Reading assignment:

MANE 4240 & CIVL 4240

Introduction to Finite Elements Chapter 2: Sections 2.1-2.5 + Lecture notes

Summary:
Introduction to the Stiffness • Developing the finite element equations for a system of
(Displacement) Method: springs using the “direct stiffness” approach
• Application of boundary conditions
Analysis of a system of springs • Physical significance of the stiffness matrix
• Direct assembly of the global stiffness matrix
Prof. Suvranu De • Problems

FEM analysis scheme

F1x F2x F3x
x
Step 1: Divide the problem domain into non overlapping regions
(“elements”) connected to each other through special points k1 k2
(“nodes”)

Step 2: Describe the behavior of each element Problem

Step 3: Describe the behavior of the entire body by putting
together the behavior of each of the elements (this is a process
known as “assembly”)
Analyze the behavior of the system composed of the two springs
loaded by external forces as shown above
Given
F1x , F2x ,F3x are external loads. Positive directions of the forces
are along the positive x-axis
k1 and k2 are the stiffnesses of the two springs

1
 F1x k1 F2x k2 F3x
F1x F2x F3x x
x
1 2 3
Element 1 Element 2
k1 k2 d1x d2x d3x
Solution Node 1
Step 1: In order to analyze the system we break it up into smaller Solution
parts, i.e., “elements” connected to each other through “nodes” Step 2: Analyze the behavior of a single element (spring)
F1x k1 F2x k2 F3x
x
1 2 3
Element 1 Element 2
d1x d2x d3x © 2002 Brooks/Cole Publishing / Thomson Learning™

Node 1 Two nodes: 1, 2

Unknowns: nodal displacements d1x, d2x, d3x, Nodal displacements: d̂1x d̂ 2x
Nodal forces: f̂1x f̂ 2x
Spring constant: k

Behavior of a linear spring (recap)

F x
k k
1 F
d k
d

F = Force in the spring

© 2002 Brooks/Cole Publishing / Thomson Learning™
d = deflection of the spring
Local (x̂ , ŷ ,ẑ) and global (x,y,z) coordinate systems k = “stiffness” of the spring
Hooke’s Law
F = kd

2
 Note T
f̂1x f̂ 2x 1. The element stiffness matrix is “symmetric”, i.e. k̂ = k̂
2. The element stiffness matrix is singular, i.e.,
© 2002 Brooks/Cole Publishing / Thomson Learning™

Hooke’s law for our spring element det ( k̂ ) = k 2 − k 2 = 0

f̂ 2x = k (d̂ 2x − d̂1x ) Eq (1) The consequence is that the matrix is NOT invertible. It is not
Force equilibrium for our spring element (recap free body diagrams) possible to invert it to obtain the displacements. Why?
f̂1x + f̂ 2x = 0 The spring is not constrained in space and hence it can attain
Eq (2) multiple positions in space for the same nodal forces
⇒ f̂1x = −f̂ 2x = − k (d̂ 2x − d̂1x ) e.g.,
Collect Eq (1) and (2) in matrix form ⎧⎪f̂1x ⎫⎪ ⎡ 2 - 2⎤ ⎧1 ⎫ ⎧- 2⎫
⎧⎪f̂1x ⎫⎪ ⎡ k - k ⎤ ⎧⎪d̂1x ⎫⎪ ⎨ ⎬=⎢ ⎥⎨ ⎬ = ⎨ ⎬
f̂ = k̂ d̂ ⎨ ⎬=⎢ ⎥⎨ ⎬ ⎪⎩f̂ 2x ⎪⎭ ⎣- 2 2 ⎦ ⎩2⎭ ⎩ 2 ⎭
⎪⎩f̂ 2x ⎪⎭ ⎣- k k ⎦ ⎪⎩d̂ 2x ⎪⎭
Element force Element nodal 
 

⎧⎪f̂1x ⎫⎪ ⎡ 2 - 2⎤ ⎧3⎫ ⎧- 2⎫
vector
Element displacement f̂ k̂ d̂ ⎨ ⎬=⎢ ⎥⎨ ⎬ = ⎨ ⎬
stiffness vector ⎪⎩f̂ 2x ⎪⎭ ⎣- 2 2 ⎦ ⎩4⎭ ⎩ 2 ⎭
matrix

Solution To assemble these two results into a single description of the

Step 3: Now that we have been able to describe the behavior of response of the entire structure we need to link between the local
each spring element, lets try to obtain the behavior of the original and global variables.
structure by assembly
Question 1: How do we relate the local (element) displacements
Split the original structure into component elements back to the global (structure) displacements?
Element 2 F1x k1 F2x k2 F3x
Element 1 x
1 k1 2 2 k2 3
1 2 3
(1) Element 1 Element 2
f̂1x(1) d̂1x
(1) f̂ 2x(1) d̂ 2x (2)
f̂1x(2) d̂1x
(2)
f̂ 2x(2) d̂ 2x d1x d2x d3x
Node 1
(1) ⎫
⎧⎪f̂1x(1) ⎫⎪ ⎡ k1 - k1 ⎤ ⎧⎪d̂1x ⎪ (2) ⎫
⎧⎪f̂1x(2) ⎫⎪ ⎡ k 2 - k 2 ⎤ ⎧⎪d̂1x ⎪
⎨ (1) ⎬ = ⎢ ⎨ (1) ⎬
(1)
d̂1x = d1x
⎥
⎪⎩f̂ 2x ⎪⎭ ⎣- k1 k1 ⎦ ⎪⎩d̂ 2x ⎪⎭ ⎨ (2) ⎬ = ⎢ ⎥ ⎨ (2) ⎬ Eq (5)
⎪⎩f̂ 2x ⎪⎭ ⎣- k 2 k 2 ⎦ ⎪⎩d̂ 2x ⎪⎭

 

2x = d̂1x = d 2x
d̂ (1) (2)
(1) (1) (1)

 

( 2)
f̂ k̂ d̂ ( 2) (2)

2x = d 3x
f̂ k̂ d̂
Eq (3)
d̂ (2)
Eq (4)

3
Hence, equations (3) and (4) may be rewritten as Question 2: How do we relate the local (element) nodal forces back
to the global (structure) forces? Draw 5 FBDs
F1x F2x k2 F3x
⎧⎪f̂1x(1) ⎫⎪ ⎡ k1 - k1 ⎤ ⎧d1x ⎫ ⎧⎪f̂1x(2) ⎫⎪ ⎡ k 2 - k 2 ⎤ ⎧d 2x ⎫
k1
x
⎨ (1) ⎬ = ⎢ ⎥⎨ ⎬ ⎨ (2) ⎬ = ⎢ ⎨ ⎬
⎪⎩f̂ 2x ⎪⎭ ⎣- k1 k1 ⎦ ⎩d 2x ⎭ ⎪⎩f̂ 2x ⎪⎭ ⎣- k 2 k 2 ⎥⎦ ⎩d 3x ⎭ 1 A B 2 C D3
d1x d2x d3x
Or, we may expand the matrices and vectors to obtain
2 3
⎧f̂1x(1) ⎫ ⎡ k1 − k1 0⎤ ⎧d1x ⎫ ⎧ 0 ⎫ ⎡0 0 0 ⎤ ⎧ d1x ⎫
⎪ (1) ⎪ ⎢ ⎥⎪ ⎪ ⎪f̂ (2) ⎪ ⎢ ⎥⎪ ⎪
⎨f̂ 2x ⎬ = ⎢- k1 k1 0⎥ ⎨d 2x ⎬ ⎨ 1x ⎬ = ⎢0 k 2 − k 2 ⎥ ⎨d 2x ⎬ F1x f̂1x(1) f̂ 2x(1) f̂1x(2) f̂ 2x(2) F3x
⎪0 ⎪ ⎢ 0 ⎪f̂ (2) ⎪ ⎢0 - k k ⎥ ⎪⎩d 3 x ⎪⎭
F2x
0⎦⎥ ⎪⎩
d 3x ⎪⎭ ⎩ 2x ⎭

⎣ 2
⎦ 

© 2002 Brooks/Cole Publishing / Thomson Learning™

⎩
⎭  ⎣  0

2

(1) e (1 ) e d f̂
( 2) e (2)e d At node 1 : F1x - f̂1x(1) = 0
f̂ k̂ k̂

Eq (6) Eq (7) At node 2 : F2x - f̂ 2x(1) − f̂1x(2) = 0

(1) e
k̂
(1) e
Expanded element stiffness matrix of element 1 (local) At node 3 : F3x - f̂ 2x(2) = 0
f̂ Expanded nodal force vector for element 1 (local)
d Nodal load vector for the entire structure (global)

In vector form, the nodal force vector (global)

But we know the expressions for the expanded local force vectors
⎧ F1x ⎫ ⎧ f̂1x ⎫
(1)
from Eqs (6) and (7)
⎪ ⎪ ⎪ (1) (2) ⎪
F = ⎨F2x ⎬ = ⎨f̂ 2x + f̂1x ⎬ f̂
(1) e
= k̂
(1)e
d and f̂
( 2)e
= k̂
(2)e
d
⎪F ⎪ ⎪ f̂ (2) ⎪
⎩ 3x ⎭ ⎩ 2x ⎭ Hence

d = ⎛⎜ k̂ ⎞⎟d
(1) e ( 2)e (1)e (2)e (1)e (2)e
Recall that the expanded element force vectors were F = f̂ + f̂ = k̂ d + k̂ + k̂
⎝ ⎠
⎧f̂ (1) ⎫ ⎧0 ⎫
⎪ ⎪ ⎪ (2) ⎪
1x
F=Kd
(1) e ( 2)e
f̂ = ⎨f̂ (1)
2x ⎬ and f̂ = ⎨f̂1x ⎬
⎪0 ⎪ ⎪f̂ (2) ⎪
⎩ ⎭ ⎩ 2x ⎭ F = Global nodal force vector
Hence, the global force vector is simply the sum of the expanded
element nodal force vectors ⎧ F ⎫ d = Global nodal displacement vector
K = Global stiffness matrix
1x
⎪ ⎪ (1) e ( 2)e
F = ⎨F2x ⎬ = f̂ + f̂
⎪F ⎪ = sum of expanded element stiffness matrices
⎩ 3x ⎭

4
 For our original structure with two springs, the global stiffness The system equations F=Kd imply
matrix is
⎧ F1x ⎫ ⎡ k1 − k1 0 ⎤ ⎧d1x ⎫
⎡ k 1 − k 1 0 ⎤ ⎡0 0 0 ⎤ ⎪ ⎪ ⎢ ⎪ ⎪
⎨F2x ⎬ = ⎢- k1 k1 + k 2 − k 2 ⎥⎥ ⎨d 2x ⎬
⎢ ⎥ ⎢
K = ⎢- k1 k1 0⎥ + ⎢0 k 2 − k 2 ⎥⎥ ⎪F ⎪ ⎢ 0
⎩ 3x ⎭ ⎣ - k2 k 2 ⎥⎦ ⎪⎩d 3x ⎪⎭
⎣⎢ 0 0


0⎦⎥ ⎣⎢0 - k 2 k 2 ⎦⎥
 

(1) e (2)e
k̂ k̂
F1x = k1d1x − k1d 2x
⎡ k1 − k1 0 ⎤
⇒ F2x = -k1d1x + (k1 + k 2 )d 2x − k 2 d 3x
= ⎢⎢- k1 k1 + k 2 − k 2 ⎥⎥
F3x = -k 2 d 2x + k 2 d 3x
⎢⎣ 0 - k2 k 2 ⎥⎦

NOTE These are the 3 equilibrium equations at the 3 nodes.

1. The global stiffness matrix is symmetric
2. The global stiffness matrix is singular

F1x k1 F2x k2 F3x

x
1 A B2 C D3 Notice that the sum of the forces equal zero, i.e., the structure is in
d1x d2x d3x static equilibrium.
2 3
F1x + F2x+ F3x =0
(1) (1) (2)
F1x f̂1x f̂ 2x F2x f̂
1x f̂ 2x(2) F3x Given the nodal forces, can we solve for the displacements?
© 2002 Brooks/Cole Publishing / Thomson Learning™

F1x = k1 (d1x − d 2x ) = f̂ (1)

To obtain unique values of the displacements, at least one of the
At node 1 : F1x - f̂1x(1) = 0 1x

F2x = -k1d1x + (k1 + k 2 )d 2x − k 2 d 3x nodal displacements must be specified.

At node 2 : F2x - f̂ (1)
− f̂ (2)
=0
2x 1x
= − k1 (d1x − d 2x ) + k 2 (d 2x − d 3x )
At node 3 : F3x - f̂ (2)
=0
2x
= f̂ 2x(1) + f̂1x(2)
F3x = -k 2 (d 2x − d 3x ) = f̂ 2x(2)

5
 Direct assembly of the global stiffness matrix
Node element connectivity chart : Specifies the global node
number corresponding to the local (element) node numbers
Global
F1x k1 F2x k2 F3x ELEMENT Node 1 Node 2 Local node number
x
2 3 1 1 2 Global node number
1
Element 1 Element 2
d1x d2x d3x
2 2 3
Local
Element 2
Element 1
1 k1 2 2 k2 3

(1) (2)
f̂1x(1) d̂1x
(1) f̂ 2x(1) d̂ 2x (2)
f̂1x(2) d̂1x f̂ 2x(2) d̂ 2x

Stiffness matrix of element 1 Stiffness matrix of element 2

Example 2.1
d1x d2x d2x d3x
(1) ⎡k - k1 ⎤ d1x ( 2) ⎡k - k 2 ⎤ d2x
=⎢ 1 =⎢ 2
k 2 ⎥⎦ d3x
k̂ ⎥ k̂ 22 3 4
⎣- k1 k1 ⎦ d2x ⎣- k 2
© 2002 Brooks/Cole Publishing / Thomson Learning™

Global stiffness matrix

Compute the global stiffness matrix of the assemblage of
d1x d2x d3x springs shown above
⎡ k1 - k1 0 ⎤ d1x d1x d2x d3x d4x
K = ⎢⎢- k1 k1 + k 2 - k 2 ⎥⎥ d2x ⎡ 1000 −1000
⎢−1000 1000+ 2000
0 0 ⎤ d1x
( ) −2000 0 ⎥⎥ d2x
⎢⎣ 0 - k2 k 2 ⎥⎦ d3x K= ⎢
⎢ 0 −2000 ( 2000+3000) −3000⎥ d3x
⎢ ⎥
Examples: Problems 2.1 and 2.3 of Logan ⎣ 0 0 −3000 3000 ⎦ d
4x

6
 Example 2.3 Imposition of boundary conditions
Consider 2 cases
Case 1: Homogeneous boundary conditions (e.g., d1x=0)
Case 2: Nonhomogeneous boundary conditions (e.g., one of the
nodal displacements is known to be different from zero)
3
Homogeneous boundary condition at node 1
© 2002 Brooks/Cole Publishing / Thomson Learning™
k1=500N/m k2=100N/m F3x=5N
Compute the global stiffness matrix of the assemblage of 1 x
springs shown above 3
2
Element 1 Element 2
⎡ k1 -k1 0 ⎤ d1x=0 d2x d3x
⎢ ⎥
K = ⎢ -k1 k1 + k 2 + k 3 - ( k 2 + k3 )⎥
⎣⎢ 0 - ( k2 + k3 ) ( k 2 + k 3 ) ⎦⎥

System equations Eq(2) and (3) are used to find d2x and d3x by solving
0
⎡ 600 −100 ⎤ ⎡ d 2 x ⎤ ⎡ 0 ⎤
⎡ 500 -500 0 ⎤ ⎡ d1x ⎤ ⎡ F1x ⎤ ⎢ −100 100 ⎥ ⎢ d ⎥ = ⎢ 5 ⎥
⎢ -500 600 -100 ⎥ ⎢ d ⎥ = ⎢ 0 ⎥ ⎣ ⎦ ⎣ 3x ⎦ ⎣ ⎦
⎢ ⎥ ⎢ 2x ⎥ ⎢ ⎥ ⎡ d 2 x ⎤ ⎡ 0.01 m ⎤
⎣⎢ 0 -100 100 ⎦⎥ ⎢⎣ d 3 x ⎦⎥ ⎣⎢ 5 ⎦⎥ ⇒⎢ ⎥=⎢ ⎥
⎣ d3 x ⎦ ⎣ 0.06 m ⎦
Global Stiffness Nodal Nodal
matrix disp load
NOTICE: The matrix in the above equation may be obtained from
vector vector
the global stiffness matrix by deleting the first row and column
Note that F1x is the wall reaction which is to be computed as part
of the solution and hence is an unknown in the above equation ⎡ 500 -500 0 ⎤
⎡ 600 −100 ⎤
⎢-500 600 -100 ⎥ ⎢ −100 100 ⎥
Writing out the equations explicitly ⎢ ⎥ ⎣ ⎦
-500d 2x = F1x Eq(1) ⎣⎢ 0 -100 100 ⎥⎦
600d 2 x − 100d 3 x = 0 Eq(2)
Note use Eq(1) to compute F1x =-500d 2x = −5 N
−100d 2 x + 100d 3 x = 5 Eq(3)

7
 Imposition of boundary conditions…contd.
NOTICE:
Nonhomogeneous boundary condition: spring 2 is pulled at
1. Take care of homogeneous boundary conditions node 3 by 0.06 m)
by deleting the appropriate rows and columns from the k1=500N/m k2=100N/m
1 x
global stiffness matrix and solving the reduced set of
equations for the unknown nodal displacements. 2 3
Element 1 Element 2
d1x=0 d2x d3x=0.06m
2. Both displacements and forces CANNOT be known at
the same node. If the displacement at a node is known, the
reaction force at that node is unknown (and vice versa)

System equations Now use only equation (2) to compute d2x

0
⎡ 500 -500 0 ⎤ ⎡ d1x ⎤ ⎡ F1x ⎤ 600d 2 x = 100(0.06)
⎢ -500 600 -100 ⎥ ⎢ d ⎥ = ⎢ 0 ⎥ ⇒ d 2 x = 0.01m
⎢ ⎥ ⎢ 2x ⎥ ⎢ ⎥
⎣⎢ 0 -100 100 ⎦⎥ ⎢⎣ d 3 x ⎦⎥ ⎣⎢ F3 x ⎦⎥
0.06
Note that now F1x and F3x are not known. Now use Eq(1) and (3) to compute F1x =-5N and F3x=5N

Writing out the equations explicitly

-500d 2x = F1x Eq(1)
600d 2 x − 100(0.06) = 0 Eq(2)
−100d 2 x + 100(0.06) = F3 x Eq(3)

8
 Recap of what we did…contd.
Recap of what we did
2. Relate the local forces of each element to the global forces at
Step 1: Divide the problem domain into non overlapping regions the nodes (use FBDs and force equilibrium).
F = ∑ f̂
e
(“elements”) connected to each other through special points
(“nodes”) Element Finally obtain
nodal
Step 2: Describe the behavior of each element ( f̂ = k̂ d̂ ) displacement F=Kd
vector

Step 3: Describe the behavior of the entire body (by “assembly”). Where the global stiffness matrix

K = ∑k
e
This consists of the following steps

1. Write the force-displacement relations of each spring in

expanded form Global
e
f̂ = k̂ e d̂ nodal
displacement
vector

Recap of what we did…contd.

Physical significance of the stiffness matrix
Apply boundary conditions by partitioning the matrix and vectors F1x k1 F2x k2 F3x
x
3
⎡K11 K12 ⎤⎧d1 ⎫ ⎧F1 ⎫ 1 2
⎢K K ⎥⎨d ⎬ = ⎨F ⎬ Element 1 Element 2
d3x
⎣ 21 22 ⎦⎩ 2 ⎭ ⎩ 2 ⎭ d1x d2x
In general, we will have a ⎡ k11 k12 k13 ⎤
Solve for unknown nodal displacements
stiffness matrix of the form
K = ⎢k k 22 k 23 ⎥⎥
(assume for now that we do not ⎢ 21
K22d2 = F2 − K21d1 know k11, k12, etc) ⎢⎣ k 31 k 32 k 33 ⎥⎦
Compute unknown nodal forces
The finite element ⎡ k11 k12 k13 ⎤ ⎧ d1 ⎫ ⎧ F1 ⎫
F1 = K11d1 + K12d2 ⎢k ⎪ ⎪ ⎪ ⎪
force-displacement ⎢ 21 k 22 k 23 ⎥⎥ ⎨d 2 ⎬ = ⎨F2 ⎬
relations: ⎢⎣ k 31 k 32 k 33 ⎥⎦ ⎪⎩d 3 ⎪⎭ ⎪⎩F3 ⎪⎭

9
 Physical significance of the stiffness matrix Physical significance of the stiffness matrix

The first equation is In general

Force equilibrium
k11d1 + k12 d 2 + k13d 3 = F1 equation at node 1
k ij = keeping
Force at node ‘i’ due to unit displacement at node ‘j’
all the other nodes fixed
Columns of the global stiffness matrix This is an alternate route to generating the global stiffness matrix
What if d1=1, d2=0, d3=0 ? e.g., to determine the first column of the stiffness matrix
While nodes 2 and 3 are held fixed Set d1=1, d2=0, d3=0
F1 = k11 Force along node 1 due to unit displacement at node 1 F1 k1 F2 k2 F3
x
F2 = k 21 Force along node 2 due to unit displacement at node 1
1 2 3
F3 = k 31 Force along node 3 due to unit displacement at node 1
Element 1 Element 2
d1 d2 d3
Similarly we obtain the physical significance of the other Find F1=?, F2=?, F3=?
entries of the global stiffness matrix

Physical significance of the stiffness matrix Physical significance of the stiffness matrix
Force equilibrium at node 1
For this special case, Element #2 does not have any contribution. F1
Look at the free body diagram of Element #1 F1 =fˆ1x(1) = k1

d̂ (1) d̂ (1)
2x
f̂1x(1)
1x
x Force equilibrium at node 2
f̂1x(1) k1 f̂ 2x(1)
F2
F2 =fˆ2x(1) = − k1
(1)
fˆ
(1)
2x = k1 (dˆ − dˆ ) = k1 (0 − 1) = − k1
(1)
2x
(1)
1x
f̂ 2x F1 = k1d1 = k1=k11
Of course, F3=0 F2 = -F1 = -k1=k21
fˆ1x(1) = −fˆ2x(1) = k1
F3 = 0 =k31

10
 Physical significance of the stiffness matrix Physical significance of the stiffness matrix
Hence the first column of the stiffness matrix is To obtain the third column of the stiffness matrix, calculate the
⎧ F1 ⎫ ⎧ k1 ⎫ nodal reactions at nodes 1, 2 and 3 when d1=0, d2=0, d3=1
⎪ ⎪ ⎪ ⎪
⎨ F2 ⎬ = ⎨− k1 ⎬
⎪F ⎪ ⎪ 0 ⎪ Check that
⎩ 3⎭ ⎩ ⎭
To obtain the second column of the stiffness matrix, calculate the ⎧ F1 ⎫ ⎧ 0 ⎫
⎪ ⎪ ⎪ ⎪
nodal reactions at nodes 1, 2 and 3 when d1=0, d2=1, d3=0 ⎨ F2 ⎬ = ⎨− k2 ⎬
⎪F ⎪ ⎪ k ⎪
Check that ⎩ 3⎭ ⎩ 2 ⎭
⎧ F1 ⎫ ⎧ − k1 ⎫
⎪ ⎪ ⎪ ⎪
⎨ F2 ⎬ = ⎨k1 + k2 ⎬
⎪ F ⎪ ⎪ −k ⎪
⎩ 3⎭ ⎩ 2 ⎭

Steps in solving a problem

Step 1: Write down the node-element connectivity table

linking local and global displacements

Step 2: Write down the stiffness matrix of each element

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

11