Professional Documents
Culture Documents
M. Vable Notes For Finite Element Method: Intro To FEM
M. Vable Notes For Finite Element Method: Intro To FEM
Vable
8-1
M. Vable
Rayleigh Ritz
Rayleigh Ritz method is an approximate method of finding displacements that is based on the theorem of minimum potential energy.
The method is restricted to conservative systems that may be linear or
non-linear.
Potential energy function (): = U W
U is the strain energy
W is the work potential of a force.
Internal virtual work: W int = U
External virtual work: W ext = W
Virtual work principle: W int W ext = U W = = 0
The zero virtual variation in the potential energy function occurs
where the slope of the potential energy function with respect to the
parameters defining the potential function is zero.
The parameters defining the potential function are the generalized displacements.
The zero slope condition can occur where the potential function is:
maximumUnstable equilibrium
saddle pointNeutral equilibrium
minimumStable equilibrium
8-2
M. Vable
u(x) =
Ci fi ( x )
i=1
= 0
Ci
i = 1, n
8-3
M. Vable
U a = --- EA
d x
2
n
Ci fi ( x )
Displacement Approximation: u ( x ) =
i=1
n
du
=
dx
df j
Cj d x
j=1
n
n
df
df k
1
du
du
1
j
Cj
U a = --- EA = --- EA
Ck
2
dx dx
2
dx
dx
j = 1
k = 1
n
n
df j df k
1
Cj Ck
U a = --- EA
2
d x d x
j=1 k=1
1
U A = U a dx = --2
0
j=1 k=1
df j df k
C j C k EA dx
d x d x
0
df j df k
Define: K jk = EA dx ----Stiffness matrix
d x d x
0
EA can be function of x.
K jk = K kj Stiffness Matrix is symmetric
8-4
M. Vable
1
Axial Strain Energy: U A = --2
C j C k K jk
j=1 k=1
Work potential
px(x)
axial distributed force
Fq
concentrated axial forces applied at m xq points
The work potential of these forces can be written as:
L
WA =
px ( x )u ( x ) dx +
0
Fq u ( xq )
q=1
WA =
Cj px ( x )fj ( x )
j=1
px ( x )fj ( x )
Fq Cj fj ( xq )
q = 1j = 1
Define: R j =
dx +
dx +
F q f j ( x q ) ---Load Vector
q=1
n
WA =
Cj Rj
j=1
1
A = U A W A = --2
j=1 k=1
C j C k K jk
Cj Rj
j=1
8-5
M. Vable
1
A = U A W A = --2
C j C k K jk
j=1 k=1
1
--2
C j
j=1 k=1
Note:
C k
C + Cj
K
Ci k
C i jk
C j
1
=
Ci
0
j=1
A
Ci
Cj Rj
= 0
C j
Ci Rj
= 0
i = 1, n
j=1
i = j
ij
n
n
1
--- C k K ik + C j K ji R i = 0
i = 1, n
2
k = 1
j=1
n
1
--- C j ( K ij + K ji ) R i = 0
i = 1, n
2
j = 1
Kij Cj
= Ri
i = 1, n or [ K ] { C } = { R }
j=1
At Equilibrium only
n
n
n
1
A = --- C j K jk C k C j R j
j=1
k=1
j=1
n
1
A = --2
j=1
Cj Rj
j=1
1
C j R j = ---
2
Cj Rj
j=1
8-6
M. Vable
Cj Rj
at equilibrium
j=1
More the degree of freedom, lower is the potential energy (more negative), higher is the strain energy stored in the structure.
With greater degree of freedom the structure becomes less stiff and
hence can store more energy.
In FEM mesh refinement schemes try to get a more uniform distribution of strain energy across the structure.
8-7
M. Vable
d v
1
Bending strain energy density:U b = --- EI zz
2
2
dx
Displacement Approximation:v ( x ) =
Ci fi ( x )
i=1
Work Potential:
py(x)
Fq
Mq
distributed load
concentrated transverse forces applied at m1 xq points
concentrated transverse forces applied at m2 xq points
m1
WA =
m2
py ( x ) ( x ) dx +
0
Fq v ( xq ) +
q=1
2
Mq
dv
(x )
dx q
q=1
2
d f d f
j
k
dx
Stiffness Matrix: K jk = ( EI zz )
2 2
dx dx
0
L
Load Vector:
m1
Rj =
py ( x )fj ( x )
0
dx +
q=1
m1
Fq fj ( xq ) +
Mq
df j
dx
( xq )
q=1
8-8
M. Vable
8.6 For the beam shown in Fig. P8.6 determine one and two parameter
Rayleigh-Ritz displacement solution using the approximation for the
bending displacement given below. Calculate the potential energy function for the one and two parameters.
v(x) =
w kips / in
ix
Ci sin ------2L
i=1
L in
L in
Fig. P8.6
8-9
M. Vable
(i)
i=1
8-10
M. Vable
Lagrange polynomials
Polynomial functions that yield function value at finite number of
points.
Linear Approximation: u ( x ) = C 1 + C 2 x
u1
x
x1
Node 1
Conditions: u ( x 1 ) = u 1
u2
Node 2
x2
u ( x2 ) = u2
x x2
x x1
u ( x ) = u 1 ----------------- + u 2 -----------------
x 1 x 2
x 2 x 1
x x2
Define: L1 ( x ) = -----------------
x 1 x 2
x x1
L 2 ( x ) = -----------------
x 2 x 1
2
u ( x ) = u1 L1 ( x ) + u2 L2 ( x ) =
ui Li ( x )
i=1
L1(x)
L2(x)
x
1
8-11
M. Vable
u2
u1
Node 1
u3
Node 3
Node 2
x2
x3
n
ui Li ( x ) [
Alternative Approach: u ( x ) =
i=1
n
u ( xj ) =
ui Li ( xj ) = uj
i=1
1
Li ( xj ) =
0
i = j
ij
L1 ( x ) = a1 ( x x2 ) ( x x3 )
x x2 x x3
L 1 ( x ) = ----------------- -----------------
x 1 x 2 x 1 x 3
L2 ( x ) = a2 ( x x3 ) ( x x1 )
x x1 x x3
L 2 ( x ) = ----------------- -----------------
x 2 x 1 x 2 x 3
L3 ( x ) = a3 ( x x1 ) ( x x2 )
x x1 x x2
L 3 ( x ) = ----------------- -----------------
x 3 x 1 x 3 x 2
8-12
M. Vable
Node 2
Node 1
Node 3
L1(x)
x
1
L2(x)
x
1
L3(x)
j = 1
( x xj )
-------------------( xi xj )
i = 1, ( n + 1 )
ij
u2
u1
1
element 1
u2
u1
u3
element 1
u3
element 2
u4
u5
element 2
Lagrange polynomials ensure that the function these are used to represent will be continuous at the element ends.
The continuity of the derivative of the function cannot be ensured at
the element end irrespective of the order of polynomials when
Lagrange polynomials are used for representing the function.
8-13
M. Vable
Axial elements
Notation: superscript to designate quantities at element level and no
superscript at the global level.
(2)
(1)
actual structure.
Displacements: u
(e)
(e)
ui Li ( x )
i=1
(e)
du
Axial strains: xx =
dx
(e)
( e ) dL i
ui
dx
i=1
(e)
Axial Stress: xx = E
(e) (e)
(e)
xx = E
( e ) dL i
ui
dx
i=1
(e)
(e) (e)
(e) (e)
= A xx = E A
( e ) dL i
ui
dx
i=1
(e)
In Equations for Rayleigh-Ritz substitute:C i = u i and f i ( x ) = L i ( x )
L
(e)
( e ) ( e ) dL j dL k
Element stiffness matrix: K jk = E A
dx
d
x
d
x
Rj
(e)
(e)
px ( x )Lj ( x ) dx + F1 Lj ( x1 ) + Fn Lj ( xn )
0
For p x = 0
(e)
R1
(e)
= F1
(e)
Rn
(e)
= Fn
(e)
Rj
= 0
j = 2, ( n 1 )
8-14
M. Vable
(e)
(e)
Fj
x
A
P/2
1
Element 1
Element 2
(m)
(m)
(m)
ui
(m)
u i
i=1
n
n
(m)
(m) (m)
(m)
(m)
=
u i
( K ij u j R i )
i=1
j=1
(1)
u ( 1 )
1
= u (21 )
u ( 1 )
3
K( 1 ) K( 1 ) K( 1 )
21 22 23
(1) (1) (1)
K 31 K 32 K 33
u( 1 )
(1)
1 F 1
(1)
u 2 0
( 1 )
u ( 1 ) F 3
8-15
M. Vable
(2)
u ( 2 )
1
= u (22 )
u ( 2 )
3
K( 2 ) K( 2 ) K( 2 )
21 22 23
(2) (2) (2)
K 31 K 32 K 33
u( 2 )
(2)
1 F 1
(2)
u 2 0
( 2 )
u ( 2 ) F 3
P/2
x
A
P/2
1
Element 1
(1)
u1
(2)
u1
Element 2
(1)
= u1
u2
= u3
u2
(2)
(1)
(1)
u ( 1 )
1
= u (21 )
u ( 1 )
3
T
u 1
u 2
u 3
u 4
u 5
(1)
= u2
u3
= u4
u3
(2)
K( 1 ) K( 1 ) K( 1 )
21 22 23
(1) (1) (1)
K 31 K 32 K 33
K( 1 ) K( 1 ) K( 1 )
11 12 13
(1) (1) (1)
K 21 K 22 K 23
(1) (1) (1)
K 31 K 32 K 33
0
0
0
0
0
0
= u3
= u5
u( 1 )
(1)
1 F 1
(1)
u 2 0
( 1 )
u ( 1 ) F 3
0 0
0 0
0 0
0 0
0 0
u1 F( 1 )
1
u2 0
u3 F( 1 )
3
u4 0
u5 0
8-16
M. Vable
(2)
u 1
u 2
u 3
u 4
u 5
0 0
0 0
u 1
u 2
u 3
u 4
u 5
0
0
0
0
(2)
(2)
(2)
(2)
(2)
(2)
(2)
(2)
(2)
0 0 K 11 K 12 K 13
0 0 K 21 K 22 K 23
0 0 K 31 K 32 K 33
0
0
(1)
u1 0
u2 0
(2)
u3 F1
u4 0
(2)
u5 F3
(2)
(1)
K( 1 ) K( 1 )
K 13
0
0
11 12
(1) (1)
(1)
K 23
0
0
K 21 K 22
K( 1 ) K( 1 ) ( K( 1 ) + K( 2 ) ) K( 2 ) K( 2 )
33
11
12
13
31 32
(2)
(2)
(2)
0
K 21
K 22 K 23
0
(2)
(2)
(2)
0
K 31
K 32 K 33
0
(1)
u1
F1
u2
0
u 3 F (31 ) + F (12 )
u4
0
(2)
u5
F3
The element stiffness matrix and element load components that add
corresponds to the degree of freedom associated with the shared node
of the elements.
x
A
P/2
1
Element 1
(1)
The force F 3
(2)
The force F 1
Element 2
(1)
is in the direction of u 3
(2)
is in the direction of u 1
8-17
M. Vable
F1
= RA
T
u 1
u 2
u 3
u 4
u 5
(1)
(2)
F3 + F1
= P
(1)
(1) (1)
K 13
0
0
K 11 K 12
(1) (1)
(1)
K 23
0
0
K 21 K 22
K( 1 ) K( 1 ) ( K( 1 ) + K( 2 ) ) K( 2 ) K( 2 )
33
11
12
13
31 32
(2)
(2)
(2)
0
K 21
K 22 K 23
0
(2)
(2)
(2)
0
K 31
K 32 K 33
0
(2)
F3
= RC
u1
u2
u3
u4
u5
R A
0
P
0
R C
u
2
= u
3
u 4
(1)
K( 1 ) K( 1 )
K 13
0
0
11 12
(1) (1)
(1)
K 23
0
0
K 21 K 22
K( 1 ) K( 1 ) ( K( 1 ) + K( 2 ) ) K( 2 ) K( 2 )
33
11
12
13
31 32
(2)
(2)
(2)
0
K 21
K 22 K 23
0
(2)
(2)
(2)
0
K 31
K 32 K 33
0
0 R A
u 2 0
u 3 P
u 4 0
0 R C
or
T
(1)
K( 1 )
K
0
23
u 2 22
u 2 0
= u K ( 1 ) ( K ( 1 ) + K ( 2 ) ) K ( 2 ) u P
3 32
33
11
12 3
u
(2)
( 2 ) u 4 0
4
K 21
K 22
8-18
M. Vable
(1)
K 22
(1)
(1)
K 23
(1)
0
(2)
(2)
K 32 K 33 + K 11 K 12
0
(2)
K 21
(2)
K 22
u
2
0
=
u
3
P
0
u4
Reaction Forces:
(1)
(1)
R A = K 12 u 2 + K 13 u 3
(2)
(2)
R C = K 31 u 3 + K 32 u 4
8-19
M. Vable
(e) (e)
uj Rj
j=1
For p x = 0
(e)
R1
(e)
(e)
= F1
Rn
(e)
= Fn
(e)
Rj
= 0
j = 2, ( n 1 )
(e)
(e) (e)
1 (e) (e)
U A = --- ( u 1 F 1 + u n F n )
2
(e)
(e)
(e)
(e)
(e)
Fn
(e)
= N1
(e)
Fn
(e)
(e)
u1
un
F1
F1
(e)
(e)
u1
N1
un
(e)
Nn
(e)
= Nn
(e)
(e)
For p x = 0
N1 = Nn = N
(e)
(e) (e)
(e) (e)
1 (e) (e)
1 (e)
U A = --- ( u n N n u 1 N 1 ) = --- ( u n u 1 )N
2
2
Mesh Refinement
Elements with high strain energy identify the region of the body
where mesh should be refined.
The h-method of mesh refinement reduces the size of element.
The p-method of mesh refinement increases the order of polynomial in
an element.
The r-method of mesh refinement relocates the position of a node.
Combinations: hr-method, hp-method, hpr-method
8-20
M. Vable
Coordinate transformation
yG
(1)
v G2
(1)
v G1
(1)
v G3
(1)
u G3
(1)
u2
(1)
u G2
(1)
u3
(1)
u1
(1)
u G1
xG
(e)
(e)
(e)
= u G1 i + v G1 j
Axial displacement: u 1
(e)
(e)
(e)
= D 1 e L = u G1 cos + v G1 sin
(e)
(e)
u G1
u G1
(e)
(e)
v G1
v G1
(e)
u
(e)
(e)
1
cos sin 0
0
0
0 u G2
(e)
u G2
u2 =
0
0 cos sin 0
0 (e) = [T] (e)
v
v
0
0
0
0
cos
sin
G2
(e)
G2
u
3
(e)
(e)
u G3
u G3
(e)
(e)
v G3
v G3
The [T] matrix in the above equation is 3 x 6 matrix relating the local
and the global coordinate system.
For n nodes on the element the matrix [T] would be n x 2n size.
8-21
M. Vable
(e)
(e)
} = [ T ] uG
(e)
= { u
(e) T
} ([K
(e)
]{u
(e)
} {R
(e)
} ) or
(e)
(e)
(e)
T (e)
(e)
= u G [ T ] [ K ] [ T ] u G { R } or
(e)
(e)
(e)
T (e)
T
(e)
= u G [ T ] [ K ] [ T ] u G [ T ] { R } or
Define:
(e)
[ KG ] = [ T ] [ K
(e)
(e)
(e)
T
(e)
RG = [ T ] { R }
][T]
T
( e ) ( e ) ( e ) ( e )
= u G [ K G ] u G R G
8-22
M. Vable
Linear: u
(e)
(e)
(e)
( x ) = u 1 L 1 ( x ) + u 2 L2 ( x )
x x2
L 1 ( x ) = -----------------
x 1 x 2
u( e )
1
(e)
{u } =
u( e )
2
x x1
L 2 ( x ) = -----------------
x 2 x 1
[K
(e)
F( e )
p
L
(e)
o 1 1
{ R } = --------- +
2 1 ( e )
F
2
EA
] = -------- 1 1
L 1 1
Quadratic: u
(e)
(e)
(e)
(e)
( x ) = u 1 L 1 ( x ) + u 2 L2 ( x ) + u3 L3 ( x )
x x2 x x3
L 1 ( x ) = ----------------- -----------------
x 1 x 2 x 1 x 3
u( e )
1
(e)
{ u } = u( e )
2
(e)
u3
x x1 x x3
L 2 ( x ) = ----------------- -----------------
x 2 x 1 x 2 x 3
x x1 x x2
L 3 ( x ) = ----------------- -----------------
x 3 x 1 x 3 x 2
[K
(e)
F( e )
1 1
po L
{ R } = --------- 4 + 0
6
1 F( e )
3
7 8 1
EA
] = -------- 8 16 8
3L
1 8 7
(e)
= { u
(e) T
} ([K
(e)
]{u
(e)
} {R
(e)
})
8-23
M. Vable
8-24
M. Vable
1 in
4 in A
h(x)
x
10 in
P = 10 kips
50 in
Fig. P8.8
8-25
M. Vable
110o
A
P
Fig. P8.10
8-26
M. Vable
dv
dx
must be
Deflection Approximation: v ( x ) = C 1 + C 2 x + C 3 x + C 4 x
(e)
(e)
dv
( x1 ) = 1
dx
(e)
(e)
dv
( x2 ) = 2
dx
v ( x1 ) = v1
Conditions:
v ( x2 ) = v2
(e)
(e)
(e)
(e)
v ( x ) = f 1 ( x )v 1 + f 2 ( x ) 1 + f 3 ( x )v 2 + f 4 ( x ) 2
x x1 2
x x1 3
f 1 ( x ) = 1 3 -------------- + 2 --------------
L
L
x x1
x x1 3
x x1 2
f 3 ( x ) = 3 -------------- 2 --------------
L
L
x x1 2
x x1 3
f 4 ( x ) = L -------------- + --------------
L
L
8-27
M. Vable
(e)
(e)
(e)
(e)
v ( x 1 ) = v 1 = f 1 ( x 1 )v 1 + f 2 ( x 1 ) 1 + f 3 ( x 1 )v 2 + f 4 ( x 1 ) 2
f1 ( x1 ) = 1
f2 ( x1 ) = 0
f3 ( x1 ) = 0
f4 ( x1 ) = 0
df 1
( e ) df 2
( e ) df 3
( e ) df 4
(e)
dv
( x1 ) =
( x 1 )v 1 +
( x 1 ) 1 +
( x 1 )v 2 +
( x 1 ) 2
dx
dx
dx
dx
dx
(e)
= 1
df 1
dx
( x1 ) = 0
df 2
dx
( x1 ) = 1
df 3
dx
( x1 ) = 0
df 3
dx
( x1 ) = 0
Node 2
zero slope
f1(x)
f2(x)
zero slope
x
x
zero slope
Node 2
Node 1
f3(x)
x
1
8-28
M. Vable
d f d f
j
k
dx
K jk = ( EI zz )
2 2
dx dx
0
L
Assume:
the area moment of inertia (I) is evaluated for the cross-section at the
mid point of the element.
the modulus of elasticity (E) is constant over the element.
(e)
v1
(
e
)
1
(e)
{v } =
(
e
)
v
2
(e)
2
6
[K
(e)
2EI
] = --------3
L
3L
3L
3L 2L 3 L L
6 3 L 6 3 L
3L L
3 L 2L
Rj =
py ( x )fj ( x )
dx +
m1
Fq fj ( xq ) +
q=1
Mq
df j
dx
( xq )
q=1
Assume point forces and moments can only be applied at element end
nodes.
L
(e)
Rj
(e)
(e)
( e ) df j
py ( x )fj ( x ) dx + F1 fj ( x1 ) + F2 fj ( x2 ) + M1
dx
( e ) df j
( x1 ) + M2
dx
( x2 )
8-29
M. Vable
the distributed load py is evaluated at the mid point of the element and
has a uniform value po.
(e)
F1
6
(
e
)
M
p
L
(e)
0 L 1
{ R } = --------- +
12 6 ( e )
F
L 2
(e)
M2
(e)
(e)
F1
v2
(e)
(e)
2
M1
(e)
v1
(e)
1
(e)
= { v
(e) T
} ([K
(e)
]{v
(e)
(e)
M2
} {R
(e)
(e)
F2
})
8-30
M. Vable
8.23 Using a single beam element for AB and a single beam element for
BC in Fig. P8.23, determine (a) the slope at B and C (b) reaction force
and moment at A.
w
A
2L
Fig. P8.23
8-31