Professional Documents
Culture Documents
Approximate Methods For Multi-Dof and Continuous Systems
Approximate Methods For Multi-Dof and Continuous Systems
Lecture 1
Approximate methods
In exact method difficulties arises in
Solving roots of the characteristic equation. Except for very simple boundary
conditions, one has to go for numerical solution.
In determining the normal modes of the system
Determination of steady state response
So for quick determination of the natural frequencies of a system, when a very accurate
result is not of much importance one should go for an approximate method.
Modeling
Approximation
. Series solution
Approximate method where approximation error should be within acceptable limits one may
assume a series solution as
�
u ( x, t ) = �f n (t ) fn ( x) (1)
n =1
where fn ( x ) is the normal modes and f n (t ) is the time function which depends upon initial
conditions and forcing function. There are certain difficulties that limit the application of
classical analysis of continua to a very simple geometry only.
The infinite series sometimes converge very slowly and it is difficult to estimate how
many terms are needed for engineering accuracy.
The formulation and computation efforts are prohibitive for systems of engineering
complexity.
The special methods (as approximate methods) treat the continuous systems, for vibration
analysis purpose, as discrete systems. This can be done with one of the following methods.
Taking only n natural modes and considering them as generalized coordinates
and then computing the n weighing functions fn(t) to best fit the initial
conditions or the forcing functions.
Considering n known functions fn(x) that satisfy the geometric boundary
conditions of the system and then computing the functions f n(t) to best fit the
differential equation, the remaining boundary conditions, and the initial
conditions of the forcing functions.
Taking the physical coordinates of n number of points of the system q, (q1, L qn)
as the generalized coordinates, considering them as functions of time, and
computing them to fit the differential equation; and the initial and boundary
conditions.
The main advantage of all theses methods is that instead of dealing with one or more
partial differential equations, one deals with a larger number of ordinary differential
equations usually liner with constant coefficients, which are particularly suitable for
solution in fast computing machines.
Rayleigh’s Method
-Rayleigh method gives a fast and rather accurate computation of the fundamental frequency
of the system.
-It applies for both discrete and continuous systems.
Consider a discrete, conservative system described by the matrix equation
Mx Kx = 0 (2)
The equation above is satisfied by a set of n eigenvalues i and normalized eigenvectors f i ,
2
we have
f�i K fi
i2 = (4)
f�
i M fi
If we know the eigenvector f i , we can obtain the corresponding eigenvalue i by eqn(4).
2
However in general, the eigenvectors are not known and one has to find it for the particular
system. Suppose that we consider an arbitrary vector Z in eqn (4). So eqn(4) can be written as
z�
Kz
2 = R ( z ) = . (5)
z�
Mz
Here R(z) depends on the vector z and is called Rayleigh’s quotient . When the vector z
From vector algebra it is known that vector z can be expressed as a linear combination of
independent vectors.
n
z = c1 z1 c2 z2 L cn zn = �ci zi = Zc (6)
i =1
where Z is a square modal matrix [ z1 z2 L zn ] and c = [c1c2 L cn ] . If the vector zi have been
normalized so that
Z T MZ = I , then Z T KZ = diag [1 , 2 ........n ] = P
2 2 2
(7)
using (7) in (6) and using orthogonal Property
n
c��
Z KZc c�
Pc �c i
2
i
R( z ) = = = i =1
(8)
c��
Z MZc c�
Ic n
�ci 2
i =1
Equation (6) is similar to the free vibration response of a system which contains all the normal
modes. In the assumed function let the r th mode deviates from the actual mode. So the cr will
ci
be larger in comparison to the other ci ’s. Now taking = ei , ei = 1 , equation (8) can be
cr
written as
2
n
�c � n
�� i �i2 2r �ei 2 i2
2
r
i =1 �cr �
Rr ( z ) = 2
= n
i =1
i �r (9)
n
�ci �
� � � � ei 2
i =1 � cr � i =1
For r =1,
n
1 �ei 2 i2
R1 ( z ) = 12 i=2
n (10)
1 �ei 2
i =1
From equation (9) it is apparent that the Rayleigh quotient differs from the eigenvalues r by
2
second order terms of the error of the eigenvector. In other words, if the error in selecting the
nonexact eigenvector is e , then the error in the eigenvalues is e . So a 20% error in the form of
eigen vector will yield a 4% error in the eigenvalues and a corresponding 10 % error in eigen
value will result in 1% error in eigenvector. From eqn(10), it may also be concluded that
R1 ( z ) �12 (11)
Hence the Rayleigh’s quotient is never lower than the smallest eigenvalues. Hence this method
gives the upper bound approximation of the fundamental frequency of the system.
Rayleigh’s principle can be stated as in a conservative system the frequency of vibration has a
stationary value in neighbourhood of a natural mode.
Example: Using Rayleigh quotient method, find the fundamental frequency for a cantilever
beam assuming the approximate function as the static deflection curve.
Static deflection of a cantilever beam can be found using bending equation as follows.
x d2y
M = wx. = EI where w is the load per unit length.
2 dx 2
dy 1 wx 3
or, EI = C1
dx 2 3
1 wx 4
EIy = C1 x C 2
6 4
y=0
1 3
BCs x = l dy C1 = wl
=0 6
dx
1 wx 4
C 2 = C1 x x = l
6 4
wl 4 4 wl 4
= wl =
24 6 8
y=
w
24 EI
x 4 4l 3 x 3l 4 (Deflection from free end)
y =
w
24 ET
(l x ' ) 4 4l 3 (l x ' ) 3l 4
=
w
24 EI
l 3 x 3 3l 2 x 3lx 2 4l 3 replacing x’ by x
=
w
24EI
3l 4 x 3 l 3l 3 x 3l 2 x 2 3l 3 x x 4 3l 2 x 2 3lx 3 3l 4
=
w
24 EI
x 4 4lx 3 6l 2 x 2
2
1 l ��2 y �
Potential energy = � EI � 2 �dx
2 0 �� x �
1 2 l
2 �
Kinetic energy T = m( x) y 2 dx
o
1 l w
2 �
0
EI .
2 EI
( x l )4 dx
I1
� 2 = 2
=
1 l w dx I2
�
2 o
m( x 4 4lx 3 6l 2 x 2 )2 2 2 2
24 E I
o t x = t dt = dx
1 EIw 4
I1 = t dt x = 0, t = l
2 l 2 EI x=l t=o
o
1 w t5
= E
2 2 5 l
=
1 w
2 10
o l 5
=
wl 5
20
1 mw 2 l
I2 = (x 16l 2 x 6 36l 4 x 4 2 4lx 7 48l 2 x 5 12lx 6 )dx
8
2 (24 EI ) 2 0
l
1 x 9 16 2 7 36 4 5 8 lx 8 8 4 8 l 3 x 6 12l 2 x 7 w2
= m l x l x
2 9 7 5 8 6 7 o (24 EI ) 2
1 l 9 16l 9 36l 9 8 l 9 12 w 2
= m 8l 9 l 9
2 9 7 5 8 7 (24 EI ) 2
1 l 9 1 16 36 12 w2
= m 1 8 2
2 9 17 5 7 (24 EI )
0.96657 ml 9 w 2
=
2 (24 EI ) 2
I1 l 5 �2 �(24 EI ) 2
� 2 = =
I 2 20 �mw2 �0.96657t 9
59.591963( EI ) 2
= 2
Ans
ml 5
Fig2
g g
The max potential energy = (m1 y1 m1 y1 ......) = �my
2 2
1 1 1
Max Kinetic Energy. = m1v12 m2v2 2 m3v32 ......
2 2 2
1 1 1
= m1 ( y1 ) 2 m2 ( y2 ) 2 m3 ( y3 ) 2 ......
2 2 2
1
= 2 �my 2
2
Now equating the maximum kinetic energy to the maximum potential energy
g 1
=>
2
�my = 2 �my 2
2
g �my
=> n 2 =
�my 2
Unlike Dunkerley’s formula, which is valid for lateral vibration of shafts only, Rayleigh’s
method is valid for a system performing oscillatory motion in any manner i.e., bending,
torsional or longitudinal motions.
Example2: Find the fundamental frequency of the simply supported beam of length l carrying
three discs of mass m, 2m and m equidistantly placed from the left end.
Figure 3.
Solution
Consider the shaft carrying three discs as shown in the figure. The influence coefficients are,
3l 3 l3 3l 3
a11 = , a22 = , a33 =
256 EI 48 EI 256 EI
Influence Coefficients:
aij deflection at station i due to
unit load at station j
Using Dunkerley’s formula
1 3ml 3 1 2ml 3 1 3ml 3
= , = , =
12 256 EI 2 2 48 EI 32 256 EI
EI
n = 3.9191
ml 3
Rayleigh method:
Flexibility influence coefficient
displacement at i due to unit load
9l 3 11l 3 7l 3 at j with all other forces equal to
a11 = , a12 = , a13 = zero.
768 EI 768EI 768 EI
16l 3 11l 3 9l 3
a22 = , a23 = , a33 =
768 EI 768 EI 768EI
m1 = m3 = m and m2 = 2m
EI
( which is slightly higher than the exact value 16.199 )
ml3
Example:
A steam turbine blade of length l, can be considered as a uniform cantilever beam, mass
m per unit length with a tip mass M. The flexural rigidity of the blades is EI. Determine
fundamental bending frequency.(Use Rayleigh Method)
Sol: g
n =
D
Assuming Y(x,t)=Y(x)cos t
g g .256 EI 256 EI
=> n 2 = = =
px D 3mgl 3 3mgl 3
Y(x)=A(1-cos )
2l
� px
Y( x, t ) = Y(x)sin t=- A(1-cos )sin t
2l
l
1 �2
K .E. = T = �m y dx
0
2
l
1 px
= �m 2 (1 cos ) 2 A2 dx
0
2 2l
mA2 2 px
l
=
2 0 �
(1 cos ) 2 dx
2l
� � �p x ��
l
l�
� 2sin px �
mA2 2 � l � �2l �� x l sin 2( )
x � � �� 2l �
2 �0 � p � 20 p �
� � 2l � 2( ) �
� �0 2 0
� �
mA2 2 �
�3 � 2l l �
= ��
� l .2(1 0) (0 0) �
2 � �2 � p p �
mA2 2 �3 4l � 2 2 3 2
= � l �= mA l ( )
2 �2 p � 4 p
1 � 1 1
M ( y (l )) = M { A} = M 2 A2
2
The K.E. of tipmass =
2 2 2
2 2 � 3 2�1
The K.E. of the system = mA l � � M A
2 2
�4 p � 2
2
1
l
�d 2 y � p 4 EI 2
P.E. V= EI � � � dx = A
2 0 �dx 2 � 64 l 3
Strain Energy
Lecture 2
n1, n2 ...... = Frequency of transverse vibration when each of W1, W2,W3…. act alone.
1 1 1 1
2
= 2 2 .......... 2
fn f n1 f n2 f ns
EI
=> n = np 4
r L4
for simply supported beam with uniformly distributed load, maximum deflection occur at
midpoint.
5WL3
D= W = total weight
384 EI
= 5 r gL4 / 384 EI
EI 5g
So, =
r L 384D
4
5g
Hence, n2 = n2p 2
384D
Similarly for a fixed-fixed beam with loading the maximum deflection can be given by
r gl 4 EI g
D= � 4 =
384 EI rl 384D
EI
In this case for the first mode n = 22.4
2
r L4
g g
So, n2 = 22.4 = 1.143
384D 384D
For Cantilever Beam
r gl 4 EI g
D= � 4 =
8 EI rl 8D
EI
In this case for the first mode n = 3.52
2
r L4
g g
So, n2 = 3.52 = 1.2445
8D D
In case of concentrated loading the natural frequencies can be determined from the relation
g
n2 = , where D is the deflection under that load. One may note for the commonly used
D
cases.
a b
a b a b
Wa 2b 2 Wa 3
D= Wa 3b3
3EIl D= D=
3EIl 3 3EI
Example2: Find the fundamental frequency of the simply supported beam of length l carrying
three discs of mass m, 2m and m equidistantly placed from the left end.
Figure 3.
Solution
Consider the shaft carrying three discs as shown in the figure. The influence coefficients are,
3l 3 l3 3l 3
a11 = , a22 = , a33 =
256 EI 48 EI 256 EI
Influence Coefficients:
Using Dunkerley’s formula aij deflection at station i due to
1 3ml 3 1 2ml 3 1 3ml 3 unit load at station j
= , = , =
12 256 EI 2 2 48 EI 32 256 EI
EI
n = 3.9191
ml 3
Example 3:
Beam m1
M
x
L
The natural frequency of a cantilever beam of negligible mass with a concentrated mass M
attached is
1 3EI
2 = mL3
11
And
the natural frequency of a cantilever beam of mass m1 is
1 EI
= 12.7
2
m1 L3
22
1 3EI 12.7 EI
2
= mL3
+
m1 L3 = 0 .412
mL3
EI
1
In the case of transverse vibration of beams, if n functions are chosen for approximating the
deflection W ( x) , can be written as
w( x ) = c1w1 ( x) c2 w2 ( x) L cn wn ( x)
Where, w1 ( x), w2 ( x),L wn ( x) are linear independent functions of the spatial coordinate x which
satisfy the boundary condition of the problem, and c1 , c2 L , cn are the coefficient to be found.
As the Rayleigh quotients have stationary value near the natural mode by differentient by
differentiating the Rayleigh quotient with respect to these coefficients will yield a set of
homogeneous algebraic equations, which can be solved to obtain the frequencies.
Example: Find the modal frequencies of a tapered cantilever beam of maximum height h,
length l with unity width.
l
Solution
Tapered cantilever beam
maximum height h, width =unity., length l
hx
Area of cross section A( x ) =
l
3
1 �hx �
Moment of inertia at any section I ( x) = � �
12 �l �
By using one term approximation, same result comes out as in case o f Rayleigh’s method
l 2
�d 2 w( x) �
� EI ( x ) � 2 �dx
� dx � = X
R= =
2 0
2 --------------1(e)
l Y
r A( x) w( x) dx
2
� 0
X Y
Y X
( )2
c1 c1
= = 0 -------------1(h)
c1 Y 2
and
X Y
Y X
( )2
c 2 c 2 --------------1(i)
= =0
c 2 Y 2
Substituting the equations 1(f,g) into the equations 1(h,i), can be written as
��1 1 � �1 2 � �
��2 15 l � �5 105 l �
��c � ��
0
�� � � �
��1 �= ��--------------1(j)
��1 2 � �1 1 � ��c2 0
�
�� l� � l��
��5 105 � �5 140 � �
32r l 4
Where l =
Eh 2
Exercise Problem : Find the first three mode frequencies of the tapered beamby using
Rayleigh-Ritz method considering the following approximate function.
2
x
w1 ( x) = 1 ---------1(a)
l
2
x x
w2 ( x ) = 1 ------------------1(b)
l l
2 2
x x
w3 ( x) = 1 --
l l
Lecture 3
4. Galerkin’s method
In Galerkin’s method the residue obtained by using the assumed mode in the governing
differential equation is minimized. Let the assumed shape function of the system be written as
n
( x ) = �ci i ( x) (a)
i =1
where i (x ) is the approximate solution of the differential equation. For example considering
the lateral vibration of a beam the differential equation of motion can be written as
d 4 f( x) m2
L( x) = f( x) = 0 .(b)
dx 4 EI
Similarly for torsional vibration of rod, longitudinal vibration of rod and lateral vibration of
taut string one use the following equation.
2
d 2 f( x) � �
L( x ) = � �f( x) = 0 (c)
dx 2 �c �
Here f ( x ) is the eigenfunction of the system. When an approximate function i (x ) is taken,
then it will not satisfy the above equation.
For each function, making the residual ( Ri ) equal to zero one may obtain the frequencies.
l
Ri = �
ci i ( x)dx = 0 (d)
0
Example
X=0 X=L
d 4 f( x) m2 d 4f( x) 4
L( x) = f( x ) = b f( x)
dx 4 EI dx 4
L( x) = b4 x 4l 4 6 x 6l 2 4 x 5l 3 x8 4 x 7l 48 x3l 24 x 4 24 x 2l 2
l
1 49 4 5
So, residual R = �
L( x)f( x)dx = b l l =0
0
630 5
b4l 4 = 504
b l = 4.738
EI EI
Hence 1 = bl
2
4
= 22.45
ml ml 4
2px 4px
Now taking two admissible functions as 1 ( x) = 1 cos and 2 ( x) = 1 cos , writing
L L
� 2px � � 4px �
= c11 ( x) c2 2 ( x) = c1 �
1 cos � c2 �
1 cos �
� L � � L �
l
Ri = �
L( x)i ( x)dx = 0
0
l
2px
R1 = �
0
c1 (r 2 b2 ) cos rx b4 c2 (16r 2 b 2 ) cos 2rx b 4 (1 cos
L
)dx
l
4px �
R2 = � �
c1 (r 2 b2 ) cos rx b 4 c2 (16r 2 b 2 ) cos 2rx b4 �
1 cos
� L �
dx
�
0
By integration
�r 4 �
R1 = � 1.5b4 �
c1 b4 c2
� 2 �
R2 = b4 c1 (1.5b4 0.5r 4 )c2 = 0
� p4 3 4 �
8 4 b
� b4 ��c� �c�
�L 2 � 1 = 0 � A �1 �= 0
4 � �
� 3 4 p �c c2 �
b 128 4 ��2 � �
� b4
� 2 L �
Hence frequency equation are for b L = 4.741 and b L = 11 .140 for both ends fixed beam
1/ 2
�EI �
1 = 22.48 � 4 � , c1 = 23, c2 = 1 ,
�ml �
1/ 2
�EI �
and 2 = 124.1 � 4 � , c1 = 0.69, c2 = 1 .
� ml �
The two modes are
2px � 4px �
1 ( x) = 23(1 cos )�1 cos �
l � l �
2px � 4px �
2 ( x) = 0.69(1 cos )� 1 cos �
l � l �
mode shape. From this equation it may be noted that any normal mode when multiplied
with the dynamic matrix will reproduce itself. In matrix iteration method, assumed
displacement of the masses are used to get the calculated displacement. This is repeated till
equation (1) is satisfied.
Assume a value of the modal vector. (for example 3:2:1 for 3 dof system)
Substitute the assumed value in left hand side of equation (1) and simplify to obtain
a ratio (for example the obtained value is 4:3:1).
If the value obtained in step II is same as the assumed value, then it is accepted as
the correct modal value. Otherwise, the obtained value is substituted as the trial
value and the second step is repeated till the correct modal value is obtained.
After getting the modal values, from equation (1) the corresponding eigenvalue can
be obtained.
In general matrix iteration method would converge to the fundamental mode. If the
assumed system of displacements does not include the fundamental mode then the matrix
iteration will converge to the next higher mode contained in the assumed system of
displacements. Orthogonality principle is used to sweep out the unwanted modes from
assumed displacements.
In case of semi-definite systems, rigid body mode (zero frequency) is also present. For such
cases constraint matrices can be constructed to sweep out rigid body component of the
absolute motion.
� m �x � m �x � �
�0 2 � 2 � 3 �3 � �
� m1 �x1 � m1 �x1 � �
where S = 0� 1 0 �
� �
�0 0 1 �
� �
� �
S is known as the sweeping matrix, which will eliminate the presence of the fundamental
mode. Similarly one should substitute c1 = c2 = 0 to eliminate the first two normal modes
from the assumed vibration to determine the 3rd mode frequency.
Example
Consider a long beam with three masses as shown in figure 1. Determine the mode shapes
and natural frequencies of the system using matrix iteration method.
Solution
To Determine natural frequency and mode shape of multi degree of freedom system, the
influence coefficients are obtained as
1
a11 = a21 = a31 =
4k m x3
1
a11 = a21 = a31 = a12 = a13 = k
4k
1 1 3 x2
a22 = = 2m
4k 2 k 4 k
3 2k
a22 = a23 = a32 =
4k
1 1 1 7 3m x1
a33 = = 4k
4k 2k k 4k
The flexibility influence coefficient matrix can be written as
1 1 1�
�
1 �
a= 1 3 3�
� �.
4k
�
3 3 7�
� �
�3 0 0�
�0 2 0�
The mass matrix can be written as � �
�
�0 0 1 �
�
Hence the displacements at different position due to inertia forces are
x1 = a11 3mx&
1 a12 2mx2 a13 mx3
& &
& &&
x2 = a21 3mx&
1 a22 2mx2 a23mx3
& &
& &&
x3 = a31 3mx&
&1 a32 2mx2 a33mx3
&
& &&
Now substituting xi = w xi
2
�x1 � 3 2 1 ��x1 �
�
� � m2 � � �
�x2 �= 3 6 3�
� ��x2 �
�x � 4k 3 6 7 ��x3 �
�
����
3
Now taking
�x1 � 3 2 1 ��x1 �
�
� � m2 � � �
�x2 �= 3 6 3�
� ��x2 �
�x � 4k 3 6 7 ��x3 �
�
����
3
� 1 � 3 2 1 �� 1 �
� 11.7521 �
�
� � m2 � �� � m2 � �
�2.4893�= 3 6 3 ��
� 2.4893�= �29.2565�
� 4k 4k �
3.7735�
����� 3 6 7 ��
� 3.7735� 44.3503�
� 1 � � 1 �
� � 11.7521m2 � �
�2.4893 �= 2.4895 �
�
� 4k
3.7735�
��
�
3.7738 �
Hence, it may be noted that the assumed mode shape matches with the obtained mode
shape upto 3rd decimal. So one can take the first normal mode X 1 as
� 1 �
� �
X1 = �2.4895�
�
3.7738 �
�
11.7521m2
and =1
4k
k
or, l1 = 1 = 0.3404
2
m
To find the second mode, using the sweeping matrix
� m �x � m �x � �
�0 2 � 2 � 3 �3 � �
� m1 �x1 � m1 �x1 � � � 0 1.659 1.2579 �
S= � 0 1 0 �
�= 0 1 0 �
� � � �
�0 0 1 � ��0 0 1 �
�
� �
� �
So the new equation for second mode iteration is
�x1 � 0 1.659 1.2579 ��x1 �
3 2 1 ��
�
� � m2 � �� � �
�x2 �= 3 6 3��
� 0 1 0 � ��x2 �
�x � 4k �
3 6 7 ��
0 0 1 ��x3 �
������
3
� 1.0 � � 1.0 �
� � 2.7756m2 � �
�0.8367 �= �0.8361 �
� 4k
1.8996 �
��
�1.8988�
As the left hand side vector and right hand side vector are matching upto 3rd decimal we can
take the 2nd normal mode as
�x1 � � 1.0 �
� � � �
�x2 �= �0.8361 �
�x � �1.8988�
��3
2.7756m2
=1
4k
4 k k
Hence the 2nd eigenvalues is l 2 = 2 = = 1.4411
2
2.7756 m m
k
Or, the second mode frequency = 2 = 1.2
m
For the determination of third mode, one may impose the condition c1 = c2 = 0 .
3
c1 = �mi xi 1 xi = 0
i =1
3
c2 = �mi xi 2 xi = 0
i =1
c1 = m1 x1 1 x1 m2 x2 1 x2 m3 x3 1 x3
c2 = m1 x1 2 x1 m2 x2 2 x2 m3 x3 2 x3
c1 = 3x1 4.979 x2 3.7738 x3
or
c2 = 3x1 1.6722 x2 1.8998 x3
x1 = 1.5896 x3
x2 = 1.7157 x3
�x1 � �0 0 1.5896 ��x1 �
� � � � �
0 0 1.7157 �
�x2 �= � ��x2 �
�x � � 0 0 1 ��x3 �
����
3
Hence one may use the following sweeping matrix to eliminate the first two modes
0 0 1.5896 �
�
S2 = �
0 0 1.7157 �
� �
�
0
� 0 1 �
�
So for third mode the matrix iteration equation will be
�x1 � 3 2 1 ��
� 0 0 1.5896 ��x1 �
� � m2 � �� ��
�x2 �= � 0 0 1.7157 �
3 6 3 �� ��x2 �
�x � 4k �
3 6 7 ��
0 0 1 ��x3 �
������
3
It can immediately be observed that the third mode converges to the last column of the
sweeping matrix.
�1.5896 � �1.5896 �
� � m2 � �
�1.7157 �= 1.4746 �1.7157 �
� 1 � 4k � 1 �
��
�1.5896 �
� �
X3 = �1.7157 �
� 1 �
�
�1.5896 �
m2 � �
1.4746 �1.7157 �= 1
4k � �
� 1
4 k k
l 3 = 32 = = 2.7133
1.4742 m m
k
So, 3 = 1.6472 .
m