1 Boundary Problem
This is Boundary problem for mixed mode (I+II) crack propagation time dependent:
0 = div (1z) 2 [u] +f(x, t)x
z (x)
2 = div((x)z) z+[u] :e[u](1z) x
t + (1.1)
+ B.C.
+ I.C.
Where we have Boundary Condition (B.C.):
u=g(x, t)on D
[u]n= 0on N (1.2)
z
= 0on
n
and Iniatial Condition (I.C.)
u(x,0) =u 0 (x)x
(1.3)
z(x,0) =z 0 (x)[0,1]x
Suppose that:
1.is a bounded two dimensional domain and using a Cartesian coordinate, wherex=
(x1 , x2 )R 2 is parallel to the plate andx 3 the coordinate perpedicular to the plate.
2.u(x, t) represent the small anti-plane displacement at the positionxand timet0
3. The functionz(x, t) is called the phaseled for crack shape, where variablez(x, t)
satises 0z(x, t)1 inand represents the crack shape.
4.f(x, t) is external force
5.g(x) is a given anti-plane displacement on the boundary D
6.q(x) is a given boundary load in thex 3 -direction on the boundary N
2
Remark 1
In this case, we want to simulation mode III crack growth with condition;f(x, t) = 0,
2 = 103 , and= 10 3 .
3
2 Semi Discretization in Time and Semi-discrete Scheme
To solve boundary problem in equation (1.1), we have to do a semi discretization in time.
t >0 time increment,wheret k :=kt
0 =t 0 < t1 < t2 <
So, approximate with time disretization:
u k(x) u(x, kt)
gk (x)g(x, kt)
(2.1)
zk z k1 z
=z t(x, kt)
t t
Using semi dicretization in time (equation 2.1), we can dene equation (1.1) and than can
ndndu k (x) and kz(x).
0 =div (1z k1 2
) [u k ]
kz z k1
k
(x) k k1 k1 k
2 = div (x)z z +[u ] :e[u ](1z )
t +
z k := max(kz, z k1 )
(2.2)
uk =g k (x, t) on D
[u k ]n= 0 on N
zk
=0 on
u
Because equation (2.2) is a couple equation, so tondu k
(x) and kz(x) we have to solve
equation (2.2) simultaneously.
3 Weak Form and Finite Element Method
3.1 Variableu k (x)
Weak Form
V u
:={wH 1 ()|w= 0 on D }u We
choose
allwV theu test functionwV for
0 =div (1z k1 2
) [u k ]
0 = div (1 z k1 2
) [u ] k w dx
4
0 = j (1 z ) ij [uk ]
k1 2
wi dx(3.1)
5
To solve equation (3.1), recall Divergence Theorem:
Divergence Theorem
(j fj )g dx= fj nj g ds fj (j g)dx
so we get:
0 = (1z k1 2
) ij [u ] nj wi ds
k
(1 z ) ij [u ] (j wi )dx
k1 2 k
0 = (1 z k1 2
) ij [uk ] (j wi )dx (1 z ) ij [uk ] nj wi ds(3.2)
k1 2
or equation (3.2):
(1 z ) ij [uk ](j wi )dx=
k 1 2
(1 z ) ij [uk ]nj wi ds(3.3)
k1 2
If we use Boundary Condition (equation 1.2) andz k1
= 0 on N , so equation (3.3) can be:
(1 z k1 2
) ij [uk ](j wi )dx= 0 (3.4)
33
ifR sym
ij [uk ](j wi ) =[u k
] : (T w)
1 T T
ij [u ](j wi )
k
=[u k
]: w+w (3.5)
2
k
=[u ] :e[w]
Using equation (3.3), equation (3.6) is writen:
(1 z k1 2
) [u k ] :e[w]dx= 0 (3.6)
Remark 2
Because the material is homogeneous and isotropic. According Hookes Law,[u] can
be dened:
6
[u] =(divu)I+ 2e[u]
[u] =(tr(e[u]))I+ 2e[u]
where:
1. divu=tr(e[u]) =I:e[u]
2.andare Lames constanta
7
By Remark 2, equation (3.6) be:
k1 2
(1 z ) ((divu k )I+ 2e[u k ]) :e[w] dx= 0
(1 z ) (divu k )I:e[w] + 2e[u
k1 2 k
] :e[w] dx= 0
(1z k1 2
) (divu k )(divw) + 2e[u k ] :e[w] dx= 0
So can be dene:
k 1 2 k k
a
u (u, w) = (1 z k
) (divu )(divw) + 2e[u ] :e[w] dx
lk
u (w) = 0ds
k
Findu V u (g k ) (k= 1,2, ) such that:
au (uk , wh ) =l k
u (w) (wV u
) (3.7)
Finite Element Method
We chooseV h V=H (), dim(Vh )<
u u
We deneV h :=V h V .
Letg k
V h be an approximation ofg k and we dene thatV u
(gk ) :=V u
+g k
h h h h h
k u k
Findu V (g ) (k= 1,2, ) such that
h h h
ak (uk , wh ) =l k
(w h V h )
u
u h u (wh )
where:
ak k k1 2 k k
u (uh , wh ) = (1 z ) (divu h )(divw h ) + 2e[uh ] :e[w h ] dx
lk
u (wh ) = 0
3.2 Variablezk (x)
Weak Form
8
V z :={vH 1 ()|v= 0 on 1N }
We choose the test functionvV
9
for allvV
z (x)
2 = div((x)z) z+[u] :e[u](1 z)
t
k k1
z z k (x) k k 1 k 1 k
2 = div((x)z ) z +[u
] :e[u
](1z )
t
Here, we set thatv= min(z k ,0)
k1 (x)
kz z
2 v dx= div((x)z ) k
z +[u
k k1
] :e[u k1 k
](1z ) v dx
t
z z
k k1
(x)
v dx= div((x)z k ) v dx kz v dx
2
t
II
I
+ [u k1 ] :e[u k1 ](1zk ) v dx (3.8)
III
from equation (3.8), we can solve:
1. Part I
div((x) z k ) v dx= div((x)zk ) v dx
= ((x) zk ) n v ds ((x)zk )v dx
k
z
= (x)v ds ((x)zk )v dx (3.9)
n
z
According Boundary Condition that = 0, so equation (3.9) be:
n
div((x) z k ) v dx= ((x) zk ) v dx(3.10)
2. Part II
1
0
(x) 1
z v dx=
k
(x)zk v dx(3.11)
3. Part III
[u k1 ] :e[u k1 ](1 zk ) v dx
k1 k1 k1
if[u ] =(divu )I+ 2e[u ], so, the above equation be:
((divu k1 )I+ 2e[u k1
]) :e[u k1 ] 1 zk v dx
(divu k1 )I:e[u k1
] + 2e[uk1 ] :e[u k1 ] 1 zk v dx
2
(divu k1 )2 + 2 e[uk1 ] 1 zk v dx(3.12)
1
1
Using equation (3.10), (3.11), and (3.12), equation (3.9) can be:
k
z z k1 1
v dx= ((x)zk )v dx (x)zk v dx
2
t
2
+ (divu k1 2
) + 2 e[uk1 ] 1zk v dx
t t
k
(z z k1
)v dx= ((x)z )v dx k
(x)zk v dx
2 2
t 2
+ (divu k1 2
) + 2 e[uk1 ] 1zk v dx
2
Separate the above equation: 2
t (x) t
1+ + (divu k1 )2 + 2 e[uk1 ] kzv dx+ ((x) zk ) v dx
2 2
2
t
= (divu k1 2
) + 2 e[uk1 ] z k1 v dx
v dx+
where:
t (x) 2
k 1 2 1
a t
z (z, v) = 1+ + (divu k
) + 2 e[uk ] kzv dx+ ((x)zk )v dx
2 2
t 2
lk 1 2 k1 k1
z (v) = (divu k
) + 2 e[u ] v dx+ z v dx
2
k z
Findz V (k= 1,2,3, ) such that:
h
az (z k , vh ) =l kz (v) (vV z
) (3.13)
Finite Element Method
1
2
We chooseV h V=H (), dim(Vh )<
z z
We deneV h :=V h V .
Letg k
V h be an approximation ofg k and we dene thatV z (g k ) :=V z
+g k
h h h h h
k z k
Findz V (g ) (k= 1,2, ) such that
h h h
ak (z k , vh ) =l k
(v h V h )
z
z h z (vh )
and we dene that:
t (x)
t
2 k
ak k 1 2 k1
k
z (zh , vh ) = 1+ + (divu ) + 2(e[u ]) hzvh dx+ ((x)zk )v dx
2 2
h
t 2
lk 1 2 k1 k1
z (vh ) = 2 (divu k
) + 2 e[u ] vh dx+ zh vh dx
1
3
4 Program of FreeFEM++
In this problem we set the boundary condition as= (1,1)(1,1) and D ={(x
(1,1), 1 , x2 )|x1
andt0.x 2 =1}(seeg. 1). The boundary condition foruis given byg= (1,1,0) forx D
D1 D1
(1,1) (1,1)
(1,1) (1,1)
G2 G2
N G3
G1 N N G3
G1 N
G4 G4
(1,1) (1,1) (1,1) (1,1)
D2 (a) D2
Case I (b) Case II
Figure 1: Computational domain of mode I + II crack growth with D1 (g= (u 1 , u2 )) and
D2 (g= (u 1 ,u 2 )) (a) and D1 (g= (u 1 , u2 )) and D2 (g= (u 1 ,u 2 )) (b)
where:
G1 : =t + (1t) (t: 01)
y 1 1
x 1 1
G2 : =t + (1 t) (t: 01)
y 1 1
x 1 1
G3 : =t + (1t) (t: 01)
y 1 1
x 1
G4 : =t 1 (t: 01)
y 1 + (1 t)
1
1
4
