Professional Documents
Culture Documents
MIT OpenCourseWare
Contents
1 Large displacement analysis of solids/structures
1.1
Project Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2
1.2.1
Mathematical model/problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.2
1.2.3
1.2.4
Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1
2.2
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10
14
19
23
6.1
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
6.1.1
F.E. model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
6.1.2
Higher-order elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
7 Isoparametric elements
28
33
9 u/p formulation
37
41
45
49
MIT 2.094
Contents
53
57
15 Field problems
61
65
71
76
19 Slender structures
81
85
90
Fall 08
1.1
MIT OpenCourseWare
Project Example
Physical problem
Reading:
Ch. 1 in
the text
analytical solution
F.E. solution(s)
You will use ADINA in your projects (and homework) for structures and uid ow.
3
MIT 2.094
1.2
Lagrangian formulations:
Total Lagrangian formulation
Updated Lagrangian formulation
1.2.1
Reading:
Ch. 6
Mathematical model/problem
Given
Calculate
the
the
the
the
the
Question Is there a unique solution? Yes, for innitesimal small displacement/strain. Not necessarily
for large displacement/strain.
For example:
Snap-through problem
MIT 2.094
Not physical
1.2.2
I. Equilibrium of stresses (Cauchy stresses, forces per unit area in t V and on t Sf ) with the applied
body forces t f B and surface tractions t f Sf
II. Compatibility
III. Stress-strain law
1.2.3
MIT 2.094
1.2.4
Notation
ij
i, j = 1, 2, 3
ij = t ji
(1.1)
Fall 08
MIT OpenCourseWare
Assume that on t Su the displacements are zero (and t Su is constant). Need to satisfy at time t:
Equilibrium of Cauchy stresses t ij with applied loads
t T
= t 11 t 22 t 33 t 12 t 23 t 31
(2.1)
(For i = 1, 2, 3)
ij,j + t fiB
t
ij nj
t Sf
fi
(e.g.
0 in t V (sum over j)
(2.2)
t Sf
fi on t Sf (sum over j)
t
i1 t n1 + t i2 t n2 + t i3 t n3
(2.3)
(2.4)
And: t 11 t n1 + t 12 t n2 = t f1 f
ij = function t uj
(2.5)
Reading:
Ch. 1, Sec.
6.1-6.2
MIT 2.094
2.1
ij t eij d t V =
tV
tV
t B
fi
ui d t V +
tS
t Sf
fi
u i f d t Sf
(2.6)
where
t eij
with ui
t
1
2
ui
uj
+ t
t xj
xi
(2.7)
= 0
(2.8)
Su
2.2
Example
t
t
11 t e11 d V =
tV
tV
t B
f1
u1 d V +
tS
Pr u 1 f d t S f
(2.9)
Derivation of (2.9)
t
t
or
11,1 + t f1B = 0
11,1 + t f1B u1 = 0
by (2.2)
(2.10)
(2.11)
MIT 2.094
Hence,
tV
11,1 + t f1B u1 d t V = 0
tS
11 u1 t Sf
u
t
tV
u1 f t11 tSf
(2.12)
u1,1 d V +
11
tV
u1 t f1B
d t V = 0
(2.13)
te11
Therefore we have
t
t
e
d
V
=
t 11
11
tV
u1 t f1B d t V + u1 f t Pr t Sf
tV
(2.14)
From (2.12) to (2.14) we simply used mathematics. Hence, if (2.2) and (2.3) are satised, then (2.14)
must hold. If (2.14) holds, then also (2.2) and (2.3) hold!
Namely, from (2.14)
tS
u1,1 t11 d tV = u1 t11 tSf
u
tV
tV
u1 t11,1 d tV =
tV
u1 tf1B d tV + u1 f
tPr tSf
(2.15)
or
u1
tV
S
11,1 + tf1B d tV + u1 f tPr t11 tSf = 0
Now let u1 = x 1
x
tL
(2.16)
11,1 + tf1B = 0
(2.17)
Pr = t11
(2.18)
Fall 08
MIT OpenCourseWare
Reading:
Sec. 6.1-6.2
(i = 1, 2, 3) in t V
t Sf
ij tnj = fi
(3.1)
(i = 1, 2, 3) on t Sf
(3.2)
Compatibility
Stress-strain law(s)
Principle of virtual displacements
t
ij teij d tV =
ui tfiB d tV +
tV
1
teij =
2
tV
ui
uj
+ t
t xj
xi
tS
f
ui | t S
t Sf
fi
d tSf
(3.3)
(3.4)
If (3.3) holds for any continuous virtual displacement (zero on tSu ), then (3.1) and (3.2) hold and
vice versa.
10
MIT 2.094
Major steps
I. Take (3.1) and weigh with ui :
t
ij,j + t fiB ui = 0.
(3.5a)
ij,j + t fiB ui d t V = 0
(3.5b)
tV
III. Use divergence theorem. Obtain a boundary term of stresses times virtual displacements on t S =
t
Su t Sf .
IV. But, on t Su the ui = 0 and on t Sf we have (3.2) to satisfy.
Result: (3.3).
Example
t
tV
11 te11 d tV =
t S
tS
f
ui f1 f d tSf
(3.6)
11
MIT 2.094
1
1
(1 + r) u1 + (1 r) u2
2
2
1
1
t
t
u(r) = (1 + r) u1 + (1 r) t u2
2
2
1
1
u(r) = (1 + r) u1 + (1 r) u2
2
2
u(r) =
(3.7)
(3.8)
(3.9)
tV
u
= t = B (1)
x
u1
u2
te11t 11 d V
(3.10)
[u1
u2 ]
tV
= U1
u2
B (1)
11 d t V
(3.11)
= t F (1)
u2 ] t F (1)
t (1)
F
U 2 U 3
[u1
(3.12)
(3.13)
u1
where
t
(1)
(1)
F1 = t F2
t (1)
F2
(3.14)
t (1)
F1
(3.15)
= U 1
U2
u2
U3
0
t
(2)
F
(3.16)
u1
R.H.S.
U1
U2
U3
0
t Sf
t
Sf f
1
(3.17)
Now apply,
U
0 0
(3.18)
1 0
(3.19)
0 1
(3.20)
then,
U
then,
U
12
MIT 2.094
This gives,
F (1)
0
F (2)
unknown reaction
0
=
t tSf t
f1 Sf
(3.21)
We write that as
t
F = tR
t
F = fn t U1 , t U2 , t U3
(3.22)
(3.23)
13
Fall 08
MIT OpenCourseWare
Reading:
Ch. 4
The exact solution of the mathematical model must satisfy the conditions:
Equilibrium within t V and on t Sf ,
Compatibility
Stress-strain law(s)
I. Dierential formulation
II. Variational formulation (Principle of virtual displacements) (or weak formulation)
We developed the governing F.E. equations for a sheet or bar
We obtained
t
F = tR
(4.1)
F = K tU
(4.2)
We want to establish,
KU (t) = R(t)
(4.3)
14
MIT 2.094
Consider
T =
U
U1
V1
W1
U2
WN
(N nodes)
(4.4)
T = U1
U
U2
U3
Un
(n = 3N )
(4.5)
We now assume
(m)
u
= v
w
,
u(m) = H (m) U
u(m)
(4.6a)
is n x 1.
where H (m) is 3 x n and U
(m) = B (m) U
(4.6b)
(m) =
e.g. xy
xx yy
v
u
=
+
x y
zz
xy
yz
zx
We also assume
u(m)
(m)
H (m) U
(4.6c)
(m)
(4.6d)
15
MIT 2.094
T
T
f B dV
dV =
U
V
(4.7)
V (m)
(m) T
fB
(m)
dV (m)
(4.8)
V (m)
T
(m) T
(m)
(m)
U
B
dV
=
V (m)
(m) T
B (m)
(4.9)
dV
(m)
V (m)
Finally,
(m) T
(4.10)
(m)
(m)
dV
(m)
=
U
V (m)
(m) T
(4.11)
fB
(m)
dV (m)
V (m)
with
T
B (m)
(m) = U
(4.12)
= RB
KU
(4.13)
where K is n x n, and RB is n x 1.
Direct stiness method:
K=
K (m)
(4.14)
RB =
(m)
RB
(4.15)
(m)
(m)
RB
=
V
(m)
(m)
H (m) f B
(m)
(4.16)
dV (m)
(4.17)
16
MIT 2.094
E = Youngs Modulus
Mathematical model
F.E. model
U1
U = U2
U3
(4.18)
Element 1
u(1) (x) =
x
100
100
H (1)
U1
0 U2
U
3
(4.19)
17
MIT 2.094
U1
0 U2
U
3
(4.20)
x
80
(4.21)
(1)
xx (x) =
1
100
100
B (1)
Element 2
u(2) (x) =
(2)
xx (x)
x
1 80
H (2)
1
1
80
80
B (2)
(4.22)
Then,
1 1 0
0
13E
E
1 1 0 +
0
K=
240
100
0
0 0
0
0
0
1 1
1 1
(4.23)
where,
<
=0
AE
E(1)
L
100
13 E
E 13
=
3 80
3 80
A
A
< A
(4.24)
(4.25)
(4.26)
=80
18
Fall 08
MIT OpenCourseWare
Reading:
Ch. 4
Dierential formulation
Variational formulation (Principle of Virtual Displacements)
(5.1a)
u(m) = H (m) U
K=
(5.1b)
K (m)
(5.1c)
(m)
(5.1d)
R=
RB
(m)
(m) = B U
T
K (m) =
B (m) C (m) B (m) dV (m)
(m)
V
(m)
T
(m)
RB =
H (m) f B
dV (m)
(5.1e)
(5.1f)
(5.1g)
(5.1h)
V (m)
Surface loads
Sf T
f Sf dSf
surface loads =
U
(5.2)
Sf
uS
H
(m)
S (m)
= HS
(m)
= H (m)
(5.3)
(5.4)
19
MIT 2.094
(m) T
(m)
T
U
HS
fS
dS (m)
(5.5)
S (m)
Rs(m) =
HS
(m) T
fS
(m)
dS (m)
(5.6)
S (m)
Need to add contributions from all surfaces of all loaded external elements.
KU = RB + RS + Rc
(5.7)
KU = R
Kaa
Kba
Kab
Kbb
Ua
Ub
Ra
Rb
(5.8)
Solve for Ua :
Kaa Ua = Ra Kab Ub
(5.9)
where Ub is known!
Then use
Kba Ua + Kbb Ub = Rb + Rr
(5.10)
1
xx
E
yy =
1 2
xy
0
1
0
0
0
1
2
xx
yy
xy
20
(5.11)
MIT 2.094
u(x, y)
=H
v(x, y)
u1
u2
u3
u4
v1
v2
v3
v4
(5.12)
If we can set this relation up, then clearly we can get H (1) , H (2) , H (3) , H (4) .
u(m) = H (m) U
(5.13)
(5.14)
v(x, y) = b1 + b2 x + b3 y + b4 xy
(5.15)
(e.g.) u(1, 1) = a1 + a2 + a3 + a4 = u1 .
h2 (x, y) = 14 (1 x)(1 + y)
21
MIT 2.094
h3 (x, y) = 14 (1 x)(1 y)
h4 (x, y) = 14 (1 + x)(1 y)
u(x, y) = h1 u1 + h2 u2 + h3 u3 + h4 u4
(5.16)
v(x, y) = h1 v1 + h2 v2 + h3 v3 + h4 v4
(5.17)
u(x, y)
v(x, y)
h1
0
h2
0
h3
0
h4
0
0
h1
0
h2
0
h3
H (2x8)
h4
u1
u2
u3
u4
v1
v2
v3
v4
(5.18)
We also want,
xx
h1,x
yy = 0
xy
h1,y
h2,x
0
h2,y
h3,x
0
h3,y
h4,x
0
h4,y
0
h1,y
h1,x
0
h2,y
h2,x
B (3x8)
u
x
v
=
y
u v
+
=
x
y
0
h3,y
h3,x
0
h4,y
h4,x
u1
u2
u3
u4
v1
v2
v3
v4
(5.19)
xx =
(5.20)
yy
(5.21)
xy
(5.22)
22
Fall 08
6.1
MIT OpenCourseWare
Example
Reading:
Ex. 4.6 in
the text
t = 0.1,
KU = R;
K=
K (m) ;
R = RB + R s + R c + R r
T
(m)
K
=
B (m) C (m) B (m) d V (m)
(m)
RB ;
6.1.1
(6.1)
(6.2)
V (m)
RB =
E, plane stress
(m)
RB
(6.3)
V (m)
F.E. model
el. (2)
u1 u2 u3 u4
= .
..
v1 v2 v3 v4
23
u1
..
.
(6.4)
MIT 2.094
In practice,
u1
..
.
u4
=B
v1
.
..
v4
K =
B T CB dV ;
el
(6.5)
..
..
..
.
.
.
..
..
..
.
.
.
K
=
assemblage
..
..
..
.
.
.
.
.
..
.
.
.
.
.
U18
U1
..
.
..
.
U11
..
.
..
.
U18
(6.6)
Consider,
HS = H
H S f S d S;
RS =
S
H=
u=
h1
0
u1
u2
..
.
u4
v1
..
.
h2
0
h3
0
h4
0
0
h1
0
h2
(6.7)
on surface
0
h3
0
h4
u(x, y)
v(x, y)
(6.8)
(6.9)
v4
24
MIT 2.094
HS = H
y=+1
1
2 (1 + x)
=
0
(6.10)
1
2 (1
x) 0
0
0
0
0
1
2 (1
+ x)
1
2 (1
x)
(6.11)
From (6.7);
+1
RS =
1
1
2 (1
1
2 (1
+ x)
x)
0
0
0
0
(0.1) dx
1
p(x)
2 (1 + x)
1
thickness
(1
x)
2
(6.12)
RS =
p0 (0.1)
p0 (0.1)
6.1.2
(6.13)
Higher-order elements
Want h1 , h2 , h3 , h4 , h5
u(x, y) =
i=1
hi u i .
25
MIT 2.094
1
h5
2
1
h5
2
1
(1 + x)(1 + y)
4
1
h2 = (1 x)(1 + y)
4
1
h3 = (1 x)(1 y)
4
1
h1 = (1 + x)(1 y)
4
h1 =
(6.14)
(6.15)
(6.16)
(6.17)
Note:
We must have
u(x, y) =
hi = 1
h i ui
(6.18)
u(x, y) =
h i u = u
h i = u
(6.19)
From (6.1),
(m)
U =R
(6.20)
U =R
(6.21)
(m) T
(m)
(m)
dV
(m)
V (m)
(6.22)
(m)
(m)
= R;
(m)
=
V
(m)
Two properties
I. The sum of the F (m) s at any node is equal to the applied external forces.
26
(6.23)
MIT 2.094
T F (m) = U
T
U
=(m)
(6.24)
(m)
(6.25)
V (m)
=0
(6.26)
27
Fall 08
We want K =
MIT OpenCourseWare
B T CB dV , RB =
Reading:
Sec. 5.1-5.3
x=
hi xi
(7.1)
hi yi
(7.2)
i=1
y=
i=1
where
1
(1 + r)(1 + s)
4
h2 = (1 r)(1 + s)
4
...
h1 =
u(r, s) =
(7.3)
(7.4)
hi ui
(7.5)
hi vi
(7.6)
i=1
v(r, s) =
i=1
28
MIT 2.094
7. Isoparametric elements
u
T =
= Bu
u
x
v4
(7.7)
= Bu
v
x
(7.8)
x
r
y
r
x
s
y
s
u2
u1
v
y
u
y
= J 1
(7.9)
(7.10)
J must be non-singular which ensures that there is unique correspondence between (x, y) and (r, s).
Hence,
1 1
K=
B T CB t det(J ) dr ds
(7.11)
1 1
dV
Also, RB =
1
H T f B t det(J ) dr ds
(7.12)
K
=t
T
Bij
CBij det(Jij ) (weight i, j)
(7.13)
(7.14)
(7.15)
29
MIT 2.094
7. Isoparametric elements
9-node element
x=
hi xi
(7.16)
hi yi
(7.17)
hi u i
(7.18)
hi vi
(7.19)
i=1
y=
i=1
u=
i=1
v=
i=1
J=
3 = 26
0
(7.20)
3
2
Then, we can determine the number of appropriate integration points by investigating the maximum
order of B T CB.
For a rectangular element, 3x3 Gauss integration gives exact K matrix. If the element is distorted,
a K matrix which is still accurate enough will be obtained, (if high enough integration is used).
30
MIT 2.094
7. Isoparametric elements
T C dV = R(u)
(7.21)
Find u, solution, in V, vector space (any continuous function that satises boundary conditions),
satisfying
R(v)
Example:
a(uh , vh ) = (f , vh ) vh Vh
(7.23)
> 0 for w V
0)
(w =
Also,
a(wh , wh ) > 0 for wh Vh
(Vh V, wh = 0)
31
Reading:
Sec. 5.5.5,
4.3
MIT 2.094
Property I
7. Isoparametric elements
Dene: eh = u uh .
(7.24)
(7.25)
Hence,
a(u uh , vh ) = 0
(7.26)
a(eh , vh ) = 0
(7.27)
(7.28)
Proof:
a(u, u) = a(uh + eh , uh + eh )
(7.29)
0 by Prop. I
2a(u
= a(uh , uh ) +
h , eh )
+ a(eh , eh )
(7.30)
a(u, u) a(uh , uh )
(7.31)
32
Fall 08
MIT OpenCourseWare
(A) Find
u V such that a(u, v) = (f , v) v V (Mathematical model)
(8.1)
a(v, v) > 0 v V,
(8.2)
v=
0.
Find
a(vh , vh ) > 0
vh Vh ,
vh = 0
(8.3)
(8.4)
Properties eh = u uh
(I) a(eh , vh ) = 0
vh Vh
(8.5)
(8.6)
33
MIT 2.094
e.g. Mesh
h1
h1
is contained in Mesh
h2
h2
(8.7)
uh minimizes! (Recall eh = u uh )
Proof: Pick wh Vh .
0
a(eh + wh , eh + wh ) = a(eh , eh ) +
2a(e
+ a(wh , wh )
h , wh )
(8.8)
a(eh , eh ) a(eh + wh , eh + wh )
= a(u uh + wh , u uh + wh )
(8.9)
(8.10)
Take wh = uh vh .
a(eh , eh ) a(u vh , u vh )
(8.11)
Using property (III) and (C), we can say that we will converge monotonically, from below, to a(u, u):
34
MIT 2.094
(Ch. 4.3)
(8.12)
(8.13)
(8.14)
Hence,
E Eh = C h2k
(8.15)
By theory,
log (E Eh ) = log C + 2k log h
(8.16)
35
MIT 2.094
By experiment, we can evaluate log(E Eh ) for dierent meshes and plot log(E Eh ) vs. log h
V = vol. strain
(8.17)
(8.18)
or
= bulk modulus
(8.19)
p nite number
V 0
(8.20)
(Small error in V results in huge error on pressure as , the constant C in (8.15) can be very large
locking)
36
Fall 08
MIT OpenCourseWare
We want to solve
Reading:
Sec. 4.4.3
I. Equilibrium
(9.1)
II. Compatibility
III. Stress-strain law
Use the principle of virtual displacements
T C dV = R
(9.2)
(V = xx + yy + zz )
E
=
3(1 2)
p = V must be accurately computed
(9.3)
(9.4)
(9.5)
Solution
ij = V ij + 2Gij
(9.6)
where
1
ij = Kronecker delta =
0
i=j
i=
j
(9.7)
Deviatoric strains:
V
ij = ij
ij
3
ij = pij + 2Gij
(9.8)
p=
kk
3
(9.9)
(9.2) becomes
T
C dV +
V V dV = R
V
V
T
C dV
TV p dV = R
V
(9.10)
(9.11)
37
MIT 2.094
9. u/p formulation
p + V = 0
(9.12)
p (p + V ) dV = 0
V
p
p V +
dV = 0
(9.13)
(9.14)
For an element,
u = Hu
= BD u
V = BV u
(9.16)
(9.17)
p = Hp p
(9.18)
(9.15)
Example
V = xx + yy
xx 31 (xx + yy )
yy 1 (xx + yy )
3
=
xy
13 (xx + yy )
Reading:
Ex. 4.32 in
the text
(9.19)
(9.20)
0!
Note: zz = 0 but zz =
p = Hp p = [1]{p0 }
(9.21)
p(x, y) = p0
(9.22)
Kuu Kup
u
R
=
Kpu Kpp
p
0
(9.23)
T
BD
C BD dV
V
=
BVT Hp dV
V
=
HpT BV dV
V
1
=
HpT Hp dV
Kuu =
(9.24a)
Kup
(9.24b)
Kpu
Kpp
(9.24c)
(9.24d)
38
MIT 2.094
9. u/p formulation
In practice, we use elements that use pressure interpolations per element, not continuous between
elements. For example:
Then, unless = 0.5 (where Kpp = 0), we can use static condensation on the pressure dofs.
equations.
Use p equations to eliminate p from the u
1
=R
Kuu Kup Kpp
Kpu u
(9.25)
(9.26)
Reading:
Sec. 4.5
=V
Vol qh vh dVol
inf
sup
>0
qh vh
qh Qh vh Vh
(9.27)
for normalization
Qh : pressure space.
If this holds, the element is optimal for the displacement assumption used (ellipticity must also be
satised).
Note:
inmum = largest lower bound
For example,
inf {1, 2, 4} = 1
sup {1, 2, 4} = 4
39
MIT 2.094
9. u/p formulation
For
(entry [3,1] in matrix) assume the circled entry is the minimum (inf) of
u
|
+
0
ph |j = Rh |i (from the top equation)
h i
=0
no equation for ph |j
spurious pressure! (any pressure satises equation)
Case 2 h = small =
uh |i = 0
uh |i = 0
ph |j + uh |i = Rh |i
Rh |i
ph |j =
displ. =
pressure
0
large
as is small
The behavior of given mesh when bulk modulus increases: locking, large pressures. See Example
4.39 textbook.
40
Fall 08
MIT OpenCourseWare
We developed
t
ij teij d tV = tR
Reading:
Ch. 6
(10.1)
tV
1
teij =
2
ui
uj
+ t
txj
xi
(10.2)
ij t eij d tV = tR
t
tV
1
t eij =
2
(ui ) (uj )
+
t xj
t xi
(10.3)
( t eij )
(10.4)
In FEA:
t
F = tR
(10.5)
In linear analysis
t
F = K t U KU = R
(10.6)
In general nonlinear analysis, we need to iterate. Assume the solution is known at time t
t
x=
x + tu
(10.7)
F =
t+t
(10.8)
41
MIT 2.094
F = tF + F
(10.9)
t+t
U = tU + U
(10.10)
F
= tK U ,
(10.11)
From (10.8),
t
K U =
t+t
R tF
(10.12)
We use this to obtain an approximation to U . We obtain a more accurate solution for U (i.e.
using
t+t
K (i1) U (i) =
t+t
(i)
t+t
R t+tF (i1)
t+t
(i1)
+ U
t+t
F)
(10.13)
(i)
(10.14)
Also,
t+t
F (0) = tF
t+t
(0)
(10.15)
= K
(10.16)
U (0) = tU
(10.17)
t+t
U (i) < D
2
t+t
t+t (i1)
R
F
< F
2
Note:
a2 =
(ai )
(i)
=U
i=1,2,3...
Process
Increase the displacement t Ui by , with no increment for all t Uj , j = i
calculate
t+
tF
.
tUi
42
(10.18)
(10.19)
MIT 2.094
So, perform
Pictorially,
..
.
t
K = ...
..
.
t+t
= 0tSij + 0 Sij
t+t
0 Sij ,
The strain measure energy-conjugate to the 2nd P-K stress 0tSij is the Green-Lagrange strain 0tij
Then,
0V
t
0 Sij
Also,
0V
0tij d 0V = tR
t+t
0 Sij
t+t0ij d 0V =
(10.21)
t+t
(10.22)
Example
43
MIT 2.094
F = tR
t+t
F =
(10.23)
t+t
(10.24)
K (0) U (1) =
t+t
t+t
K (1) U (2) =
t+t
(1)
t+t
(10.25)
(1)
(10.26)
R t+tF (1)
(10.27)
(0)
+ U
i = 2,
t+t
t+t
(2)
t+t
(1)
+ U
(2)
(10.28)
44
Fall 08
MIT OpenCourseWare
t
t
ij teij d V =
tV
Reading:
Ch. 6
t
0 Sij
0V
0tij d 0V = tR
(11.1)
tx
tx
tx
1
X
=
0
0x1
0x2
0x3
x2
0x1
x2
0x2
x2
0x3
tx3
0x1
tx3
0x2
tx3
0x3
(11.2)
d tx1
d tx = d tx2
d tx3
0
d x1
d 0x = d 0x2
d 0x3
(11.3)
(11.4)
Implies that
d tx = 0tX d 0x
(11.5)
= 0tX T 0tX
Some applications
45
(11.6)
MIT 2.094
(11.7)
1
d 0s = d 0xT d 0x 2
(11.8)
0 T t Tt
0
t 2
d x 0X
0X d x
=
,
d 0s d 0s
from (11.5)
(11.9)
Express
d 0x = d 0s 0n
(11.10)
t =
(11.11)
(11.12)
0 T t 0 12
n 0C n
(11.13)
Also,
d tx
T t t t
d x = d s d s cos t,
From (11.5),
T
T 0tX
d 0x
cos t =
t
0X
d 0x
d ts d ts
0 0T t 0
d s n 0 C n d 0s
=
d ts d ts
cos t =
(a b = ab cos )
(11.14)
t
0X
0tX
(11.15)
(11.16)
0T t 0
n 0C n
t
t
(11.17)
Also,
0
t
det 0tX
(11.18)
Example
Reading:
Ex. 6.6 in
the text
46
MIT 2.094
1
(1 + 0x1 )(1 + 0x2 )
4
h1 =
(11.19)
..
.
t
xi = 0xi + tui
=
hk txki ,
(11.20)
(i = 1, 2)
(11.21)
k=1
where txki are the nodal point coordinates at time t ( tx11 = 2, tx12 = 1.5)
Then we obtain
1
5 + 0x2
t
1
0X =
4 2 (1 + 0x2 )
At 0x1 = 0, 0x2 = 0,
1 5
t
=
0 X 0
4 12
xi = 0x2 =0
1 + 0x1
1
0
2 (9 + x1 )
(11.22)
(11.23)
9
2
1 t
1 t T t
0X 0X I =
0C I
2
2
txi
=
0xj
(11.24)
xi + tui
tui
=
+
ij
0xj
0xj
We nd that
1 t
t
t
t
t
0ij =
0ui,j + 0uj,i + 0uk,i 0uk,j ,
2
(11.25)
sum over k = 1, 2, 3
(11.26)
where
t
0ui,j
tui
0xj
(11.27)
47
MIT 2.094
= 0tR 0tU
(11.28)
=I
(11.29)
t
0X
3
2
1
2
2
3
2
4
3
(11.30)
3
2
Then,
2
= 0tX T 0tX = 0tU
1 t 2
t
I
0 =
0U
2
t
0C
(11.31)
(11.32)
This shows, by an example, that the components of the Green-Lagrange strain are independent of a
rigid-body rotation.
48
Fall 08
MIT OpenCourseWare
t
0X
txi
=
0xj
t
0C
= 0tX T 0tX
d tx = 0tX d 0x,
d 0x =
(12.1)
(12.2)
d 0x = 0t X d tx
t 1 t
dx
0X
where 0t X =
t 1
xi
=
0X
txj
1 t T t
t
0 =
0X 0X I =
0C I
2
2
(12.3)
(12.4)
Polar decomposition:
t
0X
= 0tR0tU 0t =
1 t 2
I
0U
2
(12.5)
Note in FEA
0
xi =
hk 0xki
k
for an element
ui =
hk tuki
(12.6)
xi = 0xi + tui
(12.7)
t
xi =
hk 0xki +
hk tuki
k
(12.8)
hk
0 k t k
xi + ui
(12.9)
hk txkk
(12.10)
49
MIT 2.094
E.g., k = 4
Then
0V
0 t 0 T
X tX
t t
t
0 Sij
0tij d 0V =
tV
(12.11)
(12.12)
ij t
eij d tV = tR
t+t
0S
t+t
0 Sij
t+t
0
t+t
0ij
= 0tS + 0 S
(12.13)
= 0tSij + 0 Sij
(12.14)
=
=
t
0 + 0
t
0ij + 0ij
(12.15)
(12.16)
Assume the solution is kown at time t, calculate the solution at time t + t. Hence, we apply (12.12)
at time t + t:
t+t
t+t
0
t+t
R
(12.17)
0 Sij
0ij d V =
0V
Look at 0tij :
1 t
t
t
t
0ui,j + 0uj,i + 0uk,i 0uk,j
2
1
ui
uj
uk tuk
tuk uk
t
0ij =
+ 0 + 0 0 + 0 0
2 0xj
xi
xi xj
xi xj
1
0tij =
0ui,j + 0uj,i + 0uk,i 0tuk,j + 0tuk,i 0uk,j
2
0tij =
(12.18a)
(12.18b)
(12.18c)
We have
t+t
0ij
0tij = 0ij
0ij
(12.19)
= 0eij + 0 ij
(12.20)
50
MIT 2.094
where 0eij is the linear incremental strain, 0 ij is the nonlinear incremental strain, and
0eij
t
t
u + u + u u + u u
2 0 i,j 0 j,i 0 k,i 0 k,j 0 k,i 0 k,j
(12.21)
1
u u
0 ij =
2 0 k,i 0 k,j
(12.22)
where
0uk,j
uk
,
0xj
uk =
t+t
uk tuk
(12.23)
Note
t+tij = 0ij
(12.24)
From (12.17):
0eij + 0 ij d 0V
0 Sij + 0 Sij
0V
=
0V
t
0
t
0 Sij 0eij + 0 Sij 0eij + 0 Sij 0 ij + 0 Sij 0 ij d V
(12.25)
t+t
Linearization
t
t
0 KN L U
0 KL U
t 0
0 Sij 0eij + 0 Sij 0 ij d V =
(12.26)
t+t
0V
0V
t
0
0 Sij 0eij d V
t
0F
(12.27)
We use,
0 Sij
0 Cijrs 0ers
(12.28)
t
t+t
R 0tF
0 KL + 0 KN L U =
where U is the nodal displacement increment.
51
(12.29)
MIT 2.094
Left hand side as before but using (k 1) and right hand side is
t+t (k1) 0
t+t
= t+tR
d V
0 Sij
0ij
(12.30)
0V
gives
t+t
t+t (k1)
0F
t+t (k1)
t+t (k1)
+
U (k) =
0 KL
0 KN L
(12.31)
t+t
52
t+t (k1)
0F
(12.32)
Fall 08
MIT OpenCourseWare
t+t
ij t+teij d t+tV =
t+tV
0V
t+t
0 Sij
t+t
0ij
0ij
t+t
t+t
0
0 Sij
0ij V
t+t
(13.1)
t+t
(13.2)
R
R
= 0tSij + 0 Sij
=
t
0ij
(13.3)
+ 0ij
(13.4)
= 0eij + 0 ij
(13.5)
where 0tSij and 0tij are known, but 0 Sij and 0ij are not.
1
t
t
0ui,j + 0uj,i + 0uk,i 0uk,j + 0uk,j 0uk,i
2
1
0 ij =
0uk,i 0uk,j
2
0eij
t
0
0eij 0 Cijrs 0ers d 0V +
0 Sij 0 ij d V =
0V
t
0 KL + 0 KN L U =
(13.6)
(13.7)
t+t
0V
t+t
R 0tF
R
0V
0eij 0tSij d 0V
(13.8)
(13.9)
53
MIT 2.094
t
0 KL
t
0 KN L
0V
0V
t T
t
0
0 BL 0 C 0 BL d V
(13.10)
t T
0 BN L
(13.11)
t
t
0
0 S 0 BN L d V
matrix
t
0F
=
0V
t T
0 BL
t
0
0S d V
(13.12)
vector
t+t (i1)
t+t (i1)
+
U (i) =
0 KL
0 KN L
t+t
U (i) =
t+t
t+t
U (i1) + U (i)
t+t (i1)
0F
(13.13)
(13.14)
(p. 545)
u1
tu
uk
+ 0 k 0
0
x1
x1 x1
1 uk uk
=
2 0x1 0x1
=
(13.15)
(13.16)
We are after
u11
u12 t
= 0tBL
u21 = 0 BL u
u22
0e11
ui =
(13.17)
hk uki
(13.18)
k=1
ui =
hk tuki
(13.19)
k=1
54
MIT 2.094
u1
tu1 u1
tu2 u2
+
+
0x
0x1 0x1
0x1 0x1
0 1
t 2
0
u1 = L + L cos L
t 2
u2 = 0L + L sin
0e11
0e11
1
1
=0
L
0 1
(13.20a)
(13.20b)
(13.20c)
L + L
1
+
cos
0L 1
0L
tu1
0x1
L + L
1
+
sin
0L 0
0L
(13.20d)
tu2
0x1
=0tBL u
(13.20e)
Hence,
0
0e11
L + L
cos
2
( 0L)
sin
cos
sin
(13.20f)
where the boxed quantity above equals 0tBL . In small strain but large rotation analysis we assume
L 0L,
0e11
0 11
0 11
cos
0L
1
=
2
sin
cos
u1 u1
u2 u2
+ 0
0
0
x1 x1
x1 0x1
sin
(13.20g)
(13.21a)
1 u1 u1
u1 u1
u2 u2
u2 u2
=
+ 0
+ 0
+ 0
2 0x1 0x1
x1 0x1
x1 0x1
x1 0x1
u1 u1
u2 u2
=
+ 0
0x1 0x1
x1 0x1
t
0 S11 0 11
u1
0x1
u2
0x1
tS
0
0 11
t
0
0 S11
u1
0x1
u2
0x1
(13.21b)
(13.21c)
(13.21d)
t
0S
u1
0x1
u2
0x1
1
0L
1
0
0
1
BN L
1 0
0 1
(13.21e)
55
MIT 2.094
0C
t
0S
=E
(13.22)
= 0tS11
(13.23)
EA
0L
cos2
sym
cos sin
sin2
1
P
0
+0
1
L
0
cos2
sin cos
cos2
t
0 KL
0 1
0
1
0
1
1 0
t
0 KN L
cos sin
sin2
sin cos
sin2
0
1
0
1
When = 0, 0tKL doesnt give stiness corresponding to u22 , but 0tKN L does.
56
(13.24)
Fall 08
MIT OpenCourseWare
t+t
ij t+teij d t+tV =
t+tV
0V
0V
t+t
t+t
0
0 Sij
0ij V
(14.1)
t+t
(14.2)
R
R
linearization
t
0
t+t
R
0 Sij 0 ij V =
t+t
V +
0V
0V
t
0
0 Sij 0eij V
(14.3)
Note:
0eij = 0tij
varying with respect to the conguration at time t.
F.E. discretization
0
xi =
hk 0xki
xi =
k
t
ui =
t+t
hk txki
xi =
hk tuki
t+t
ui =
t
0 KL + 0 KN L U =
hk t+txki
(14.4a)
hk uik
(14.4b)
hk t+tuik
ui =
t+t
R 0tF
(14.5)
57
MIT 2.094
Truss
L
0L
E 0A
0L
cos2
cos sin
cos2
2
cos sin
sin sin cos
cos2
cos2 cos sin
2
cos sin
sin
sin cos
1
0 1
0
t
P
0
0
1
1
0L 1
0
1
0
0
1
0 1
cos sin
sin2
sin cos
sin2
(14.6)
u1
v1
cos
sin
sin
cos
u1
v1
(14.7)
E 0A
0L
1
0
1
0
(14.8)
0
0
0
0
1 0
0 0
+
1 0
0 0
0L
1
0
1
0
0 1
0
0 1
1
0
1
0
1
0
1
58
(14.9)
MIT 2.094
Q 0L = tP
Q=
0L
(14.10)
t
E A
0L
0L
0L
0L
is important.
cos
t
t sin
0 F = P cos
sin
2D/3D (e.g. Table 6.5)
(14.11)
2D:
2
2
1
t
t
u
+
u
=
u
+
u
u
+
u
u
+
0 2,1
0 11
0 1,1
0 1,1 0 1,1
0 2,1 0 2,1
0 1,1
0e11
(14.12)
0 11
022
(14.13)
012
(14.14)
= ?
(14.15)
(Axisymmetric)
033
1 t 2
I
0U
2
0
0
t 2
0 0
0U =
t 2
( )
t
0
2 0x1 + tu1
d ts
= 0 =
d s
2 0x1
t
u
=1+ 0 1
x1
(14.16)
(14.17)
t
033
2
u1
1+ 0
1
x1
2
t
u
1 tu1
= 0 1 +
x1
2 0x1
1
=
2
(14.18)
(14.19)
59
MIT 2.094
t+t
033
033 =
u + u1
1
= 10
+
x1
2
t+t
033
0t33 =
t
2
u 1 + u1
0x
1
t
u 1 u1
1
u1
0x + 0x 0x + 2
1
1
1
(14.20)
2
u1
0x
1
(14.21)
Find u uh and h .
References
[1] T. Sussman and K. J. Bathe. Studies of Finite Element Procedures - on Mesh Selection. Computers
& Structures, 21:257264, 1985.
[2] T. Sussman and K. J. Bathe. Studies of Finite Element Procedures - Stress Band Plots and the
Evaluation of Finite Element Meshes. Journal of Engineering Computations, 3:178191, 1986.
60
Reading:
Sec. 4.3.6
Fall 08
MIT OpenCourseWare
Reading:
Sec. 7.2-7.3
Dierential formulation
Variational formulation
Incremental formulation
F.E. discretization
15.1
Heat transfer
15.1.1
Dierential formulation
qy = k
y
qz = k
z
(15.1)
(15.2)
III. Compatibility: temperatures need to be continuous and satisfy the boundary conditions.
61
MIT 2.094
k
+
k
+
k
= q B
x
x
y
y
z
z
(15.3)
unit cross-section
dV = dx (1)
(15.4)
q |x q |x+dx + q B dx = 0
qx
q|x q|x +
dx + q B dx = 0
x
(15.5)
(15.6)
+ qB
=0
dx
dx
(15.7)
= q B
(15.8)
= qS
n
(15.9)
on Sq .
15.1.2
k
x
+ + q
=0
(15.10)
( S = 0 and to be continuous.)
k
+ + q B dV = 0
x
(15.11)
62
MIT 2.094
T
Sq
k dV =
q dV +
q S dSq
heat ow
(15.12)
Sq
(15.13)
k
k= 0
0
0
k
0
0
0
k
(15.14)
qS = h e S
(15.15)
4
4
q S = (r ) S
2 r
2
= (r ) + S
+ S r S
= r S
(15.16)
(15.17)
(15.18)
T t+t t+t
S
t+t B
k
dV =
q dV +
t+tq S dSq
V
t+t
= t +
Let
t+t (i)
or
t+t (0)
with
t+t (i1)
(15.19)
Sq
(15.20)
+
(i)
(15.21)
(15.22)
From (15.19)
T
(i)
t+tk(i1) dV
V
T
t+t (i1)
=
t+tq B dV
t+tk(i1)
dV
V
V
(i1)
(i)
S
+
t+th(i1) t+t e t+t S
+ S
dSq
(15.23)
Sq
where the S
(i)
S
t+th t+t e t+t S dSq
(15.24)
Sq
63
MIT 2.094
F.E. discretization
= H1x4 t+t4x1
t+t
2x1 = B2x4 t+t4x1
t+t S
= H S t+t
t+t
(15.25)
(15.26)
(15.27)
For (15.23)
gives
(i)
t+tk(i1) dV =
B T t+tk(i1)
B dV
(i)
V
T
4x2
t+tq B dV
2x4
2x2
(15.28)
4x1
H T t+tq B dV
(15.29)
dV
B T t+tk(i1) BdV
S T t+t (i1)
t+t e
t+t S (i1)
t+t (i1)
+ S
(i)
dSq
known
(15.30)
Sq
Sq
15.2
(15.31)
S T t+t (i1)
h
H S t+te t+t(i1) +
H
(i) dSq
4x1
1x4
4x1
4x1
4x1
=0
y
x
or
Reading:
Sec. 7.3.2
(15.32)
(incompressible)
(15.33)
(irrotational)
(15.34)
vy =
(15.35)
2 2
+ 2 = 0 in V
x2
y
(15.36)
64
Fall 08
MIT OpenCourseWare
in V
(16.1)
= qS
n Sq
Sq
(16.2)
is prescribed, k
S
S Sq = S
S Sq =
(16.3)
S
B
,i k,i dV =
q dV +
q S dSq
V
(16.4)
Sq
65
MIT 2.094
(16.5)
In general 3D, we have an additional term for the left hand side of (16.1):
v ) cp (v )
(cp v) = cp (v) = cp (
(16.6)
term (A)
v = vi,i = div(v) = 0
(16.7)
So (16.1) becomes
,i k,i dV +
(cp ,i vi ) dV =
q B dV +
V
(16.8)
q S dSq
(16.9)
Sq
Navier-Stokes equations
Dierential form
ij,j + fiB = vi,j vj
(16.10)
vj
1 vi
ij = pij + 2eij
+
eij =
2 xj
xi
(16.11)
ij nj = fi
on Sf
(16.12)
vn
n
vt
n
= 0 (outow or
MIT 2.094
Variational form
v i vi,j vj dV +
eij ij dV =
v i fiB dV +
V
v i f fi f dSf
(16.13)
Sf
p vdV = 0
(16.14)
F.E. solution
We interpolate (x1 , x2 , x3 ), vi , v i , , , p, p. Good elements are
: linear pressure
: biquadratic velocities
(Q2 , P1 ), 9/3 element
9/4c element
Example:
For Sf e.g.
nn = 0,
vt
= 0;
n
(16.15)
67
MIT 2.094
and
vn
t
Reading:
Sec. 7.4
Re =
vL
Pe =
vL
k
cp
Reading:
Sec. 7.4.3
(16.16)
Dierential equations
|x=0 = L
k = cp v
(16.17)
|x=L = R
(16.18)
In non-dimensional form
1
=
Pe
Reading:
p. 683
exp Pe
L
L x 1
=
R L
exp (Pe) 1
(16.19)
(16.20)
68
MIT 2.094
F.E. discretization
= Pe
(16.21)
dx + Pe
0
(16.22)
1
2
(h )
Pe =
(i+1 2i + i1 ) =
Pe
(i+1 i1 )
2h
vL
(16.23)
(16.24)
Dene
Pee = Pe
h
vh
=
L
Pee
1
2
(16.25)
i1 + 2i +
Pee
1 i+1 = 0
2
(16.26)
what is happening when Pee is large? Assume two 2-node elements only.
i1 = 0
i+1 = 1
1
i =
2
(16.27)
(16.28)
Pee
1
2
(16.29)
69
MIT 2.094
i =
1
2
Pee
2
(16.30)
70
Fall 08
17.1
MIT OpenCourseWare
Abstract body
Reading:
Sec. 7.4
Fluid Flow
Sv , Sf
Sv Sf = S
Sv Sf = 0
17.2
Heat transfer
S , Sq
S Sq = S
S Sq = 0
71
MIT 2.094
17.3
Basic equations
P.V. velocities
v i vi,j vj dV +
ij eij dV =
v i fiB dV +
V
v i f fi f dSf
(17.1)
Sf
Continuity
pvi,i dV = 0
(17.2)
P.V. temperature
cp ,i vi dV +
,i k,i dV =
q B dV +
V
q S dS
(17.3)
Sq
F.E. solution
xi =
hk xki
(17.4)
vi =
hk vik
(17.5)
hk k
(17.6)
p=
k pk
h
(17.7)
F (u) = R
17.4
u = p nodal variables
(17.8)
Model problem
1D equation,
cp v
d
d2
=k 2
dx
dx
(17.9)
d
d2
= 2
dx
dx
(17.10)
72
MIT 2.094
Pe =
vL
,
k
cp
(17.11)
is non-dimensional
exp PeLx 1
L
=
exp (Pe) 1
R L
d
d d
Pe dV +
dV = 0
dx
V
V dx dx
(17.12)
(17.13)
h =
() =
1 i1 + i
h
h
h
L
(17.14)
For node i:
i1
Pee
Pee
i1 + 2i i+1 +
i+1 = 0
2
2
(17.15)
where
Pee =
vh
h
= Pe
L
(17.16)
=
2 (i+1 2i + i1 )
i
(h )
i+1 i1
=
2h
i
73
(17.17)
(17.18)
MIT 2.094
1 (Pee /2)
2
(17.19)
Physically unrealistic solution when Pee > 2. For this not to happen, we should rene the mesha very
ne mesh would be required. We use upwinding
d
i i1
=
dx i
h
(17.20)
The result is
(1 Pee ) i1 + (2 + Pee ) i i+1 = 0
Very stable, e.g.
i1 = 0
i+1 = 1
i =
(17.21)
1
2 + Pee
(17.22)
Unfortunately it is not that accurate. To obtain better accuracy in the interpolation for , use the
function
x
exp Pe L
1
(17.23)
exp (Pe) 1
The result is Pee dependent:
This implies ow-condition based interpolation. We use such interpolation functionssee references.
References
[1] K.J. Bathe and H. Zhang. A Flow-Condition-Based Interpolation Finite Element Procedure for
Incompressible Fluid Flows. Computers & Structures, 80:12671277, 2002.
[2] H. Kohno and K.J. Bathe. A Flow-Condition-Based Interpolation Finite Element Procedure for
Triangular Grids. International Journal for Numerical Methods in Fluids, 51:673699, 2006.
74
MIT 2.094
17.5
FSI briey
f
f
f
=
+
v
(17.24)
t
x
particle
where v is the particle velocity. For a mesh point,
f
f
f
=
+
vm
t
x
mesh point
(17.25)
f
= f
+
(v vm )
x
particle
mesh point
(17.26)
Use (17.26) in the momentum and energy equations and use force equilibrium and compatibility at
the FSI boundary to set up the governing F.E. equations.
References
[1] K.J. Bathe, H. Zhang and M.H. Wang. Finite Element Analysis of Incompressible and Compressible
Fluid Flows with Free Surfaces and Structural Interactions. Computers & Structures, 56:193213,
1995.
[2] K.J. Bathe, H. Zhang and S. Ji. Finite Element Analysis of Fluid Flows Fully Coupled with
Structural Interactions. Computers & Structures, 72:116, 1999.
[3] K.J. Bathe and H. Zhang. Finite Element Developments for General Fluid Flows with Structural
Interactions. International Journal for Numerical Methods in Engineering, 60:213232, 2004.
75
Fall 08
MIT OpenCourseWare
In structures,
Reading:
Sec. 8.4
F (u, p) = R.
(18.1)
In heat transfer,
F () = Q
(18.2)
In uid ow,
F (v, p, ) = R
(18.3)
In structures/solids
F =
F (m) =
0V (m)
Elastic materials
Example
p. 590 textbook
76
(18.4)
MIT 2.094
Material law
t
0 S11
0t11
=E
(18.5)
In isotropic elasticity:
=
E
t
0
E (1 )
,
(1 + ) (1 2)
( = 0.3)
1 t 2
1
=
U
I
0t11 =
2 0
2
(18.6)
L + tu
0L
1
1 =
2
u
1+ 0
L
2
1
(18.7)
0
X t 0X T
t t 11 11 t 11
(18.8)
with
0
0
t X11
0L
L
,
+ tu
0tS11 =
L
0L
0 0
L = t tL
(18.9)
0 2
0
L
L
t
11 = t t11
tL
L
L
1
t t11 = E
L
2
u
1+ 0
L
(18.10)
2
1
(18.11)
t
t
A
E
u
u
11 A = P =
1+ 0
1 1 +
0
2
L
L
t
(18.12)
This is because of the material-law assumption (18.5) (okay for small strains . . . )
Hyperelasticity
t
0W
1 0tW
0tW
t
+ t
0 Sij =
2 0tij
0ji
0tSij
1 0 Sij
C
=
+
0 ijrs
2 0trs
0tsr
77
(18.13)
(18.14)
(18.15)
MIT 2.094
Plasticity
yield criterion
ow rule
hardening rule
t
tt
(18.16)
tt
R t+tF
f
t+t (i1)
U t+tU (i1) + H.O.T.
=f
U
+
U t+tU (i1)
where
t+t
(18.17)
(18.18)
t+t
t+tF
t+t
t+t (i1)
0=
R
F
U (i)
U t+tU (i1)
(18.19)
We obtain
t+t
K (i1) U (i) =
t+t
(i1)
t+t
R t+tF (i1)
(18.20)
t+tF
F
=
=
U t+tU (i1)
U t+tU (i1)
(18.21)
Physically
t+t
(i1)
K11
t+t
(i1)
F1
(18.22)
u
78
MIT 2.094
i = 1;
t+t
i = 2;
Convergence
K u(1) =
t+t
(1)
t+t
(2)
R tF
R
(18.23)
t+t
(1)
(18.24)
Use
U (i) 2 <
2
a2 =
(ai )
(18.25)
(18.26)
But, if incremental displacements are small in every iteration, need to also use
t+tR t+tF (i1) 2 < R
18.1
(18.27)
Slender structures
t
1
Li
(18.28)
79
MIT 2.094
Beam
e.g.
t
L
1
100
(4-node el.)
(9-node el.)
We do not have a shear (better)
But, still for thin structures, it has problems
like ill-conditioning.
We need to use beam elements. For curved structures also spurious membrane strain can be
present.
80
Fall 08
Beam analysis,
t
L
MIT OpenCourseWare
1 (e.g.
t
L
1
1
100 , 1000 , )
Reading:
Sec. 5.4,
6.5
(plane stress)
L
0
2
J=
0 2t
1
h2 = (1 r) (1 + s)
4
1
h3 = (1 r) (1 s)
4
(19.1)
(19.2)
(19.3)
u 2 = u2 2
2
B =
(19.4)
(19.5)
(19.6)
u2
v2
u3
v3
41 (1 + s) L2
14 (1 s) L2
0
1
4 (1
r) 2t
1
4 (1
r) 2t
14 (1 + s) L2
14 (1
0
14 (1 r) 2t
81
r) 2t
14 (1 s) L2
etc
(19.7)
MIT 2.094
u2
Bbeam
L1
v2
2
t
2L s
L1
21 (1 r)
u
x
u
y
v
x
(19.8)
1
(1 r)v2
2
1
st
(19.9)
(19.10)
at r = 1,
v(1) = v2
(19.11)
st
u(1) = 2 + u2
2
(19.12)
Kinematics is
u(r) =
1
(1 r)u2
2
(19.13)
results into xx
xx =
u 2
1
=
r L
L
(19.14)
st
(1 r)2
4
(19.15)
u(r, s) =
results into xx , xy
xx =
xy =
v(r) =
st
2L
u 2
1
= (1 r)
s t
2
1
(1 r)v2
2
(19.16)
(19.17)
(19.18)
results into xy
xy =
82
1
L
(19.19)
MIT 2.094
1
1
v2 (1 r)2 = 0
L
2
(19.20)
for all r! Impossible (except for v2 = 2 = 0) So, the element has a spurious shear strain!
Beam kinematics (Timoshenko, Reissner-Mindlin)
dw
dx
1 3
I=
bt
12
(19.21)
(19.22)
EI
0
d d
dx + AS G
dx dx
dw
dx
L
dw
dx =
pwdx
dx
0
As = kA = kbt
(19.23)
(19.24)
To calculate k
2
V
1
1
2
(a ) dA =
dAs
A
2G
2G
s
A
AS
Reading:
p. 400
(19.25)
t
y2
3 V
2
a =
t 2
2 A
(19.26)
Reading:
Ex. 5.23
5
6
(19.27)
83
MIT 2.094
Now interpolate
w(r) = h1 w1 + h2 w2
(19.28)
(r) = h1 1 + h2 2
(19.29)
1+r
w1
2
1+r
=
1
2
w=
(19.30)
(19.31)
Shearing strain
=
1+r
w1
1
L
2
(19.32)
Shear strain is not zero all along the beam. But, at r = 0, we can have the shear strain = 0.
1
w1
can be zero
L
2
(19.33)
Namely,
1
w1
2
= 0 for 1 = w1
L
L
2
(19.34)
84
Fall 08
MIT OpenCourseWare
The ber moves up and rotates and its length does not change.
Principle of virtual displacement
EI
dx + (Ak)G
(Linear Analysis)
dw
dx
L
dw
dx =
wT pdx
dx
0
(20.1)
Two-node element:
Three-node element:
= w1 wq 1 q
u
w = Hw u
= H u
Hw = h1 hq 0 0
H = 0 0 h1 hq
dx
J=
dr
(20.2)
(20.3)
(20.4)
(20.5)
(20.6)
(20.7)
85
MIT 2.094
dw
= J 1 Hw,r u
dx
(20.8)
= J 1 H,r u
dx
(20.9)
Bw
Hence we obtain
1
EI
BT B det(J )dr + (Ak)G
1
(20.10)
=R
Ku
(20.11)
K is a result of the term inside the bracket in (20.10) and R is a result of the right hand side.
For the 2-node element,
w1 = 1 = 0
(20.12)
w2 , 2 = ?
(20.13)
1+r
w2
2
L
2
(20.14)
We cannot make equal to zero for every r (page 404, textbook). Because of this, we need to use
about 200 elements to get an error of 10%. (Not good!)
Recall almost or fully incompressible analysis: Principle of virtual displacements:
T
C dV +
v (v )dV = R
V
(20.15)
u/p formulation
T
C dV
v pdV = R
V
V
p
+ v dV = 0
(20.16)
(20.17)
But now we needed to select wisely the interpolations of u and p. We needed to satisfy the inf-sup
condition
qh vh dVol
>0
(20.18)
inf sup Vol
qh vh
qh Qh vh Vh
86
MIT 2.094
4/1 element:
We can show mathematically that this element does not satisfy inf-sup condition. But, we can also
show it by giving an example of this element which violates the inf-sup condition.
v1 = ,
above
v2 = 0 vh for both elements is positive and the same. Now, if I choose pressures as
qh vh dVol = 0,
(20.19)
Vol
9/3 element
satises inf-sup
9/4-c
satises inf-sup
EI
dx + (AkG)
0
dw
AS dx = R
dx
(20.20)
AS AS dx = 0
(20.21)
dw
,
dx
(20.22)
where
=
MIT 2.094
(20.23)
L
dw
dx =
dx
AS
AS
AS
dx
0
+1
AS =
1+r
2
2
Reading:
Sec. 4.5.7
(20.24)
L
dr + w2 = AS L
2
(20.25)
w2 L2 2
L
(20.26)
AS (shear strain) is equal to the displacement-based shear strain at the middle of the beam.
EI
0
L
2
2
(20.27)
dx = M x=L
(20.28)
hence
w2 =
2
1
L 2 = M
EI
L
2 =
ML
,
EI
w2 =
(20.29)
M L2
2EI
(20.30)
(exact solutions)
88
MIT 2.094
Plates
Reading:
Fig. 5.25,
p. 421
u = zx (x, y)
(20.31)
For any particle in the plate with coordinates (x, y, z), the expressions in (20.31) hold!
We use
w=
hi wi
(20.32)
i=1
x =
y = +
i=1
q
hi yi
(20.33)
hi xi
(20.34)
i=1
where q equals the number of nodes. Then the element locks in the same way as the displacement-based
beam element.
89
Fall 08
MIT OpenCourseWare
A = area of mid-surface
p = load per unit area on mid-surface
w = w(x, y)
(21.1)
w(x, y, z) = w(x, y)
(21.2)
(21.3)
(21.4)
x
u
= z
x
x
y
v
= z
=
y
y
u v
x
y
=
+
= z
+
y
x
y
x
xx =
(21.5)
yy
(21.6)
xy
x
x
xx
yy = z
xy
y
y
x
y
(21.7)
y
x
(21.8)
90
MIT 2.094
w u
w
x
+
=
x
z
x
w v
w
y
=
+
=
y
z
y
xz =
yz
1
xx
E
yy =
1 2
xy
0
1
0
(21.9)
(21.10)
0
0
1
2
xx
yy = C
xy
(21.11)
(plane stress)
xz
yz
E
=
2(1 + )
xz
yz
= G
(21.12)
+ 2t
xx
yy
xy
2t
+ 2t
k
A
2t
1
E
1 2
0
xz
yz
xx
yy dzdA+
1
xy
2
1 0
xz
G
dzdA =
wpdA
0 1
yz
A
1
0
0
0
(21.13)
Consider a at element:
Kb
w1
x1
y1
.
..
, also Kpl.
str.
u1
v
str. 1
...
(21.14)
is 8x8.
91
MIT 2.094
For a at element:
Kb
0
0
Kpl. str.
w1
x1
y1
..
.
4
y =
u1
v1
..
.
v4
(21.15)
u=
hi u i
(21.16)
v=
hi vi
(21.17)
w=
hi wi
x =
hi yi
y =
hi xi
(21.18)
(21.19)
(21.20)
From (21.13)
1
3
Et
T
12 (1 2 )
A
0
1
0
0
0
1
2
dA +
T Gt k
1
0
0
1
dA
(21.21)
w w x
x , y , x ,
This displacement-based element locks in shear. We need to change the transverse shear interpola
tions.
yz =
1
1
A
C
(1 r)yz
+ (1 + r)yz
2
2
(21.22)
where
A
yz
=
w
y
y
evaluated at A
(21.23)
92
MIT 2.094
w
x
w=
1
1
r(1 + r)w1 r(1 r)w2 + 1 r2 w3
2
2
But,
w
1
1
= (1 + 2r)w1 (1 2r)w2 2rw3
r 2
2
w2 and w3 aect w
r 1 . This results in diculties to develop a good
element based on Kirchho theory.
With Reissner-Mindlin theory, we independently interpolate rotations such that this problem does
not arise.
For at structures, we can superimpose the plate bending and plane stress element stiness. For
shells, curved structures, we need to develop/use curved elements, see references.
References
[1] E. Dvorkin and K.J. Bathe. A Continuum Mechanics Based Four-Node Shell Element for General
Nonlinear Analysis. Engineering Computations, 1:7788, 1984.
[2] K.J. Bathe and E. Dvorkin. A Four-Node Plate Bending Element Based on Mindlin/Reissner Plate
Theory and a Mixed Interpolation. International Journal for Numerical Methods in Engineering,
21:367383, 1985.
[3] K.J. Bathe, A. Iosilevich and D. Chapelle. An Evaluation of the MITC Shell Elements. Computers
& Structures, 75:130, 2000.
[4] D. Chapelle and K.J. Bathe. The Finite Element Analysis of Shells Fundamentals. Springer,
2003.
93
MIT OpenCourseWare
http://ocw.mit.edu
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.