FINITE ELEMENT METHOD

Dr. NGUYỄN HỒNG ÂN

Head
Department of Mechanics of Materials & Structures
Tel: 0909.48.58.38
Email: [email protected]
Summary:

• Stiffness matrix of a bar/truss element

• Coordinate transformation
• Stiffness matrix of a truss element in 2D space
•Problems in 2D truss analysis (including multipoint
constraints)
•3D Truss element
Trusses: Engineering structures that are composed only
of two-force members. e.g., bridges, roof supports
Actual trusses: Airy structures composed of slender
members (I-beams, channels, angles, bars etc) joined
together at their ends by welding, riveted connections or
large bolts and pins

A typical truss structure

Gusset plate
Ideal trusses:
Assumptions
• Ideal truss members are connected only at their ends.
• Ideal truss members are connected by frictionless pins (no
moments)
• The truss structure is loaded only at the pins
• Weights of the members are neglected

A typical truss structure

Frictionless pin
These assumptions allow us to idealize each truss
member as a two-force member (members loaded only
at their extremities by equal opposite and collinear
forces)
member in
compression

member in
tension

Connecting pin
 FEM analysis scheme
Step 1: Divide the truss into bar/truss elements connected to
each other through special points (“nodes”)

Step 2: Describe the behavior of each bar element (i.e. derive its
stiffness matrix and load vector in local AND global coordinate
system)

Step 3: Describe the behavior of the entire truss by putting

together the behavior of each of the bar elements (by assembling
their stiffness matrices and load vectors)

Step 4: Apply appropriate boundary conditions and solve

 Stiffness matrix of bar element

E, A

© 2002 Brooks/Cole Publishing / Thomson Learning™

L: Length of bar
A: Cross sectional area of bar
E: Elastic (Young’s) modulus of bar
û(x̂) :displacement of bar as a function of local coordinate x̂ of bar
The strain in the bar at x̂ dû
ε(x̂) =
dx̂
The stress in the bar (Hooke’s law)
 (x̂) = E ε(x̂)
 d̂ 2x
Tension in the bar
T(x̂) = EAε  x̂  x̂
x̂ û(x̂) = 1 − d̂1x + d̂ 2x
x̂ d̂1x  L L

L
Assume that the displacement û(x̂) is varying linearly along the bar
 x̂  x̂
û(x̂) = 1 − d̂1x + d̂ 2x
 L L
dû d̂ 2x − d̂1x
Then, strain is constant along the bar: ε= =
dx̂ L
Stress is also constant along the bar:
E
(
 = Eε = d̂ 2x − d̂1x
L
)
Tension is constant along the bar: T = EAε =
EA
L
(
d̂ 2x − d̂ 1x )

k

The bar is acting like a spring with stiffness k = EA

L
Recall the lecture on springs
E, A

© 2002 Brooks/Cole Publishing / Thomson Learning™

Two nodes: 1, 2
Nodal displacements: d̂1x d̂ 2x
Nodal forces: f̂1x f̂ 2x
Spring constant: k = EA
L
Element stiffness matrix in local coordinates
f̂1x   k - k  d̂1x 
f̂ = k̂ d̂  =   
f̂ 2x  - k k  d̂ 2x 
Element force Element nodal     
Element displacement f̂ k̂ d̂
vector
stiffness vector
matrix
What if we have 2 bars?
E1, A1
E2, A2

L2
L1

This is equivalent to the following system of springs

E1A1 E 2A 2
k1 = k2 =
L1 L2
x
Element 1 2 Element 23
1
d1x d2x d3x

PROBLEM
Problem 1: Find the stresses in the two-bar assembly loaded as
shown below
E, 2A
E, A
1 P
2 3
L
L
Solution: This is equivalent to the following system of springs
2EA EA
k1 = k2 =
L L x
Element 1 2 Element 23
1
d1x d2x d3x

We will first compute the displacement at node 2 and then the

stresses within each element
 The global set of equations can be generated using the technique
developed in the lecture on “springs”
 k1 −k1 0   d1x   F1x 
 −k k + k     
 1 1 2 −k2  d 2 x  =  F2 x 
 0 −k2 k2   d3 x   F3 x 
here d1x = d3 x = 0 and F2 x = P

Hence, the above set of equations may be explicitly written as

− k1d 2 x = F1x (1)
(k1 + k2 )d 2 x = P (2)
− k2 d 2 x = F3 x (3)
P PL
From equation (2) d 2 x = =
k1 + k2 3EA
To calculate the stresses:
For element #1 first compute the element strain
d 2 x − d1x d 2 x P
 (1)
= = =
L L 3EA
and then the stress as
P
 (1)
= E (1)
= (element in tension)
3A
Similarly, in element # 2
d3 x − d 2 x d2 x P
 =
(2)
=− =−
L L 3EA
P
 = E = −
(2) (2)
(element in compression)
3A
© 2002 Brooks/Cole Publishing / Thomson Learning™

Inter-element continuity of a two-bar structure

Bars in a truss have various orientations

member in
compression

member in
tension

Connecting pin
 d 2y , f 2y
x̂
dˆ 2y , fˆ2y = 0 d̂ 2x , f̂ 2x
y ŷ
d 2x , f 2x
d1y , f1y
dˆ 1y , fˆ1y = 0 θ
d̂1x , f̂1x d1x , f1x

x
At node 1: At node 2:
d̂1y d̂ 2y

d1y d 2y
θ d̂1x θ d̂ 2x
d1x d 2x f̂ 2y = 0
f̂1y = 0

f1y f 2y
θ f̂1x θ f̂ 2x
f1x f 2x
 In the global coordinate system, the vector of nodal
displacements and loads
d1x  f1x 
d  f 
 1y   1y 
d =  ; f = 
d f
 2x   2x 
d 2y  f 2y 

Our objective is to obtain a relation of the form

f = k d
41 44 41

Where k is the 4x4 element stiffness matrix in global coordinate

system
 The key is to look at the local coordinates
x̂
y ŷ dˆ 2y , fˆ2y = 0
d̂ 2x , f̂ 2x

f̂1x   k - kd̂1x 
θ  =   
dˆ 1y , fˆ1y = 0

f̂ 2x 
 - k k  
d̂ 2x 

d̂1x , f̂1x
EA
x k=
L
Rewrite as f̂ 
k 0 - k 0 d̂1x 
    
1x

f̂1y   0 0 0 0 d̂1y  f̂ = k̂ d̂
 =  
f̂ 2x  - k 0 k 0 d̂ 2x 
f̂   0  
0 0 0 d̂
 2y   2y 
NOTES

1. Assume that there is no stiffness in the local ^y direction.

2. If you consider the displacement at a point along the local x

direction as a vector, then the components of that vector along the
global x and y directions are the global x and y displacements.

3. The expanded stiffness matrix in the local coordinates is

symmetric and singular.
NOTES
5. In local coordinates we have f̂ = k̂ d̂
41 44 41

But or goal is to obtain the following relationship

f = k d
41 44 41

Hence, need a relationship between d̂ and d

and between f̂ and f d̂1y
d̂1x  Need to understand
d1x   
d1y
how the components
d  θ d̂1x
  d̂1y  of a vector change
d =  1y  d̂ =   d1x
d̂ 2y with coordinate
d 2x
  d̂ 2x  transformation
d 2y  d̂  d 2y
 2y  θ d̂ 2x
d 2x
 Transformation of a vector in two dimensions

y vx v y cos θ Angle q is
ŷ x̂
v̂ x measured positive
v̂ y
θ in the counter
v v x sin θ clockwise direction
vy
from the +x axis)
θ
v y sin θ x
v x cos θ

The vector v has components (vx, vy) in the global coordinate system
and (v^x, v^y) in the local coordinate system. From geometry
v̂ x = v x cos θ + v y sin θ
v̂ y = − v x sin θ + v y cos θ
In matrix form

 v̂ x   cos θ sin θ   v x 
 =   
 y   − sin θ cos θ   v y 
v̂
Or Direction cosines
 v̂ x   l m   v x  l = cos q
 =  v  where
 y 
v̂ − m l  y m = sin q
Transformation matrix for a single vector in 2D
 l m =
*
T = v̂ T v
*
 relates
− m l 
 v̂ x   v x  are components of the same
where v̂ =   and v =  
 v̂ y   v y  vector in local and global
coordinates, respectively.
Relationship between d̂ and d for the truss element
d̂1y
At node 1 d̂1x  * d1x 
 =T   d1y
d̂1y  d1y  θ d̂1x
d1x d̂ 2y
At node 2 d̂ 2x  * d 2x 
 =T   d 2y
d̂ 2y  d 2y  θ d̂ 2x
Putting these together d̂ = Td d 2x

d̂1x   l m 0 0  d1x 
  
d̂1y  − m l 0 0  d1y  T =
T * 0 
*
 = d  44
 0 T 
d̂ 2x   0 0 l 
m  2x 
d̂   0  d 
0 − m l   2y 
  
2y
T d
d̂
 Relationship between f̂ and f for the truss element
At node 1 f̂1x  * f1x 
f̂1y
 =T  
f̂1y  f1y  f1y
θ f̂1x
f̂ 2y
At node 2 f̂ 2x  * f 2x 
f1x
 =T  
f̂ 2y  f 2y 
f 2y
θ f̂ 2x
Putting these together f̂ = Tf f 2x

f̂1x   l m 0 0  f1x 
  
f̂1y  − m l 0 0  f1y  T =
T * 0 
*
 = f  44
 0 T 
f̂ 2x   0 0 l 
m  2x 
f̂   0  f 
0 − m l   2y 
  
2y
T f
f̂
Important property of the transformation matrix T

The transformation matrix is orthogonal, i.e. its inverse is its

transpose

−1
T =T
T

Use the property that l2+m2=1

 Putting all the pieces together

x̂ f̂ = T f
y ŷ d̂ 2y , f̂ 2y
d̂ 2x , f̂ 2x d̂ = T d
θ f̂ = k̂ d̂
d̂1y , f̂1y
d̂1x , f̂1x
 Tf = k̂ Td
x
(
 f = T k̂ T d

 
−1
)
The desired relationship is f = k d k
41 44 41

k = T k̂ T is the element stiffness matrix in the

T
Where
44 44 44 44 global coordinate system
  l m 0 0  k 0 - k 0
− m l  0
0 0 0 0 0
T=  k̂ = 
 0 0 l m  - k 0 k 0
   
 0 0 −m l  0 0 0 0

 l2 lm − l2 − lm 
 2 
EA  lm m 2
− lm − m 
k = T k̂ T =
T

L  − l 2 − lm l2 lm 
 
− lm − m m 
2 2
lm
Computation of the direction cosines

2 (x2,y2)
x2 − x1 L
l = cos q =
L
y2 − y1 θ
m = sin q =
L 1 (x ,y )
1 1

What happens if I reverse the node numbers?

x1 − x2 L
1 (x1,y1)
l ' = cos q = = −l
L
y −y
m' = sin q = 1 2 = −m θ
L 2 (x ,y )
2 2
Question: Does the stiffness matrix change?
 Example Bar element for stiffness matrix evaluation
© 2002 Brooks/Cole Publishing / Thomson Learning™ E = 30 106 psi
A = 2 in2
L = 60 in
q = 30 3
l = cos 30 =
2
1
m = sin 30 =
 3 3 3 3  2
 − − 
 4 4 4 4 
 3 1 3 1 
k=
(
30 10 6 (2 )  4 ) 4
−
4
−
4  lb
60  3 3 3 3  in
− − 
 4 4 4 4 
 3 1 3 1 
− − 
 4 4 4 4 
 Computation of element strains
© 2002 Brooks/Cole Publishing / Thomson Learning™

Recall that the element strain is

d̂ 1x 
 
d̂ 2x − d̂ 1x 1 d̂ 1y 
ε= = − 1 0 1 0 
L L d̂ 2x 
d̂ 
 2y 
= − 1 0 1 0 d̂
1
L
= − 1 0 1 0 Td
1
L
  l m 0 0 
− m l 0 0 
ε = − 1 0 1 0 
1 d
L  0 0 l m 
 
 0 0 −m l 

= − l − m l md
1
L
d 1x 
d 
 
= − l − m l m 1y 
1
L d
 2x 
d 2y 
Computation of element stresses stress and tension

Recall that the element stress is

 = Eε =
E
L
( )
d̂ 2x − d̂1x = − l − m l md
E
L
Recall that the element tension is

EA
T = EAε =  −l −m l m d
L
Steps in solving a problem

Step 1: Write down the node-element connectivity table

linking local and global nodes; also form the table of
direction cosines (l, m)
Step 2: Write down the stiffness matrix of each element in
global coordinate system with global numbering

Step 3: Assemble the element stiffness matrices to form the

global stiffness matrix for the entire structure using the
node element connectivity table

Step 4: Incorporate appropriate boundary conditions

Step 5: Solve resulting set of reduced equations for the unknown

displacements
Step 6: Compute the unknown nodal forces
Node element connectivity table

ELEMENT Node 1 Node 2

1 1 2
2 2 3

3 3 1

1 2 (x2,y2)
L
El 1 60 El 3
θ
2 60 60 3 1 (x ,y )
1 1
El 2
Stiffness matrix of element 1 Stiffness matrix of element 2
d1x d1y d2x d2y d2x d2y d3x d3y
  d1x   d2x
   
=  d1y =  d2y
(1) ( 2)
k k
  d2x   d3x
   
  d2y   d3y

Stiffness matrix of element 3

There are 4 degrees of
d3x d3y d1x d1y
freedom (dof) per
  d3x element (2 per node)
 
=  d3y
( 3)
k
  d1x
 
  d1y
 (1)
k
Global stiffness matrix
d1x d1y d2x d2y d3x d3y
  d1x
  ( 2)
  d1y k
  d2x
K=  d2y
 
  d3x
( 3)
k
  d3y
  66

How do you incorporate boundary conditions?

 Example 2 The length of bars 12 and 23 are equal (L)
y E: Young’s modulus
3 A: Cross sectional area of each bar
El#2 P2
Solve for
P1 (1) d and d
2x 2y

El#1 (2) Stresses in each bar

2
45o
x
1 Solution

Step 1: Node element connectivity table

ELEMENT Node 1 Node 2

1 1 2
2 2 3
Table of nodal coordinates
Node x y
1 0 0
2 Lcos45 Lsin45

3 0 2Lsin45

Table of direction cosines

ELEMENT Length x2 − x1 y −y
l= m= 2 1
length length
1 L cos45 sin45
2 L -cos45 sin45
Step 2: Stiffness matrix of each element in global coordinates
with global numbering
Stiffness matrix of element 1
 l2 lm −l 2−lm 
 2 
EA  lm m2 −lm −m 
=
(1)
k
L  −l 2 −lm l2 lm 
 
 −lm − m 2
lm m 2

d1x d1y d2x d2y

 1 1 −1 −1  d1x
 1 1 −1 −1 
EA   d1y
=
2L  −1 −1 1 1  d2x
 
 −1 −1 1 1  d2y
 Stiffness matrix of element 2

d2x d2y d3x d3y

 1 −1 −1 1  d2x
 −1 1 1 −1 
EA   d2y
=
(2)
k
2L  −1 1 1 −1  d3x
 
 1 −1 −1 1  d3y
Step 3: Assemble the global stiffness matrix

 1 1 −1 −1 0 0 
 1 1 −1 −1 0 0 
 
EA  −1 −1 2 0 −1 1 
K=  
2L  −1 −1 0 2 1 −1
 0 0 −1 1 1 −1
 
 0 0 1 −1 −1 1 

The final set of equations is Kd = F

Step 4: Incorporate boundary conditions
 0 
 0 
 
d2 x 
d = 
d 2 y 
 0 
 
 0 

Hence reduced set of equations to solve for unknown

displacements at node 2

EA  2 0   d 2 x   P1 
 = 
2L 
0

2   d 2 y   P2 
Step 5: Solve for unknown displacements

 P1L 
d2 x  
 EA  
 = 
d
 2 y   P2 L 

 EA  
Step 6: Obtain stresses in the elements 0

For element #1:  d1x  0

d 
E 1 1 1 1   1y 
 = −
(1)
−   
L 2 2 2 2  d2 x 
 d 2 y 
E P+P
= (d 2 x + d 2 y ) = 1 2
2L A 2
For element #2: d2 x 
d 
E 1 1 1 1   2y 
 (2) =  − −    0
L 2 2 2 2   d3 x  0
 d3 y 
E P1 − P2
= (d 2 x − d 2 y ) =
2L A 2
 Multi-point constraints

© 2002 Brooks/Cole Publishing / Thomson Learning™

Figure 3-19 Plane truss with inclined boundary

conditions at node 3 (see problem worked out in class)
Problem 3: For the plane truss
P=1000 kN,
y L=length of elements 1 and 2 = 1m
P El#2 3 E=210 GPa
A = 6×10-4m2 for elements 1 and 2
2 = 6 2 ×10-4 m2 for element 3
El#1
El#3
Determine the unknown displacements
45o and reaction forces.
x
1
Solution
Step 1: Node element connectivity table
ELEMENT Node 1 Node 2
1 1 2
2 2 3
3 1 3
Table of nodal coordinates
Node x y
1 0 0
2 0 L

3 L L

Table of direction cosines

ELEMENT Length x2 − x1 y −y
l= m= 2 1
length length
1 L 0 1
2 L 1 0

3 L 2 1/ 2 1/ 2
Step 2: Stiffness matrix of each element in global coordinates
with global numbering
Stiffness matrix of element 1
 l2 lm −l 2−lm 
 2 
EA  lm m2 −lm −m 
=
(1)
k
L  −l 2 −lm l2 lm 
 
 −lm − m 2
lm m 2

d1x d1y d2x d2y

0 0 0 0  d1x
-4 
(210 10 )(6 10 ) 0 1
9
0 −1  d1y
=
1 0 0 0 0  d2x
 
0 −1 0 1  d2y
Stiffness matrix of element 2 d2x d2y d3x d3y
1 0 −1 0  d2x
-4 
(210 10 )(6 10 )  0
9
0 0 0  d2y
=
(2)
k
1  −1 0 1 0  d3x
 
0 0 0 0  d3y
Stiffness matrix of element 3
d1x d1y d3x d3y
 0.5 0.5 −0.5 −0.5  d1x
 
(210 109 )(6 2 10-4 )  0.5 0.5 −0.5 −0.5  d1y
=
(3)
k
2  −0.5 −0.5 0.5 0.5  d3x
 
 −0.5 −0.5 0.5 0.5  d3y
Step 3: Assemble the global stiffness matrix

 0.5 0.5 0 0 −0.5 −0.5 

 0.5 1.5 0 −1 −0.5 −0.5 
 
 0 0 1 0 −1 0 
K = 1260 10 
5
 N/m
 0 −1 0 1 0 0 
 −0.5 −0.5 −1 0 1.5 0.5 
 
 −0.5 −0.5 0 0 0.5 0.5 

The final set of equations is Kd = F Eq(1)

Step 4: Incorporate boundary conditions y
x
 0  y
 0  3
  P El#2

d 2 x 

d =  2
 0  El#1
 d3 x  El#3
  o

 3y 
d  45
x
1

Also, d 3y = 0 in the local coordinate system of element 3

How do I convert this to a boundary condition in the global (x,y)

coordinates?
 y
x
 F1x  y
F  3
 1y  P El#2

 P  
F =  2
F
 2y  El#1
 F3 x  El#3
 

 F3 y 
 45o
x
1

Also, F 3x = 0 in the local coordinate system of element 3

How do I convert this to a boundary condition in the global (x,y)

coordinates?
 Using coordinate transformations

d 3x   l m   d3 x  1
 =  −m    l=m=
d 3 y 
   l  d3 y  2

 1 1   1 

d 3x   2 2   d  
 2
( d 3x + d 3 y )

  =    3x  =  
  − 1
d 3 y  1  d3 y   1
( d − d ) 

 2 2  2

3 y 3 x



d 3y = 0 (Multi-point constraint)

 d 3y =
1
2
( d3 y − d3 x ) = 0

 d3 y − d3 x = 0 Eq (2)
 Similarly for the forces at node 3

 F 3x 
  l m   F3 x  1
 =  −m    l=m=

F 3 y 
  n   F3 y  2

 1 1   1 
  
 F 3x  2 2   F  
 2
( F3x + F3 y )

  =    3x  =  
 −
1   F3 y 

F 3y 

1

 1
 2
( F3 y − F3 x ) 

 2 2  

F 3x = 0
 F 3x =
1
2
( F3 y + F3 x ) = 0

 F3 y + F3 x = 0 Eq (3)
Therefore we need to solve the following equations simultaneously

Kd = F Eq(1)
d3 y − d3 x = 0 Eq(2)

F3 y + F3 x = 0 Eq(3)

Incorporate boundary conditions and reduce Eq(1) to

1 −1 0  d 2 x   P 
   
1260  105 
 −1 1.5 0.5
  3x  =
d  3x 
F
 0.5 d  F 
0 0.5   3y   3y 
 Write these equations out explicitly

1260 105 ( d 2 x − d 3 x ) = P Eq(4)

1260 105 ( − d 2 x + 1.5d3 x + 0.5d3 y ) = F3 x Eq(5)
1260 105 (0.5d3 x + 0.5d3 y ) = F3 y Eq(6)

Add Eq (5) and (6)

1260 105 (−d 2 x + 2d3 x + d3 y ) = F3 x + F3 y = 0 using Eq(3)

 1260 105 (−d 2 x + 3d3 x ) = 0 using Eq(2)

 d 2 x = 3d 3 x Eq(7)
 1260  105 (3d 3 x − d 3 x ) = P
Plug this into Eq(4)
 2520  105 d 3 x = 106
  d3 x = 0.003968m
d 2 x = 3d3 x = 0.0119m

Compute the reaction forces

 F1x  0 −0.5 −0.5
F  0
 1y 
 −0.5 −0.5 d 
  2x 
 
 F2 y  = 1260  10  0 0   d3 x 
5
0
F   
 −1 1.5 0.5   d
 3y 

 3x


 F3 y 
 
0 0.5 0.5 

 −500 
 −500 

 

=  0  kN
 −500 
 

 500 
Physical significance of the stiffness matrix

In general, we will have a stiffness matrix of the form

 k11 k12 k13 

K = k 21 k 22 k 23 
k 31 k 32 k 33 

And the finite element force-displacement relation

 k11 k12 k13   d1   F1 
k     
 21 k 22 k 23  d 2  = F2 
k 31 k 32 k 33  d 3  F3 
 Physical significance of the stiffness matrix

The first equation is

Force equilibrium
k11d1 + k12 d 2 + k13 d 3 = F1 equation at node 1

Columns of the global stiffness matrix

What if d1=1, d2=0, d3=0 ?

While d.o.f 2 and 3 are held fixed
F1 = k11 Force along d.o.f 1 due to unit displacement at d.o.f 1
F2 = k 21 Force along d.o.f 2 due to unit displacement at d.o.f 1
F3 = k 31 Force along d.o.f 3 due to unit displacement at d.o.f 1

Similarly we obtain the physical significance of the other

entries of the global stiffness matrix
In general

k ij = keeping
Force at d.o.f ‘i’ due to unit displacement at d.o.f ‘j’
all the other d.o.fs fixed
Example The length of bars 12 and 23 are equal (L)
y E: Young’s modulus
3 A: Cross sectional area of each bar
El#2 P2
Solve for d2x and d2y using the “physical
P1 interpretation” approach
El#1 2
45o
x
1 Solution

Notice that the final set of equations will be of the form

 k11 k12   d 2 x   P1 
k   = 
 21 k22  d 2 y   P2 
Where k11, k12, k21 and k22 will be determined using the
“physical interpretation” approach
  k11  d = 1
To obtain the first column
1   apply 2 x

y  2 = 1.cos(45) =
2  k 21  y d2 y = 0
3 F2y=k21
F2y=k21
F2x=k11 T2
El#2
2 F2x=k11
2’
El#1 T1 2
x 1 x
1 1 = 1.cos(45) =
2
d2x=1

Force equilibrium Force-deformation relations

F
EA
x = k11 − T1 cos(45) − T2 cos(45) = 0 T1 = 1
L
F y = k21 − T1 sin(45) + T2 sin(45) = 0 EA
T2 =  2
L
Combining force equilibrium and force-deformation relations
( T1 + T2 ) EA
k11 = = (1 +  2 )
2 2L
(T1 − T2 ) EA
k21 = = (1 −  2 )
2 2L
Now use the geometric (compatibility) conditions (see figure)
1
1 = 1.cos(45) =
2
1
 2 = 1.cos(45) =
2
Finally
EA EA 2 EA
k11 = (1 +  2 ) = ( )=
2L 2L 2 L
EA
k21 = (1 −  2 ) = 0
2L
  k12  d = 0
To obtain the second column
  apply 2 x

y  k22 y d2 y = 1
3 1
 2 = −1.cos(45) = − F2y=k22
2
2’ T2
El#2 d2y=1
2 F2x=k12
El#1 T1 2
x 1 x
1 1 = 1.cos(45) =
2

Force equilibrium Force-deformation relations

F
EA
x = k12 − T1 cos(45) − T2 cos(45) = 0 T1 = 1
L
F y = k22 − T1 sin(45) + T2 sin(45) = 0 EA
T2 =  2
L
Combining force equilibrium and force-deformation relations
( T1 + T2 ) EA
k12 = = (1 +  2 )
2 2L
(T1 − T2 ) EA
k22 = = (1 −  2 )
2 2L
Now use the geometric (compatibility) conditions (see figure)
1
1 = 1.cos(45) =
2
1
 2 = −1.cos(45) = − This negative is due to compression
2
Finally
EA
k12 = (1 +  2 ) = 0
2L
EA EA 2 EA
k22 = ( 1 2)
 −  = ( ) =
2L 2L 2 L
© 2002 Brooks/Cole Publishing / Thomson Learning™ 3D Truss (space truss)
In local coordinate system f̂ = k̂ d̂
f̂ 1x  d̂ 1x 
  k 0 0 −k 0 0  
f̂ 1y   0 0 0 0 0 0 d̂ 1y 
    
f̂ 1z   0 0 0 0 0 0 d̂ 1z 
 =  
f̂ 2x  − k 0 0 k 0 0 d̂ 2x 
f̂   0 0 0 0 0 0 d̂ 
 2y     2y 
f̂   0 0 0 0 0 0  
 2z  d̂ 2z 
 The transformation matrix for a single vector in 3D
d̂ = T d
*

 l1 m1 n1  l1, m1 and n1 are the direction cosines of x^

T = l 2 n2 
*
m2
l1 = cos q x
l3 m3 n3  m1 = cos q y

© 2002 Brooks/Cole Publishing / Thomson Learning™

n1 = cos q z
Transformation matrix T relating the local and global
displacement and load vectors of the truss element

d̂ = Td
T * 0
T = *
f̂ = Tf
66
0 T 

Element stiffness matrix in global coordinates

k = T k̂ T
T

66 66 66 66

  l1 2 l1 m1 l1 n1 − l1
2
− l1 m1 − l1 n1 
 
m1 n1 − l1 m1 − m1 − m1 n1 
2 2
 l1 m1 m1

EA 1 1 l n m n n
2
l n m n − n
2 
k = T k̂ T =  
T 1 1 1 1 1 1 1 1

L  − l1 2
− l1 m1 − l1 n1 l1
2
l1 m1 l1 n1 
− l m − m 2 − m n l m m
2
m n 
 1 1 1 1 1 1 1 1 1 1

 − l1 n1 − m1 n1 − n1 n1 
2 2
l1 n1 m1 n1

Notice that the direction cosines of only the local ^x axis enter the
k matrix