You are on page 1of 26

Basic Matrix Operations

•Transpose: [ A] T , A ijT = A ji
{ x} T =  x
( [ A][ B] ) T = [ B] T [ A] T

•Symmetric matrix: [ A] = [ A] T
1 0 . . 0
 0 1 . . 0
 
•Identity matrix: [ I] =  . . . . . 
 
 . . . . . 
0 0 . . 1
Determinants
N N
•Definition:: det[ A ] = ∑ A ri cof ( A ri ) = ∑ A iccof ( A ic )
i =1 i =1
( )
where cof A ij = ( − 1) i + j M ij
•Properties: det[ A ] = det[ A ] T
det[ A ] = 0, if [ A ] is sin gular .
det ( [ A ][ B ] ) = det[ A ] det[ B ]

Matrix Inverse

•Definition: [ A] −1
=
adj[ A ]
det[ A ]
( )
, adj[ A ] = cof [ A ] T

•Property: [ A][ A] −1 = [ A] −1 [ A] = [ I ]
System of Linear Algebraic Equations
•Definition: A11x1 + A12 x 2 + . . A1N x N = b1
A 21x1 + A 22 x 2 + . . A 2N x N = b2
. . . . . = .
. . . . . = .
A N1x1 + A N 2 x 2 + . . A NN x N = bN
or
[ A]{ x} = { b}
•Matrix inverse: { x} = [ A] −1{ b}
•Cramer’s rule: Ex. 9.6.2

•Gauss elimination: Ex. 9.7.1


Introduction to the Displacement (Stiffness) Method
Nodal
Force Nodal
Node 1 Node 2 displacement
•Spring element: X1 , u1 X2 , u2
Axial motion only
•Force equilibrium (linear elastic): X1 = k ( u1 − u 2 ) = − X 2
ΣFx = X1 + X2 = 0 X1 .1 k(u1 - u2) .2 X2
•In matrix notation:  X1   k − k  u1 
 =  
X2  − k k  u 2 
or
{ Q} = [ k ]{ q}
Element
Element displacement
load vector Element stiffness vector (unknown)
(known) matrix (known)
•Primary unknown: displacement
Example: Assembly of the Stiffness Matrix
1 1 2 2 3
X1 , u1 X3 , u3
k1 X2 , u2 k2
•Step 1: Define the element stiffness matrices.
[K ] = k −11
1
1
− 1
1 
, [K ] = k
2  1 − 1
2 
− 1 1 

•(Step 2: Align the element coordinates with the global coord.)


•Step 3: Expand into the global DOF.
 1 − 1 0 0 0 0   k1 − k1 0 
[ K ] = k 1 − 1 1 0 + k 2 0 1 − 1 = − k 1 k 1 + k 2 − k 2  Banded:
 0 Bandwidth
0 0 0 − 1 1   0 − k2 k 2 
of 3
•Force equilibrium and displacement compatibility is maintained.
Example: Spring force calculation
1 1 2 2 3
R P
k1 k2
•The stiffness equation:  k1 − k1 0  u1   X1  R 
− k k + k      
Either displacement or load  1 1 2 − k 2  u 2  =  X 2  =  0 
(not both) is known at each DOF  0 − k2 k 2  u 3   X 3   P 
•Apply the prescribed displacement or the boundary condition
u1 = 0
•The stiffness equation becomes
k 1 + k 2 − k 2  u 2   0 
 −k  = 
 2 k 2  u 3  P 

•Solve the system of equations to get (the fundamental solution)


P P P
u2 = , u3 = +
k1 k1 k 2
Example: Spring force calculation (cont.)
1 1 2 2 3
R P
k1 k2
•The stiffness equation:  k1 − k1 0  u1   X1  R 
− k k + k      
 1 1 2 − k 2  u 2  =  X 2  =  0 
 0 − k2 k 2  u 3   X 3   P 
P
•The reaction: R = k 1u 1 + ( − k 1 ) u 2 = ( − k 1 ) = −P
k1
•Internal spring forces: X1 X2 = P
•Spring 1: P
1 1 1
 X1  k − k  u 1   k1 − k 1  u 1   k 1 − k 1  0  − P 
  =   =  =  = 
X 2  − k k  u 2  − k 1 k 1  u 2   − k 1 k 1  P k 1   P 
•Spring 2:
2 2 2 P k 1 
 X1   k − k  u 1   k2 − k 2  u 2   k 2 − k 2   − P 
  =  − k k  u  = − k   =  P P = 
X
 2    2  2 k 2  u 3   − k 2 k 2  +   P 
k
 1 k 2
Solution Procedure Using the FE Displacement Method

•Assemble the stiffness equation for the problem.


•Define the element stiffness matrices
•Align them with the global coordinate using a coordinate
transformation
•Assemble them into the global stiffness matrix
•Define the load vector
•Apply prescribed displacements or displacement boundary
conditions

•Solve the equation for the unknown displacements.


•Conduct the post-processing as needed.
•Internal element loads and stresses
•Support reactions
Basic Formulation of the FEM
•The constitutive equations in a matrix form

{ ε} = [ D]{ σ} + { εT } or { σ} = [ E] ( { ε} − { εT } )
where
 σx   εx  αT 
σ  ε  αT 
 y  y  
σ  ε  αT 
{σ} =  z , { ε} =  z , { εT } =  
τxy   γ xy   0 
 τxz   γ xz   0 
     

 τ yz 
 
 γ yz 
  0 

 1 ν ν 
 E − − 0 0 0 
E E
 ν 1 ν 
− − 0 0 0 
 E E E 
− ν −
ν 1
0 0 0 
[ D] =  E E E
2( 1 + ν )

 = [ E]
−1

 0 0 0 0 0 
 E 
 0 2( 1 + ν )
0 0 0 0 
 E 
 2( 1 + ν ) 
 0 0 0 0 0
E 
Basic Formulation of the FEM (cont.)
•Stiffness equation: Q j+1
Finite element
Structure with N DOF Qj

y,v
Grids q i +1
x,u
qi
The stiffness equation
[ K ]{q} = {Q} relates the applied loads to
the nodal displacements.

•Shape functions (assumed): Nodal displacements

{ue } = [Ne ]{qe } Shape functions relate the


interior displacements of a
Displacements Shape
within the element functions finite element to the nodal
displacements approximately.
Basic Formulation of the FEM (cont.)
{ } [ ]{ }
•For a 2D finite element with N DOF (N/2 nodes): u e = N e q e

{ }
e u ( x , y ) 
u =
v ( x , y )  [ ]
e N11 ( x, y ) N12 ( x, y ) . . . N1N ( x, y ) 
N = 
 N 21 ( x, y ) N 22 ( x, y ) . . . N 2N ( x, y )
 
•Strain-displacement relationship assuming a plane stress condition:
∂u  ∂N11 ∂N12 ∂N1N  e
εx = = . . . { q}
∂x  ∂x ∂x ∂x 
∂v  ∂N 21 ∂N 22 ∂N 2 N  e
εy = = . . . {q}
∂y  ∂y ∂y ∂y 
γ xy =
∂u ∂v  ∂N11 ∂N 21 
+ = 
∂ x ∂ y   ∂y
+ 
∂x 
 ∂N12 ∂N 22 

 ∂y
+
∂x 
 ∂N
 . . .  1N +
 ∂y
∂N 2 N  
∂x   { }
 q e

or
{ε } = [B ]{q }
e e e
Material
properties
(E, G, ν)
Shape
functions
•Stress-displacement relationship (approx.)

{σe } = [Ee ] {εe } = [Ee ][Be ]{qe }


Basic Formulation of the FEM (cont.)
•For { δq} ≠ 0



[ ] [ ][ ] { } { }
T 
 ∫∫∫ Be Ee Be dV  qe = Qe

 V 
or
[K ]{q } = {Q }
e e e

where the element stiffness matrix is defined as

[K ] = ∫∫∫[B ] [E ][B ]dV


e eT e e
e
Ve

•The illustrated procedure is applicable to general problems.


Orthogonal Coordinate Transformations
• Vector v y,j y,j
• in xyz coord.: v = vxi + v y j + v zk
• in xyz coord.: v = vx i + v y j + v z k
v
x, i
• Coordinate transformation
x,i
 v x   cos φxx cos φxy cos φxz   v x 
    
 v y  = cos φyx cos φyy cos φyz   v y  z,k
 v   cos φ cos φzy cos φzz   v z 
 z  zx
z,k
 v x   l1 m1 n1   v x 
    
 v y  = l 2 m2 n 2   v y  or { v} = [ R ]{ v}

 v  l m3 n 3   v z 
 z  3
• Properties of the coordinate transformation matrix, [R]
[ R ][ R ] = [ R ] [ R ] = I
T T

[ R ] −1 = [ R ] T (Orthogonal matrix)
Element Stiffness Matrix for a Rod x u

X1 , u1 X 2 , u2
•Apply the boundary conditions, u( 0) = u1 , u( L ) = u 2 1 2
L
u( 0) = co = u1 co = u1
u( x ) = co + c1x
u( L ) = co + c1L = u 2 (
c1 = u 2 − u1 / L )
 x x Shape function
u( x ) =  1 − u1 + u 2 for a rod element
 L L

u( x ) = 1 −
x x   u1 
 
 L L  u 2 
[ ]
u( x ) = Ne q 
e
 
•The element stiffness matrix for a rod:  K  = [B ] [E ][B ] dV
e eT e e
  ∫∫∫ e

[N ] = − 1
Ve

[B ] = ddx [E ] = E
e
e 1 e
,
 L L Element stiffness matrix
 1 in the local coord.
L −
K e  = L E − 1 1 AE  1 − 1
  ∫  1   L  Adx =
0  L L  − 1 1 
 L 
Element Stiffness Equation for a Rod in the Global Coordinate
Y2 , v 2
• 2D Example X 2 , u2

cal X 2 , u2
Lo 2

Y1 , v1
X1 , u1 θ
Global
X1 , u1
1
•The coordinate transformation for the displacement:
 π    u1 
cos θ cos − θ  0 0   v 
 u1    2 
 =  1 
u 2   0 π  u
0 cos θ cos − θ    2 
 2    v 2 
[ ]{ }
q e  = Λe q e
 
 
 u1 
cos θ sin θ 0 0   v1 
=  
 0 0 cos θ sin θ u 2 
 v 2 
Element Stiffness Equation for a Rod in the Global Coordinate
(cont.)
•Coordinate transformation for the load:
 X1  cos θ 0 
 Y   sin θ 0   X1 
 1 
 =
X 2   0
 
cos θ  X 2 
{Q } = [Λ ]
e e T e
Q 
 
 Y2   0 
sin θ 

•Substitute the stiffness equation in the local coord, Q  =  K  q 


e e e
     

{Q } = [Λ ]
e e T
K  q 
e
   
e

q e  = Λe q e
•Substitute the displ. coord. transformation,  
 
[ ]{ }
{Qe } = [Λe ]T K e  [Λe ] {qe }
or

{Qe } = [K e ] {qe } Element stiffness equation in the global coord.


Example: Problem 10.1
1m 0.8 m
A1 = 10 −4 m 2
1 2 A 2 = 0.25x10− 4 m 2
E = 70x109 N / m 2
1 2,000 N
2 3
[ K ] = [Ω1 ] [K ][Ω ] + [Ω ] [K ][Ω ]
T 1 1 2T 2 2
Step1: Assemble the stiffness equation
•Define the element stiffness matrices:  9187500 - 7000000 - 2187500 
[ ] 1  A E   1 − 1
K 1 = K  =  1  
   L1   − 1 1 

=  - 7000000 7000000
 - 2187500 0
0 

2187500

[K ] = K
2
=
 A 2E   1 − 1
2
 
  L   − 1 1 
 2  
• Define the system load vector:

•Assemble system stiffness matrix: •Apply the BC’s: 2000 


 u1  {Q} =  X 2 
{ } u  
q1 =  2  = 
0 1 0
u1  1 0 0 u 


 
[ ]
u 2  = Ω { q}
1
 X3 
 
 3 Step2: Solve the stiffness equation:
 u1  9,187,500 u1 = 2,000
{ }  u  
q2 =  1  = 
1 0 0 

 
[ ]
u 2  = Ω { q}
u 3  0 0 1 u 
2
u1 = 2.1769e - 4 m
 3
Example: Problem 10.1 (cont.)
1m 0.8 m
A1 = 10 −4 m 2
1 2 A 2 = 0.25x10− 4 m 2
E = 70x109 N / m 2
2 1 2,000 N
3
Step 3: Post-processing

•Stresses:
[ ][ ]
σ1 = E B1 Ω1 { q} = 1.5238e7 N/m 2
σ = E[B ][Ω ]{ q} = - 1.9048e7 N/m
2 2 2 2

where
[B ] = − L1 L1  and [B ] = − L1
1
1 1
2
2
1 

L2 
•Reaction forces:
X 2 
{ Qs } =   = [ K sf ]{ qf } + [ K ss ]{ qs }
X
 3
 - 7,000,000 7,000,000 0  0 - 1523.8 
=  ( 2. 1769e − 4 ) +   0 =  - 476.2 ( N )
 - 2,187,500   0 2,187 ,500    
A Rod Element with a Higher Order Polynomial Shape Function:
The p-version Finite Element x u

•A rod element with constant cross sectional area (A) X1 , u1 X 2 , u2 X 3 , u3


and homogeneous, isotropic, and linear elastic 1 2 3
material and with three nodes:
L
 u1   X1 
•Element displacement and load vectors, q e  =   Qe  =  
respectively, in the local coordinate:   u 2    X 2 
       
u 3  X 3 
2
•Assume a quadratic axial displacement within the rod as: u( x ) = c o + c1x + c 2x

•Apply the boundary conditions, u( 0 ) = u1 , u( L / 2) = u 2 , and u( L ) = u 3


u( 0) = co = u1
2 co = u1
L
u( x ) = c o + c1x + c 2x 2 L
u( L / 2) = co + c1 + c 2   = u 2
2 2
( )
c1 = − 3u1 + 4u 2 − u 3 / L
c1 = ( 2u1 − 4u 2 + 2u 3 ) / L2
u( L ) = c o + c1L + c 2L2 = u 3
 u1 
 3x 2x 2
[ ]
2 2
4x 4x x 2x   
u( x ) = Ne q 
e
u( x ) = 1 − + − − +  u 2 
 L L2 L L2 L L2     
u 3 
A Rod Element with a Higher Order Polynomial Shape Function:
The p-version Finite Element (cont.)
•The element stiffness matrix for a rod: K e  =
  ∫∫∫
[ ] [ ][ ]
B eT e
E Be dVe
V
[N ] = − 3 + 4x
e

[B ] = ddx [E ] = E
e
e 4 8x 1 4x  e
− − + ,
2 L L2 L L2 
 L L
 3 4x 
− L + 2 
 L   
L
K e  = AE  − x   − 3 + 4x
4 8 4 8x

1 4x 
− +  dx =
AE  
  ∫  L L2   L 2 L L2 L L2  L  
0  L
 
 1 4x 
− L + 2 
 L 
Pros and Cons of Using Higher Order Polynomial Shape Functions

Pros Cons

• Better approximation of element •Complicated element stiffness


interior displacement for rods with matrix definition with a larger
non-uniform cross-sections. number of grids per element.
•Smaller number of finite elements •Unnecessary complication for a rod
to achieve the same accuracy. with a constant cross-section.
Stiffness Matrix for a Beam Element y
2
θ z1 θz2
•Shape function to satisfy the 4 boundary
conditions ( v 1 θ z1 v 2 θ z 2 ) 1
v(x)
v2
v1

v( x ) = a 0 + a1 x + a 2 x 2 + a 3 x 3
x
x
L
•Apply the BC’s to find:
dv dv
v( 0 ) = v 1 = a 0 , θ z ( 0) = = θ z1 = a 1 , where = a 1 + 2a 2 x + 3a 3 x 2
dx x = 0 dx
dv
v ( L ) = v 2 = a 0 + a1L + a 2 L2 + a 3 L3 , θz ( L) = = θ z 2 = a 1 + 2a 2 L + 3a 3 L2
dx x = L

•Solve for the unknown coefficients ( a 0 a 1 a 2 a 3 ) to get:


a 0 = v1 v ( x ) = { N }{ q}
a 1 = θ z1 where  2 x 3 3 x 2 
3 2 3 1  L3 − + 1
a2 = − v 1 − θ z1 + v 2 − θz2 L2
 x3 2 x2   v1 
L2 L L2 L  2− + x θ 
L   z1 
a3 =
2
v +
1
θ −
2
v +
1
θ { N } T
=  3
L
2  , { q} =  
3 1 2 z1 3 2 2 z2 2 x
− 3 + 2  3 x  v 2 
L L L L
 L L  
 θ z 2 
Substitute  x 3
x 2


 L2 L 
v( x ) = a 0 + a1 x + a 2 x 2 + a 3 x 3
Stiffness Matrix for a Beam Element (cont.)
•The stiffness matrix becomes: •For a constant EI z

 12 6L − 12 6L 
 6L 4L2 − 6L 2L2 
[ K ] = ∫ EI z {B̂}T {B̂}dx
L
[ K ] = EI3z  
0 L  − 12 − 6L 12 − 6L 
 2 
 6L 2L − 6L 4L2 
•Local element load vector corresponding to {q}: Symmetric [K]

{ Q} = { Y1 M z1 Y2 M z2 } T , { q} = { v 1 θ z1 v2 θz2 } T

y
θz2
2
θ z1
y
1 Y2 , v 2
Y1 , v 1 v(x)
x z
M z1 x M z2
L
A cross-section
Summary
y θz2
{ Q} = [ K ] { q}
2
θ z1
 12 6L − 12 6L 
1 Y2 , v 2  2
− 6L 2L2 
EI z  6L 4L
Y1 , v 1 v(x)
[K] = 3
x L  − 12 − 6L 12 − 6L 
M z2  2 
M z1 x
 6L 2L − 6L 4L2 
L

y  Y1   v1 
M  θ 
z
{ Q} =  z1 , { q} =  z1 
 Y2   v2 
M z 2  θ z 2 
A cross-section

Useful Relationships for a Beam Element


•The axial stress:
 d 2v 
σ x = Eε x = E  − y 2  = − Eyv′′
 dx 
•The moment resultant at a cross-section:

{ }
M z = − ∫∫ yσ x dA = ∫∫ Ey 2v′′dA = EI z v′′ = EI z Bˆ { q}
A A
Distributed Load Definition for the Beam Element
p( x ) = p 1 +
( p 2 − p1 ) x
•Linearly varying load:
L
p( x ) : Force / Length

p2 Y1 Y2
p1

2 M z1 M z2
1 1 2
L

•Equivalent nodal loads:


L  2x 3 3 x 2  
∫  − + 1 p( x ) dx    7p 1 3p 2  


0 L 3
L 2    + L   pL 
 20 20
 L 3       2 
2 
  x − 2x + x p( x ) dx    p 1 p 2  2   2 
For p 1 = p 2 = p
 ∫  L2    + L   pL 
{ }
 0
Q eq = 
L  


= 
20 30  

{Qeq } 
=  12 
pL

L 3 2  3p 7 p   
  2 x 3 x p( x ) dx    1 + 2
L   2 

  − +    20 20  
 0 L
3
L2    p
 pL2 
−
1 p2  2   12 

 L x 3
x 
2   − − L 
 ∫  − p( x ) dx    30 20  

 0  L 2 L   
Distributed Load Definition for the Beam Element (Cont.)
•Concentrated load:

P
Y1 Y2

a 2 M z1 M z2
1 1 2
L

L  2x 3 3 x 2  
∫  − 
+ 1 Pδ(a )dx   2a 3 3a 2 

0 L 3
L 2    3 − 2 + 1
 L L 
 L 3 2  
  x − 2x + x  Pδ(a )dx  a 3
2a 2   P/2 

  L2 L    − + a  For a = L / 2  PL / 8 
{
Q eq =  }
 0
L 3 2
 

 2
= P L 3
L
2 

{ } 
Q eq = 
P / 2


  2 x 3 x Pδ(a )dx  − 2a 3 a   
∫ − + +
    L3 L2  − PL / 8
 0 L
3
L2    3 2
 L x 3 x 2    a −a 
 ∫  − Pδ(a )dx   L2 L 

 0  L 2 L  

2-D Frames (Sect. 10.4)
•Rod element (axial stiffness only): •2D frame element (both
axial and bending stiffness):
1 − 1
[ K ] rod = AE 
L  − 1 1 
 AE AE 
 L 0 0 − 0 0 
L
 12EI z 6EI z 12EI z 6EI z 
 0 − 
•Beam element (bending stiffness only):  L3 L2 L3 L2 
 4EI z 6EI z 2EI z 
 0 − 
 12 6L − 12 6L  [ ]
K =

L
AE
L2 L 
 6L 4L2 − 6L 2L2  0 0 
EI z  
[ K ] beam =   
L
12EI z 6EI z 
L3  − 12 − 6L 12 − 6L   − 
 2   L3 L2 
 6L 2L − 6L 4L2   4EI z 
sym . L 
•A 2-D frame element in the local coordinates:
y

Y1 , v 1 Y2 , v 2
EI z : bending stiffness
X1 , u 1 EA : axial stiffness
X2 ,u2
x
M z1 , θ z1 1 M z2 , θ z2 2
L

You might also like