics
11 Tectonic motions
12 Deep drilling
Hyper velocity impact
13 Impact and fragmentation Molecular Dynamics simulation
4/34

2 Railroad contacts
4 Breaking systems
5 Tire-road contact
Impact crater, Arizona
6 Metal forming www.MrEclipse.com et maps.google.com
7 Crash tests
8 Biomechanics
11 Tectonic motions
12 Deep drilling
Hyper velocity impact
13 Impact and fragmentation Molecular Dynamics simulation
14 MEMS
4/34
Involved physical phenomena

Primary phenomena (atomic / electronic interactions)
1 Contact (non-penetration)
2 Friction (tangential resistance)
3 Adhesion (resistance to tension)
4 Quantum effects (tuneling, Casimir effect)
Secondary phenomena
1 Wear
2 Lubrication (fluid-structure interaction)
3 Environmental effects (oxidation)
4 Aging, diffusion
5 Heat production
6 Heat and electricity transfer
5/34
Contact complexity: physics and mathematics

Particular difficulties related to contact problems:
multiphysical aspects, mathematical aspects
Fractality of surfaces
Chemical reactions
Generation and diffusion of heat
Multiscale nature of friction
6/34
Contact complexity: physics and mathematics

Particular difficulties related to contact problems:
multiphysical aspects, mathematical aspects
Lack of standard optimization

problem
Non-convexity and
non-differentiability
Non-continuous character
Bad scalability
6/34
Particularities: mesh and convergence

Strong mesh refinement is required
• especially at unknown edges of contact zones
Typical mesh for fretting analysis [L. Sun, H. Proudhon, G. Cailletaud, 2011]
2D ∼ 30 000 DoFs, 3D ∼ 5 000 000 DoFs
7/34

Infinite contact pressure and/or its derivative
7/34

Slow change of boundary conditions:
• strong non-linearities of contact / friction problems
• non-uniqueness of solution for frictional problems
Infinite looping
Initial guess R(x0 , f0 ) = 0
7/34

Infinite looping
Too rapid change in boundary conditions R(x0 , f1 ) , 0
7/34

Infinite looping
Iterations of Newton-Raphson method

−1
∂R
δx = 0 → δx = − ∂R

R(x0 , f1 ) + ∂x x0
R(x0 , f1 ) → x1 = x0 + δx
∂x x 0
7/34

Infinite looping

−1
R(x1 , f1 ) + ∂R δx = 0 → δx = − ∂R

∂x x1
R(x1 , f1 ) → x2 = x1 + δx
∂x x1
7/34

Infinite looping
Infinite looping
7/34

Convergence to a “false” solution
Initial guess R(x0 , f0 ) = 0
7/34

Too rapid change in boundary conditions R(x0 , f1 ) , 0
7/34


−1
∂R
δx = 0 → δx = − ∂R

R(x0 , f1 ) + ∂x x0
R(x0 , f1 ) → x1 = x0 + δx
∂x x 0
7/34


−1
R(x1 , f1 ) + ∂R δx = 0 → δx = − ∂R

∂x x1
R(x1 , f1 ) → x2 = x1 + δx
∂x x1
7/34

Convergence, but is it a “true” solution ?
7/34
2. Governing equations
Notations
Vectors and tensors
• a, α scalars •a·b=c scalar (dot) product
•b vectors •a×b=c vector (cross) product
• C, β 2nd order tensors •a⊗b=C tensor product
== =
4
•D 4th order tensors • AT transposition
= =
• ∇a = B gradient operator •∇·a=c divergence operator

=
•∇×a=B rotor operator • I = ei ⊗ ei unitary 2nd order tensor
= =
Mechanics
•σ Cauchy stress tensor •ε Small strain tensor
= =
• g, gn gap, normal gap • ξ position vector in parent space
∼
• penalty parameter •n outward unit normal vector
∂% ∂%
• λ, λn , λt lagrange multipliers • ∂ξ1
, ∂ξ2 surface tangent vectors
• σn = (σ · n) · n contact pressure •µ Coefficient of friction
=
9/34
Equilibrium and contact conditions

Balance of momentum
∇ · σ + fv = 0 in Ω1,2

=




σ · n = t0 on Γf f


=

1
u = u0 on Γu





u

on Γ

 ?
c 1
c
2
Frictionless contact c
conditions (intuitive) 2
f
u
1 No penetration
2 No adhesion
3 No shear transfer
10/34

Balance of momentum
∇ · σ + fv = 0 in Ω1,2

=

f



σ · n = t0 on Γf


=
 1
u = u0 on Γu


u


1


c
on Γ

 ?
c
n
Frictionless contact v
conditions (intuitive) c
2
1 No penetration 2
f
u
2 No adhesion
3 No shear transfer
10/34

Balance of momentum
∇ · σ + fv = 0 in Ω1,2

=

f



σ · n = t0 on Γf


=
 1
u = u0 on Γu


u




on Γ

 ?
c 1
c
2
conditions (intuitive)
1 No penetration 2
f
u
2 No adhesion
3 No shear transfer
10/34

Balance of momentum
∇ · σ + fv = 0 in Ω1,2

=

f



σ · n = t0 on Γf


=
 1
u = u0 on Γu


u




on Γ

 ?
c 1
c
2
1 No penetration 2
f
u
2 No adhesion
3 No shear transfer
10/34

Balance of momentum
∇ · σ + fv = 0 in Ω1,2

=

f



σ · n = t0 on Γf


=
 1
u = u0 on Γu


u




on Γ

 ?
c 1
c
2
1 No penetration 2
f
u
2 No adhesion
3 No shear transfer
10/34
Gap function
Gap function g
gap = – penetration g>0
asymmetric function non-contact
n
defined for n g=0
• separation g > 0 n contact
• contact g = 0
• penetration g < 0 g<0
penetration
governs normal contact
Gap between a slave point and a master surface
Master and slave split
Gap function is determined for all
slave points with respect to the
master surface
11/34
Gap function
Gap function g
gap = – penetration g>0
asymmetric function non-contact
n
defined for n g=0
• separation g > 0 n contact
• contact g = 0
• penetration g < 0 g<0
penetration
governs normal contact
Gap between a slave point and a master surface
Master and slave split
Gap function is determined for all
rs
slave points with respect to the n
master surface
( (
Normal gap
h i
gn = n · rs − ρ(ξπ ) ,
n is a unit normal vector, rs Definition of the normal gap
slave point, ρ(ξπ ) projection
point at master surface
11/34
Frictionless or normal contact conditions

n
No penetration
0
Always non-negative gap
g
g≥0
No adhesion
Always non-positive contact pressure
σ∗n ≤ 0
Complementary condition Scheme explaining normal
Either zero gap and non-zero pressure, or contact conditions
non-zero gap and zero pressure
g σn = 0
No shear transfer (automatic)
σ ∗∗
t =0
σ∗n = (σ · n) · n = σ : (n ⊗ n)
= =
σ ∗∗
t == σ · n − σn n = σ · I − n ⊗ n
= =
12/34

n
No penetration
0 non-contact
Always non-negative gap
g > 0, n = 0 g
g≥0
g = 0, n < 0
contact
No adhesion
restricted
Always non-positive contact pressure regions
σ∗n ≤ 0
Complementary condition Improved scheme explaining
Either zero gap and non-zero pressure, or normal contact conditions
non-zero gap and zero pressure
g σn = 0
No shear transfer (automatic)
σ ∗∗
t =0
σ∗n = (σ · n) · n = σ : (n ⊗ n)
= =
σ ∗∗
t == σ · n − σn n = σ · I − n ⊗ n
= =
12/34

In mechanics: n
Normal contact conditions 0 non-contact
≡ g > 0, n = 0 g
g = 0, n < 0
Frictionless contact conditions
contact
≡
Hertz1 -Signorini2 conditions restricted
≡ regions
Hertz1 -Signorini2 -Moreau3 conditions
also known in optimization theory as
Karush4 -Kuhn5 -Tucker6 conditions Improved scheme explaining
normal contact conditions
g ≥ 0, σn ≤ 0, gσn = 0
1 Heinrich Rudolf Hertz (1857–1894) a German physicist who first formulated and solved the frictionless contact
problem between elastic ellipsoidal bodies.
2 Antonio Signorini (1888–1963) an Italian mathematical physicist who gave a general and rigorous mathematical
formulation of contact constraints.
3 Jean Jacques Moreau (1923) a French mathematician who formulated a non-convex optimization problem based
on these conditions and introduced pseudo-potentials in contact mechanics.
4 William Karush (1917–1997), 5 Harold William Kuhn (1925) American mathematicians,
6 Albert William Tucker (1905–1995) a Canadian mathematician.
12/34
Relative sliding
Recall:
2 2
• Convective coordinate in parent 4
17 3 6
space ξi ∈ (−1; 1) 3 ( *1, *2) 5
• Mapping to real space } *1, *2}

8 6 8
7
-1 1
e3
8 1 1
X 4
ρ(ξ1 , ξ2 , t) = N (ξ1 , ξ2 )ρ (t)
i i
e2
1
-1
5 2 e1
i=1
13/34
Relative sliding
Recall:
2 2
• Convective coordinate in parent 4
17 3 6
space ξi ∈ (−1; 1) 3 ( *1, *2) 5
• Mapping to real space } *1, *2}

8 6 8
7
-1 1
e3
8 1 1
X 4
ρ(ξ1 , ξ2 , t) = N (ξ1 , ξ2 )ρ (t)
i i
e2
1
-1
5 2 e1
i=1
Tangential slip velocity vt

must take into account:
• only tangential component
• relative rigid body motion
• master’s deformation
∂ρ ∂ρ
vt = ξ̇1 + ξ̇2
∂ξ1 ∂ξ2
where ∂ρ/∂ξi are the tangent vectors
of the local basis and ξi are the
convective coordinates.
Relative slip between a slave point and a
deformable master surface 13/34
Relative sliding: example
A
Consider a one-dimensional example:
P is a projection of A on segment BC.
xP = ξxC + (1 − ξ)xB (1) B P C
Velocity of the projection point
ẋP = ξẋC + (1 − ξ)ẋB + (xC − xB )ξ̇
x
| {z } | {z }
∂xP ∂xP
∂t ∂ξ
ξ̇
Example of a one-dimensional
Substract the velocity of point xP for fixed ξ relative slip
∂xP ∂x
vt = ẋP − ∂t
= (xC − xB )ξ̇ = ∂ξ
ξ̇
Compute tangential slip increment

∂x
∆gn+1
t = ∂ξ ξn
(ξn+1 − ξn )
14/34
A
xP = ξxC + (1 − ξ)xB (1) B P C
ẋP = ξẋC + (1 − ξ)ẋB + (xC − xB )ξ̇
x
| {z } | {z }
∂xP ∂xP
∂t ∂ξ
ξ̇
∂xP ∂x
vt = ẋP − ∂t
= (xC − xB )ξ̇ = ∂ξ
ξ̇

∂x Ship-river analogy
∆gn+1
t = ∂ξ ξn
(ξn+1 − ξn )
14/34
A
xP = ξxC + (1 − ξ)xB (1) B P C
ẋP = ξẋC + (1 − ξ)ẋB + (xC − xB )ξ̇
x
| {z } | {z }
∂xP ∂xP
∂t ∂ξ
ξ̇
∂xP ∂x
vt = ẋP − ∂t
= (xC − xB )ξ̇ = ∂ξ
ξ̇

∆gn+1
t = ∂ξ ξn
(ξn+1 − ξn )
14/34
A
xP = ξxC + (1 − ξ)xB (1) B P C
ẋP = ξẋC + (1 − ξ)ẋB + (xC − xB )ξ̇
x
| {z } | {z }
∂xP ∂xP
∂t ∂ξ
ξ̇
∂xP ∂x
vt = ẋP − ∂t
= (xC − xB )ξ̇ = ∂ξ
ξ̇

∆gn+1
t = ∂ξ ξn
(ξn+1 − ξn )
Li derivative: the change of a
vector field along the change of
another vector field
14/34
Amontons-Coulomb’s friction
t
No contact g > 0, σn = 0
Stick |vt | = 0 n
Inside slip surface / Coulomb’s cone
f = |σ t | − µ|σn | < 0
Slip |vt | > 0
On slip surface / Coulomb’s cone 0 vt
f = |σ t | − µ|σn | = 0 t
Complementary condition 1
Either zero velocity and negative
slip criterion, or non-zero velocity
and zero slip criterion

|vt | |σ t | − µ|σn | = 0
0 n
Scheme explaining frictional contact
conditions
15/34
t
Stick |vt | = 0 n slip
restricted
stick
f = |σ t | − µ|σn | < 0 regions
Slip |vt | > 0
On slip surface / Coulomb’s cone 0 vt
f = |σ t | − µ|σn | = 0 t
Complementary condition 1 restricted

Either zero velocity and negative region
sli
slip criterion, or non-zero velocity
p
and zero slip criterion
stick
|vt | |σ t | − µ|σn | = 0
0 n
Improved scheme explaining
frictional contact conditions
15/34
Stick |vt | = 0
f = |σ t | − µ|σn | < 0
Scheme of 2D frictional contact
Slip |vt | > 0 vt n
On slip surface / Coulomb’s cone t2
f = |σ t | − µ|σn | = 0 n
t1
Complementary condition
Either zero velocity and negative
slip slip
slip criterion, or non-zero velocity t2
stick
stick
and zero slip criterion n

|vt | |σ t | − µ|σn | = 0 t1
Scheme of 3D frictional contact

|vt | ≥ 0, |σ t | − µ|σn | ≤ 0, |vt | |σ t | − µ|σn | = 0
15/34
More friction laws

• Static criteria
t t t t
slip max t
sli
p max t
sli
stick stick stick sli stick p
p
0 n 0 n 0 n 0 n
(a) (b) (c) (d)
(a) Tresca (b) Amontons-Coulomb (c) Coulomb-Orowan (d) Shaw
• Kinetic criteria
t t t t
n n n n
s s s slip s
k slip k slip k slip
stick
stick
stick
stick
0 vt 0 vt 0 log( vt +v0 ) 0 gt
(a) (b) (c) (d)
(a,b) velocity weakening (c) velocity weakening-strengthening
(d) Linear slip weakening
• µs static and µk kinetic coefficients of friction.
16/34
Rate and state friction and regularization

• Rate and state friction law
0.5
Slip velocity
0.4
Rate vt = |vt | – relative slip velocity
Slip velocity
0.3
0.2
State θ – ≈ internal time 0.1
0.8
Dieterich–Ruina–Perrin (1979, 83, 95) 0.7
Frictional resistance
0.6
0.5
Frictional resistance 0.4
0.3
σct = |σn | µs + bθ + a ln(vt /v0 )

Resistance
0.2
500 600 700 800 900 1000 1100 1200
Time
Evolution of the state variable

h i Rate and state friction law
θ̇ = − vLt θ + ln vv0t
500
Contact pressure
400
• Prakash-Clifton friction law (1992,2000)
Contact pressure
300
200
100
500
Resistance
Viscous type evolution of frictional 400
300
resistance σt 200
100
σ̇t = − vLt (σt + µσn ) 0

500 600 700 800
Time
900 1000 1100 1200
Prakash-Clifton regularization
17/34

0.5
Slip velocity
0.4
Slip velocity
0.3
0.2
0.1
0.8
0.7
0.6
0.5
0.4
0.3
Resistance
0.2
500 600 700 800 900 1000 1100 1200
Time
17/34

0.5
Slip velocity
0.4
Rate vt = |vt | – relative slip velocity
Slip velocity
0.3
0.2
State θ – ≈ internal time 0.1
0.8
Dieterich–Ruina–Perrin (1979, 83, 95) 0.7
0.6
0.5
Frictional resistance 0.4
0.3
σct = |σn | µs + bθ + a ln(vt /v0 )

Resistance
0.2
500 600 700 800 900 1000 1100 1200
Time
Evolution of the state variable

h i Rate and state friction law
θ̇ = − vLt θ + ln vv0t
500
Contact pressure
400
Contact pressure
300
200
100
500
Resistance
Viscous type evolution of frictional 400
300
resistance σt 200
100
σ̇t = − vLt (σt + µσn ) 0

500 600 700 800
Time
900 1000 1100 1200
Prakash-Clifton regularization
17/34

500
Contact pressure
400
Contact pressure
300
200
100
500
Resistance
400
300
200
100
0
500 600 700 800 900 1000 1100 1200
Time
17/34

500
Contact pressure
400
Contact pressure
300
200
100
0
Time
2
Normalized frictional resistance
Normalized resistance
1.5
0.5
0
500 600 700 800 900 1000 1100 1200
Time
17/34
Strong to weak form

• Balance of momentum and boundary conditions
∇ · σ + f v = 0 in Ω = Ω1 ∪ Ω2 + B.C.
= f
u
1
c
2
c
2
f
u
Two solids in contact
18/34
Strong to weak form

∇ · σ + f v = 0 in Ω = Ω1 ∪ Ω2 + B.C.
= f
1
• Balance of virtual works
u
Z Z h 1
i c
n · σ · δu dΓ + f v · δu − σ ·· δ∇u dΩ = 0 2
= = c
∂Ω Ω 2
f
u
18/34
Strong to weak form

∇ · σ + f v = 0 in Ω = Ω1 ∪ Ω2 + B.C. f
=
1
• Balance of virtual works u

1
c
n
Z Z h i
n · σ · δu dΓ = f v · δu − σ ·· δ∇u dΩ = 0 v
= =
2
∂Ω Ω c
2
f
Z Z Z u
n·σ· δρ dΓ1c + ν·σ· δr dΓ2c + σ 0 · δu dΓf

= =
Γf
Γc1 Γc2
18/34
Strong to weak form

∇ · σ + f v = 0 in Ω = Ω1 ∪ Ω2 + B.C. f
=
1

1
c
n
Z Z h i
n · σ · δu dΓ ⇒ f v · δu − σ ·· δ∇u dΩ = 0 v
= =
2
∂Ω Ω c
2
f
Z Z Z u
n·σ· δρ dΓ1c + ν·σ· δr dΓ2c = σ 0 · δu dΓf

= =
Γf
Γc1 Γc2
Z Z
= n · σ · δ(ρ − r) dΓ1c = σ Tt δξ dΓ1c
σn δgn + ∼
= ∼
Γc1 Γc1
18/34
Strong to weak form

∇ · σ + f v = 0 in Ω = Ω1 ∪ Ω2 + B.C. f
=
1

1
c
n
Z Z h i
n · σ · δu dΓ ⇒ f v · δu − σ ·· δ∇u dΩ = 0 v
= =
2
∂Ω Ω c
2
f
Z Z Z u
n·σ· δρ dΓ1c + ν·σ· δr dΓ2c = σ 0 · δu dΓf

= =
Γf
Γc1 Γc2
Z Z
= n · σ · δ(ρ − r) dΓ1c = σ Tt δξ dΓ1c
σn δgn + ∼
= ∼
Γc1 Γc1
Z Z Z Z
σ · ·δ∇u dΩ + σ Tt δξ dΓ1c =
σn δgn + ∼ σ 0 · δu dΓ + f v · δu dΩ
= ∼
Ω Γc1 Γf Ω
| {z }
Contact term 18/34
Strong to weak form

∇ · σ + f v = 0 in Ω = Ω1 ∪ Ω2 + B.C. f
=
1

1
c
n
Z Z
σ · ·δ∇u dΩ + σ Tt δξ dΓ1c =
σn δgn + ∼ v
= ∼ 2
Ω Γc1 c
2
| {z }
| {z } f
Change of the internal energy u
Contact term
Z Z
σ 0 · δu dΓ + f v · δu dΩ Two solids in contact
Γf Ω
| {z } | {z }
Virtual work of external forces Virtual work of volume forces
• Functional space
u ∈ H1 (Ω) Hilbert space of the first order
and u satisfy boundary and contact conditions.
18/34
Towards variational inequality

Contact term
Z
σ Tt δξ dΓ1c
σn δgn + ∼
∼
Γc1
Z
σn δgn dΓ1c ≤ 0
Γc1
Contact configuration σn δgn = 0, σn ≤ 0

Z Z Z Z
σ · ·δ∇u dΩ +
=
σ Tt δ
∼ ξ dΓ1c ≥ σ 0 · δu dΓ + f v · δu dΩ,
∼
Ω Γc1 Γf Ω
n o
K = δu ∈ H1 (Ω) δu = 0 on Γu , gn (u + δu) ≥ 0 on Γc
19/34

Contact term
Z
σ Tt δξ dΓ1c
σn δgn + ∼
∼
Γc1
Z
Γc1
Virtual change of the configuration

Z Z Z Z
σ · ·δ∇u dΩ +
=
σ Tt δ
∼
Ω Γc1 Γf Ω
n o
19/34

Contact term
Z
σ Tt δξ dΓ1c
σn δgn + ∼
∼
Γc1
Z
Γc1
Normal term in separation δgn > 0

Z Z Z Z
σ · ·δ∇u dΩ +
=
σ Tt δ
∼
Ω Γc1 Γf Ω
n o
19/34

Contact term
Z
σ Tt δξ dΓ1c
σn δgn + ∼
∼
Γc1
Z
Γc1
Normal term in sliding δgn = 0

Z Z Z Z
σ · ·δ∇u dΩ +
=
σ Tt δ
∼
Ω Γc1 Γf Ω
n o
19/34
Back to variational equality (unconstrained)

• Constrained minimization problem
Z Z Z Z
σ · ·δ∇u dΩ + σ t δξ dΓc ≥
T 1
σ 0 · δu dΓ + f v · δu dΩ,
= ∼ ∼
Ω Γc1 Γf Ω
n o
• Use optimization theory to convert to
Z Z Z Z
σ · ·δ∇u dΩ + C(σn , σt , gn , ξ , δu) dΓc =
1
σ 0 · δu dΓ + f v · δu dΩ,
= ∼
Ω Γ1c | {z } Γf Ω
Contact term∗
n o
Unconstrained functional space K = δu ∈ H1 (Ω) δu = 0 on Γu
Contact term∗ is defined on the potential contact zone Γ1c .
20/34
3. Optimization methods
Optimization methods: recall

Functional to be minimized F(x) under constraint g(x) ≥ 0
Penalty method
Lagrange multipliers method
Augmented Lagrangian method
22/34

Penalty method
• New functional

2 0,
 if g(x) ≥ 0 non-contact
Fp (x) = F(x) + −g(x) = F(x) + 


g2 (x),
 if g(x) < 0 contact
where is the penalty parameter.
• Stationary point must satisfy
∇Fp (x) = ∇F(x) + 2 −g(x) ∇g(x) = 0

• Solution tends to the precise solution as → ∞


x, if x ≥ 0
Macaulay brackets hxi = 

0, otherwise
22/34

2
Penalty method Fp (x) = F(x) + −g(x)


• New functional called Lagrangian
L(x, λ) = F(x) + λg(x)
• Saddle point problem
min max{L(x, λ)} −→ x∗ ←− min {F(x)}
x λ g(x)≥0
• Stationary point
∇x F(x) + λ∇x g(x)
" #
∇x,λ L = = 0 need to verify λ ≤ 0
g(x)

x, if x ≥ 0

0, otherwise
22/34

2
Penalty method Fp (x) = F(x) + −g(x)

Lagrange multipliers method L(x, λ) = F(x) + λg(x)

[Hestnes 1969], [Powell 1969], [Glowinski & Le Tallec 1989], [Alart & Curnier 1991], [Simo & Laursen 1992]
• New functional, augmented Lagrangian


 λg(x) + g2 (x) , if λ + 2g(x) ≥ 0, contact

La (x, λ) = F(x) + 

− 1 λ2 ,

if λ + 2g(x) < 0, non-contact
4
• Stationary point
 
∇x F(x) + λ∇x g(x) + 2g(x)∇g(x)
 = 0, if contact



 

  g(x)
∇x,λ La = 

 
∇x F(x)

 λ  = 0,


 if non-contact
−

x, if x ≥ 0

0, otherwise 22/34
Optimization methods: example
Functional : f (x) = x2 + 2x + 1
Constrain : g(x) = x ≥ 0
Solution : x∗ = 0
23/34
Optimization methods: example
Functional : f (x) = x2 + 2x + 1
Constrain : g(x) = x ≥ 0
Solution : x∗ = 0
23/34
Penalty method: example
F(x) = x2 + 2x + 1, g(x) = x ≥ 0, x∗ = 0
Penalty method
Fp (x) = F(x) + −g(x) 2

24/34
F(x) = x2 + 2x + 1, g(x) = x ≥ 0, x∗ = 0
Penalty method
Fp (x) = F(x) + −g(x) 2

=0
24/34
F(x) = x2 + 2x + 1, g(x) = x ≥ 0, x∗ = 0
Penalty method
Fp (x) = F(x) + −g(x) 2

=1
24/34
F(x) = x2 + 2x + 1, g(x) = x ≥ 0, x∗ = 0
Penalty method
Fp (x) = F(x) + −g(x) 2

= 10
24/34
F(x) = x2 + 2x + 1, g(x) = x ≥ 0, x∗ = 0
Penalty method
Fp (x) = F(x) + −g(x) 2

= 50
24/34
F(x) = x2 + 2x + 1, g(x) = x ≥ 0, x∗ = 0
Penalty method
Fp (x) = F(x) + −g(x) 2

Advantages , Drawbacks /
simple physical interpretation practically non-smooth
simple implementation functional
no additional degrees of freedom solution is not exact:
“mathematically” smooth too small penalty →
functional large penetration
too large penalty →
ill-conditioning of the
tangent matrix
user has to choose penalty
properly or automatically and/or
adapt during convergence
24/34
Lagrange multipliers method: example
F(x) = x2 + 2x + 1, g(x) = x ≥ 0, x∗ = 0
L(x, λ) = F(x) + λg(x) → Saddle point → min max L(x, λ)
x λ
Need to check that λ ≤ 0
-1 λ
1
0 -2
-1
λ
2
-2 1 -3
0
-3
-2
-1 X
-2 -1 0 1 2
X
25/34
Lagrange multipliers method: example
F(x) = x2 + 2x + 1, g(x) = x ≥ 0, x∗ = 0
L(x, λ) = F(x) + λg(x) → Saddle point → min max L(x, λ)
x λ
Need to check that λ ≤ 0
exact solution Lagrangian is not smooth
no adjustable parameters additional degrees of freedom
not fully unconstrained: λ ≤ 0
25/34
Augmented Lagrangian method: example
F(x) = x2 + 2x + 1, g(x) = x ≥ 0, x∗ = 0


La (x, λ) = F(x) + 

− 1 λ2 ,

4
-1 λ
1
0 -2
-1
λ
2
-2 1 -3
0
-3
-2
-1 X
-2 -1 0 1 2
X
Yellow line separates contact and non-contact regions
26/34
F(x) = x2 + 2x + 1, g(x) = x ≥ 0, x∗ = 0


La (x, λ) = F(x) + 

− 1 λ2 ,

4
-1 λ
1
0 -2
-1
λ
2
-2 1 -3
0
-3
-2
-1 X
-2 -1 0 1 2
X
26/34
F(x) = x2 + 2x + 1, g(x) = x ≥ 0, x∗ = 0


La (x, λ) = F(x) + 

− 1 λ2 ,

4
-1 λ
1
0 -2
-1
λ
2
-2 1 -3
0
-3
-2
-1 X
-2 -1 0 1 2
X
26/34
F(x) = x2 + 2x + 1, g(x) = x ≥ 0, x∗ = 0


La (x, λ) = F(x) + 

− 1 λ2 ,

4
exact solution additional degrees of freedom
smooth functional (!) quite sensitive to parameter
fully unconstrained need to adjust during
convergence
26/34
Application to contact problems: weak form

Z Z Z Z
σ · ·δ∇u dΩ + C dΓ1c = σ 0 · δu dΓ + f v · δu dΩ,
=
Ω Γf Ω
|{z}
Γ1c
Contact term
n o
K = δu ∈ H1 (Ω) δu = 0 on Γu
Penalty method

gt , if stick |σt | < µ|σn |

Pressure: σn = gn , Shear: σ t = 

µgn δg /|δg |,
 if slip |σt | = µ|σn |
t t
Contact term
C = C(gn , g , δgn , δg ) = σn δgn + σ t · δg
t t t
27/34
Application to contact problems: weak form

Z Z Z Z
σ · ·δ∇u dΩ + C dΓ1c = σ 0 · δu dΓ + f v · δu dΩ,
=
Ω Γf Ω
|{z}
Γ1c
Contact term
n o
K = δu ∈ H1 (Ω) δu = 0 on Γu
Contact term
C = C(gn , g , λn , λt , δgn , δg , δλn , δλt )
t t
 1


− λn δλn − λt · δλt , if non-contact λn + gn ≥ 0






λ̂n δgn + gn δλn + λ̂t · δgt + gt · δλ̂t , if stick |λ̂t | ≤ µ|σ̂n |


C=





λ̂t λ̂ 

  
λt + µσ̂n t  · δλ ,

λ̂n δgn + gn δλn + µσ̂n − µσ̂n |λ̂ | · δgt − if slip |λ̂t | ≥ µ|σ̂n |
1

 


|λ̂ |
   t
t t
where λ̂n = λn + gn and λ̂t = λt + g .

t
28/34
Application to contact problems: linearization

• Non-linear equation
R(u) = 0
• Contains δgn , δg
t
• Use Newton-Raphson method • Initial state at step i
R(ui ) = 0
• Want to have the same at step i + 1
R(ui+1 ) = R(ui + δu) = 0
• Linearize
∂R(u)
R(ui + δu) = R(ui ) + δu = 0
∂u
• Finally
#−1
∂R(u)
"
δu = − R(ui )
∂u
| {z }
contains ∆δgn , ∆δg
t
29/34
4. Examples and conclusion
Examples of contact problems

With analytical solution
∗ ∗ ? linear elasticity
∗ ∗ ? with/without friction
From literature
∗ ∗ ? post-buckling 2D
∗ ∗ ? finite strains
∗ ∗ ? elasticity / plasticity
New
∗ ∗ ? multi-contacts
∗ ∗ ? post-buckling 3D
∗ ∗ ? finite strains
∗ ∗ ? elasticity / plasticity
31/34
Self-contact problem
Finite element analysis of a post-buckling behavior of a thin walled tube

Collection of non-linearities: buckling instability, self-contact, finite strain plasticity
32/34
Reading
It’s just a tip of the

“Computational Contact
Mechanics” iceberg
Contact detection
Contact discretization and
integration
Smoothing techniques
Energy conservative methods for
dynamics
33/34
Thank you for your attention!
hvladislav.yastrebov@mines-paristech.fri

CompContactMech Yastrebov

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CompContactMech Yastrebov

Uploaded by

Copyright:

Available Formats

NONLINEAR COMPUTATIONAL MECHANICS

Industrial and natural contact problems

Aircraft’s engine GP 7200

Industrial and natural contact problems

High speed train TGV www.sncf.com

Industrial and natural contact problems

[1] F. Massi, J. Rocchi, A. Culla, Y. Berthier

Industrial and natural contact problems

Helical gear www.tpg.com.tw

Industrial and natural contact problems

Industrial and natural contact problems

Tire-road contact www.michelin.com

[1] M. Brinkmeier, U. Nackenhorst, S. Petersen,

Industrial and natural contact problems

[1] G. Rousselier, F. Barlat, J. W. Yoon

Industrial and natural contact problems

[1] O. Klyavin, A. Michailov, A. Borovkov

Industrial and natural contact problems

J. A. Weiss, University of Utah

Industrial and natural contact problems

E. Azema et al, LMGC90

Industrial and natural contact problems

Simulation of electric current

Industrial and natural contact problems

[1] J.D. Garaud, L. Fleitout, G. Cailletaud

Industrial and natural contact problems

Industrial and natural contact problems

Industrial and natural contact problems

Involved physical phenomena

Contact complexity: physics and mathematics

Contact complexity: physics and mathematics

Lack of standard optimization

Particularities: mesh and convergence

Particularities: mesh and convergence

Infinite contact pressure and/or its derivative

Particularities: mesh and convergence

Initial guess R(x0 , f0 ) = 0

Particularities: mesh and convergence

Too rapid change in boundary conditions R(x0 , f1 ) , 0

Particularities: mesh and convergence

Iterations of Newton-Raphson method

Particularities: mesh and convergence

Iterations of Newton-Raphson method

Particularities: mesh and convergence

Particularities: mesh and convergence

Initial guess R(x0 , f0 ) = 0

Particularities: mesh and convergence

Too rapid change in boundary conditions R(x0 , f1 ) , 0

Particularities: mesh and convergence

Iterations of Newton-Raphson method

Particularities: mesh and convergence

Iterations of Newton-Raphson method

Particularities: mesh and convergence

Convergence, but is it a “true” solution ?

• ∇a = B gradient operator •∇·a=c divergence operator

Equilibrium and contact conditions

Equilibrium and contact conditions

Equilibrium and contact conditions

Equilibrium and contact conditions

Equilibrium and contact conditions

Frictionless or normal contact conditions

Frictionless or normal contact conditions

Frictionless or normal contact conditions

• Mapping to real space } 1, 2}

• Mapping to real space } 1, 2}

• Solution tends to the precise solution as → ∞