International Workshop on Computational Methods in Vibration and Acoustics

Department of Mechanical Engineering

Indian School of Mines (ISM), Dhanbad
16-21, March 2015

Novel Boundary Characteristic Orthogonal Polynomials Satisfying All Boundary Conditions to Study
Vibration of Structures
Rama B. Bhat
Professor
Department of Mechanical and Industrial Engineering
Faculty of Engineering and Computer Science
Concordia University, Montreal, Quebec, Canada
Abstract
Boundary Characteristic Orthogonal Polynomials (BCOP) were constructed by the author in 1985 to
study the vibration of plates with different boundary conditions using the Rayleigh Ritz method.
Products of BCOP along the two directions of the plate were used as deflection shapes. The first
member of the orthogonal BCOP set was constructed to satisfy both the geometrical and natural
boundary conditions of the vibrating beam. However, the higher members of the set satisfied only the
geometrical boundary conditions. In the present study, a modified Gram-Schmidt orthogonalization
procedure is used where all the members of the orthogonal set of polynomials satisfy all the boundary
conditions. The exact solution of the beam differential equation is expressed in the form of a
generalized Fourier series in terms of the set of newly constructed BCOP which results in an eigenvalue
problem. The solution of this eigenvalue problem provides the natural frequencies and the
corresponding normal modes of the beam more accurately.
1. Introduction:
Bhat [1] used Boundary characteristic orthogonal polynomials as assumed deflection shapes in the
Rayleigh Ritz method to study the vibrations of plates. Many investigations were reported using the
BCOP to study the vibration of structures and some of them can be found in [2-4]. A review paper on the
use of BCOP to study vibration problems until 1999 is given in [5]. Use of orthogonal functions as
assumed deflection shapes rendered the mass matrix diagonal and the stiffness matrix had large
diagonal terms making the matrices well-conditioned making it easier to compute the natural
frequencies and the natural modes.
The first member of the BCOP set is constructed in [1] satisfying all the geometrical boundary
conditions. However, the higher members of the set did not satisfy all the boundary conditions.
Although the deflection can be expressed as a linear combination of the BCOP, which is a Generalized
Fourier series representation of the deflection, the solution is still approximate since the natural
boundary conditions are not satisfied. However, Kim [4] proposed a generating function which will help
in the higher polynomials of the set also satisfying all the boundary conditions, which can construct the
exact solution. His formulation was modified by replacing his generating function with an evolution
function that is different for different members of the BCOP set.

2. Construction of Novel Boundary Characteristic Orthogonal Polynomials Satisfying All the

Boundary Conditions:
The starting function in the BCOP set is constructed as the lowest order polynomial that satisfies four
arbitrary beam boundary conditions. It is expressed as

( ) a0 (1 a1* a2* 2 a3* 3 a4* 4 )

where a k
*

(1)

ak
, k 1,2,3,4 , and x / L is the non-dimensioned length coordinate of the beam.
a0

The second and the higher members of the BCOP set are constructed using the Gram-Schmidt
orthogonalization procedure as

1 ( ) [ g 0 ( ) B0 ]0 ( )
k 1 ( ) [ g k ( ) Bk ]k ( ) C k k 1 ( )

(2)

g k ( ) Ak1 Ak 2 2 Ak 3 3 Ak 4 4 5

(3)

where

The member specific constants Bk and Ck ensure orthogonality of the newly constructed polynomial
with respect to the two previous member polynomials. The member specific constants Akj ensure that
the boundary conditions of the newly constructed polynomial are satisfied. The number of constants
Akj depends on the number and nature of the boundary conditions.
The k-th member is constructed using two members immediately below, thus forming a three term
recurrence relation. Using the condition that the BCOP are orthogonal to each other, the constants

B0 , Bk and Ck are
1

B0

w( ) g ( ) ( )d

( )d
2
0

2
0

Bk

w( ) g ( ) ( )d

( )d
2
k

Ck

w( ) g ( ) ( )
( )d
0

(4)

2
k

k 1

( )d

2
k 1

where w( ) is a weight function and g k ( ) is an evolution function. For a uniform beam, the weight

w( ) 1. For simplicity the beam would be assumed to be uniform in the following analysis. The set of
orthogonal polynomials will have infinite members with k = 1,2,3... . The member polynomials and
their derivatives with respect to the independent variable are

k 1 ( ) [ g k ( ) Bk ]k ( ) Ckk 1 ( )
k1 ( ) [ g k ( ) Bk ]k ( ) g k ( )k ( ) Ckk1 ( )
k1 ( ) [ g k ( ) Bk ]k( ) 2 g k ( )k ( ) g k ( )k ( ) Ckk1 ( )
k1 ( ) [ g k ( ) Bk ]k( ) 3g k ( )k( ) 3g k ( )k ( ) g k( )k ( ) Cik1 ( )
(5)
The boundary conditions and the corresponding conditions on g k () and its derivatives in order to
satisfy all the boundary conditions by the BCOP are presented in Table 1. The evolution function
g k () differs from the generating function defined by Kim [4]. While Kims generating function is
constant for all the members, the evolution function depends on the previous member of the BCOP set.

Table 1. Beam Boundary Conditions and the Corresponding Conditions on g k ( ) and its Derivatives
C-C

k (0) k (0) 0
k (1) k (1) 0

0 ( ) 2 2 3 4 ,
g k (0) 0,

g k (1) 0,

g k ( ) 1.5 2 3
C-S

k (0) k (0) 0
k (1) k(1) 0

0 ( ) 1.5 2 2.5 3 4 ,
g k (0) 0, g k (1) 0
g k ( ) 1.5 2 3

S-S

k (0) k(0) 0
k (1) k(1) 0

0 ( ) 2 3 4 ,
g k (0) 0,

g k (1) 0,

g k ( ) 1.5 2 3

C-F

k (0) k (0) 0
k(1) k(1) 0

0 ( ) 2 2 3 4 ,
2 g k (1) k (1) g k(1) k (1) 0,
3g k(1) k (1) g k(1) k (1) 0,
Dk 0 k (0), Ek 0 k (0), Dk 1 k (1), Ek 1 k (1),
Ak 1 0.
T22 2 En1 Dn1
T23 3En1 3Dn1
T32 En1
T33 3En1 Dn1
f 2 (4 En1 6 Dn1 )
f 3 (6 En1 4 Dn1 )
Ak 2

f 2T33 f 3T23
,
T22T33 T32T23

Ak 3

T22 f 3 T32 f 2
,
T22T33 T32T23

g k ( ) Ak 2 2 Ak 3 3 4
F-F

k(0) k(0) 0
k(1) k(1) 0

0 ( ) 1,
2 g k (0) k (0) g k(0) k (0) 0, 3g k(0) k (0) g k(0) k (0) 0,
2 g k (1) k (1) g k(1) k (1) 0,
3g k(1) k (1) g k(1) k (1) 0,
Dk 0 k (0), Ek 0 k' (0), Dk 1 k (1), Ek 1 k (1),
2

E
E
T11 En1 n 0 (2 En1 Dn1 ) n 0 (3En1 3Dn1 )
Dn 0
Dn 0
T12 4 En1 6 Dn1
2

E E
T21 2 n 0 n 0 (6 En1 2 Dn1 )
Dn 0 Dn 0
T22 12 En1 8Dn1
f1 (5En1 10 Dn1 )
f 2 (20 En1 20 Dn1 )
Ak 1

T22 f1 T12 f 2
,
T11T22 T21T12

Ak 2
2

(0)
(0)
Ak 3 k
A2 k
,
k (0)
k (0)

Ak 4

g k ( ) A1 A2 2 A3 3 A4 4 5

k (0)
A,
k (0) 1

T22 f1 T12 f 2
,
T11T22 T21T12

The BCOP set must be constructed such that the evolution function g k ( ) satisfies the conditions in
Table 1. The construction of the evolution function g k () is illustrated below for the different cases of
the boundary conditions given in Table 1.
(i) Clamped-Clamped Conditions, C-C:
The boundary conditions are:

k (0) k (0) 0, k (1) k (1) 0

The first polynomial is the lowest order polynomial that satisfies these conditions and is
given by

0 () 2 23 4

(6)

Corresponding to these boundary conditions on the evolution function

are g k (0) 0, g k (1) 0 .
The evolution function, g k () , is the simplest polynomial that satisfies these conditions and
is given by

g k ( ) 1.5 2 3

(7)

(ii) Clamped-Simply Supported Conditions, C-S:

The corresponding boundary conditions are

k (0) k (0) 0
k (1) k(1) 0

(8)

The first polynomial is the lowest order polynomial that satisfies all these conditions and is
given by

0 () 1.52 2.53 4

(9)

The corresponding conditions on the evolution function, g k () , are given by

g k (0) 0, g k (1) 0

(10)

The evolution function, g k () , is the simplest polynomial that satisfies these conditions and
is given by

g k ( ) 1.5 2 3

(11)

(iii) Simply Supported-Simply Supported Conditions, S-S:

The corresponding boundary conditions are

k (0) k(0) 0
k (1) k(1) 0

(12)

The first polynomial is the lowest order polynomial that satisfies all these conditions and is
given by

0 () 23 4

(13)

The corresponding conditions on the evolution function, g k () , are given by

gk (0) 0,

gk (1) 0

(14)
and the evolution function is the simplest polynomial that satisfies these conditions and is
given by

g k () 1.52 3

(15)

(iv) Clamped-Free Conditions, C-F:

The associated boundary conditions are:

k (0) k (0) 0
k(1) k(1) 0

(16)

The first polynomial is the lowest order polynomial that satisfies all these conditions and is
given by

0 ( ) 6 2 4 3 4

(17)

The corresponding conditions on the evolution function, g k () , are given by

2 gk (1)k (1) gk(1)k (1) 0,

3gk(1)k (1) gk(1)k (1) 0

(18)

The evolution function, g k () , is given by the simplest polynomial that satisfies these
conditions and is given by

Dk 0 k (0), Ek 0 k (0), Dk 1 k (1), Ek 1 k (1),

Ak 1 0.
T22 2 En1 Dn1
T23 3En1 3Dn1
T32 En1
T33 3En1 Dn1
f 2 (4 En1 6 Dn1 )
f 3 (6 En1 4 Dn1 )
Ak 2

f 2T33 f 3T23
,
T22T33 T32T23

Ak 3

T22 f 3 T32 f 2
,
T22T33 T32T23

g k ( ) Ak 2 2 Ak 3 3 4

(19)

(v) Free-Free Conditions, F-F:

The associated boundary conditions are:

k(0) k(0) 0
k(1) k(1) 0

(20)

The first polynomial is the lowest order polynomial that satisfies all these conditions and is
given by

0 () 1

(21)

The corresponding conditions on the evolution function, g k () , is given by

2 gk (0) k (0) gk(0) k (0) 0, 3gk(0) k (0) gk(0) k (0) 0,

2 gk (1) k (1) gk(1) k (1) 0,
3gk(1) k (1) gk(1) k (1) 0

(22)

The evolution function, g k () , is given by the simplest polynomial that satisfies these
conditions and is given by

Dk 0 k (0), Ek 0 k' (0), Dk 1 k (1), Ek 1 k (1),

2

E
E
T11 En1 n 0 (2 En1 Dn1 ) n 0 (3En1 3Dn1 )
Dn 0
Dn 0
T12 4 En1 6 Dn1
2

E E
T21 2 n 0 n 0 (6 En1 2 Dn1 )
Dn 0 Dn 0
T22 12 En1 8Dn1
f1 (5En1 10 Dn1 )
f 2 (20 En1 20 Dn1 )
Ak 1

T22 f1 T12 f 2
,
T11T22 T21T12

Ak 2
2

(0)
(0)
Ak 3 k
A2 k
,
k (0)
k (0)

Ak 4

k (0)
A,
k (0) 1

T22 f1 T12 f 2
,
T11T22 T21T12

g k ( ) Ak 1 Ak 2 2 Ak 3 3 Ak 4 4 5

(23)

3. Solution of the Vibrating Beam

The differential equation for the beam vibration can be written as

EI

d4y
d2y

m
0
dx 4
dt 2

(24)

Free vibration response is assumed in the form of a sinusoidal function in time. Accordingly,

y( x, t ) Y ( x) sin t

(25)

where the axial coordinate x is non-dimensionalized as x / L , and L is the length of the beam.
Accordingly the beam equation can be written as

y '''' 4 y 0

(26)

where 4

m 2
and the ( ) indicates differentiation with respect to .
EI

Assume the solution of the vibrating beam in the form

N

y ( ) ckk ( )

(27)

i 0

which is a generalized Fourier series representation of the solution in terms of the BCOP, with N .
Substituting Eq. (35) in Eq. (34) we have
N

y( ) cii( )

(28)

i 0

The fourth derivative of the assumed shape functions are expressed in Fourier series in terms of the
BCOP as
i

k ( ) d kll ( )

(29)

j 0

Multiplying both sides of Eq. (37) by k and integrating from 0 to 1, we get

1

d d d
0

kl

2
l

(30)

where
1

d kl

d
d
k

2
l

(31)

Substituting Eqs. (35-39) into Eq. (34) results in

c0 [d 000 d 011 d 02 2 d 033 ...........] 4 c00

c1 [d100 d111 d12 2 d133 ...........] 4 c11
c 2 [d 200 d 211 d 22 2 d 233 ...........] 4 c 2 2
c3 [d 300 d 311 d 32 2 d 333 ...........] 4 c33

Eq. (40) can be reorganized in the form

(32)

0 [c0 (d 00 4 ) c1 d10 c 2 d 20 c3 d 30 .......]

1 [c0 d 01 c1 (d11 4 ) c 2 d 21 c3 d 31 c 4 d 41 .......]
2 [c0 d 02 c1 d12 c 2 (d 22 4 ) c3 d 23 c 4 d 24 .......]
3 [c0 d 02 c1 d12 c 2 (d 22 ) c3 d 23 c 4 d 24 .......]

(33)

................
................
This can be cast in the form of an eigenvalue problem as

D I c 0
4

(34)

where

(d 00 4 )
d 01
d 02

4)
(d11 )
d 21
d10
4
D I d 20
d 21
(d 22 4 )

d
d N1
dN2
N0

d0N

d1N
d 2N

d NN

and the column vector of generalized coordinates is given by

c1
c
2
c c3


c N

(35)

The formulation leads to the same final form as that obtained in a Galerkin solution.
4. Discussion of the Results
The boundary characteristic orthogonal polynomials form a complete set, and are also linearly
independent in view of Eq. (2), and hence the solution can be expressed as a generalized Fourier series
in terms of the BCOP. When the beam boundary conditions are applied, solution to the resulting
eigenvalue problem provides an exact solution to the beam differential equation. While the
exponential solution provides an exact solution in terms of the sinusoidal and hyperbolic sinusoidal
functions, the present method provides exact solutions in terms of the BCOP whose order steadily
changes monotonically. The exact results due to Young and Felgar [6] provide up to 5 modes only, while
n = 6 terms are used in computing the results in the present study.

Table 2 presents natural frequency coefficients,

ml 2
2 , for the 5 cases shown in Table 1. As
EI

can be seen, the natural frequency coefficients are very close to the exact solutions, even with the low
number of terms used, N = 6. The results are very close to the exact results for the cases of C-S, S-S and
C-F beams.
Table 2. Natural frequency coefficients for the different cases of beam boundary conditions
Cases

Exact

Rayleigh Ritz with old

BCOP

C-C

22.373286
61.672822
120.903391
199.859448
298.555533

22.373286
61.672882
121.130621
201.126049
353.433320
531.828852

Exact with BCOP

Satisfying all boundary
conditions
22.375426
61.677967
121.006004
200.181319
299.228923
418.152978

C-S

15.418206
49.964862
104.247697
178.269730
272.030971

15.418213
49.964964
104.573236
181.682883
335.966285
783.178283

15.418733
49.969845
104.296435
178.532507
272.772401
387.969555

S-S

9.869604
39.478418
88.826440
157.913671
246.740110

C-F

3.5160154
22.034492
61.697214
120.901916
199.859530

F-F

0
0
22.373286
61.672822
120.903391
199.859448

9.869616
39.478501
89.474383
161.807027
454.582023
775.394745
3.516015
22.034498
61.8224
127.349730
227.787287
847.556596
0
0
22.564205
63.537311
223.362552
455.678395

9.869625
39.478731
88.833054
157.927750
247.921703
357.336481
3.516037
22.036892
61.753280
121.289338
206.176927
335.222642
0
0
22.373352
61.673159
121.066814
223.733524

5. Conclusions
The new boundary characteristic orthogonal polynomials are constructed so as to satisfy both the
geometrical and natural boundary conditions of the different beam problems. These boundary
characteristic orthogonal polynomials built to satisfy all the boundary conditions form a complete set of
orthogonal functions. The solution expressed in a generalized Fourier series provides an exact solution
to the beam differential equation. Natural frequency coefficients are obtained for different cases of
boundary conditions. The solutions will tend towards the accurate values when sufficiently large number
of terms are used in the generalized Fourier series. Products of these novel BCOP in the x and y
directions can be used as assumed deflection shapes in the Rayleigh Ritz method for rectangular plates
in order to obtain their natural frequencies and normal mode shapes.
6. References

1. R. B. Bhat, Natural frequencies of rectangular plates using characteristic orthogonal

polynomials in Rayleigh-Ritz method, J. Sound and Vib., 102 (1985) 493-499.
2. B. Singh and S. Chakraverty, Transverse vibration of completely free elliptic and circular plates
using orthogonal polynomials in Rayleigh Ritz Method, Int. J. Mech. Sci., 114 (1992), 65-71.
3. K. Y. Lam, K. M. Liew K. M. and S. T. Chow, Use of two dimensional orthogonal polynomials for
vibration of circular and elliptic plates, J. Sound and Vib., 154 (1992) 261-270.
4. C. S. Kim, The vibration of beams and plates studied using orthogonal polynomials, Ph. D. Thesis,
Department of Mechanical Engineering, The University of Western Ontario, 1988.
5. S. Chakraverty, R. B. Bhat and I. Stiharu, Recent research on vibration of structures using
boundary characteristic orthogonal polynomials in the Rayleigh-Ritz method, The Shock and Vib.
Digest, 31 (1999) 187-194.
6. D. Young and R. P. Felgar Jr., Tables of characteristic functions representing normal modes of
vibration of a beam, The Univ. Texas, Austin, Texas, Engineering Research Series No. 44, 1949.