FINITE ELEMENT ANALYSIS

IN STRUCTURES

Chapter 10
DYNAMIC ANALYSIS

Zahit Mecitoğlu

© 2006 İstanbul Technical University

INTRODUCTION
„ Static analysis holds when the loads are slowly
applied.
„ When the loads are suddenly applied, or when
the loads are of a variable nature, the mass and
acceleration effects come into the picture.
„ The types of vibration:
… Free Vibrations
… Forced Vibrations
„ Harmonic Response
„ Transient Vibrations
FREE VIBRATION
„ If a solid body, such as an engineering structure,
is deformed elastically and suddenly released, it
tends to vibrate about its equilibrium position.
„ This periodic motion due to the restoring strain
energy is called free vibration.
„ The number of cycles per unit time is called the
frequency.
„ The maximum displacement from the equilibrium
position is the amplitude.
DAMPING
„ In the real word, the vibrations subside with time
due to damping action. In the simplest vibration
model, the damping effects are neglected.
„ The undamped free vibration model of a
structure gives significant information about its
dynamic behavior.
„ We present here the considerations needed to
apply finite elements to the analysis of
undamped free vibrations of structures.
FORMULATION
We define the Lagrangean L by
T : kinetic energy
L =T −Π Π : the potential energy.

Hamilton’s principle: For an arbitrary time interval from t1 to

t2, the state of motion of a body extremizes the functional
t2
I = ∫ Ldt
t1
FORMULATION
If L can be expressed in terms of the generalized
variables
(Q1, Q2 ,..., Qn , Q1, Q 2 ,..., Q n ) Q i = dQi dt

then the equations of motion are given by

d ⎛ ∂L ⎞ ∂L
⎜  ⎟− =0 ( i = 1, 2,..., n )
dt ⎝ ∂Qi ⎠ ∂Qi
FORMULATION
V The kinetic energy is given by

{ } {δ}dV
T
v, v T= 1
2 ∫ ρ δ
u , u V
w, w
y
dV where ρ is the density (mass
z x per unit volume) of the material

{δ} is the velocity vector of point as P, given by

{δ} = ⎢⎣u v w ⎥⎦
 T
FORMULATION
We express {δ} in terms of the nodal displacements {q},
using shape functions [N]

{δ} = [ N]{q}
In dynamic analysis, the elements of {q} are dependent on
time, while [N] represents (spatial) shape functions defined on
a master element. The velocity vector is then given by

{δ} = [ N ]{q }
FORMULATION
Substituting the velocity vector into the kinetic energy
expression for an element, we obtain
⎡ ⎤
{q } ⎢ ∫ ρ [ N ] [ N ] dV ⎥ {q }
T T
Te = 1
2
⎢⎣V e ⎥⎦
where the bracket expression is the element mass matrix
[m ] = ∫ ρ [ N ] [ N ] dV
T

Ve

This mass matrix is consistent with the shape functions

chosen and is called the consistent mass matrix.
FORMULATION
On taking summation over all the elements, we get
E E
T = ∑ Te = ∑ {q } [m ]{q } = { } [M ]{Q }
T
1 T 1 
Q
2 2
e=1 e=1

The potential energy is given by

Π= 1
2 {Q} [ K ]{Q} − {Q} {F}
T T

Using the Lagrangean L = T – P, we obtain the equations of

motion:
[M ]{Q } + [K ]{Q} = {F}
 FORMULATION
For the steady-state condition, starting from the equilibrium
state, we set
{Q} = {A} sin ωt
where {A} is the vector of nodal amplitudes of vibration and ω
(rad/s) is the circular frequency (=2πf, f= cycles/s or Hz).
[K ]{A} = ω 2 [M ]{A}
This is the generalized eigenvalue problem
[K ]{A} = λ [M ]{A}
The frequency f in hertz (cycles per second) can be
expressed in terms of the circular frequency ω.
f = ω / 2π
 CONSISTENT MASS MATRICES
AXIAL ELEMENT
L⎡
⎧ N1 ⎫ N12 N1 N 2 ⎤
[m ] = ∫ ρ [ N ] [ N ] dV = ρ ∫ ⎨ N ⎬ ⎣⎢ N1
T
N 2 ⎦⎥ dV =ρ ∫ ⎢ ⎥ Ae dx
Ve ⎩ 2⎭ 0⎢ N 2 ⎥⎦
2
Ve ⎣ N1N 2
where
x x
N1 = 1 − and N2 =
L L
On carrying out the integration of each term in the matrix, we
find
ρ Ae L ⎡ 2 1 ⎤
[m ] = ⎢1 2⎥
6 ⎣ ⎦
 CONSISTENT MASS MATRICES
TRUSS ELEMENT
We have four translations and the shape function matrix is
⎡ N1 0 N2 0 ⎤
[N] = ⎢ 0 N 2 ⎥⎦
⎣ N1 0

The mass matrix can be found as

⎡2 0 1 0⎤
⎢ 1 ⎥⎥
ρ Ae Le ⎢ 0 2 0
[m ] =
6 ⎢1 0 2 0⎥
⎢ ⎥
⎣0 1 0 2⎦
 CONSISTENT MASS MATRICES
BEAM ELEMENT
y v y
θyz

dA y
dA
u
x x

z z

For the beam or flexural element, a typical cross-section of

this type of member translates in the y direction and rotates
about its neutral axis.
 CONSISTENT MASS MATRICES
The translational inertia terms are much important than the
rotational terms.
The kinetic energy for an element can be written as
Le
Te = ∫ Ae dx
ρ 2
1
2

v
0
The mass matrix can be obtained using the relation
Le
[m t ] = ∫ ρ [ N ] [ N ] Ae dx
T

where [ N] = [ N1 N2 N3 N4 ]
T
CONSISTENT MASS MATRICES

⎡ 156 22 L 54 −13L ⎤
⎢ 2⎥
ρ AL ⎢ 22 L 4 L2
13L −3L ⎥
[mt ] =
420 ⎢ 54 13L 156 −22 L ⎥
⎢ ⎥
⎢⎣ −13L −3L2 −22 L 4 L2 ⎥⎦
 CONSISTENT MASS MATRICES
The kinetic energy of an element for the rotational behavior
can be written as Le
Te = 12 ∫ ρ Iθz2 dx
0

where I is the moment of inertia of the cross section about the

neutral axis. The rotational velocity can be written as
∂v ∂
θz = = [ N ]{q } = ⎡⎣ N , x ⎤⎦ {q }
∂x ∂x
where d
⎡⎣ N ,x ⎤⎦ = [ N ]
dx
CONSISTENT MASS MATRICES
The mass matrix can be obtained as
Le Le
( )
T
Te = 1
2 ∫ ρ Iθ 2 dx =
z
1
2 ∫ ρ I ⎡⎣ N, x ⎤⎦ {q } ⎡⎣ N, x ⎤⎦ {q } dx
0 0
⎡ Le ⎤
{q }T ⎢ ∫ ρ I ⎡⎣ N , x ⎤⎦ ⎡⎣ N , x ⎤⎦ dx ⎥ {q }
T
= 1
2
⎢⎣ 0 ⎥⎦
After integrations
⎡ 36 3L −36 3L ⎤
⎢ 2⎥
ρ I ⎢ 3 L 4 L −3 L − L ⎥
L 2
[mr ] = ∫ ρ I ⎡⎣ N,x ⎤⎦ ⎡⎣ N,x ⎤⎦ dx =
T
30 L ⎢ −36 −3L 36 −3L ⎥
0 ⎢ ⎥
⎢⎣ 3L − L2 −3L 4 L2 ⎥⎦
CONSISTENT MASS MATRICES
TORSIONAL ELEMENT
The kinetic energy of a torsional element can be written as
Le
Te = 1
2 ∫ ρ Jθ 2 dx
x
The polar moment of inertia is J = ∫ ( y 2
+ z 2
)dA
Ae
0
The twist, θx, and its time derivative can be expressed as

θ x = [ N ]{q} and θx = [ N ]{q }

CONSISTENT MASS MATRICES
The mass matrix can be deduced as
Le ⎡ Le ⎤
∫ ρ Jθ 2 dx = {q } ⎢ ∫ ρ J [ N ] [ N ] dx ⎥ {q }
T T
Te = 1
2 x
1
2
0 ⎢⎣ 0 ⎥⎦

[ N ] = ⎢⎣ N1 N 2 ⎥⎦

L
ρ JL ⎡ 2 1 ⎤
[m ] = ∫ ρ J [ N ] [ N ] dx =
T

0
6 ⎢⎣1 2 ⎥⎦
CONSISTENT MASS MATRICES
CST ELEMENT
For the plane stress, plane strain conditions for the CST
element, the shape function matrix is
⎡ N1 0 N2 0 N3 0 N4 0 ⎤
[N] = ⎢ 0 N 4 ⎥⎦
⎣ N1 0 N2 0 N3 0

The element stiffness matrix is then given by

[m ] = ρ te ∫ [ N ] [ N ] dA
T

Ae
CONSISTENT MASS MATRICES
Noting that, ∫A N12dA = 16 Ae , ∫A N1N 2dA = 121 Ae , etc. We have
e e

⎡2 0 1 0 1 0⎤
⎢0 2 0 1 0 1 ⎥⎥
⎢
ρ tA ⎢1 0 2 0 1 0⎥
[ m] = ⎢ ⎥
12 ⎢ 0 1 0 2 0 1⎥
⎢1 0 1 0 2 0⎥
⎢ ⎥
⎣0 1 0 1 0 2⎦
 CONSISTENT MASS MATRICES
TETRAHEDRAL ELEMENT
The shape function matrix is

⎡ N1 0 0 N2 0 0 N3 0 0 N4 0 0 ⎤
[ N ] = ⎢⎢ 0 N1 0 0 N2 0 0 N3 0 0 N4 0 ⎥⎥
⎢⎣ 0 0 N1 0 0 N2 0 0 N3 0 0 N 4 ⎥⎦

N1 = ξ , N2 = η, N3 = ζ , N4 = 1− ξ −η − ζ
CONSISTENT MASS MATRICES
The mass matrix of the element is then given by,
⎡2 0 0 1 0 0 1 0 0 1 0 0⎤
⎢ 2 0 0 1 0 0 1 0 0 1 0⎥
⎢ ⎥
⎢ 2 0 0 1 0 0 1 0 0 1⎥
⎢ ⎥
⎢ 2 0 0 1 0 0 1 0 0 ⎥
⎢ 2 0 0 1 0 0 1 0⎥
⎢ ⎥
ρV
[m ] = e ⎢⎢ ⎥
2 0 0 1 0 0 1
20 2 0 0 1 0 0⎥
⎢ ⎥
⎢ 2 0 0 1 0⎥
⎢ 2 0 0 1⎥
⎢ ⎥
⎢ 2 0 0⎥
⎢ ⎥
⎢ 2 0 ⎥
⎢⎣ 2 ⎥⎦
 EIGENVALUE PROBLEM
Free vibration problem is an eigenvalue problem.
[ K ]{Q} = λ [ M]{Q}
λ= ω2 is the eigenvalue
Q is the eigenvector and indicates the mode shape.

[K] and [M] are symmetric matrices.

[K] is positive definite for properly constrained problems.
 PROPERTIES OF EIGENVECTORS
For a positive definite symmetric stiffness matrix of size n,
there are n real eigenvalues and corresponding
eigenvectors.
The eigenvalues may be arranged in ascending order:
0 ≤ λ1 ≤ λ2 ... ≤ λn

If {Q}1 ,{Q}2 ,...,{Q}n are the corresponding eigenvectors,

we have
[ K ]{Q}i = λi [ M ]{Q}i
 PROPERTIES OF EIGENVECTORS
The eigenvectors posses the property of being orthogonal
with respect to both the stiffness and mass matrices:
{Q}i [M]{Q} j = 0
T
if i ≠ j

{Q}i [K ]{Q} j = 0
T
if i ≠ j
The lengths of eigenvectors are generally normalized so that
{Q}i [M ]{Q}i = 1
T

The above normalization of the eigenvectors leads to the relation

{Q}i [K ]{Q}i = λi
T

In many codes, other normalization schemes are also used. The length of an
eigenvector may be fixed by setting its largest component to a preset value, say unity.
 EIGENVALUE-EIGENVECTOR
EVALUATION
The eigenvalue-eigenvector evaluation procedures fall
into the following basic categories:
 Characteristic polynomial technique
 Vector iteration methods
 Transformation methods
Characteristic polynomial technique
([K ] − λ [M ]){Q} = {0}
If the eigenvector is to be nontrivial, the required condition is
det ([ K ] − λ [ M ]) = 0
EXAMPLE
Determine the eigenvalues and eigenvectors for the
stepped bar shown in the Figure. E = 200 GPa, ρ = 7800
kg/m3.
A1 = 600 mm2
A2 = 300 mm2

L1 = 300 mm L2 = 150 mm
 Solution
Gathering the stiffness and mass values corresponding to the
degrees of freedom Q2 and Q3, we get the eigenvalue problem:
⎡⎛ A1 A2 ⎞ A2 ⎤
⎢⎜ + ⎟ − ⎥
3 ⎥ ⎧Q2 ⎫ ρ ⎡ 2 ( A1L1 + A2 L2 ) A2 L2 ⎤ ⎧Q2 ⎫
E ⎢⎝ 1 2⎠
L L L
⎨ ⎬ = λ ⎢ ⎥⎨ ⎬
⎢ A2 ⎥
A2 ⎩ 3 ⎭
Q 6 ⎣ A L 2 A2 2 ⎦ ⎩Q3 ⎭
L
⎢ −
2 2
⎥
⎣ L2 L2 ⎦

Substituting the values we get

−3 −3 ⎤ −4
⎡ 4 x10 − 2 x10 ⎧Q ⎫ ⎡ 4.5 x10 4.5 x10−5 ⎤ ⎧Q2 ⎫
⎥ ⎨ ⎬ = 1300λ ⎢
2
200 x109 ⎢ ⎥⎨ ⎬
−3 −3
⎢⎣ −2 x10 2 x10 ⎥⎦ ⎩Q3 ⎭ ⎢⎣ 4.5 x10−5 −5
9 x10 ⎥⎦ ⎩Q3 ⎭
 Solution
The characteristic equation is
⎡ 8 x108 − 0.585λ −4 x108 − 0.0585λ ⎤
det ⎢ ⎥=0
⎢⎣ −4 x108 − 0.0585λ 4 x108 − 0.117λ ⎥⎦
which simplifies to
0.065λ 2 − 3.744 x108 λ + 16 x1016 = 0
The eigenvalues are
λ1 = 4.649 x108
λ2 = 52.95 x108

Note that λ = ω2, where ω is the circular frequency given by 2πf, f = frequency in hertz.
 Solution
The above frequencies are
f1 = 3432 Hz
f 2 = 11581 Hz
The eigenvector for λ1 is found from
([K ] − λ1 [M ]){Q}1 = {0}
which gives
8⎡ 5.280−4.272 ⎤ ⎧Q2 ⎫ ⎧0 ⎫
10 ⎢ ⎥ ⎨ ⎬ =⎨ ⎬
⎣ −4.272 3.456 ⎦ ⎩Q3 ⎭1 ⎩0 ⎭
 Solution
The two equations above are not independent since the
determinant of the matrix is zero. This gives
5.280Q2 = 4.272Q3
Thus,
{Q}1 = ⎢⎣Q2 1.236Q2 ⎥⎦
T

For normalization, we set

{Q}1 [M]{Q}1 = 1
T

On substituting {Q}1 for we obtain

{Q}1 = ⎢⎣1.049 1.297 ⎥⎦ T
 Solution
The eigenvector for λ2 is found from
([ K ] − λ2 [M ]){Q}2 = {0}
which gives
⎡ −22.98 −7.097 ⎤ ⎧ A2 ⎫ ⎧0 ⎫
⎨ ⎬ =⎨ ⎬
8
10 ⎢ ⎥
⎣ −7.097 −2.195 ⎦ ⎩ A3 ⎭2 ⎩0 ⎭
This matrix equation gives
−22.98Q2 = 7.097Q3
Hence, we obtain the mode shape
{Q}2
T
= ⎢⎣Q2 −3.238Q2 ⎥⎦

After the normalization {Q}2 = ⎢⎣0.835 −2.705⎥⎦ T

 Solution
The mode shapes of vibrations
Mode 1

Mode 2
 REDUCTION METHODS
 The basic idea of matrix condensation is simply Gaussian
elimination of chosen displacements to reduce the size of
a problem.
 In static analysis no loss of accuracy results from such a
reduction because the dependent displacements are
recovered (exactly) in the back-substitution phase.
 In dynamic (or vibrational) analysis is smilar type of
condensation can be used to reduce the number of
degrees of freedom, but a new type of approximation is
involved.
REDUCTION METHODS
Let us discuss with stiffness (or static) reduction and rewrite
the equation of motion in expanded form, as follows,

⎡[ M AA ] [M AB ]⎤ ⎧⎪{Q A}⎫⎪ + ⎡[K AA ] [K AB ]⎤ ⎪⎧{Q A}⎪⎫ = ⎪⎧{0}⎪⎫

⎢
⎣[ M BA ] [ M BB ]⎥⎦ ⎪⎩⎨{Q B }⎪⎭⎬ ⎢⎣[K BA ] [K BB ]⎥⎦ ⎨⎪⎩{Q B }⎬⎪⎭ ⎨⎩⎪{0}⎬⎭⎪

In this equation the subscript A denotes the displacements

that are to be eliminated, while the subscript B refers to
those that will be retained.
REDUCTION METHODS
Let the accelerations {Q
 } and {Q
A
 } be null, and write the
B
remaining static equations as two sets:
[K AA ]{Q A} + [K AB ]{Q B } = {0}
[K BA ]{Q A} + [K BB ]{Q B } = {0}
Solve the first equation for the vector {Q A}
−1
{Q A} = − [K AA ] [ K AB ]{Q B }
Substitute into the second equation,
−1
− [ K BA ][ K AA ] [K AB ]{Q B } + [K BB ]{Q B } = {0}
 REDUCTION METHODS
The equations of motion have been reduced to a smaller
set, having the same order as [KBB].
⎡K *BB ⎤ {Q B } = {0}
⎣ ⎦
in which
⎡K*BB ⎤ = [ K BB ] − [ K BA ][ K AA ]−1 [ K AB ]
⎣ ⎦

Moreover, the following equation can be used to find vector

{QA} from {QB}.
−1
{Q A } = − [ K AA ] [ K AB ]{Q B }
 REDUCTION METHODS
Turning next to the vibrational problem, we shall now
consider mass (or dynamic) reduction.

As a new approximation, assume that the acceleration

vector {Q }
A is dependent upon {Q B } in the same manner

that the vector {Q A} is related to {Q B }

{ }
 = − [ K ]−1 [ K ] Q
Q A AA AB

B{ }
REDUCTION METHODS
Then equate the virtual work done by the inertial actions in
the reduced system to that of the inertial actions in the
original system. Thus,
∗
δ {Q B } [ M BB ] {Q B } = δ [Q ] [ M ][Q ]
T T

= ⎢δ {Q A }
T T ⎥ ⎡[ M AA ]
δ {Q B } ⎢
[M AB ]⎤ ⎧⎪{Q A}⎫⎪
⎣ ⎦ ⎣[ M BA ] [M BB ]⎥⎦ ⎨⎪⎩{Q B }⎬⎪⎭
But we can write from static condensation,
−1
δ {Q A} = − [ K AA ] [ K AB ]δ {Q B }
Let us substitute the expressions for {Q
 } and δ {Q } into
A A
the above equation.
REDUCTION METHODS
⎡M*BB ⎤ = [ TB ]T [ M ][ TB ]
⎣ ⎦
In this equation the transformation matrix [TB] is
⎡ − [ K ]−1 [ K ]⎤
[TB ] = ⎢ AA AB
⎥
⎢⎣ [I B ] ⎥⎦
where [IB] is an identity matrix of the same order as [MBB].
Due to the virtual work equality in the previous slide, the
mass terms in matrix [M*BB] are energy-equivalent to those
in the original mass matrix [M]. However, the reduction in
size caused by the above equation represents an additional
approximation inherent to the method.
EXAMPLE
The cantilever beam in the figure is composed of two
prismatic flexural elements, both of which have the same
flexural rigidity EI. Elimininate the rotational degrees of
freedom. Solve the natural frequencies and mode shapes.
y

3 5
1
1 2 3
1 2 x
2
4 6

l l
z
L
 Solution
For this arrangement of elements, the nodal stiffness matrix
for the unrestrained beam is
⎡ 12 6l −12 6l 0 0 ⎤ 1
⎢ ⎥
⎢ 6l 4l
2
−6l 2l 2 0 0 ⎥ 2
⎢ −12 6l ⎥
EI ⎢ −12 −6l 24 0
[K ] = 3 ⎢ ⎥
3
2⎥
(a)
l 6l 2l 2 0 8l 2
−6l 2l 4
⎢ ⎥
⎢ 0 0 −12 −6l 12 −6l ⎥ 5

⎢ ⎥
⎣ 0 0 6l 2l 2 −6l 4l 2 ⎦ 6

1 2 3 4 5 6
Solution
Similarly, the nodal mass matrix takes the form

⎡ 156 22l 54 − 13l 0 0 ⎤ 1

⎢ 22l 4l 2 13l − 3l 2 0 0 ⎥⎥
⎢ 2

ρAl ⎢ 54 13l 312 0 54 − 13l ⎥

[M ] =
3
⎢ ⎥ (b)
420 ⎢− 13l − 3l 2 0 8l 2 13l − 3l 2 ⎥ 4
⎢ 0 0 54 13l 156 − 22l ⎥ 5
⎢ 2 ⎥
⎢⎣ 0 0 − 13l − 3l 2 − 22l 4l ⎥⎦ 6

1 2 3 4 5 6
 Solution
The objective of this example is to show how we can
eliminate the rotational degrees of freedom while retaining
the translational dof in a beam.
As the first step, we remove the fifth and sixth rows and
columns from matrices [K] and [M] because displacements
five and six in the figure are restrained by supports. Then
the remaining 4 x 4 arrays are rearranged to put the
rotational terms before the rotational terms before the
translational terms, as follows
Solution
⎡ 4l 2 2l 2 6l −6l ⎤ 2
⎢ ⎥
⎡[ K AA ] [K AB ] = EI ⎢ 2l 2 8l 2
⎤ 0 ⎥
[K ] = ⎢ K 6l 4
(c)
⎣[ BA ] [K BB ]⎦⎥ l 3 ⎢⎢ 6l 6l 12 −12 ⎥
⎥ 1

⎢⎣ −6l 0 −12 24 ⎥⎦ 3

2 4 1 3

⎡ 4l 2 −3l 2 13l ⎤
22l 2
⎢ ⎥
⎡[ M AA ] [M AB ]⎤ = ρ Al ⎢ −3l 2 8l 2 −13l 0 ⎥
[M ] = ⎢ M
4

⎣[ BA ] [M BB ]⎥⎦ 420 ⎢⎢ 22l −13l 156 54 ⎥

⎥ 1
(d)
⎢⎣ −13l 0 54 312 ⎥⎦ 3

2 4 1 3
Solution
The inverse of submatrix [KAA] taken from Eq. (c), is
−1 l ⎡ 4 −1⎤
[K AA ] = ⎢ −1 2 ⎥ (e)
14 EI ⎣ ⎦
Substituting this array and the other submatrices of [K] from
Eq. (e) the following equation
⎡ K *BB ⎤ = [ K BB ] − [ K BA ][ K AA ]− 1 [ K AB ]
⎣ ⎦

and then we obtain

⎡K *BB ⎤ = 12 EI ⎡ 1 −1⎤ 18 EI ⎡ 4 −3⎤ 6 EI ⎡ 2 −5 ⎤ 1
⎣ ⎦ ⎢ −1 2 ⎥ − 3 ⎢ −3 4 ⎥ = 3 ⎢ −5 16 ⎥ (f)
l3 ⎣ ⎦ 7l ⎣ ⎦ 7l ⎣ ⎦ 3

which is the reduced stiffness matrix. 1 3

 Solution
To reduce the mass matrix, we form the transformation
matrix [TB].
−1 3 ⎡ −3 4 ⎤
− [ K AA ] [K AB ] = ⎢ (g)
7l ⎣ −1 −1⎦⎥
⎡ −9 7l 12 7l ⎤
⎡ − [ K ]−1 [ K ]⎤ ⎢ −3 7l −3 7l ⎥
[TB ] = ⎢ AA AB = ⎢
⎥ ⎥
⎢⎣ [I B ] ⎥⎦ ⎢ 1 0 ⎥ (h)
⎢ ⎥
⎣ 0 1 ⎦

In this case the submatrix [IB] in the lower partition of matrix [TB] is of order 2
because there are two translational displacements remaining (numbers 1 and 3).
Solution
Using the matrix [M] from Eq. (d) and [TB] from Eq. (h), we
obtain the reduced mass matrix
⎡M BB ⎤ = [ TB ] [ M ][ TB ] ==
* T ρ Al ⎡ 5652 3615 ⎤ 1
⎣ ⎦ ⎢
20580 ⎣3615 18336⎦ 3 ⎥ (i)
1 3
.
Now, let us write the eigenvalue problem as follows
6 EI ⎡ 2 − 5⎤ ⎧Q1 ⎫ ρAl ⎡5652 3615 ⎤ ⎧Q1 ⎫
7l 3 ⎢− 5 16 ⎥ ⎨ ⎬ = λ 20580 ⎢3615 18336⎥ ⎨ ⎬ (j)
⎣ ⎦ ⎩Q3 ⎭ ⎣ ⎦ ⎩Q3 ⎭

in which λ = ω 2
 Solution
Let us define a parameter
ρAl 4
λ = ω2 (k)
EI
and arrange the equation
⎡ 1.7143 − 0.27464λ − 4.2857 − 0.17566λ ⎤ ⎧Q1 ⎫ ⎧0⎫
⎢ . ⎥⎨ ⎬=⎨ ⎬
⎣− 4.2857 − 0.17566λ 13.714 − 0.89096λ ⎦ ⎩Q3 ⎭ ⎩0⎭ (l)

The nontrivial solutions of Eq. (l) exist only if the following

condition is fulfilled
⎡ 1.7143 − 0.27464λ − 4.2857 − 0.17566λ ⎤
det ⎢ ⎥=0 (m)
⎣− 4.2857 − 0.17566λ 13.714 − 0.89096λ ⎦
Solution
The determinant gives the following characteristic equation
0.21384 λ 2 − 6.7994 λ + 5.1427 = 0 (n)
The solution of Eq. (n) is
λ1 = 12.4045 λ2 = 496.40 (o)
.

Substitution of these values and l = L/2 into Eq. (k) gives

1 EI 1 EI
ω1 = 3.522 ω2 = 22.28
L2 ρA L2 ρA

These angular frequencies are in error by +0.17 percent and +1.1 percent, respectively.
Solution
The corresponding mode shapes may be found by
substituting and into the homogeneous equations [see Eq.
(m)]. Thus,

⎡1.0000 1.0000 ⎤
[Q ] = [{Q}
1 {Q}2 ].= ⎢
− ⎥
⎣ 0 . 3395 0 . 6991⎦

In this modal matrix each modal vector is listed columnwise

and normalized with respect to its larger component.
Solution
Sketches of the mode shapes

First mode

0.3395

Second mode
1

-0.6991