Professional Documents
Culture Documents
Matrix Structural Analysis - The Stiffness Method
Matrix Structural Analysis - The Stiffness Method
1/25
Mechanics of Structures,
A. Mansour
2/25
Mechanics of Structures,
Input Data
The coordinates of the nodes are given below:
Node number
1
2
3
4
Label of Fig.1
A
D
C
B
X coordinate - m
0.0
0.002
0.004
0.008
We should inform the program of the nodes associated with each element.
Element number
1
2
3
Label of Fig. 1
AD
DC
CB
1st node
1
2
3
2nd node
2
3
4
The previous two tables give the information required to calculate the length
of each element. For instance, the length of element (2), L(2) = 0.004 0.002
= 0.002 m. By the same token L(3) = 0.008 0.004 = 0.004 m.
We should specify the material of each element or the relevant properties for
each element.
Element number
1
2
3
4
u
0.0
0.0
Fx - (N)
+24 000
0.0
Fx2 is positive because it is in the positive x direction. Usually if u for any node
is known in advance, then F for that node is unknown, and vice versa.
3/25
Mechanics of Structures,
Having a full description of the problem, computer programs can determine all
the nodal displacements and forces. The relationship among these variables
is given below.
Stiffness Matrix
A typical element (e) is shown in Fig. 2a. The x-displacement of nodes 1 and
2 are u1 and u2. The nodal forces are fx1 and fx2. Of course, fx1 = -fx2.
However, in order to have a systematic representation, we will keep a
separate name for each nodal force.
The element is elastic and by consulting Fig. 2b,
fx2 = k(e) (u2 u1) = k(e) (-u1 + u2)
Where, k(e) = EA / L ; the elemental stiffness.
Fig. 2c shows that
fx1 = k(e) (u1 u2)
Where, fx1 is a compressive force and (u1 u2) represents a corresponding
contraction of the length of the element.
The following matrix equation represents the previous two equations.
f x1
k
=
f x2 e k
k u1
k e u 2
( f ) e = [k ]e (u)
or
Temperature Effect
We need to include the effect of temperature rise T = T T0. Fig. 2b gives:
u2 u1 = fx2 / k(e) + L T
In addition, Fig. 2c gives
4/25
Mechanics of Structures,
u1 u2 = fx1 / k(e) - L T
where, (u1 u2) implies that node 1 moves in the positive x direction (the right
direction). On the other hand, L T implies that node 1 moves to the left to
allow for the increase in length due to T. This explains why ( - L T ) must
be used.
1 f
k
EA T + x1 =
1 f x2 e k
k u1
k e u 2
Degrees of Freedom
Each node can move in the x direction only. Therefore, each node has only
one degree of freedom. Computer programs would address the
displacements by their degrees of freedom (DOF). The displacements of
nodes 1, 2, 3 and 4 correspond to degrees of freedom 1 up to 4. In addition,
fx1 up to fx4 corresponds to degrees of freedom 1 up to 4.
Mechanics of Structures,
k1 = E1 A1 / L1 =
k2 = E2 A2 / L2 =
k3 = E3 A3 / L3 =
For element 1:
f x1
4 4 u1
= 10 9
f x 2 (1)
4 4 (1) u 2
or
[ k ] (1)
1
2 DOF
(
4
)
(
4)
1
= 10 9
2
( 4) (4)
= 10 9
4 4 ( 2) u3
f x3 ( 2)
or
[ k ] ( 2)
2
3
DOF
2
((4)) (( 4))
= 10 9
3
(( 4)) ((4))
For element 3:
f x3
4 4 u 3
= 10 9
4 4 ( 3) u 4
f x 4 ( 3)
or
[ k ] ( 3)
3
4
DOF
3
(((4))) ((( 4)))
= 10 9
4
((( 4))) (((4)))
Fx 2 K 21
F = K
x 3 31
F
x 4 K 41
2
K 12
K 22
K 32
K 42
3
K 13
K 23
K 33
K 43
4 DOF
K 14
1
u1
K 24
2
u2
K 34
u
3
3
K 44 STRUCTURE u 4
4
Where the coefficients of the structure matrix Kij are constructed from the
coefficients of the individual stiffness matrices. We place each entry
according to its associated DOF, as shown below:
6/25
Mechanics of Structures,
1
2
3
4 . DOF
( 4)
1
Fx1
u1
(4)
(4) + ((4))
(( 4))
2
Fx 2
u2
9 ( 4)
F = 10
u
(( 4))
((4)) + (((3))) ((( 3))
3
x3
3
F
((( 3)))
(((3))) STRUCTURE u 4
4
x4
In the above structure stiffness matrix, empty entries show up because there
is no element connecting nodes 1 and 3, 1 and 4, and 2 and 4. These entries
must be replaced by zeroes as follows.
Fx1
0
u1
4 4 0
Fx 2
8 4 0
u2
9 4
=
10
u
0 4 7 3
Fx 3
Fx 4
0 3 3 STRUCTURE u 4
0
= 10 9
4 7 u3
Fx 3 = 0
Solve these equations to get u2 = 4.2 10-6 m and u3 = 2.4 10-6 m.
Now, we get the forces from equations 1 and 4:
7/25
Mechanics of Structures,
= 109
6
4 4 u2 = 4.2 10 16800
f x2
Element 2
f x2
4 4 u2 = 4.2 10 6 7200
=
= 109
6
4 4 u3 = 2.4 10 7200
f x3
and element 3
f x3
3 3 u3 = 2.4 10 6 7200
=
= 109
u4 = 0
3 3
f x4
7200
Fig. 4 shows that the forces acting on each element are indeed in equilibrium.
The external forces at any node also must be in equilibrium with the forces
transmitted to the bar. Fig. 4 shows the equilibrium of node 2. We can see
that the external force Fx2 = 24.0 kN is in equilibrium with the elemental
(internal) forces ( fx2 (1) + fx2 (2) = 16.8 + 7.2 = 24.0 kN ).
8/25
Mechanics of Structures,
Example (1)
.
The bar of Fig. 5 is subjected to F3 = 15 kN, and T = 20C. A(1)= 100 mm2, A(2) = 75, A(3)
= 50, E = 200 GPa, and = 12 x 10-6 / C
Determine the nodal displacements, the reactions, and the forces transmitted through
each element. Compare between the developed states of stresses for T = 20C and
T = 0C
Solution
The stiffness of each element is:
k1 = ( 200 x 109) (100 x 10-6) / 0.2 = 100 x 106 N/m
k2 = 100 x 106 N/m
k3 = 80 x 106 N/m
We should calculate the thermal terms ( EAT ) for each element:
( EAT )(1) = 4800 N
( EAT )(2) = 3600 N
( EAT )(3) = 2400 N
The element equations are:
f x1
4800
100 100 u1
+
= 10 6
100 100 u 2
f x 2 (1) 4800
f x2
3600
100 100 u 2
+
= 10 6
100 100 u 3
f x 3 ( 2 ) 3600
9/25
Mechanics of Structures,
f x3
2400
100 100 u 3
+
= 10 6
100 100 u 4
f x 4 (3) 2400
Assemble the equations:
( 4800)
0
0 u1
Fx1
100 100
200 100
0 u 2
Fx 2 (4800) + (( 3600))
6 100
F + ((3600)) + ((( 2400))) = 10 0
100 180 80 u 3
x3
F
(((2400)))
0
80 80 u 4
0
x4
Boundary conditions
u1 = 0, u4 = 0 ( we may ignore rows and columns 1 and 4 )
Fx2 = 0, Fx3 = 15000. Hence,
1200
200 100 u 2
= 10 6
16200
100 180 u 3
(The above stiffness equation is the reduced stiffness matrix after applying the boundary
conditions.)
Solve the equations to get
u2 = 0.000070615
u3 = 0.00012923
m
m
The reactions are obtained from rows 1 and 4. The following table shows the reactions
at the support when T = 20C as well as when T = 0C.
Fx1
Fx4
kN
kN
T = 20C
-2.2615
-12.7384
T = 0C
-5.76923
-9.23072
We can use the element matrix equations to get the forces acting on each element.
Element (1)
fx1 = -2.2615 kN
fx2 = 2.2615
Element (2)
fx2 = -2.2615
fx3 = 2.2615
Element (3)
fx3 = 12.7
fx4 = -12.7
Normal stresses are obtained by dividing each normal force by the corresponding cross
10/25
Mechanics of Structures,
sectional area. Elements 1 and 2 are subjected to tensile forces and element 3 is
subjected to a compressive force.
Elemental stresses
(1)
(MPa)
(2)
(MPa)
(3)
(MPa)
T = 20C
23
30
-255
T = 0C
58
77
-185
= 11
f x 2 e k 21
k12 u1
k 22 e u 2
or
( f ) e = [k ] e (u)
The matrix equation is valid for any combination of u1 and u2. Take u1 = 1.0
and u2 = 0.0 (Fig. 6). Then fx1 = k11 and fx2 = k21. However, from elementary
mechanics fx1 = EA / L u1 = k and fx2 = -k. Therefore, k11 = k and k21 = -k.
Taking u1 = 0 and u2 = 1 yields the expressions for the remainder coefficients (
k21 = -k and k22 = k ).
11/25
Mechanics of Structures,
f x1
c2
f y1
cs
= k 2
c
f x2
f y 2
cs
cs
s2
cs
s2
c2
cs
c2
cs
cs u1
s 2 v1
cs u 2
s 2 v 2
The matrix has a size of 4 x 4, because there are four degrees of freedom.
The angle is measured in the counter clockwise direction. Hence, is
negative when measured in the clockwise direction. We can use or (
180 ) and still get the same stiffness matrix.
The stiffness matrix is symmetric and singular. Diagonal terms are 0. For
each column, the sum of the coefficients in odd rows (as well as those in even
rows) is equal zero.
Derivation of [k]
The general expression is:
12/25
Mechanics of Structures,
f x1
k11
k
f y1
= k 21
f x2
k 31
f y2
k 41
k12
k 22
k 32
k 42
k13
k 23
k 33
k 43
We would first determine the coefficients of the first column. Take u1= 1 and
v1 = u2 = v2 = 0
u1
v1
u2
v2
1
0
0
0
Then fx1 = k11, fy1 = k21, fx2 = k31, and fy2 = k41.
However, we can determine these nodal forces independently.
Due to the imposed displacement u1 = 1, the bar contracts by = u1 cos =
cos , Fig. 8. Then f = k = k cos k c.
The force f is inclined by an angle Resolve f into fx1 and fy1 to get fx1 = f cos
= k c2 and fy1 = f sin = k c s. Thus, k11 = k c2 and k12 = k c s.
In addition, by resolving f acting at node 2, we can show that k31 = - k c2 and
k41 = - k c s.
We could determine the coefficients of columns 2, 3, and 4 by using the
following displacement states.
Column no.
2
3
4
u1
0
0
0
v1
1
0
0
13/25
u2
0
1
0
v2
0
0
1
Mechanics of Structures,
Example (2)
Construct the reduced
stiffness matrix of the
shown truss, Fig. 9.Then
determine
the
nodal
displacements and the
normal stress in element
3. L(1) = L(2) = 2 m, L(3) =
22, A(1) = A(2) = A(3) = 80
mm2, and E = 200 GPa.
Solution
Calculate the following quantities:
K = EA / L N/m
c2
s2
cs
Element (1)
8 x 106
0
1
0
0
Element (2)
8 x 106
270 (or -90)
0
1
0
Element (3)
5.65685 x 106
225 (-135 or 45)
0.5
0.5
0.5
0 0
0 v1
f y1
6 0
f = 10 8 0 (8) (0) u
2
x2
f
v
0
0
(
0
)
(
0
)
2
y2
Element (2)
f x1
0 u1
0 0 0
8 0 8 v1
f y1
60
f = 10 0 0 0
0 u3
x3
f
0 8 0 ((8)) v 3
y3
Element (3)
f x2
(((2.8284))) (((2.8284))) 2.8284 ((( 2.8284))) u 2
f
(((
2
.
8284
)))
(((
2
.
8284
)))
2
.
8284
(((
2
.
8284
)))
v3
y3
The brackets identify the coefficients that contribute to the reduced stiffness matrix.
14/25
Mechanics of Structures,
Boundary conditions:
u1 = 0
v1 = 0
Fx2 = 2000 N Fy2 = -3000
u3 = 0
Fy3 = 0
The structure stiffness matrix has a size of 6 x 6. The reduced stiffness matrix has a
size of 3 x 3. We construct the reduced stiffness matrix by ignoring the rows and
columns corresponding to u1, v1, and u3.
Fx 2
2.8284 u 2
(8) + 2.8284 (0) + 2.8284
6
2.8284 v 2 =
F y 2 = 10 (0) + 2.8284 (0) + 2.8284
F
2.8284
2.8284
((8)) + 2.8284 v 3
y3
2.8284 2.8284 u 2
10.8284
10 2.8284
2.8284 2.8284 v 2
2.8284 2.8284 10.8284 v 3
6
v2
-2.061 x 10-3
v3
-0.375 x 10-3
f x2
1 1 1 1 u 2 = 0.625 (10 3 ) 3 10 3
f y2
1 1 1 v 2 = 2.061 (10 3 ) 3 10 3
6 1
= 2.8284.10
=
f x3
1 1 1 1
u
=
0
3 10 3
3
3
3
f y 3
1 1 1 1 v3 = 0.375 (10 ) 3 10
15/25
Mechanics of Structures,
Degrees of Freedom
Fig. 11 shows a beam element. It has two degrees of freedom per node. The
element stiffness matrix has a size of 4 x 4. The sign convention used for the
moments and forces is not universal.
2
6 L 2 L2 1
m1 EI 6 L 4 L
f = L3 12 6 L
12
6 L v2
2
2
m
6 L 4 L2 2
6L 2L
2
16/25
Mechanics of Structures,
Example (3)
Get the vertical deflection and
angle of rotation (slope) at node 2,
(Fig. 12). Get the results in terms
of E, I, and L.
Solution
The boundary conditions are:
v1 = 0
1 = 0
F2 = -W
M2 = 0
By ignoring the 1st and 2nd columns and rows in the element stiffness matrix,
we get the following matrix equation.
W EI 12 6 L v 2
= 3
2
0 L 6L 4L 2
We solve these simultaneous equations to get
v2 =
(W L3 / 3EI )
2 = ( W L2 / 2EI )
The student should verify these results. Note: the slope at node 2 is
clockwise.
Distributed Loads
Fig. 13-a shows a uniformly distributed force w (N/m). This force is replaced
by equivalent nodal loads as shown in Fig. 13-b (consult a textbook for the
proof).
17/25
Mechanics of Structures,
2
f1
wl
m1 12
f + wl = [ K ]( )
2
m 2
2
2
wl
12
Where, {}T = { v1 1 v2 2 }T.
Example (4)
Solution
We shall model the beam using two elements. Each has L = 2 m.
EI / L3 = (200 x 109) (4 x wL/2 = (300)(2) / 2 = 300 N
wL2/12 = 100 Nm
10-6) / 23 = 105 N/m
The elements equations
Element (1)
18/25
Mechanics of Structures,
f1
300
12 12 12 12 v1
16
12
8 1
m1
100
5 12
f + 300 = 10 12 12
12
12 v 2
2
100
m
12
8
12
16
2
2 (1)
Element (2)
f2
300
12 12 12 12 v 2
12 16
12
8
m2
100
5
2
f + 300 = 10 12 12
12
12 v3
3
100
m
12
8
12
16
3
3 ( 2)
Boundary Conditions
v1 = 1 = v3 = 0
F2 = 0
M2 = 6000 Nm
M3 = 0
5
6000 + 100 + 100 = 10 12 12 16 + 16 8 2
0
12
100
8
16 3
or
600
24 0 12 v 2
5
6000 = 10 0 32 8 2
100
12 8 16 3
2 = 0.00246875 rad
= 0.1414
3 = -.002375 rad
= -0.1361
16
12
8
0
m1
100 150
5 12
+
=
f = 10 12 12
12
12 0.0014375 300 1537.5
2
m
12
16 0.00246875 100 2325
12 8
2 (1)
19/25
Mechanics of Structures,
Element (2)
f2
12 12 12 12 v 2 300 1537.5
16
12
8 2 100 3675
m2
5 12
+
=
f = 10 12 12
12
12 0 300 2137.5
3
m
12
16 3 100
0
12 8
3 ( 2)
The nodal loads acting on the elements are shown in Figs. 15-(a) & 15-(b).
The reaction loads acting on the beam are shown in Fig. 15-(c).
Document 2 contains the computer results to this very same problem. The
computer solution gives not only the nodal displacements but also the entire
elastic curve.
Symmetry
The following table depicts examples of symmetric beams and trusses under
static conditions Symmetry is in geometry, material properties, relevant
boundary conditions, as well as in loading. Each configuration has a plane of
symmetry. This plane virtually cut the structure into two identical parts.
Therefore, we could reduce the size of the problem by half. In doing this, we
should introduce the proper boundary conditions at the plane of symmetry
(the new edge of the reduced structure).
For beams:
The slope at the plane of symmetry is zero.
The transverse force acting along the plane of symmetry must be
halved.
For trusses:
The displacement at the plane of symmetry normal to it u is zero.
The forces at this cutting plane must be halved.
The cross sectional area of bars aligned with the axis of symmetry
must be halved.
Bars that cross the plane of symmetry at an angle must be cut by that
plane resulting in a shorter bar (and u = 0 at the intersection).
Figures 16a, 17a, and 18a can be modelled by Figs. 16b to 18b.
20/25
Mechanics of Structures,
Configuration
Min # of
elements
F = -4 kN
M=0
4
21/25
F1 = -2 kN
1 = 0
F=0;M=0
We may solve the problem
without a node in the
middle.
F1 = 0
1 = 0
Mechanics of Structures,
Configuration
Configuration
Min # of
elements
Fx1 = 0
Fy1 = -5 kN
Fx2 = 0
Fy2 = 0
Note: A(1) = 1000 mm2
u1 = 0
Fy1 = -2.5 kN
u2 = 0
Fy2 = 0
Plus
u3 = 0
F3 = 0
Note: A(1) = 500 mm2
Min # of
elements
Plane Frames
Fig. 19 shows a simple planar frame with assigned nodes and elements (or
joints and members). Each element is capable of sustaining bending
moments, shearing and axial forces.
A typical plane frame element (Fig. 20) has two nodes each has three
degrees of freedom.
The element equation is:
{f}(e) = [k](e) {}; where
22/25
Mechanics of Structures,
23/25
Mechanics of Structures,
Fig. 24
A Practical Example
Fig. 24 shows a bus frame subjected to roof load of 100 kN in order to test its
strength. This is an example of how engineers use computer programs to
solve engineering problems (Logan). In this example, 599 frame elements
and 357 nodes were used.
Comments
This introduction is elementary and limited in scope. Many topics were
omitted such as inclined rolling supports and the details of frame elements.
In order to appreciate the strength of the method, students should solve
certain assigned problems using a computer program.
References
P.P. Benham, R.J. Crawford & C.G. Armstrong (1996) Mechanics of Engineering Materials,
2nd edition, Longman, Essex.
R.G. Budynas (1999) Advanced Strength and Applied Stress Analysis, 2nd edition, McGrawHill, Boston.
H. Grandin,Jr. (1986) Fundamentals of the Finite Element Method, Macmillan, New York.
C.E. Knight (1993) The Finite Element Method in Mechanical Design, PWS-Kent, Boston.
R. L. Logan (1992) A first course in the Finite Element method, PWS-Kent, Boston.
S. Moaveni (2003) Finite Element Analysis Theory and Application with ANSYS, 2nd edition,
Pearson Education / Prentice Hall, New Jersey. ( www.prenhall.com/Moaveni )
24/25
Mechanics of Structures,
Appendix Formulas
1-Dim bar
1 f
k
EA T + x1 =
1 f x2 e k
k u1
k e u 2
Plane truss
f x1
c2
f y1
cs
=
k
c2
f x2
f y 2
cs
cs
s2
cs
s2
c2
cs
c2
cs
cs u1
s 2 v1
cs u 2
s 2 v 2
m1 EI 6 L
=
f L3 12
2
m
6L
2
wl
2
f1
wl
m1 12
f + wl =
2
m 2
2
2
wl
12
6 L 12 6 L v1
4 L2
6 L 2 L2 1
6L
12
6 L v2
2 L2
6 L 4 L2 2
[ K ]( )
25/25