Finite Elements II - Master 2

∗
Ioan R. Ionescu

November 13, 2020

Contents
1 Basics in finite elements method 4
1.1 FreeFem++ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Mesh generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1 2D meshes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.2 3D meshes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Finite element spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.1 Introducing finite element spaces . . . . . . . . . . . . . . . . . . 7
1.3.2 Finite element functions . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.3 Finite element basis . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.4 Finite element interpolation . . . . . . . . . . . . . . . . . . . . . 8
1.4 Visualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4.1 FreeFem++ Visualization . . . . . . . . . . . . . . . . . . . . . . 9
1.4.2 VTK visualization . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2 Static problems in linear elasticity 13

2.1 Boundary value problem . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1.1 Variational problem . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.2 Existence and unicity of the solution . . . . . . . . . . . . . . . . 14
2.1.3 Finite element approximation . . . . . . . . . . . . . . . . . . . . 15
2.2 Out-of-plane (scalar) two dimensional problems . . . . . . . . . . . . . . 16
2.2.1 Boundary value problem . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.2 Problème modèle: déformation d’une pièce rectangulaire . . . . . 18
∗
LSPM, University Paris 13, Sorbonne-Paris-Cité, Villetaneuse, France, [email protected]

1
 2.2.3 Cisaillement anti-plan d’un carré fissuré (mode III) . . . . . . . . 23
2.2.4 Déformation (anti-plane) d’une demi-disque . . . . . . . . . . . . 25
2.2.5 Déformation anti-plane d’un écrou épais . . . . . . . . . . . . . . 26
2.3 Vectorial two dimensional problems . . . . . . . . . . . . . . . . . . . . . 28
2.3.1 Assumptions in two spatial dimensions . . . . . . . . . . . . . . . 28
2.3.2 Model problem: Ecrasement d’un lopin . . . . . . . . . . . . . . 29
2.3.3 Piéce suspendue . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.3.4 Traction d’un carré fissuré (mode II) . . . . . . . . . . . . . . . . 35
2.3.5 Different treatment of displacement boundary conditions . . . . . 37
2.3.6 Visser un écrou épais . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.3.7 Poinçonnement d’un demi disque . . . . . . . . . . . . . . . . . . 40
2.4 Axi-symmetric problems . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.4.1 Ecrasement d’un tube . . . . . . . . . . . . . . . . . . . . . . . . 42
2.4.2 Compression triaxiale d’un cylindre creux . . . . . . . . . . . . . 43
2.5 Three dimensional problems . . . . . . . . . . . . . . . . . . . . . . . . . 43

3 Static problems in nonlinear elasticity 44

3.1 Finite elasticity mechanical model . . . . . . . . . . . . . . . . . . . . . . 44
3.1.1 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.1.2 Conservation laws . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.1.3 Constitutive elastic law . . . . . . . . . . . . . . . . . . . . . . . . 46
3.2 Approximation of the static nonlinear elastic problems . . . . . . . . . . 47
3.2.1 Lagrangian formulation of the boundary value problem . . . . . . 47
3.2.2 Variational formulation . . . . . . . . . . . . . . . . . . . . . . . . 48
3.2.3 Newton-Raphson algorithm . . . . . . . . . . . . . . . . . . . . . 48
3.2.4 Incremental loading and/or quasi-static evolution . . . . . . . . . 49
3.3 Two dimensional problems . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.3.1 Deformation of a diving board . . . . . . . . . . . . . . . . . . . 52
3.3.2 Floor with a nonuniform loading . . . . . . . . . . . . . . . . . . 56
3.3.3 Folding a semicircle . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.3.4 Squeezing an empty square . . . . . . . . . . . . . . . . . . . . . . 59
3.3.5 Wrapping a cylinder with a planar sheet . . . . . . . . . . . . . . 60
3.4 Axi-symmetric problems . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.5 Three dimensional problems . . . . . . . . . . . . . . . . . . . . . . . . . 61

2
4 Appendix 62
4.1 Macros for Nonlinear elasticity . . . . . . . . . . . . . . . . . . . . . . . . 62
4.1.1 General Macros in 2D . . . . . . . . . . . . . . . . . . . . . . . . 62
4.1.2 Macros for Saint-Venant-Kirchhoff model . . . . . . . . . . . . . . 63

3
1 Basics in finite elements method

1.1 FreeFem++
FreeFEM is a partial differential equation solver for non-linear multi-physics systems
in 2D and 3D.Problems involving partial differential equations from several branches
of physics, such as fluid-structure interactions, require interpolations of data on several
meshes and their manipulation within one program. FreeFEM includes a fast interpola-
tion algorithm and a language for the manipulation of data on multiple meshes. FreeFEM
is written in C++ and its language is a C++ idiom.
To use FreeFEM, two installation methods are available: user access (binary package)
and access developers (from the source code). Follow the section corresponding to your
type of installation. For the installation details and the FreeFEM documentation please
visit https://freefem.org/.

How to use FreeFem++ ?

1/ Edit a file "nomfichier.edp" (with : emacs, vi, medit, etc.)
2/ Run your file script : freefem++ nomfichier.edp

1.2 Mesh generation

Let D ⊂ R2 or D ⊂ R3 be a domain and let Th be a triangulation of D, i.e.
[
D= Th ,
Th ∈Th

where Th are triangles in R2 or tetrahedra in R3 .

The mesh information of a mesh called Th:

General information
• Th.nt = Nt number of triangles
• Th.nv = Nv number of vertexes
• Th.ne = Ne number of edges
• Th.measure = A aria
• Th.bordermeasure = P perimeter
• N.x, N.y = (nx , ny ) the external normal at the boundary

Information of a point of coordinates (xx,yy)

• Th(xx,yy).nuTriangle = number of the triangle T such that (xx, yy) ∈ T

4
 • Th(xx,yy).region = number of the region R such that (xx, yy) ∈ R
Information of a vertex i = 0, ..., Nv − 1:
• Th(i).x Th(i).y = coordinates of i
• Th(i).label = region of i

Information of a triangle i = 0, ..., Nt − 1:

• Th [i][j] = the global vertex of the local vertex j = 0, 1, 2
Chapter
• Th [i][j] .x 5Th [i][j] .y = coordinates of the global vertex Th [i][j]
• Th[i].area = area of i
• Th[i].region = region of i
Mesh Generation
1.2.1 2D meshes

Direct mesh generation

5.1 Commands
The command for Mesh
square triangulates Generation
the unit square [0, 1] × [0, 1] on a 10 × 10 grid
mesh Th = square(10,10);
Let us begin with the two important keywords border and buildmesh
Toexamples
All plot a mesh on the
in this screen
section andfrom
come on a the
file a.eps”
files mesh.edp and tablefunction.edp.
plot(Th,wait=1,ps="a.eps");
To construct
5.1.1 Squarea n × m grid in the rectangle [x0 , x1 ] × [y0 , y1 ]:
The real x0=1.2,x1=1.8;
command“square” triangulates the unit square. The following
real y0=0,y1=1;
mesh Th = square(4,5);
int n=5,m=20;
generates 4 ⇥ 5 grid in the unit square [0, 1]2 . The labels of the boundaries are shown in
meshaTh=square(n,m,[x0+(x1-x0)*x,y0+(y1-y0)*y]);
Fig. 5.1. To construct a n ⇥ m grid in the rectangle [x0 , x1 ] ⇥ [y0 , y1 ], proceeds as follows:

$! " # $% (

!! " # $ $! " # $
% & % )

$! " # $% '

Figure 5.1: Boundary labels of the mesh by square(10,10)

Figure 1: Boundary labels of the mesh bysquare(10,10).
write

real x0=1.2,x1=1.8; 5
real y0=0,y1=1;
int n=5,m=20;
mesh Th=square(n,m,[x0+(x1-x0)*x,y0+(y1-y0)*y]);

109
 Mesh generation by the border
Boundaries are defined piecewise by parametrized curves. The pieces can only intersect
at their endpoints, but it is possible to join more than two endpoints. The curve C
parametrized through t → (fx (t), fy (t)), with t ∈ (0, 2π) and fx (t) = cos(t), fy (t) =
sin(t), avec le label 1 will be defined by
border C(t=0,2*pi){ x=cos(t); y=sin(t);label=1;}
To plot the curve C with 50 segments
plot(C(50),wait=1,ps="C.eps" ) ;
Define a triangulation with buildmesh
mesh Th= buildmesh(C(50) );
The general expression112to define a triangulation with buildmesh is
CHAPTER 5. MESH GENERATION

mesh NameMesh= buildmesh(C1(m1)+....+CJ

!"! "!
!"! "! !"! "! (mj) ); $ #
$

where mj are positive or negative numbers

% " " ! " ! % ! & ' " # " ! " ! % !to indicate how many vertices should be on
! ! !
&& " #

!!! "!#
Cj with mj > 0 if the domain is on the left side
% " " ! " ! ' and
" # " ! " ! & mj "! ") !0 if the domain is on the right
!"! "!$
% " " ! " ! ' " #<
% $ ! "% !& ($ !'"$ ! #% !& ($ !& !"! "!#
&

side. Figure 5.4: Orientation of the boundary defined by ( x (t), y (t))

Construct a mesh withTheaorientation

interior boundary
of boundaries can be changedD
by changing the sign of m . The following j
example shows how to change the orientation. The example generates the unit disk with a
border D(t=0,2*pi){ x=0.3+0.3*cos(t); y=0.3*sin(t);label=2;}
small circular hole, and assign “1” to the unit disk (“2” to the circle inside). The boundary
label must be non-zero, but it can also be omitted.

mesh CBh= buildmesh(C(50)+D(30) );

1:
2:
border a(t=0,2*pi){ x=cos(t); y=sin(t);label=1;}
border b(t=0,2*pi){ x=0.3+0.3*cos(t); y=0.3*sin(t);label=2;}
3: plot(a(50)+b(+30)) ; // to see a plot of the border mesh

Construct a mesh with a hole 4:

5:
mesh Thwithouthole= buildmesh(a(50)+b(+30));
mesh Thwithhole = buildmesh(a(50)+b(-30));
6: plot(Thwithouthole,wait=1,ps="Thwithouthole.eps"); // figure 5.5
7: plot(Thwithhole,wait=1,ps="Thwithhole.eps"); // figure 5.6
mesh CHh= buildmesh(C(50)+D(-30) );
Note 5.2 Notice that the orientation is changed by “b(-30)” in 5th line. In 7th line,
ps="fileName" is used to generate a postscript file with identification shown on the figure.

Figure 2: Meshes CBh

Figure 5.5: mesh without holeand
Figure 5.6: meshCHh.
with hole

Note 5.3 Borders are evaluated only at the time plot or buildmesh is called so the global
variable are defined at this time andhere since r is changed between the two border calls the
following code will not work because the first border will be computed with r=0.3:
1.2.2 3D meshes

6
1.3 Finite element spaces
1.3.1 Introducing finite element spaces

Let let Th be a triangulation of D. We consider two types of finite element spaces:

Continuous finite element functions

Vh = {v : D → R ; v is continuous and v ∈ P k (T ), for all T ∈ Th },

where P k (T ) is the space of polynomial function of degree k on T , with k = 1, 2.

fespace NameSc(Namemesh, P1);
fespace NameSc(Namemesh, P2);
Discontinuous finite element functions

Vh = {v : D → R ; v ∈ P k (T ), for all T ∈ Th },

where P k (T ) is the space of polynomial function of degree k on T , with k = 0, 1, 2.

fespace NameSd(Namemesh, P0);
fespace NameSd(Namemesh, P1dc);
fespace NameSd(Namemesh, P2dc);
Note that P 1dc and P 2dc are implemented only in 2D.

c0 c0 +a1x+b1y c0 +a1x+b1y+a2x2+b2y2+d2xy

P1 P2

P0 P1dc P2dc

Figure 3: Representation of the finite element spaces.

7
1.3.2 Finite element functions

Introducing a function u ∈ Vh
NameS u;
The value u(x0 , y0 ) of u in (x0 , y0 )
real value=u(x0,y0);
Note that x, y, z are notations reserved for spatial variables.

int dim=NameS.ndof; // dim{N ameS}=Number of degree of freedom

int dimE=NameS.ndofK; // = Number of degree of freedom/element
int ndofkj=NameS(k,j); // =the j th degree of freedom of element k

1.3.3 Finite element basis

Let Φ1 , .......ΦM ∈ Vh be a base (set of shape functions) in Vh , i.e.

dim(Vh ) = M, Vh = Sp{Φ1 , .......ΦM },

where M is the number of degree of freedom and any function u ∈ Vh can be written as

M
X
u(x, y, z) = Ui Φi (x, y, z).
i=1

M = Nt for P 0, M = Nv for P 1, M = Nv + Ne for P 2,

M = 3Nt for P 1dc, M = 6Nt for P 2dc

u[] is the vector [U1 , ....UM ] and u[][i] is Ui

Defining the shape function Φk :
Vh Phik=0;
Phik[][k]=1;

1.3.4 Finite element interpolation

The operator = is the interpolation operator of FreeFem++ , The continuous finite func-
tions are extended by continuity to the outside of the domain.
int n=4;

8
 IsoValue IsoValue
-0.0526316 IsoValue -0.169591
0.0263158 -0.0467836 -0.0818713
0.0789474 0.0233918 -0.0233918
0.131579 0.0701754 0.0350877
0.184211 0.116959 0.0935673
0.236842 0.163743 0.152047
0.289474 0.210526 0.210526
0.342105 0.25731 0.269006
0.394737 0.304094 0.327485
0.447368 0.350877 0.385965
0.5 0.397661 0.444444
0.552632 0.444444 0.502924
0.605263 0.491228 0.561404
0.657895 0.538012 0.619883
0.710526 0.584795 0.678363
0.763158 0.631579 0.736842
0.815789 0.678363 0.795322
0.868421 0.725146 0.853801
0.921053 0.77193 0.912281
1.05263 0.818713 1.05848
0.935673

Mesh Φ23 in P1 Φ23 in P2 Φ27 in P2

IsoValue IsoValue
-0.0526316 -0.0565302 IsoValue IsoValue
0.0263158 0.0232146 -0.0612922 -0.170659
0.0789474 0.0763778 0.0103081 -0.0813373
0.131579 0.129541 0.0580416 -0.0217898
0.184211 0.182704 0.105775 0.0377578
0.236842 0.235867 0.153509 0.0973054
0.289474 0.289031 0.201242 0.156853
0.342105 0.342194 0.248976 0.216401
0.394737 0.395357 0.296709 0.275948
0.447368 0.44852 0.344443 0.335496
0.5 0.501684 0.392176 0.395043
0.552632 0.554847 0.43991 0.454591
0.605263 0.60801 0.487643 0.514138
0.657895 0.661173 0.535377 0.573686
0.710526 0.714336 0.58311 0.633234
0.763158 0.7675 0.630844 0.692781
0.815789 0.820663 0.678577 0.752329
0.868421 0.873826 0.726311 0.811876
0.921053 0.926989 0.774044 0.871424
1.05263 1.0599 0.821778 0.930971
0.941112 1.07984

Φ23 in P0 Φ23 in P1dc Φ23 in P2dc Φ54 in P2dc

Figure 4: Shape functions of different finite element space on a 10 × 10 mesh of the

square (0, 1) × (0, 1).

func f=sin(pi*x*n)*sin(pi*y*n);
NameS uh=f;

1.4 Visualization
There are two ways of visualization in FreeFem++ : One, the default view of FreeFem++.
Another method is to use external tools as Paraview which use the .vtk (Visualization
Tool Kit) files.

1.4.1 FreeFem++ Visualization

The FreeFem++ visualization supports the drawing of meshes, iso-values of real FE-
functions and of vector fields, all by the command plot.
plot(Th, ps="Mesh.eps",wait=1);
plot(u, fill=true, value=1, ps="uh.eps", wait=1);
NameS uh=x*x+y*y,vh=-y*y+x*x;

9
 10x10 mesh, P0 10x10 mesh, P1 10x10 mesh, P2

100x100 mesh, P0 100x100 mesh, P1 100x100 mesh, P2

Figure 5: Interpolation of f (x, y) = sin(4πx) sin(4πy) on the square (0, 1) × (0, 1) on

different meshes and finite element spaces.

plot(Th,[uh,vh], fill=true, value=1, ps="Vect.eps", wait=1);

1.4.2 VTK visualization

Creation of the .vtk file for a surface (x, y) → (x, y, uh (x, y))
fespace Rh(Th,P1dc) ;
Rh Xdes=x, Ydes=y, Zdes=uh, devdes=uh;
int id;
ofstream par("results.vtk"); // nom du fichier .vtk
par «"# vtk DataFile Version 1.0" «endl;
par «"3D Unstructured Grid of Linear Triangles" «endl;
par «"ASCII" «endl;
par «" " «endl;
par «"DATASET UNSTRUCTURED_GRID" «endl;
par «"POINTS " « Rh.ndof« " float" «endl;// Rh.ndof number of degree of
freedom

10
 Vec Value
0
0.105269
0.210537
0.315806
0.421075
0.526343
0.631612
0.736881
0.84215
0.947418
1.05269
1.15796
1.26322
1.36849
1.47376
1.57903
1.6843
1.78957
1.89484
2.00011

Figure 6: Representation ParaView of the function (x, y) → (x2 + y 2 , x2 − y 2 ) on the

square (0, 1) × (0, 1).

for (id=0;id<Rh.ndof;id++) par« Xdes[][id] «" "« Ydes[][id]«" "« uzdes[][id]«endl;

par «"CELLS " « Rh.nt« " " « (Rh.ndofK+1)*Rh.nt«endl; // Rh.nt number

of triangles
for (id=0;id<Rh.nt;id++) par«"3 "« Rh(id,0) «" "« Rh(id,1)«" "«Rh(id,2)«endl;

par «" " «endl;

par «"CELL_TYPES " « Rh.nt«endl;
for (id=0;id<Rh.nt;id++) par«"5"«endl;
par «" " «endl;
par «"POINT_DATA " « Rh.ndof«endl;
par «"SCALARS uh float"«endl;
par «"LOOKUP_TABLE default " «endl;
for (id=0;id<Rh.ndof;id++) par« devdes[][id]«endl;

11
Figure 7: Representation ParaView of the surface x, y, z = sin(4πx) sin(4πy) on the
square (0, 1) × (0, 1) on the mesh 100 × 100 for the P1dc finite element space.

12
2 Static problems in linear elasticity

2.1 Boundary value problem

Let D ⊂ Rd (d = 2 or d = 3) be the bounded domain occupied by the elastic solid

and ∂D = Γd ∪ Γc ∪ Γ1m ∪ Γ2m be a partition of its boundary. We consider the following
boundary value problem: find u : D → Rd (the displacement) solution of

1
ε(u) = (∇u + ∇T u) in D, (small perturbations assumption)
2

σ(u) = λtrace(ε(u))I + 2µε(u), in D (Hooke’s elastic law)

−divσ(u) = ρb in D, (equilibrium)

u=U on Γd , σ(u) · n = F on Γc ,

un = Un , στ = Fτ on Γ1m , uτ = Uτ , σn = Fn on Γ2m

where n is the exterior unit normal on ∂D, un = u · n, uτ = u − un n are the normal

and tangential displacement, σn = σ(u)n · n, στ = σ(u)n − σn n are the normal and
tangential stresses. The functions ρb (volume forces), U , Un , Uτ (imposed displacements),
and F , Fn , Fτ (imposed stresses) are given.

Gm 2

Gc
Gd
D

Gm 1

Figure 8: Schematic representation of the elastic domain D.

13
2.1.1 Variational problem

Let
U ad = {v : Dd → Rd ; v = U on Γd , vn = Un on Γ1m , vτ = Uτ on Γ2m , }

the set of admissible in-plane displacements and

V = {v : Dd → Rd ; v = 0 on Γd , vn = 0 on Γ1m , vτ = 0 on Γ2m , }

the associated vectorial space.

The variational formulation of the boundary value problem : find u solution of

(V P ) u ∈ U ad , a(u, v) − l(v) = 0, for all v ∈ V,

where a and l are the bilinear and the linear forms given by
Z
a(v, w) = λtrace(ε(v))trace(ε(w)) + 2µε(v) : ε(w) dx,
D

Z Z Z Z
l(w) = b · w dx + F · w da + Fτ · wτ da + Fn wn da.
D Γc Γ1m Γ2m

If we denote by J(v) = 21 a(v, v)−l(v) the elastic energy functional then the variational
problem (V P ) is equivalent with the following minimum problem for the energy

(M P ) u ∈ U ad , J(u) ≤ J(v), for all v ∈ U ad .

2.1.2 Existence and unicity of the solution

If R = {r : D → Rd ; a(r, r) = 0} is the kernel of the bilinear form a then from the

Korn’s inequality it follows that R is the space of infinitesimal rigid displacements

R = {r(x, y) = t + ω(−y, x) ; t ∈ R2 , ω ∈ R}, for d = 2,

R = {r(x) = t + ω ∧ x ; t ∈ R3 , ω ∈ R3 }, for d = 3.

The dimension of this space is finite: dim(R) = 3 for d = 2 and dim(R) = 6 for d = 3.

14
 In what it follows we denote by R0 = R ∩ V. Then we have

Theorem. If D is smooth enough and the input data ρb, U , Un , Uτ , F , Fn , Fτ are

regular then we have:

• a solution u of (VP) exists if and only if l(r) = 0 for all r ∈ R0

• if u is a solution of (VP) then for all r ∈ R0 , u + r is a solution too

• the solution u of (VP) is unique if and only if R0 = {0}

2.1.3 Finite element approximation

Let Vh be the finite element space and let U ad ad

h = Vh ∩ U , V h = Vh ∩ V. Then the
Galerkin (internal) approximation of (P V ) is

(V Ph ) uh ∈ U ad
h , a(uh , vh ) − l(vh ) = 0, for all vh ∈ Vh ,

which is equivalent with the minimization problem for the energy functional

(M Ph ) uh ∈ U ad
h , J(uh ) ≤ J(vh ), for all vh ∈ U ad
h .

Then we have the following error estimation: there exists a constant C such that

||u − uh || ≤ C inf ||u − vh ||.

vh ∈U ad
h

The L2 (D)2 -norm of absolute and relative errors are

sR sR
D
||2
||u − uh dx DR
||u − uh ||2 dx
Eabs = R , Erel = .
D
1 dx D
||u||2 dx

Matrix form of the linear variational problem

Let Φ1 , .......ΦM ∈ Vh be a base (set of shape functions) in Vh , i.e. dim(Vh ) =

M, Vh = Sp{Φ1 , .......ΦM }, with M is the number of degree of freedom. Then uh ∈ Vh
can be written as uh (x, y) = M i=1 Ui Φi (x, y) and let U = (U1 , .., UM ). Let denote by
P

15
A = (Aij )i,j=1,M the rigidity matrix, and by L = (L1 , .., LM ) the loading term (or right-
side vector) given by

Aij = a(Φi , Φj ), Li = l(Φi ), for all i, j = 1, ..M.

Then (P Vh ) is equivalent with the following linear system:

(LSh ) uh ∈ U ad
h , AU = L.

2.2 Out-of-plane (scalar) two dimensional problems

We suppose here that D = Ω × R with Ω ⊂ R2 and that the displacement field is out of
the plane Oxy, i.e.
ux = 0, uy = 0, uz = u(x, y).

In this case the stress field is

σxx = σxy = σyy = σzz = 0, σxz = µ∂x u, σyz = µ∂y u

and iwe denote by τ = (σxz , σyz ).

D
Gc
W
x

Gd

Figure 9: Schematic representation of the elastic anti-plane domain D = Ω × R.

16
2.2.1 Boundary value problem

Let ∂Ω = Γd ∪ Γc and let consider the boundary value problem: find u : Ω → R (the
anti-plane displacement) solution of

τ (u) = µ∇u, in Ω (Hooke’s elastic law)

−div(τ (u)) = ρb, in Ω (equilibrium)

u=U on Γd , τ (u) · n = f on Γc , (boundary conditions)

where ρb, f and U are given.

Variational problem

Let

U ad = {v : Ω → R ; v = U on Γd }, V = {v : Ω → R ; v = 0 on Γd }

be the set of admissible functions (displacements) and its associated vectorial space.
The variational formulation of the boundary value problem : find u solution of

(V P ) u ∈ U ad , a(u, v) − l(v) = 0, for all v ∈ V,

where a and l are the bilinear and the linear forms given by
Z Z Z
a(v, w) = µ∇v · ∇w dx, l(w) = ρbw dx + f w da.
Ω Ω Γc

If we denote by

1 1
Z Z Z
2
J(v) = a(v, v) − l(v) = µ|∇v| dx − ρbv dx − f v da
2 2 Ω Ω Γc

the elastic energy functional then the variational problem (V P ) is equivalent with the
following minimum problem for the energy

(M P ) u ∈ U ad , J(u) ≤ J(v), for all v ∈ U ad .

17
 Existence and unicity of the solution

Let R = {r : Ω → R ; a(r, r) = 0} be the kernel of the bilinear form a, then the set
of anti-plane rigid motions:

R = {r : Ω → R ; r ≡ const}.

If we denote by R0 = R ∩ V then we have

Theorem. If Ω is smooth enough and the input data ρb, U, f are regular then we
have:

• a solution u of (VP) exists if and only if l(r) = 0 for all r ∈ R0

• if u is a solution of (VP) then u + r is a solution too for all r ∈ R0

• the solution u of (VP) is unique if and only if R0 = {0}.

Finite element approximation

Let Vh be the finite element space and let Uhad = Vh ∩ U ad , Vh = Vh ∩ V. Then the
Galerkin (internal) approximation of (P V ) is

(V Ph ) uh ∈ Uhad , a(uh , vh ) − l(vh ) = 0, for all vh ∈ Vh ,

and we have the following error estimation: there exists a constant C such that

||u − uh || ≤ C inf ||u − vh ||.

vh ∈Uhad

The L2 (D)-norm of absolute and relative errors are

sR sR
(u − uh )2 dx
D R D
(u − uh )2 dx
Eabs = , Erel = R
D
1 dx D
u2 dx

2.2.2 Problème modèle: déformation d’une pièce rectangulaire

On se propose d’étudier la déformation statique d’une pièce entre deux plaques parallèles
planes et rigides. La pièce, qui est rapportée à un système d’axes orthonormés Oxyz,
est infinie dans la direction verticale Ox3 et a une section rectangulaire Ω =]0, a[×]0, b[
dans le plan Oxy. Les faces x = 0, x = a sont en contact sans frottement avec deux
plaques parallèles rigides (déplacement normal nul et contraintes tangentielles nulles).

18
 x

3
a
B

B
z

0 b y

Figure 10: The anti-plane deformation of a square: schematic representation.

On impose à la face y = b un déplacement le (0, 0, B), où B(x) = cos( πx

a
) et la face y = 0
est fixée (le déplacement est nul). Les forces volumiques (le pesanteur) sont négligeables.
Le matériau est supposée élastique linéaire, homogène et isotrope.

1) Mise en équation - Il s’agit d’écrire les équations, leur interprétation physique, leurs
limites d’applicabilité et les conditions aux bords.

Find u : D → R3 (the displacement) solution of

1
ε(u) = (∇u + ∇T u) in D, (small perturbations assumption)
2

σ(u) = λtrace(ε(u))I + 2µε(u), in D (Hooke’s elastic law)

−divσ(u) = 0 in D, (equilibrium)

u = 0 on y = 0, u = (0, 0, B(x)) on y = b, un = 0, στ = 0 on x = 0, a.

2)Simplification du problème Il s’agit d’utiliser l’hypothèse ux = 0, uy = 0, uz =

uz (x, y), d’écrire le tenseur de déformation linéarisé, l’interpretation des ses composantes
et le tenseur de contraintes afin de déduire le problème simplifié.

19
 Find u : Ω → R (the out of plane displacement) solution of

div(µ∇u) = 0, in D

u=0 on y = 0; u = B(x) on y = b; µ∂x u = 0 on x = 0, a.

2’) Trouver l’expression analytique du déplacement.

πx πy πb
u(x, y) = cos( )sinh( )/sinh( )
a a a

3) Formulation variationnelle. Il s’agit te trouver la formulation variationnelle.

U ad = {v ∈ H 1 (Ω) ; v = B on y = b, v = 0 on y = 0},

V = {v ∈ H 1 (Ω) ; v = 0 on y = 0, b}

The variational formulation of the boundary value problem : find uz solution of

(V P ) u ∈ U ad , a(u, v) = 0, for all v ∈ V,

where a is the bilinear form given by

Z
a(v, w) = µ∇v · ∇w dx.
Ω

4) Existence et unicité. Il s’agit d’étudier l’éxistence et l’unicité de la solution.

R0 = R ∩ V = {0}: the solution exists and it is unique.

5) Approximation par la méthode des éléments finis. Utiliser un code FreeFem++

afin d’obtenir la solution approximative.

(V Ph ) uh ∈ Uhad , a(uh , vh ) = 0, for all vh ∈ Vh ,

5a) Mesh.

//DEFORMATION ANTI-PLANE
verbosity=1;
//CONSTRUCTION du DOMAINE

20
 real a=1,b=1.;
border gama4(t=0,a){x=t;y=0.; label=4;};
border gama3(t=0.,b){x=a;y=t; label=3;};
border gama2(t=a,0.){x=t;y=b; label=2;};
border gama1(t=b,0.){x=0.;y=t; label=1;};
//MAILLAGE
int n1=40,n2=40, n3=40,n4=40;// nombre des segments sur les frontières
mesh Th=buildmesh(gama1(n1)+gama2(n2)+gama3(n3)+gama4(n4)); // construc-
tion
plot(Th,ps="mesh.eps", wait=1);

5a) Finite element spaces.

//ESPACE ELEMENTS FINIS

fespace Vh(Th,P2); // espace elements finis pour les déplacements
fespace Wh(Th,P2); // espace elements finis pour contraintes
Vh uh, vh;
Wh sxz,syz, dev;

5b) Input data.

//CONSTANTES DE MATERIAU
real mu=1;
//CONDITIONS aux limites en Déplacement
func real B(real tx) return cos(pi*tx/a);

5c) Find the approximative solution.

problem antiplane(uh,vh)= // variational formulation of "antiplane"

int2d(Th)(mu*(dx(uh)*dx(vh)+dy(uh)*dy(vh))) // bilinear form a
+on(1,uh=0) +on(3,uh=cos(pi*x/a)); // essential boundary conditions
antiplane; // solve the variational problem "antiplane"
sxz=mu*dx(uz); syz=mu*dy(uz); dev=sqrt(sxz*sxz+syz*syz); // compute the
stresses

5d) Visualization of the result.

Visualisation FreeFem++

plot(Th,uh, fill=true, value=1, ps="u.eps", wait=1);

21
 plot(Th,dev, fill=true, value=1, ps="Deviator.eps", wait=1);

Visualisation ParaView

fespace Rh(Th,P1dc) ;
Rh Xdes=x, Ydes=y, Zdes=uh, devdes=dev;
int id;
ofstream par("piece.vtk"); //nom du fichier .vtk
par <<"# vtk DataFile Version 1.0" <<endl;
par <<"3D Unstructured Grid of Linear Triangles"<<endl; //triangle à
trois noeuds
par <<"ASCII" «endl;
par <<" " <<endl;
par<<"DATASET UNSTRUCTURED_GRID" «endl;
par «"POINTS " « Rh.ndof« " float" «endl; // Rh.ndof number of degree
of freedom
for (id=0;id<Rh.ndof;id++){par<<Xdes[][id]<<" "<< Ydes[][id]<<" "<<Zdes[][id]<<en
}
par<<"CELLS " << Rh.nt<< " " << (Rh.ndofK+1)*Rh.nt<<endl; //
Rh.nt number of triangles
for (id=0;id<Rh.nt;id++) {par<<"3 "<< Rh(id,0) <<" "<<Rh(id,1)<<"
"<<<Rh(id,2)<<endl; }
par <<" " <<endl;
par <<"CELL_TYPES "<< Rh.nt«endl;
for (id=0;id<Rh.nt;id++){pa<<"5"<<endl; }
par <<" " <<endl;
par <<"POINT_DATA " << Rh.ndof«endl;
par <<"SCALARS Deviator float"<<endl;
par <<"LOOKUP_TABLE default " <<endl;
for (id=0;id<Rh.ndof;id++) {par<< devdes[][id]<<endl; }

6) Comparer la solution approchée obtenue avec le code FreeFem++ avec la solution

exacte. Calculer l’erreur d’une disrétisation P1 et P2 pour différents maillages. Voir le
rapport avec le temps de calcul.

func u=cos(pi*x/a)*sinh(pi*y/a)/sinh(pi*b/a);
Vh error=u-uh;
plot(Th,error, fill=true, value=1, ps="Error.eps", wait=1);

22
Figure 11: The anti-plane deformation of a square: visualization FreeFe++ (left) and
ParaView (right) of the numerical solution.

cout<<"L2 norm of the absolute error ="

<<sqrt(int2d(Th)(error*error)/int2d(Th)(1)<<endl;

2.2.3 Cisaillement anti-plan d’un carré fissuré (mode III)

On se propose d’étudier la déformation d’une pièce élastique, homogène et isotrope. La

pièce, qui est rapportée à un système d’axes orthonormés Oxyz, est infinie dans la
direction Oz et a une section Ω dans le plan Oxy (voir figure). Ω est un carré de largeur
2a = 4 avec une fissure, située sur l’axe Oy et de longueur 2b = 2. On impose aux faces
2 et 4 les contraintes F ez et −F ez respectivement avec F = θ. Les autres faces (1,3,5 et
6 sur le dessin) sont libres de contraintes et on néglige les forces volumiques.

uzh=uzh-uzh(0,b); // choice of the solution

23
 y
3
a
F
-F
b
z F
4 6 5 2
x
-F -a
a F

F
-b
-F

-F
1

Figure 12: The anti-plane deformation of a cracked square: schematic representation.

Figure 13: Computed anti-plane deformation of a cracked square. In color scale the stress
deviator.

24
 σxz(x,y)

uz(0+,y) σxz(0+,y)

σyz(0+,y)

σyz(x,y)

Figure 14: Anti-plane cracked square. Left: computed stress distributions (x, y) →
σxz (x, y) and (x, y) → σyz (x, y). Right: computed stress distributions y → σxz (0+, y)
and y → σyz (0+, y) and displacement distribution y → uz (0+, y).

2.2.4 Déformation (anti-plane) d’une demi-disque

On se propose d’étudier la déformation d’une pièce élastique, homogène et isotrope (λ =

µ = 1), est rapportée à un système d’axes orthonormés Oxyz. Elle est infinie dans la
direction Oz et sa section Ω = {r ∈ [0, R], θ ∈ [π, 2π]} dans le plan Oxy est un demi
disque de rayon R. Sur la surface du disque (r = R, notée avec 1 sur la figure) la pièce
este fixée (déplacement nul). Sur une partie de la face y = 0 ( x ∈]−a, a[, notée avec 3 sur
le dessin) on impose la force θF ez (avec F = 1) et l’autre partie (x ∈] − R, −a[ ∪ ]a, R[,
notée avec 2 et 4 sur le dessin) est libre de contraintes (vecteur contrainte nul). Les forces
volumiques (le pesenteur) sont négligeables.

25
 y

F
F F
4 2
-R -a a R
3 x

Figure 15: The anti-plane deformation of a half disque: schematic representation.

Figure 16: Computed anti-plane deformation of a half disque. In color scale the stress
deviator.

2.2.5 Déformation anti-plane d’un écrou épais

On se propose d’étudier la déformation d’une pièce élastique homogène et isotrope (λ =

µ = 1). La pièce, qui est rapporté à un système d’axes orthonormés Oxyz, est infini dans
la direction Oz et a une section Ω hexagonale de largeur 4a = 8 avec un trou dans le
plan Oxy (voir figure) de rayon R = a. La pièce est supposée homogène et isotrope. Sur
la face 3 sur la figure, la contrainte est (0, 0, θF ). On impose sur la frontière du trou, (7
sur le dessin) un déplacement nul. Les autres faces (1,2,4,5 et 6 sur le dessin) sont libres
de contraintes et on néglige les forces volumiques (le pesenteur).

26
 y
F
F
2a 4 F
F

5 F
3

x
-2a 2a

6 7 2

1
-2a

Figure 17: The anti-plane deformation of a thick nut: schematic representation.

Figure 18: Computed anti-plane deformation of a thick nut. In color scale the stress
deviator.

27
2.3 Vectorial two dimensional problems
2.3.1 Assumptions in two spatial dimensions

We will suppose that the elastic coefficients λ, µ depends only on the first two variables,
i.e.
λ = λ(x, y), µ = µ(x, y).

Plain-strain assumption

Let D = Ω × R and we will suppose that the displacements

(plain-strain) ux = ux (x, y), ux = ux (x, y), uz = 0,

and from the Hooke law we deduce

σxx = σxx (x, y), σxy = σxy (x, y), σyy = σyy (x, y), σxz = σyz = 0.

Since the stress σzz can be computed from the in-plane stresses or strains

λ
σzz = (σxx + σyy ) = λ(xx + yy ),
2(λ + µ)

we deduce that the plain-strain assumption can be easily handled by considering only the
unknowns acting in the plane Oxy (i.e. in-plane displacements, strains and stresses) which
depends only on the in-plane, variables x and y. Indeed, if we denote by u2 = (ux , uy )
the in-plane displacement and and by σ2 the in-plane stresses the

1
ε2 (u2 ) = (∇2 u2 + ∇T2 u2 ) in Ω, (small perturbations assumption)
2

σ2 (u2 ) = λtrace(ε2 (u2 ))I2 + 2µε2 (u2 ), in Ω (Hooke’s elastic law)

−div2 σ(u2 ) = ρb2 in Ω, (equilibrium)

Since these last equations could also be recovered by considering d = 2 in the formulation
of the boundary value problem, we will drop off the subscript 2 in the above notations.

Plane-stress assumption

For the plain-stress we suppose that D = Ω × (−h, h) and

(plain-stress) σxx = σxx (x, y), σxy = σxy (x, y), σyy = σyy (x, y), σxz = σyz = σzz = 0.

28
From the Hooke law we deduce

xx = xx (x, y), xy = xy (x, y), yy = yy (x, y), xz = yz = 0,

while for the zz we have

λ ν
zz = − (xx + yy ) = − (σxx + σyy ).
(λ + 2µ) E

The plain-stress assumption can be easily handled by considering only the unknowns
acting in the plane Oxy (i.e. in-plane displacements, strains and stresses) which depends
only on the in-plane, variables x and y. Indeed, if we denote by u2 = (ux , uy ) the in-plane
displacement and and by σ2 the in-plane stresses the

1
ε2 (u2 ) = (∇2 u2 + ∇T2 u2 ) in Ω, (small perturbations assumption)
2

σ2 (u2 ) = λ0 trace(ε2 (u2 ))I2 + 2µε2 (u2 ), in Ω (Hooke’s elastic law)

−div2 σ(u2 ) = ρb2 in Ω, (equilibrium),

where we have denoted by

2λµ
λ0 =
λ + 2µ
These last equations could also be recovered by considering d = 2 in the formulation of
the boundary value problem and replacing λ with λ0 . For simplicity we will drop off the
subscript 2 in the above notations.

2.3.2 Model problem: Ecrasement d’un lopin

On se propose d’étudier la déformation d’une pièce élastique (lopin) entre deux presses
parallèles planes et rigide. Le lopin, qui est rapporté à un système d’axes orthonormés
Oxyz, est infini dans la direction Oz et a une section rectangulaire ] − a, a[×] − b, b[ dans
le plan Oxy. La pièce est supposée homogène et isotrope. Les faces x = ± a sont libres
de contraintes. On impose à la face y = b (resp. y = −b) un déplacement (0, −U, 0)
(resp. (0, U, 0)) où U est donné. On néglige les forces volumiques (le pesenteur).

1) Mise en équation - Il s’agit d’écrire les équations, leur interprétation physique, leurs
limites d’applicabilité et les conditions aux bords.

Let D = Ω × R ⊂ R3 be the domain occupied by the elastic solid. We consider the

29
 y

3
-U -U b
-U -U -U -U

4 2
-a a x

U U U U U
U -b
1

Figure 19: The in-plane deformation of a rectangle: schematic representation.

following boundary value problem: find u : D → R3 (the displacement) solution of

1
ε(u) = (∇u + ∇T u) in D, (small perturbations assumption)
2

σ(u) = λtrace(ε(u))I + 2µε(u), in D (Hooke’s elastic law)

−divσ(u) = 0 in D, (equilibrium)

u = (0, U, 0) on 1, u = (0, −U, 0) on 3, σn = 0 on 2 and 4.

2) Simplification du problème -Il s’agit d’utiliser l’hypothèse des déformations planes

(ux = ux (x, y), uy = uy (x, y), uz ≡ 0) et d’écrire le problème simplifié.

We denote by u = u(x, y) = (ux (x, y), uy (x, y) the in-plane displacement and by σ
the in-plane 2 × 2 stress tensor .

1
ε(u) = (∇u + ∇T u) in Ω, (small perturbations assumption)
2

σ(u) = λtrace(ε(u))I + 2µε(u), in Ω (Hooke’s elastic law)

−divσ(u) = 0 in Ω, (equilibrium)

u = (0, U ) on 1, u = (0, −U ) on 3, σn = 0 on 2 and 4.

30
 3) Formulation variationnelle. Il s’agit te trouver la formulation variationnelle en
déformations planes.

Let
U ad = {v ∈ H 1 (Ω)2 ; v = (0, U ) on 1, v = (0, −U ) on 3}

the set of admissible in-plane displacements and

V = {v ∈ H 1 (Ω)2 ; v = 0 on 1, 3}

the associated vectorial space.

The variational formulation of the boundary value problem : find u solution of

(V P ) u ∈ U ad , a(u, v) = 0, for all v ∈ V,

where a is the bilinear and the linear forms given by

Z
a(v, w) = λtrace(ε(v))trace(ε(w)) + 2µε(v) : ε(w) dx.
D

4) Existence et unicité. Il s’agit d’étudier l’éxistence et l’unicité de la solution.

Since R0 = R ∩ V = {0} then there exists a solution of the variational problem and
it is unique.

5) Approximation par la méthode des éléments finis. Utiliser un code FreeFem++

afin d’obtenir la solution approximative.

(V Ph ) uh ∈ U ad
h , a(uh , vh ) = 0, for all vh ∈ V h ,

5a) Mesh.

//DEFORMATION PLANE
verbosity=1;
//CONSTRUCTION du DOMAINE
real a=1,b=1.;
border gam1(t=-a,a){x=t;y=-b; label =1;};
border gam2(t=-b,b){x=a;y=t; label =2;};
border gam3(t=a,-a){x=t;y=b; label =3;};
border gam4(t=b,-b){x=-a;y=t; label =4;};

31
 //MAILLAGE
int nh=40;// nombre des segments sur les frontières
mesh Th=buildmesh(gama1(nh)+gama2(nh)+gama3(nh)+gama4(nh)); // construc-
tion
plot(Th,ps="mesh.eps", wait=1);

5a) Finite element spaces.

//ESPACE ELEMENTS FINIS

fespace Vh(Th,P2); // espace elements finis pour les déplacements
fespace Wh(Th,P2); // espace elements finis pour contraintes
Vh ux, uy,vx, vy;
Wh sxy, sxx, syy, szz, pres,dev ;

5b) Input data.

//CONSTANTES DE MATERIAU
real lambda=1., mu=1.;
//CONDITIONS aux limites en Déplacement
real U=0.5;

5c) Find the approximative solution.

problem plane([ux,uy],[vx,vy])= // variational formulation of "plane"

int2d(Th)(lambda*(dx(ux)+dy(uy))*(dx(vx)+dy(vy)) +
2*mu*(dx(ux)*dx(vx)+dy(uy)*dy(vy)+0.5*(dx(uy)+dy(ux))*(dx(vy)+dy(vx)))) //
bilinear form a
+on(1,uy=U,ux=0) + on(3, ux=0.,uy=-U) ; // essential boundary conditions
plane; // solve the variational problem "antiplane"

sxx=lambda*(dx(ux)+dy(uy))+2*mu*dx(ux);
syy=lambda*(dx(ux)+dy(uy))+2*mu*dy(uy);
sxy=mu*(dx(uy)+dy(ux));
szz=lambda*(dx(ux)+dy(uy));
pres=-(sxx+syy+szz)/3.;
dev=sqrt(2*sxy∧2+(sxx +pres)∧2+(syy +pres)∧2+(szz +pres)∧2); // compute the
stresses

5d) Visualization of the result.

Visualisation FreeFem++

32
 plot(Th,[ux,uy], fill=true, value=1, ps="Displacement.eps", wait=1);
plot(Th,dev, fill=true, value=1, ps="Deviator.eps", wait=1);

Visualisation ParaView

fespace Rh(Th,P1dc) ;
real mag=1; // the magnification coefficient
Rh Xdes=x+mag*ux, Ydes=y+mag*uy, Zdes=0, devdes=dev;
int id;
ofstream par("Lopin.vtk"); //nom du fichier .vtk
par <<"# vtk DataFile Version 1.0" <<endl;
par <<"3D Unstructured Grid of Linear Triangles"<<endl; //triangle à
trois noeuds
par <<"ASCII" «endl;
par <<" " <<endl;
par<<"DATASET UNSTRUCTURED_GRID" «endl;
par «"POINTS " « Rh.ndof« " float" «endl; // Rh.ndof number of degree
of freedom
for (id=0;id<Rh.ndof;id++){par<<Xdes[][id]<<" "<< Ydes[][id]<<" "<<Zdes[][id]<<en
}
par<<"CELLS " << Rh.nt<< " " << (Rh.ndofK+1)*Rh.nt<<endl; //
Rh.nt number of triangles
for (id=0;id<Rh.nt;id++) {par<<"3 "<< Rh(id,0) <<" "<<Rh(id,1)<<"
"<<<Rh(id,2)<<endl; }
par <<" " <<endl;
par <<"CELL_TYPES "<< Rh.nt«endl;
for (id=0;id<Rh.nt;id++){pa<<"5"<<endl; }
par <<" " <<endl;
par <<"POINT_DATA " << Rh.ndof«endl;
par <<"SCALARS Deviator float"<<endl;
par <<"LOOKUP_TABLE default " <<endl;
for (id=0;id<Rh.ndof;id++) {par<< devdes[][id]<<endl; }

33
Figure 20: The plane deformation of a square: visualization FreeFe++ (left) and Par-
aView (right) of the numerical solution.

2.3.3 Piéce suspendue

-a a x

−ρg

-b
z

On se propose d’étudier la déformation d’une pièce élastique sous son propre poids.
La pièce, qui est supposée homogène et isotrope, est rapportée à un système d’axes
orthonormés Oxyz. Elle est infinie dans la direction Oz et a une section rectangulaire
] − a, a[×] − b, b[ dans le plan Oxy. Les faces x = a, y = ±b sont libres de contraintes
et la face x = −a est fixée (déplacement nul). Les forces volumiques (le pesenteur) sont
(0, −ρg, 0).

34
Figure 21: Computed plane deformation of a heavy square. In color scale the stress
deviator.

2.3.4 Traction d’un carré fissuré (mode II)

y
3
-F F
a

-F b F
z
4 6 5 2
x
-a
a
-F F

-b
-F F

On se propose d’étudier la déformation d’une pièce éastique, homogène et isotrope.

La pièce, qui est rapportée à un système d’axes orthonormés Oxyz, est infinie dans la
direction Oz et a une section Ω dans le plan Oxy (voir figure). Ω est un carré de largeur
2a = 4 avec une fissure, situé sur l’axe Oy et de longueur 2b = 2. On impose aux faces 2
et 4 les contraintes F0 ex et −F0 ex respectivement avec F0 = 1. Les autres faces (1,3,5 et 6
sur le dessin) sont libres de contraintes et on néglige les forces volumiques (la pesanteur).
real omega=-0.5*(ux(0,b)-ux(0,-b))/b;
real ax=0.5*(ux(0,b)+ux(0,-b)), ay=uy(a,0)-omega*a;

35
ux=ux-(ax-omega*y); uy=uy-(ay+omega*x); // solution choice

Figure 22: Computed plane deformation of a cracked square.

36
2.3.5 Different treatment of displacement boundary conditions

Let
ad
Ũ = {v : D → R2 continuous ; v = U on Γd }

the set of admissible in-plane displacements without the displacement boundary condi-
tions on Γ1m and Γ2m and let

Ṽ = {v : D → R continuous ; v = 0 on Γd }

the associated vectorial space.

Introducing displacement boundary conditions by penalization

If we denote by b and by m the bilinear and linear forms associated to the boundary
conditions
Z Z Z Z
b(v, w) = un vn da + uτ vτ da, m(w) = Un vn da + Uτ vτ
Γ1m Γ2m Γ1m Γ2m

then for each small parameter δ << 1 we consider

1 ˜lδ = l(w) + 1 m(w).

ãδ (v, w) = a(v, w) + b(v, w),
δ δ

The variational formulation of the penalized boundary value problem: find uδ solution of

ad
(V P )δ uδ ∈ Ũ , ãδ (uδ , v) − ˜lδ (v) = 0, for all v ∈ Ṽ,

and we have the following convergence result

uδ −→ u, for δ → 0.

real eps=0.000000000000001, Un=1, Ut=1;

// variational formulation of "planePenalization"
problem planePenalization(ux,uy,vx,vy)=
// bilinear form ãδ
int2d(Th)(lambda*(dx(ux)+dy(uy))*(dx(vx)+dy(vy))
+2*mu*(dx(ux)*dx(vx)+dy(uy)*dy(vy)
+0.5*(dx(uy)+dy(ux))*(dx(vy)+dy(vx))))
+int1d(Th,3)( 1/eps*(ux*N.x+uy*N.y)*(vx*N.x+vy*N.y))

37
 +int1d(Th,4)( 1/eps*(-ux*N.y+uy*N.x)*(-vx*N.y+vy*N.x))
// linear form ˜lδ
-int2d(Th)(bx*vx+by*vy)-int1d(Th,2)(Fx*vx+Fy*vy)
+int1d(Th,3)( -1/eps*Un*(vx*N.x+vy*N.y))
+int1d(Th,4)( -1/eps*Ut*(-vx*N.y+vy*N.x))
// essential boundary conditions
+on(1,ux=Uxd, uy=Uyd) ;

Introducing displacement boundary conditions by Lagrange multipliers

Let L = {q : Γ1m ∪ Γ2m → R continuous } be the set of Lagrange multipliers and
consider the bilinear and linear forms
Z Z Z Z
L(q, v) = qvn da + qvτ da, G(q) = qUn dl + qUτ dl,
Γ1m Γ2m Γ1m Γ2m

ãL ([v, p], [w, q]) = a(v, w) + L(v, q) − L(w, p), ˜lL ([w, q]) = l(w) + G(q).

The Lagrange variational formulation of the boundary value problem: find [u, p] solution
of

ad
(V P )L [u, p] ∈ Ũ × L, ãL ([u, p], [v, q]) − ˜lL ([v, q]) = 0, for all [v, q] ∈ Ṽ × L,

and we have the following interpretation of the multiplier p

p = σn on Γ1m , p = στ on Γ2m .

real eps=0.000000000000001, Un=1, Ut=1;

Vh p,q;
// variational formulation of "planeLagrange"
problem planeLagrange(ux,uy,p,vx,vy,q)=
// bilinear form ãL
int2d(Th)(lambda*(dx(ux)+dy(uy))*(dx(vx)+dy(vy))
+2*mu*(dx(ux)*dx(vx)+dy(uy)*dy(vy)
+0.5*(dx(uy)+dy(ux))*(dx(vy)+dy(vx))))
+ int2d(Th)(eps*p*q)
+int1d(Th,3)( q*(ux*N.x+uy*N.y)) -int1d(Th,3)( p*(vx*N.x+vy*N.y))
+int1d(Th,4)( q*(-ux*N.y+uy*N.x)) -int1d(Th,4)( p*(-vx*N.y+vy*N.x))

38
 // linear form ˜lL
-int2d(Th)(bx*vx+by*vy)-int1d(Th,2)(Fx*vx+Fy*vy)
-int1d(Th,3)( q*Un) -int1d(Th,4)( q*Ut)
// essential boundary conditions
+on(1,ux=Uxd, uy=Uyd) ;

2.3.6 Visser un écrou épais

2a

4
5

F 3
x
2a
U

6 7 2

z
1
2a

On se propose d’étudier la déformation d’une pièce élastique homogène et isotrope

entre deux plaques parallèles planes et rigides. La pièce, qui est rapporté à un système
d’axes orthonormés Oxyz, est infini dans la direction Oz et a une section hexagonale de
largeur 4a avec un trou dans le plan Oxy (voir figure). La pièce est supposée homogène
et isotrope. Sur la face 5 sur la figure le déplacement normal est nul et la contrainte
tangentielle est F . On impose sur la la frontière du trou, qui a un rayon de a, (7 sur le
w
dessin) le déplacement U = −weθ = − p (−y, x). Les autres faces (1,2,3,4 et 6
x2 + y 2
sur le dessin) sont libres de contraintes et on néglige les forces volumiques (le pesanteur).

39
Figure 23: Computed plane deformation of a thick nut. In color scale the stress deviator.

2.3.7 Poinçonnement d’un demi disque

4 -a 3 a 2
-R R x
-U -U -U

f f
z
1

On se propose d’étudier la déformation d’une pièce élastique, homogène et isotrope,

est rapportée à un système d’axes orthonormés Oxyz. Elle est infinie dans la direction
Oz et sa section Ω = {r ∈ [0, R], θ ∈ [π, 2π]} dans le plan Oxy est un demi disque de
rayon R. Sur la surface du disque (r = R, notée avec 1 sur la figure) la pièce est en
contact avec un corps rigide fixe ( déplacement normal nul) qui exerce sur elle un effort
tangentiel −f eθ . Sur une partie de la face y = 0 ( x ∈] − a, a[, notée avec 3 sur le dessin)
on impose le déplacement −U ey et l’autre partie (x ∈] − R, −a[ ∪ ]a, R[, notée avec 2 et

40
4 sur le dessin) est libre de contraintes (vecteur contrainte nul). Les forces volumiques (le
pesenteur) sont négligeables. Calculer la résultante des forces exercés sur la face y = 0.

Figure 24: Computed deformation of a half-disk : the initial mesh, the penalized solution
in wireframe and the Lagrangian solution with the stress deviator in color scale.

2.4 Axi-symmetric problems

Pour chaque des problèmes suivants il faut passer par les étapes suivantes :

1) Mise en équation - Il s’agit d’écrire les équations, leur interprétation physique, leurs
limites d’applicabilité et les conditions aux bords.
2)Simplification du problème - Il s’agit d’utiliser l’hypothèse ur = ur (r, z), uθ =
0, uz = uz (r, z) et d’écrire le problème simplifié.
3) Formulation variationnelle. Montrer que (ur , uz ) : Ω → R+ × R est solution
d’un problème variationnel. Trouver la fonctionnelle d’energie et étudier l’unicité de la
solution.
4) Approximation par la méthode des éléments finis. Il s’agit d’utiliser la méthode de
Galerkin et un espace de dimension finie construit avec la méthode des éléments finis P2.
Utilisation d’un code FreeFem++ afin d’obtenir la solution approximative.
5) Présentation des résultats numériques - Il s’agit d’une présentation graphique des
déplacements, des déformations, des contraintes et du module de leur déviateur.

41
2.4.1 Ecrasement d’un tube

3
H

pe pi
kH kH kH
pe
pi 4 2
pe pi
pe Ω
pi
pe

0 r
1
Ri
Re

On se propose d’étudier la déformation statique d’un tube entre deux plaques par-
allèles planes et rigides. La pièce, qui est rapportée à un système d’axes orthonormés
Oxyz, est definie dans les coordonnées cylindriques par {(r, z) ∈ Ω; θ ∈ [0, 2π[}, avec
Ω =]Ri , Re [×]0, H[ ooù Ri , Re sont les rayons intérieur et extérieur du cylindre et H est
sa hauteur. Les faces z = 0 et z = H sont en contact sans frottement (contraintes tan-
gentielles nulles) avec deux plaques parallèles rigides (pas de variation de déplacement
normal entre la pièce et les plaques). La plaque z = 0 est fixée et la plaque z = H a un
déplacement −hkez . Sur les parois intérieur (r = Ri ) et extérieur (r = Re ) s’exercent les
pressions pi et pe . Les forces volumiques (le pesanteur) sont négligeables et le matériau
est supposée élastique linéaire, homogène et isotrope.

i) Trouver les expressions analytiques du déplacement en supposant que ur = ur (r)

et uz = uz (z). Comparer avec la solution approchée obtenue avec le code FreeFem++.
ii) Trover la "formule du chaudronnier" pour δ = (Re − Ri )/Re  1 petit :

σrr ≈ 0, σθθ ≈ −(pe − pi )/δ, σzz ≈ Ek − ν(pe − pi )/δ.

iii) Supposons que pi = pe = 0. Calculer T > 0 le plus grand pour lequel pour
0 < k < T le comportement du materiau reste élastique suivant le critére de von Mises
(|σ D | < τ0 ).

42
2.4.2 Compression triaxiale d’un cylindre creux

F F
F H
F 3
4
p
2
p
p Rs r
0
p
p 5 Ω
p
F 6
F -H 1 F R
c
F

On se propose d’étudier la déformation statique d’une d’un cylindre circulaire creux

entre deux planes. La pièce est definie dans les coordonnées cylindriques par {(r, z) ∈
Ω; θ ∈ [0, 2π[}, avec Ω = {(r, z) ; z ∈] − H, H[, r ∈ [0, Rc [, r2 + z 2 > Rs2 } ooù Rc , Rs
sont les rayons du cylindre et de la sphére et 2H est l’hauteur du cylindre. Sur les
faces z = −H et z = H on impose les contraintes −F ez et F ez respectivement, avec
F = tF0 . Sur le paroi extérieur (r = Rc ) on exerce une contrainte −per , avec p = tp0 et
le trou r2 + z 2 = Rs2 est libre de contraintes. Les forces volumiques (le pesanteur) sont
négligeables et le matériau est supposée élastique linéaire, homogène et isotrope.

Calculer T > 0 le plus grand pour lequel pour 0 < t < T le comportement du materiau
reste élastique suivant le critére de von Mises (|σ D | < τ0 ).
......................................................................

2.5 Three dimensional problems

..........................................................

43
3 Static problems in nonlinear elasticity

3.1 Finite elasticity mechanical model

3.1.1 Kinematics

D n
x

da
N
c
dA
X
Euler configuration

D0
Lagrange configuration

Figure 25: Lagrange and Euler configurations.

Let D0 ⊂ Rd (with d = 2, 3) be the solid non-deformed (Lagrangian) configuration

and let us denote by X ∈ D0 the Lagrangian coordinates. The deformed (Eulerian)
configuration will be denoted by D while x ∈ D is the Eulerian coordinates given by the
unknown χ : D0 → D, i.e.
x = χ(X).

The deformation gradient tensor is F = F (χ), its Jacobien will be denote by J = J(χ),
and the right and left Cauchy-Green tensors will be denoted by C, B while the (large)
deformation is denoted by E, i.e.

1
F (χ) = ∇X χ, J = det(F ), C = FTF, B = FFT, E(χ) = (F T F − I),
2

∂
where ∇X = ⊗ cK is the gradient operator with respect to the Lagrangian config-
∂XK
uration. The Nanson formulae reads

dx = F dX, dv = JdV, nda = JF −T N dA,

where we have denoted by dv, dV the volume, by da, dA the aria elements and by n, N

44
the unit exterior normals, in Eulerian and Lagrangian configurations respectively.

C (right Cauchy-Green)
Lagrange configuration E (deformation tensor) Lagrange configuration

F (d
efor
mat
ion
grad
ient
)
Euler configuration B (left Cauchy-Green) Euler configuration

Figure 26: Diagram of deformation tensors as linear applications from a configuration to

another one.

3.1.2 Conservation laws

Let denote by ρ : D → R+ , ρ0 : D0 → R+ the mass densities in Eulerian and Lagrangian

configurations. Then the mass conservation law reads

J(X)ρ(x) = ρ0 (X),

and the incompressibility condition is J(X) = 1.

Let σ : D → Rd×d be the Cauchy stress tensor which gives the loads (stress) σn on
the surfaces in Eulerian configuration and the momentum balance law (equilibrium
equations) reads
divx σ + ρb = 0, σ T = σ,
∂
where divx (·) = (·ck ) is the divergence in the Eulerian variables, and b are the body
∂xk
forces.
The first Piola-Kirchhoff (PK1) stress tensor is Π : D0 → Rd×d which is related to
the Cauchy stress tensor through

Π = JσF −T , σn da = ΠN dA,

and the equilibrium equations (momentum balance law) in Lagrange variables reads

divX Π + ρ0 b = 0,

45
 ∂
where divX (·) = (·cK ) is the divergence in the Lagrangian variables.
∂XK
Finally let S : D0 → Rd×d be the second Piola-Kirchhoff (PK2) stress tensor which is
symmetric (S = S T ) and is related to the other two stress tensors through

S = JF −1 σF −T = F −1 Π, Π = F S.

S (PK2)
Lagrange configuration Lagrange configuration

P (P K
1)

Euler configuration
s (Cauchy) Euler configuration

Figure 27: Diagram of stress tensors as linear applications from a configuration to another
one.

3.1.3 Constitutive elastic law

A material is said to be elastic if the the second Piola-Kirchhoff (PK2) stress tensor S
at any particle X is a function of the deformation tensor E associated to this particle,
i.e. there exists A : D0 × Rd×d → Rd×d such that

S(X) = A(X, E(X)).

If in addition to the elasticity assumption we suppose that the material does dot dissipate
energy during cyclic deformations we find the class of hyper-elastic materials for which
there exits an elastic (energy) potential W : D0 × Rd×d → R such that

∂W
A(X, E) = .
∂E

For isotropic hyper-elastic materials the energy depends only on three scalars IE =
1
trace(E), IIE = (|E|2 − IE2 ) and IIIE = det E, the three invariants of E, i.e.
2

W (X, E) = W (X, IE , IIE , IIIE ),

46
and after some computation we find that

A(E) = c0 (IE , IIE , IIIE )I + c1 (IE , IIE , IIIE )S + c2 (IE , IIE , IIIE )S 2 ,

where the three scalar functions c0 , c1 , c2 are given by

∂W ∂W ∂W ∂W ∂W ∂W
c0 = − IE − IIE , c1 = − IE , c2 = .
∂IE ∂IIE ∂IIIE ∂IIE ∂IIIE ∂IIIE

A first example of isotropic hyper-elastic material is the Saint-Venant-Kirchhoff

material for which

1
W (E) = (λ + 2µ)IE2 + 2µIIE , A(E) = λIE I + 2µE,
2

where λ and µ are the Lamé elastic coefficients.

Another example could be the Murnaghan material, given by

1 l + 2m 3
W (E) = (λ + 2µ)IE2 + 2µIIE + IE + 2mIE IIE + nIIIE ,
2 3

where l, m and n are other three elastic coefficients.

3.2 Approximation of the static nonlinear elastic problems

3.2.1 Lagrangian formulation of the boundary value problem

Let ∂D0 = Γd ∪ Γc and let consider the boundary value problem: find χ : D0 → Rd (the
Eulerian position) solution of

1
E(χ) = (F (χ)T F (χ) − I) in D0 , (nonlinear deformation operator)
2

Π(χ) = F (χ)A(E(χ)) in D0 (elastic law)

divX Π(χ) + ρ0 b = 0, in D0 , (equilibrium)

χ = χ0 on Γd , Π(χ)N = f on Γc , (boundary conditions),

where N is the exterior unit normal on ∂D0 . The functions b (body forces), ρ0 (the mass
density), χ0 (imposed positions), and f (imposed stresses, dead loads) are given.

47
3.2.2 Variational formulation

Let
U ad = {v : D → Rd continuous ; v = χ0 on Γd }

be the set of admissible displacements and let

V = {v : D → Rd continuous ; v = 0 on Γd }

be the associated vectorial space.

The variational formulation of the boundary value problem : find χ solution of

(V P ) χ ∈ U ad , F(χ)(v) − l(v) = 0, for all v ∈ V,

where F(χ) and l are linear forms with respect to v given by

Z Z Z
T
F(χ)(v) = A(E(χ)) : (F (χ)∇X v) dV, l(v) = ρ0 b · v dV + f · v dA.
D0 D0 Γc

To formulate the finite element approximation of the above problem let Vh be the finite
element space and let U ad ad
h = Vh ∩ U , V h = Vh ∩ V. Then the Galerkin (internal)
approximation of (P V ) is

(V P )h χh ∈ Uhad , F(χh )(vh ) − l(vh ) = 0, for all vh ∈ V h .

For the simplicity we will use in the next the notations χ, v instead of χh , vh .

3.2.3 Newton-Raphson algorithm

The Newton-Raphson algorithm consists in computing an iterative sequence (χn )n≥0

which approaches the solution χ of the nonlinear problem (V P )h , i.e. χn → χ when
n → +∞.
1) Initialization. Let χ0 ∈ U ad
h be the initial guess.
2) Iterative loop n → n + 1 has two steps:
2a) Solving a linear system: find the increment un solution of

un ∈ V h , aχn (un , v) − lχn (v) = 0, for all vh ∈ V h ,

48
where the bilinear and linear forms aχ , lχ are given by

aχ (w, v) = F 0 (χ)(w, v), lχ (v) = l(v) − F(χ)(v).

2b) Update the solution from

χn+1 = χn + un .

3) Algorithm stops when the relative error en = ||un || is sufficiently small.

The derivative aχ (w, v) = F 0 (χ)(w, v) can be found from direct calculation to be

Z Z
aχ (w, v) = A(E(χ)) : (∇TX w∇X v) dV + A0 (E(χ))E 0 (χ)(w) : ∇X v dV,
D0 D0

d
where A0 (E)A = A(E + tA)|t=0 is the derivative of A(E) with respect to E while
dt
E (χ)(w) is the derivative of E(χ) with respect to χ is given by
0

1
E 0 (χ)(w) = (∇TX wF (χ) + F T (χ)∇X w).
2

3.2.4 Incremental loading and/or quasi-static evolution

Let us suppose in the following that boundary conditions and the external forces are
depending on parameter t. This parameter could be the time (quasi-static evolution) or
an artificial loading parameter. In the later case it is related to a computing strategy
which will provides a good initial to assure the convergence of the Newton algorithm.
We are looking for χ : [0, T ]×D0 → Rd (the Eulerian position at any "time" ) solution
of

1
E(χ(t)) = (F (χ(t))T F (χ(t)) − I) in D0 , (nonlinear deformation operator)
2

Π(χ(t)) = F (χ(t))A(E(χ(t))) in D0 (elastic law)

divX Π(χ(t)) + ρ0 b(t) = 0, in D0 , (equilibrium)

χ(t) = χ0 (t) on Γd , Π(χ(t))N = f (t) on Γc , (boundary conditions),

The functions b(t) (body forces) abd χ0 (t) (imposed positions), and f (t) (imposed
stresses, dead loads) are given at any "time" t.

49
 Let
U ad (t) = {v : D → Rd continuous ; v = χ0 (t) on Γd }

be the set of admissible displacements with its associated vectorial space V.

The variational formulation of the boundary value problem : find χ solution of

(V P )t χ(t) ∈ U ad (t), F(χ(t))(v) − l(t)(v) = 0, for all v ∈ V,

where the l(t) is the linear form given by

Z Z
l(t)(v) = ρ0 b(t) · v dV + f (t) · v dA.
D0 Γc

As before we can formulate the finite element approximation of the above problem by
considering U ad
h (t) = Vh ∩ U (t) at any time t. For the simplicity we will use in the next
ad

the notations χ(t), v instead of χh (t), vh .

If we take the derivative with respect to parameter t (denoted by an upper dot) in
the above variational equality we found

ad
(dV P )t χ̇(t) ∈ U̇ (t), ˙
F 0 (χ(t))(χ̇(t), v) − l(t)(v) = 0, for all v ∈ V,

where we have denoted by

ad
U̇ (t) = {v : D → Rd continuous ; v = χ̇0 (t) on Γd },
Z Z
˙
l(t)(v) = ρ0 ḃ(t) · v dV + f˙(t) · v dA.
D0 Γc

Euler discretization. Let us discretize the parameter interval [0, T ] into M subin-
tervals denoted by [tk , tk+1 ] with t0 = 0 and tM = T and let us denote by χk the
"approximation" of χ(tk ). Using the explicit Euler discretization

χk − χk−1 ˙ k )(v) ≈ lk (v) − lk−1 (v) , χ0 (tk ) − χ0 (tk−1 )

χ̇(tk ) ≈ , l(t χ̇0 (tk ) ≈
tk − tk−1 tk − tk−1 tk − tk−1

one can find an excellent guess χ0k+1 for the Newton method which will compute χk+1 ,
the solution at t = tk+1 , by solving the linear equation for χ̇(tk+1 ). To be more precisely,
let first denote by
Z Z
δlk (v) = ρ0 [b(tk ) − (b(tk−1 )] · v dV + [f (tk ) − f (tk−1 )] · v dA.
D0 Γc

50
 δU k = {v : D → Rd continuous ; v = χ0 (tk ) − χ0 (tk−1 ) on Γd },

Newton-Raphson algorithm with Euler prediction.The algorithm consists in

computing at each parameter tk an iterative sequence (χnk )n≥0 which approaches the
solution χk of the nonlinear problem (V P )tk , i.e. χnk → χk when n → +∞.
1) Parameter initialization. Let χ0 ∈ U ad (0) be a (simple) initial guess for the
initial problem at t = 0.
2) Parameter iterative loop k − 1 → k has four steps:
2a) Solve a linear system: find the "velocity" δχk solution of

δχk ∈ δU k , aχk−1 (δχk , v) = 0, for all v ∈ V,

2b) Predict the solution by given an initial guess for the Newton algorithm

χ0k = χk−1 + δχk .

2c) Iterative loop n → n + 1 (Newton corrector) has two steps

2c i) Solve the linear system: find the increment unk solution of

unk ∈ V, aχnk (unk , v) − lχnk (v) = 0, for all v ∈ V.

2c ii) Update the solution from

χn+1
k = χnk + unk .

2d) Newton Algorithm stops when the relative error enk = ||unk || is sufficiently
small, let say for n = nek .
3) Update the parameter solution

ne
χk = χk k

51
3.3 Two dimensional problems
In we focus here to the plane deformation χx = χx (X, Y ), χy = χy (X, Y ), χz = Z or
plane stress configurations.

3.3.1 Deformation of a diving board

3 b

-a a 2
x
4

1 -b

Figure 28: The in-plane deformation of a diving board: schematic representation.

We are modeling the in-plane large deformation of an elastic diving board. The diving
board, which occupies the domain D0 = Ω0 × R is infinite in the OZ direction and has a
rectangular shape Ω0 = (−a, a)×(−b, b) ⊂ R2 in the OXY plane, with a = 5 and b = 0.5.
The elastic material is homogeneous, isotropic and modeled by the Saint-Venant-Kirchoff
constitutive law with E = 1 and ν = 0.36. On the sides y = −b (label 1) and on
x = a (label 2) the loads a vanishing while on the side y = −b (label 3) the load is
f = (0, fy (x), 0), where fy (x) = 0 if x < 0.9a and fy (x) = −f if x > 0.9a, with f = 0.01.
On the side x = −a (label 4) the diving board is the clamped (vanishing displacements).
The body forces b are considered to be negligible.

1) Equation setting. Write the equations, their mechanical interpretation, and the
boundary conditions.

2) Simplifying the problem. Use the in-plane deformation assumptions χx = χx (X, Y ), χy =

χy (X, Y ), χz = Z to deduce the simplified equations.

1
E(χ) = (F (χ)T F (χ) − I) in Ω0 , (nonlinear deformation operator)
2

52
 Π(χ) = F (χ)(λIE(χ) I + 2µE(χ)) in Ω0 (elastic law)

EZZ = EXZ = EY Z = 0, SXZ = SY Z = 0, SZZ = λIE (χ)

divX Π(χ) + ρ0 b = 0, in Ω0 , (equilibrium)

χ = (X, Y ) on 4,

Π(χ)N = 0 on 1, 2, Π(χ)N = (0, fy (x), 0), on 3 (boundary conditions).

3) Variational formulation . Find the variational formulation of the simplified prob-

lem.

U ad = {v : D → R2 continuous ; v = (X, Y ) on 4},

V = {v : D → R2 continuous ; v = 0 on 4},
Z
l(v) = fy vy dA.
3

4) Finite element approximation. Use a FreeFem++ code to compute the solution.

////////////Macros for Finite elasticity

include"Macros2D.edp"
///////////Material Modeling
//////////Elastic coefficients
real E=1, nu=0.36, lambda=(E*nu)/((1+nu)*(1-2*nu)), mu=E/(2*(1+nu));
//////////Macros for Saint-Venant-Kirchhoff
include"SaintVenantKirchhoff2D.edp"
////////////Finite element spaces
fespace Vh(Th,P2);
Vh Xix,Xiy,ux,uy, vx,vy,Xixo,Xiyo;
fespace Wh(Th,P1dc);
Wh pres,dev;
//////////////////
//Vizualization Settings
fespace Rh(Th,P1dc) ;
Rh Xdes, Ydes, Zdes=0, devdes;
int id;
string NameFile; // the name of the file were we will write the results
include "MeshVisualization.edp"

53
 ///////////////////
//Newton parameters
int nN=0,NewtonMax=50;real tol=0.0000001;
/////////External force
real f=0.01;
/////////////Newton problem
problem Newton2D(ux,uy, vx,vy)=
int2d(Th)(TBL(Xix,Xiy,ux,uy, vx,vy))// bilinear term aχ (w, v) in macros
+int2d(Th)(TL(Xix,Xiy,vx,vy)) //linear term F(χ)(v) in macros
-int1d(Th,3)(-f*vy*(x>0.9*a)) // load term l(v)
+on(4,ux=0,uy=0); // boundary conditions included in V
//initialisation
Xix=x;Xiy=y;
///////Begin Newton Loop
for(nN=0;nN<NewtonMax;nN++)
//update
Xixo=Xix;Xiyo=Xiy;
////solve Newton
Newton2D;
////////////Compute the new configuration
Xix=Xixo+ux; Xiy=Xiyo+uy;
////////compute error
real error=sqrt( int2d(Th)(ux*ux+uy*uy)/
(int2d(Th)((Xix-x)*(Xix-x)+(Xiy-y)*(Xiy-y))+0.0000000000000000001));
cout «"iteration="«nN«" error="«error«endl;
///stop criterion
if(error<tol) break;
/////End Newton Loop
//////////////////////////////////

5) Presentation of numerical results. Use a graphic representation of the deformed

configuration with the mean stress and the stress deviator in color scale.

Visualization ParaView
//////////////////////////////////
////////////Final Visualization
pres=(S11(Xix,Xiy)+S22(Xix,Xiy)+S33(Xix,Xiy))/3; // mean stress

54
 dev=sqrt((S11(Xix,Xiy)-pres)*(S11(Xix,Xiy)-pres)+
(S22(Xix,Xiy)-pres)*S22(Xix,Xiy)-pres)+(S33(Xix,Xiy)-pres)*(S33(Xix,Xiy)-
pres)+2*S12(Xix,Xiy)*S12(Xix,Xiy)); /// stress deviator
NameFile="Final.ResultsBeam.vtk";
include "VisualizationPlane.edp"
////////////////////////////////////////////

6) Comparison with the linear approach. Compute the solution using the linear model
(small perturbations assumption and Hooke law) and compare it with the solution of the
nonlinear problem.

Figure 29: The plane deformation of a diving board: visualization ParaView of the mesh,
of the solution of the linear model (red) and of the nonlinear model with the deviator of
the PK2 stress tensor in color scale.

7) Force resultant. Compute the force resultant acting on boundary 4.

Z Z
Rx = (ΠxX NX + ΠxY NY ) dA, Ry = (ΠyX NX + ΠyY NY ) dA,
4 4

55
 real Rx=int1d(Th,4)(Pi11(Xix,Xiy)*N.x+Pi12(Xix,Xiy)*N.y);
real Ry=int1d(Th,4)(Pi21(Xix,Xiy)*N.x+Pi22(Xix,Xiy)*N.y);
cout«"Rx="«Rx«" Ry="«Ry«endl;

8) Incremental loading. Use the incremental loading to compute the deformation for
f = 0.02.

3.3.2 Floor with a nonuniform loading

3
b
-a a
x
4 2
-b
1

Figure 30: In-plane deformation of a floor: schematic representation.

We are modeling the in-plane large deformation of an elastic floor. The floor, which
occupies the domain D0 = Ω0 × R is infinite in the OZ direction and has a rectangular
shape Ω0 = (−a, a)×(−b, b) ∈ R2 in the OXY plane, with a = 10 and b = 0.5. The elastic
material is homogeneous, isotropic and modeled by the Saint-Venant-Kirchoff constitutive
law with E = 1 and ν = 0.36. On the side y = −b (label 1) the loads a vanishing while
on the side y = −b (label 3) the load is non-uniform f = (0, fy (X), 0), where fy (X) = tf ,
if |X| < a/5 (f = 0.1, t ∈ (0, 1)) and fy (X) = 0, if |X| > a/5. On the sides x = −a
and x = a (label 4 and 2) the floor is the clamped (vanishing displacements). The body
forces b are considered to be negligible.

56
Figure 31: The in-plane deformation of a floor with a nonuniform loading: visualization
ParaView of the mesh, of the solution of the nonlinear model with the deviator of the PK2
stress tensor in color scale for different values of the loading parameter t = 1/8, 3/8, 1.

3.3.3 Folding a semicircle

x
4 Ri 2 Re

Figure 32: The in-plane folding a semicircle: schematic representation.

We are modeling the in-plane folding of a semicircle. The semicircle, which occupies
the domain D0 = Ω0 × R is infinite in the OZ direction and has a shape of half circular
crown Ω0 = {r ∈ (Ri , Re ), θ ∈ (π, 2π)} ⊂ R2 in the OXY plane, with Ri = 1.8 and
Re = 2. The elastic material is homogeneous, isotropic and modeled by the Saint-Venant-
Kirchoff constitutive law with E = 1 and ν = 0.36. On the sides r = Ri (label 1) and
r = Re (label 3) the loads a vanishing . On the side θ = π (label 4) the semicircle is the

57
clamped (vanishing displacements) while the side θ = 0 (label 2) is gradually translated
in the −y direction by assuming that χ0 (t, X) = (X − 2tRi , 0), for t ∈ (0, 1). The body
forces b are considered to be negligible.

Figure 33: The in-plane deformation of a folding semicircle: visualization ParaView of

the mesh and of the nonlinear model with the deviator of the PK2 stress tensor in color
scale for different values of loading parameter t.

58
3.3.4 Squeezing an empty square

42 3

5
5

4 2 x

22
1

Figure 34: The in-plane squeezing of an empty square: schematic representation.

We are modeling the in-plane squeezing an empty square. The empty square, which
occupies the domain D0 = Ω0 × R is infinite in the OZ direction and has a shape of
square with a circular hole and without two corners Ω0 = (−a, a) × (−a, a) \ ({|(x, y)| <
R} ∪ {|(x, y) − (−a, a)| < r} ∪ {|(x, y) − (a, −a)| < r}) ⊂ R2 in the OXY plane, with
a = 5, r = 0.5 and R = 4. The elastic material is homogeneous, isotropic and modeled by
the Saint-Venant-Kirchoff constitutive law with E = 1 and ν = 0.36. On all sides (labels
1,2,3,4 and 5) the loads a vanishing . The two corners (label 22 and 42) are gradually
√ √
translated in the (− 2/2, 2/2) direction by assuming that for any t ∈ (0, 0.9R) the
√ √
position is given by χ0 (t, X, Y ) = (X − t 2/2, Y + t 2/2) on 22 and χ0 (t, X, Y ) =
√ √
(X + t 2/2, Y − t 2/2) on 42. The body forces b are considered to be negligible.

59
Figure 35: The in-plane squeezing of an empty square: visualization ParaView of the
mesh, a of the solution at different stages of the loading with the deviator of the PK2
stress tensor in color scale.

3.3.5 Wrapping a cylinder with a planar sheet

We are modeling the in-plane large deformation of an elastic planar sheet to wrap a
cylinder of radius R = 10. The planar sheet, which occupies the domain D0 = Ω0 × R
is infinite in the OZ direction and has a rectangular shape Ω0 = (−a, a) × (−b, b) ∈ R2
in the OXY plane, with a = πR and b = 0.4. The elastic material is homogeneous,
isotropic and modeled by the Saint-Venant-Kirchoff constitutive law with E = 1 and
ν = 0.36. On the side Y = b (label 3) the loads are vanishing while on the side Y = −b
(label 3) the boundary conditions depends on a loading increasing parameter t ∈ (0, 1).
For X < −a + D(t) the planar sheet is clamped (vanishing displacements) while for
X > −a + D(t) the loads are vanishing, where D(t) = 2a(1 − t). On the side X = −a
(label 4 ) the planar sheet is clamped (vanishing displacements) while on the side X = a
(label 2) the position

χ0 (t, X, Y ) = (R sin(2at/R) − a + D(t) − Y sin(2at/R), R(1 − cos(2at/R)) + Y cos(2at/R))

is imposed to wrap the cylinder. The body forces b are considered to be negligible.

60
 y

D(t)

-a b 3 a
x
4 -b 2
1
Figure 36: The in-plane wrapping a cylinder with a planar sheet: schematic representa-
tion.

Figure 37: The in-plane wrapping a cylinder with a planar sheet: visualization ParaView
of the mesh and of the nonlinear model with the deviator of the PK2 stress tensor in
color scale for different values of loading parameter t.

3.4 Axi-symmetric problems

......................................................................

3.5 Three dimensional problems

..........................................................

61
4 Appendix

4.1 Macros for Nonlinear elasticity

4.1.1 General Macros in 2D
//===========================
//NON-LINEAR ELASTICITY 2D
//=============================================
verbosity = 1;
//=============================================
// Macros for the gradient of a vector field ∇X u
macro grad11(u1,u2) (dx(u1)) //
macro grad12(u1,u2) (dy(u1)) //
macro grad21(u1,u2) (dx(u2)) //
macro grad22(u1,u2) (dy(u2)) //
// Macros for the deformation gradient F (u) = ∇X u
macro F11(u1,u2) ( grad11(u1,u2)) //
macro F12(u1,u2) ( grad12(u1,u2)) //
macro F21(u1,u2) ( grad21(u1,u2)) //
macro F22(u1,u2) ( grad22(u1,u2)) //
// Macro for the right Cauchy-Green strain tensor : C(u) = F (u)T F (u), CJK = FiJ FiK
macro C11(u1,u2)(F11(u1,u2)*F11(u1,u2) + F21(u1,u2)*F21(u1,u2))//
macro C12(u1,u2)(F11(u1,u2)*F12(u1,u2) + F21(u1,u2)*F22(u1,u2))//
macro C22(u1,u2)(F12(u1,u2)*F12(u1,u2) + F22(u1,u2)*F22(u1,u2))//
macro C21(u1,u2)(C12(u1,u2))//
// Macro for the trace(E(u)) //=======================
macro TrE(u1,u2) ((C11(u1,u2) + C22(u1,u2)-2.)/2.)//
////////////////////////////////////////// /////// F (u)T ∇v
macro FTNv11(u1,u2,v1,v2) (F11(u1,u2)*grad11(v1,v2)+F21(u1,u2)*grad21(v1,v2))//
macro FTNv12(u1,u2,v1,v2) (F11(u1,u2)*grad12(v1,v2)+F21(u1,u2)*grad22(v1,v2))//
//////////////////////////////////
macro FTNv21(u1,u2,v1,v2) (F12(u1,u2)*grad11(v1,v2)+F22(u1,u2)*grad21(v1,v2))//
macro FTNv22(u1,u2,v1,v2) (F12(u1,u2)*grad12(v1,v2)+F22(u1,u2)*grad22(v1,v2))//
///////////////////////////////////////////////////////////////////////////////////////////////
/////// F T (u)∇v
macro NuTNv11(u1,u2,v1,v2) (grad11(u1,u2)*grad11(v1,v2)+grad21(u1,u2)*grad21(v1,v2))//
macro NuTNv12(u1,u2,v1,v2) (grad11(u1,u2)*grad12(v1,v2)+grad21(u1,u2)*grad22(v1,v2))//
//////////////////////////////////
macro NuTNv21(u1,u2,v1,v2) (grad12(u1,u2)*grad11(v1,v2)+grad22(u1,u2)*grad21(v1,v2))//
macro NuTNv22(u1,u2,v1,v2) (grad12(u1,u2)*grad12(v1,v2)+grad22(u1,u2)*grad22(v1,v2))//
///////////////////////////////////////////////////////////////////////////////////////////////
///////////////// E 0 (u)(w) = 21 (∇TX wF (u) + F T (u)∇X w).

62
 macro Ed11(u1,u2,varu1,varu2) (0.5*(FTNv11(u1,u2,varu1,varu2)+FTNv11(u1,u2,varu1,varu2)))//
macro Ed12(u1,u2,varu1,varu2) (0.5*(FTNv12(u1,u2,varu1,varu2)+FTNv21(u1,u2,varu1,varu2)))//
///////////////////////////////////////
macro Ed21(u1,u2,varu1,varu2) (0.5*(FTNv21(u1,u2,varu1,varu2)+FTNv12(u1,u2,varu1,varu2)))//
macro Ed22(u1,u2,varu1,varu2) (0.5*(FTNv22(u1,u2,varu1,varu2)+FTNv22(u1,u2,varu1,varu2)))//
/////////////////////////////////////////////////////////////////////////////
/////////T race(E 0 (u)(w))///////////////
macro TrEd(u1,u2,varu1,varu2) (Ed11(u1,u2,varu1,varu2)+Ed22(u1,u2,varu1,varu2))//
////////////////////////////////////////////////////////////////////
// Macros for the linear form S(u) : (F T (u)∇X v)
//=====================================
macro TL(u1,u2,v1,v2)( S11(u1,u2)*FTNv11(u1,u2,v1,v2)+S12(u1,u2)*FTNv12(u1,u2,v1,v2)+
S21(u1,u2)*FTNv21(u1,u2,v1,v2)+S22(u1,u2)*FTNv22(u1,u2,v1,v2) )//
// Macros for the bilinear form
//========================================
// First term : S(u) : (∇TX w∇X v)
macro TBL1(u1,u2,varu1,varu2,v1,v2)(
S11(u1,u2)*NuTNv11(varu1,varu2,v1,v2) + S12(u1,u2)*NuTNv12(varu1,varu2,v1,v2) +
S21(u1,u2)*NuTNv21(varu1,varu2,v1,v2) + S22(u1,u2)*NuTNv22(varu1,varu2,v1,v2) )//
//=====================================================
// Second term : S 0 (u)E 0 (u)(w) : ∇X v
macro TBL2(u1,u2,varu1,varu2,v1,v2)(
SD11(u1,u2,varu1,varu2)*FTNv11(u1,u2,v1,v2)+SD12(u1,u2,varu1,varu2)*FTNv12(u1,u2,v1,v2)+
SD21(u1,u2,varu1,varu2)*FTNv21(u1,u2,v1,v2)+SD22(u1,u2,varu1,varu2)*FTNv22(u1,u2,v1,v2) )//
// Macros for the bilinear form
macro TBL(u1,u2,varu1,varu2,v1,v2) (TBL1(u1,u2,varu1,varu2,v1,v2)+TBL2(u1,u2,varu1,varu2,v1,v2))//
///////////////// Π(u) = F (u)S(u)
macro Pi11(u1,u2)(F11(u1,u2)*S11(u1,u2)+F12(u1,u2)*S21(u1,u2))//
macro Pi22(u1,u2)(F21(u1,u2)*S12(u1,u2)+F22(u1,u2)*S22(u1,u2))//
macro Pi12(u1,u2)(F11(u1,u2)*S12(u1,u2)+F12(u1,u2)*S22(u1,u2))//
macro Pi21(u1,u2)(F21(u1,u2)*S11(u1,u2)+F22(u1,u2)*S21(u1,u2))//
macro Pi33(u1,u2)(S33(u1,u2))//

4.1.2 Macros for Saint-Venant-Kirchhoff model

// SAINT-VENANT-KIRCHHOFF (SVK) : S(u) = λtrace(E(u))I + 2µE(u)
//=============================================
macro S11(u1,u2)(lambda*TrE(u1,u2)+mu*(C11(u1,u2)-1.0))//
macro S22(u1,u2)(lambda*TrE(u1,u2)+mu*(C22(u1,u2)-1.0))//
macro S12(u1,u2)(mu*C12(u1,u2))//
macro S21(u1,u2)(mu*C12(u1,u2))//
macro S33(u1,u2)(lambda*TrE(u1,u2))//

63
 ///////////////////////////////////////////////////////////////////////////////////////////
// tangent operator of SVK : S 0 (u)(E 0 (u)(w)) = λtrace(E 0 (u)(w))I + 2µE 0 (u)(w)
macro SD11(u1,u2,varu1,varu2) (lambda*TrEd(u1,u2,varu1,varu2)+2*mu*Ed11(u1,u2,varu1,varu2))//
macro SD22(u1,u2,varu1,varu2) (lambda*TrEd(u1,u2,varu1,varu2)+2*mu*Ed22(u1,u2,varu1,varu2))//
macro SD12(u1,u2,varu1,varu2) (2*mu*Ed12(u1,u2,varu1,varu2))//
macro SD21(u1,u2,varu1,varu2) (2*mu*Ed21(u1,u2,varu1,varu2))//

64