City Research Online

City, University of London Institutional Repository

Citation: Banerjee, J. R. & Ananthapuvirajah, A. (2018). Free vibration of functionally

graded beams and frameworks using the dynamic stiffness method. Journal of Sound and
Vibration, 422, pp. 34-47. doi: 10.1016/j.jsv.2018.02.010

This is the accepted version of the paper.

This version of the publication may differ from the final published version.

Permanent repository link: https://openaccess.city.ac.uk/id/eprint/20189/

Link to published version: https://doi.org/10.1016/j.jsv.2018.02.010

Copyright: City Research Online aims to make research outputs of City,

University of London available to a wider audience. Copyright and Moral Rights
remain with the author(s) and/or copyright holders. URLs from City Research
Online may be freely distributed and linked to.

Reuse: Copies of full items can be used for personal research or study,
educational, or not-for-profit purposes without prior permission or charge.
Provided that the authors, title and full bibliographic details are credited, a
hyperlink and/or URL is given for the original metadata page and the content is
not changed in any way.
City Research Online: http://openaccess.city.ac.uk/ [email protected]
 FREE VIBRATION OF FUNCTIONALLY GRADED BEAMS AND
FRAMEWORKS USING THE DYNAMIC STIFFNESS METHOD

J.R. Banerjee* and A. Ananthapuvirajah

School of Mathematics, Computer Science and Engineering, City, University of London,
Northampton Square, London EC1V 0HB, U.K.

Abstract

The free vibration analysis of functionally graded beams (FGBs) and frameworks containing

FGBs is carried out by applying the dynamic stiffness method and deriving the elements of

the dynamic stiffness matrix in explicit algebraic form. The usually adopted rule that the

material properties of the FGB vary continuously through the thickness according to a power

law forms the fundamental basis of the governing differential equations of motion in free

vibration. The differential equations are solved in closed analytical form when the free

vibratory motion is harmonic. The dynamic stiffness matrix is then formulated by relating the

amplitudes of forces to those of the displacements at the two ends of the beam. Next, the

explicit algebraic expressions for the dynamic stiffness elements are derived with the help of

symbolic computation. Finally the Wittrick-Williams algorithm is applied as solution

technique to solve the free vibration problems of FGBs with uniform cross-section, stepped

FGBs and frameworks consisting of FGBs. Some numerical results are validated against

published results, but in the absence of published results for frameworks containing FGBs,

consistency checks on the reliability of results are performed. The paper closes with discussion

of results and conclusions.

Keywords: Free vibration, functionally graded beams, dynamic stiffness method, frameworks,
Wittrick-Williams algorithm.
*
Corresponding author.
Email address: [email protected]

1
1. Introduction

In recent years interest in functionally graded material (FGM) has grown enormously. The

progress made in understanding this material has been phenomenal. One great advantage of

FGM is that the properties vary gradually in a continuous manner within the material so that

there is no abrupt change or mismatch of the properties which can cause delamination or other

problems generally associated with fibre-reinforced composites. Thus, FGM can be designed

in a way to have the properties of ceramic at one end and those of metal at the other so that

the thermal resistance of ceramic and the mechanical behaviour of metal can be exploited to

advantage to guarantee structural integrity. Consequent on this, researchers have been

continually motivated to use various techniques and methodologies to deal with this exciting

material in order to enhance its state-of-the-art. There are now excellent books [1-4] available

on the subject. As potential application of FGM, beams which are extensively used in civil,

mechanical, aeronautical and other branches of engineering as principal load carrying

structural members can be investigated for their free vibration characteristics. Investigators

have expended considerable efforts which have led to the insurgence of massive literature on

the free vibration behaviour of Functionally Graded Beams (FGBs). A number of theories and

methodologies have been proposed to study the free vibration characteristics of FGBs.

Foremost amongst these are the applications of direct analytical procedure using the governing

differential equations of motion [5-19], finite element [20-22], Rayleigh-Ritz [23], finite

volume [24-26], differential quadrature [27], differential transformation [27, 28] and transfer

function [29, 30] methods. Recently the dynamic stiffness method (DSM) has also been

proposed [31, 32]. The current paper stems from the previously published DSM theories. The

entire formulation using DSM in this paper is accomplished in the real domain as opposed to

previous formulations which used complex arithmetic when developing the element dynamic

2
stiffness matrices [31, 32]. Another important further development reported in this paper is

the derivation of explicit algebraic expressions for the dynamic stiffness elements using

symbolic computation [33-35]. The explicit expressions for the dynamic stiffness elements

are particularly useful in optimisation studies and also when some, but not all of the stiffnesses

are needed. Of particular significance of this investigation is the application of DSM to analyse

the free vibration characteristics of stepped FGBs and frameworks containing FGBs. The

substantial advantages of the DSM over the conventional finite element method (FEM) are

well known [36-38]. The DSM is often called an exact method because in sharp contrast to

chosen approximate shape function assumed in the FEM, the DSM uses exact shape function

obtained from the analytical solution of the governing differential equation of motion of the

structural element in free vibration. The uncompromising accuracy of the DSM in all

frequency ranges and its independency on the number of elements used in the analysis makes

it particularly appealing in free vibration analysis. Within this pretext, the application of the

DSM in the free vibration analysis of FGBs and frameworks containing FGBs is considered

to be a welcome development. The solution technique used in the DSM is robust, particularly

when the well-established algorithm of Wittrick and Williams [39], known as Wittrick-

Williams algorithm in the literature, is applied. The algorithm ensures that no natural

frequency of the structure is missed, and it has featured in literally hundreds of papers. It is

worth noting that earlier investigations on the free vibration of FGBs were focused on

individual FGBs except for a few isolated cases where stepped FGBs with collinear axes were

reported [40, 41]. Accordingly, the literature on the free vibration of frameworks containing

FGBs is virtually non-existent. One of the essential purposes of this paper is to fill this gap.

3
2. Theory

In a right-handed Cartesian coordinate system, Fig. 1 shows a uniform rectangular cross-

section FGB of length L, width b and thickness h. The beam material has Young’s Modulus

E and mass density ρ which can vary through the thickness direction (Z) of the cross-section

according to the following power law distribution [14, 17, 30, 32]:

k k
 z 1  z 1
E ( z ) = ( Et − Eb ) +  + Eb , ρ ( z ) = ( ρ t − ρ b ) +  + ρ b (1)
h 2 h 2

where Et and Eb are the Young’s moduli, and ρ t and ρb are the densities at the top and

bottom surfaces of the beam, respectively.

In Eq. (1), k (k ≥ 0) is the power law index parameter which indicates the material property

variation through the beam thickness. The parameter k has been extensively discussed in the

literature [14-19] and hence it is not elaborated here. However, three special cases maybe

observed. Clearly k = 1 indicates a linear variation of the composition between the top and

bottom surfaces of the beam, k = 0 represents the case when the beam is made of full material

of the top surface whereas infinite k represents the case when the beam is made of full

material of the bottom surface.

2.1 Governing differential equations of motion and solution

The classical Bernoulli-Euler theory is considered here so that the effects of shear

deformation and rotary inertia that are relevant to the Timoshenko beam theory are assumed

to be small and hence disregarded in the analysis. Referring to Fig. 1, the displacements u1 , v1

and w1 along the X, Y and Z directions of a point on the cross-section are given by [6, 30,

32]:

∂w( y, t )
u1 = 0 , v1 ( y, z , t ) = v( y, t ) − z , w1 ( y, z, t ) = w( y, t ) (2)
∂y

4
where v and w are the corresponding displacements of a point on the neutral axis of the

beam. It should be noted that due to the variation of the material properties through the

thickness, the neutral axis would no longer be at the central line of the beam cross-section [42,

43].

Using the displacement field given by Eq. (2) and through the application of Hamilton’s

principle, the governing differential equations of motion in free vibration of the FGB are given

by [30, 32]

′ + A0v′′ − A1w′′′ = 0,

− B0v + B1w  − B1v′ + B2 w
− B0 w ′′ + A1v′′′ − A2 w′′′′ = 0 (3)

where

Ai = ∫ z i E ( z )dA, Bi = ∫ z i ρ ( z )dA (i = 0, 1, 2) (4)

The natural boundary conditions from the Hamiltonian formulation [30, 32] give the following

expressions for axial force F , shear force S and bending moment M as follows:

F = − A0v′ + A1w′′, S = B1v − A1v′′ − B2 w

′ + A2 w′′′, M = A1v′ − A2 w′′ (5)

Clearly, due to the use of FGM, the axial ( v ) and bending motions ( w ) are coupled as evident

from Eqs. (3) and (5).

Assuming harmonic oscillation so that

v ( y , t ) = V ( y ) e iω t , w( y, t ) = W ( y )e iω t (6)

where ω is the angular or circular frequency, V ( y ) and W ( y ) are the amplitudes of v and

w , respectively.

5
 d
Introducing the differential operator D = and the non-dimensional length ξ as:
dξ

y
ξ= (7)
L

The differential equations of motion in Eqs. (3) can now be written as:

( B0ω 2 L3 + A0 L D 2 )V (ξ ) − ( B1ω 2 L2 D + A1 D 3 )W (ξ ) = 0 
 (8)
( B1ω 2 L3 D + A1L D 3 )V (ξ ) + ( B0ω 2 L4 − B2ω 2 L2 D 2 − A2 D 4 )W (ξ ) = 0

By combining the above two differential equations, it is possible to obtain a sixth order

ordinary differential equation satisfying both V (ξ ) and W (ξ ) to give:

(D 6 + a D 4 − b D 2 − c) H = 0 (9)

where

H = V (ξ ) or W (ξ ) (10)

and

2 A1B1 − A0 B2 − A2 B0 B0 B2ω 2 − A0 B0 − B12ω 2 B02

a= L2ω 2 , b = L4ω 2 , c = − L6ω 4 (11)
A12 − A0 A2 A12 − A0 A2 A12 − A0 A2

By assuming the solution in the form H = eλξ the characteristic or auxiliary equation of the

differential equation Eq. (9) can be expressed as

λ6 + aλ4 − bλ2 − c = 0 (12)

Equation (12) can be reduced to a cubic equation to give

µ 3 + aµ 2 − bµ − c = 0 (13)

where

µ = λ2 (14)

6
 Equation (13) can now be solved analytically for µ using standard procedure [44]. In earlier

investigations [30, 32], it was assumed that the square root of the three roots of the cubic in

Eq. (13) could be either real or complex and thus the six roots r j ( j = 1, 2,, 6) of the

characteristic equation Eq. (12) give the solutions of the differential equation (Eq. (9)) leading

to V and W in the following forms:

6 6
V (ξ ) = ∑ R j e j , W (ξ ) = ∑ Q j e
rξ r jξ
(15)
j =1 j =1

where R j and Q j ( j = 1, 2,,6) are two sets of constants which could be possibly complex.

Clearly a method utilising Eq. (15) as the solution to derive the dynamic stiffness matrix will

involve numerical operations using complex arithmetic. This cumbersome and somehow

computationally inefficient procedure is circumvented in this paper.

Using an approach similar to the one described in Refs [45, 46], it can be shown that the

three roots of the cubic equation Eq. (13) are real and the solution for H in Eq. (9) can be

expressed in terms of trigonometric and hyperbolic functions as opposed to complex

exponential functions of Eq. (15). This is advantageous when deriving the explicit expressions

for the dynamic stiffness elements of the FGB. Explicit expressions are particularly useful

when some, but not all of the stiffness elements are needed, e.g. sensitivity analysis in

optimisation studies. Thus if the roots [44] of Eq. (13) are α, β and γ, the solution for H is

given by

H (ξ ) = C1 cosh αξ + C 2 sinh αξ + C3 cos βξ + C 4 sin βξ + C5 cos γξ + C6 sin γξ (16)

7
where
1
 1
2
  q  2 φ a
α = 2  cos −
 3 3 3
 
1
 1
2
  q  2 π −φ a
β = 2  cos + (17)
 3 3 3
 
1
 1
 2
q π +φ a
γ = 2  cos + 
2

 3 3 3
 

with
a2
q =b+ (18)
3

and
 
 27c − 9ab − 2a 3 
φ = cos−1
(19)
 
( )
3

 2 a 2
+ 3b 2


𝐻𝐻(𝜉𝜉) of Eq. (16) represents the solution for both axial displacement V(ξ) and bending
displacement 𝑊𝑊(ξ), containing different sets of constants as follows

V (ξ ) = Q1 cosh αξ + Q2 sinh αξ + Q3 cos βξ + Q4 sin βξ + Q5 cos γξ + Q6 sin γξ (20)

W (ξ ) = R1 cosh αξ + R2 sinh αξ + R3 cos βξ + R4 sin βξ + R5 cos γξ + R6 sin γξ (21)

The two different sets of constants 𝑄𝑄1 − 𝑄𝑄6 and 𝑅𝑅1 − 𝑅𝑅6 can be related with the help of any
one of the two of Eqs. (8) to give

k   kβ   kγ 
Q1 =  α  R2 , Q3 =   R4 , Q5 =   R6 ,
 L   L   L
k   kβ   kγ 
Q2 =  α  R1 , Q4 = −  R3 , Q6 = −  R5 (22)
 L   L   L

where

8
 α (A1α 2 + B1ω 2 L2 ) β (A1β 2 − B1ω 2 L2 ) γ (A1γ 2 − B1ω 2 L2 )
kα =
(A0α 2 + B0ω 2 L2 ) , k β = (A0 β 2 − B0ω 2 L2 ) , kγ = (A0γ 2 − B0ω 2 L2 ) (23)

With the help of Eqs. (5), (20) and (21), the expressions for bending rotation 𝜃𝜃(𝜉𝜉), axial force
F(ξ), bending moment 𝑀𝑀(𝜉𝜉), and shear force 𝑆𝑆(𝜉𝜉) can be obtained after some simplification,
as
W ' (ξ )  1 
θ (ξ ) = =  {R1α sinh αξ + R2α cosh αξ − R3 β sin βξ + R4 β cos βξ
L L (24)
− R5γ sin γξ + R6γ cos γξ }

 ' ''

 
A  e
F (ξ ) = − 0 V − 1 W  =  0 {− R1 α cosh αξ − R2 α sinh αξ + R3 β cos βξ
A A e e
L A0 L   L  L L L
  (25)
 
eβ e e
+ R4 sin βξ + R5 γ cos γξ + R6 γ sin γξ}
L L L

A2  '' A1L '  A 

M (ξ ) = − W −
2 
V  = − 22 {− R1 gα cosh αξ − R2 gα sinh αξ + R3 g β cos βξ
L  A2  L  (26)
+ R4 g β sin βξ + R5 gγ cos γξ + R6 gγ sin γξ }

A2  B2 L2ω 2 A1L B1L3ω 2 

S (ξ ) =  W ′′′ + W ′ − V ′′ − V 
L3  A2 A2 A2  (27)
= 32 {R1 fα sinh αξ + R2 fα cosh αξ + R3 f β sin βξ − R4 cos βξ + R5 sin γξ − R6 cos γξ }
A
L

where

eα = αkα − A1α 2 / A0 , eβ = βk β − A1β 2 / A0 , eγ = γkγ − A1γ 2 / A0 (28)

gα = α 2 − αkα A1 / A2 , g β = β 2 − βk β A1 / A2 , gγ = γ 2 − γkγ A1 / A2 (29)
fα = α 3 − α 2 kα A1 / A2 + αω 2 L2 B2 / A2 − kα ω 2 L2 B1 / A2
f β = β 3 − β 2 k β A1 / A2 − βω 2 L2 B2 / A2 + k β ω 2 L2 B1 / A2 (30)
fγ = γ 3 − γ 2 kγ A1 / A2 − γω 2 L2 B2 / A2 + kγ ω 2 L2 B1 / A2

2.2 Dynamic Stiffness Formulation

The dynamic stiffness matrix of the FGB can now be formulated by applying natural

boundary conditions for displacements and forces at the ends of the beam. Referring to the

9
sign convention for positive axial force, shear force and bending moment shown in Fig. 2, the

boundary conditions for displacements and forces, see Fig. 3, are:

At ξ = 0 : V = V1 , W = W1 , θ = θ1 , F = F1 , S = S1 , M = M 1 (31)

At ξ = 1 : V = V2 , W = W2 , θ = θ 2 , F = − F2 , S = − S 2 , M = − M 2 (32)

The displacement vector δ and the force vector P can be expressed as:

δ = [V1 W1 θ1 V2 W2 θ2 ]T , P = [ F1 S1 M 1 F2 S 2 M 2 ] T (33)

where the upper script T denotes a transpose.

The relationship between the displacement δ and the constant vector R can be obtained by

using Eqs. (20)-(24) and Eqs. (31)-(32) to give,

δ=BR (34)

where

 kα kβ kγ 
 0 L
0
L
0
L 
 1 0 1 0 1 0
 α β γ 
 0 0 0
B =  k S
L L L 
kα C hα kβ Sβ k β Cβ kγ S γ k γ Cγ  (35)
α hα
 L − −
L L L L L 
 C S hα Cβ Sβ Cγ Sγ 
 hα 
 α S hα α C hα βS β βC β γSγ γCγ 
 L − −
L L L L L 

with

Chα = cosh α , S hα = sinh α , C β = cos β , S β = sin β , Cγ = cos γ , Sγ = sin λ (36)

Similarly, the relationship between the force vector P and the constant vector R can be

obtained using Eqs. (25)-(27) and Eqs. (31)-(32) to give

P = AR (37)

10
where

 W1eα W1eβ W1eγ 

 − L 0
L
0
L
0 
 0 W3 fα 0 − W3 f β 0 − W3 f γ 
 
 − W2 gα 0 W2 g β 0 W2 g γ 0 
A=
W1eγ S γ 
(38)
W1eβ C β W1eβ S β W1eγ Cγ
 W1eα C hα W1eα S hα
− − − − 
 L L L L L L 
− W3 fα S hα − W3 fα C hα − W3 f β S β W3 f β C β − W3 f γ Sγ W3 f γ Cγ 
W g C W2 gα S hα − W2 g β C β − W2 g β S β − W2 g γ Cγ − W2 g γ Sγ 
 2 α hα

By eliminating the constant vector R from Eqs. (34) and (37), P and δ can be related to give

the dynamic stiffness matrix relationship of the FGB as

P=Kδ (39)

where

K = A B −1 (40)

is the required dynamic stiffness matrix with elements kij (i = 1, 2, 3..6; j=1,2, 3,..6).

With the help of symbolic computation [33-35], the matrix B of Eq. (35) was inverted

algebraically and the inverted matrix was pre-multiplied by the matrix A of Eq. (38) in order

to generate the explicit expressions for each of the elements of the dynamic stiffness matrix

K. The stiffness expressions are simplified very considerably by means of symbolic

computations. They are not necessarily in the shortest form, but they are surprisingly concise.

The twelve independent terms of the dynamic stiffness matrix K are given by

11
  A  Φ 
K1,1 = K 4, 4 =  0  1 
 L  ∆ 
 A  Φ 
K1, 2 = − K 4,5 =  22  2 
 L  ∆ 
 A  Φ 
K1,3 = K 4, 6 =  2  3 
 L  ∆ 
  Φ 
A
K1,4 =  0  4 
 L  ∆ 
 A  Φ 
K1,5 = − K 2, 4 =  22  5 
 L  ∆ 
 A  Φ 
K1, 6 = K 3, 4 =  2  6  (41)
 L  ∆ 
 A  Φ 
K 2, 2 = K 5,5 =  32  7 
 L  ∆ 
 A  Φ 
K 2,3 = − K 5, 6 =  22  8 
 L  ∆ 
 A  Φ 
K 2,5 =  32  9 
 L  ∆ 
 A  Φ 
K 2, 6 = − K 3,5 =  22  10 
 L  ∆ 
 A  Φ 
K 3,3 = K 6 ,6 =  2  11 
 L  ∆ 
 A  Φ 
K 3,6 =  2  12 
 L  ∆ 

where

12
 Φ1 = ν 1 S β Cγ S hα +ν 3C β S γ S hα −ν 2 S β S γ C hα + σ 3C β Cγ C hα +ν 3C β +ν 1Cγ −ν 2 C hα
Φ 2 = −τ 3 S β Sγ S hα − λ3 S hα (1 − Cβ Cγ ) − λ1S β (1 − Cγ Chα ) − λ2 Sγ (1 − Cβ Chα )
Φ 3 = ρ1S β Cγ S hα + ρ3Cβ Sγ S hα − ρ 2 S β Sγ Chα + σ 2Cβ Cγ Chα + ρ3Cβ + ρ1Cγ − ρ 2Chα
Φ 4 = −ν 1S β S hα − ν 3 Sγ S hα + ν 2 S β Sγ + ν 3Cγ Chα − ν 2C β Cγ + ν 1C β Chα + σ 3
Φ 5 = −τ 1{αS hα (Cβ − Cγ ) − βS β (Cγ − Chα ) + γSγ (Cβ − Chα )}
Φ 6 = − ρ1S β S hα − ρ3 Sγ S hα + ρ 2 S β Sγ + ρ3Cγ Chα − ρ 2Cβ Cγ + ρ1Cβ Chα + σ 2 (42)
(
Φ 7 = −τ1 µ1S β Cγ Shα − µ2 S β Sγ Chα + µ3Cβ Sγ Shα )
Φ 8 = τ 2 S β Sγ S hα + λ3 S hα (1 − Cβ Cγ ) + λ1S β (1 − Cγ Chα ) + λ2 Sγ (1 − Cβ Chα )
Φ 9 = τ 1 (µ1S β S hα − µ 2 S β Sγ + µ3 Sγ S hα )
Φ10 = −τ 1{kα S hα (Cβ − Cγ ) − k β S β (Cγ − Chα ) + kγ Sγ (Cβ − Chα )}
Φ11 = −ξ1S β Cγ S hα + ξ 2 S β Sγ Chα − ξ 3Cβ Sγ S hα + σ 1Cβ Cγ Chα − ξ 3Cβ − ξ1Cγ + ξ 2Chα
Φ12 = ξ1S β S hα − ξ 2 S β Sγ + ξ 3 Sγ S hα − ξ 3Cγ Chα + ξ 2Cβ Cγ − ξ1Cβ Chα + σ 1

with

µ1 = αkβ − βkα , µ 2 = βkγ − γk β , µ3 = γkα − αkγ

ε1 = fα k β + f β kα , ε 2 = f β kγ − fγ k β , ε 3 = fγ kα + fα kγ
ζ 1 = f α β + f β α , ζ 2 = f β γ − f γ β , ζ 3 = f γ α + fα γ
η1 = gα + g β , η 2 = g β − gγ ,η3 = gγ + gα
η1 = eα + eβ , η 2 = eβ − eγ , η3 = eγ + eα
λ1 = µ1ε 2 − µ 2ε1 , λ2 = µ2ε 3 + µ3ε 2 , λ3 = µ3ε1 − µ1ε 3
λ1 = µ1ζ 2 − µ 2ζ 1 , λ2 = µ 2ζ 3 + µ3ζ 2 , λ3 = µ3ζ 1 − µ1ζ 3
τ1 = fα µ 2 − f β µ3 − fγ µ1 , τ 2 = µ1ε1 + µ 2ε 2 − µ3ε 3 , τ 3 = µ1ζ 1 + µ 2ζ 2 − µ3ζ 3 (43)
ξ1 = kα µ3η 2 − k β µ 2η3 , ξ 2 = k β µ1η3 − kγ µ3η1 , ξ3 = kγ µ2η1 + kα µ1η2
ξ1 = kα µ2η3 + k β µ3η2 , ξ2 = kβµ3η1 − k γ µ1η3 , ξ3 = kγ µ1η 2 − kα µ 2η1
ρ1 = αµ3η2 − βµ 2η3 , ρ 2 = βµ1η3 − γµ3η1 , ρ3 = γµ2η1 + αµ1η2
ρ1 = αµ 2η3 + βµ 3η2 , ρ 2 = βµ 3η1 − γµ1η3 , ρ3 = γµ1η 2 − αµ 2η1
ν 1 = αµ3η 2 − βµ 2η3 , ν 2 = βµ1η3 − γµ3η1 ,ν 3 = γµ2η1 + αµ1η 2
ν 1 = αµ 2η3 + βµ 3η 2 , ν 2 = βµ 3η1 − γµ1η3 , ν 3 = γµ1η 2 − αµ 2η1
σ 1 = −kα µ2η 2 − k β µ3η3 + kγ µ1η1 , σ 2 = αµ 2η2 + βµ3η3 − γµ1η1 , σ 3 = αµ 2η2 + βµ3η3 − γµ1η1

and

13
 ( ) ( ) ( ) (
∆ = µ12 − µ 2 2 + µ32 S β Sγ S hα − 2 µ1µ 2 S β 1 − Cγ Chα − 2 µ 2 µ3Sγ 1 − C β Chα + 2 µ1µ3S hα 1 − C β Cγ ) (44)

The above 6×6 frequency dependent dynamic stiffness matrix K of Eq. (40) can now be used

to compute the natural frequencies and mode shapes of either an individual FGB or an

assembly of FGBs for different boundary conditions. A reliable and accurate method of

solving the problem is to apply the Wittrick-Williams algorithm [39] which is well suited for

the application of DSM. The algorithm uses the Sturm sequence property of the dynamic

stiffness matrix to ensure that no natural frequencies of the structure analysed are missed. The

Wittrick-Williams algorithm has featured in hundreds of papers in the literature and its details

are not repeated here. Basically the algorithm [39] gives the number of natural frequencies of

a structure that lie below an arbitrarily chosen trial frequency specified by the user. As

successive trial frequencies can be chosen by the user, this simple feature of the algorithm can

be exploited to bracket any natural frequency between its upper and lower bounds to any

desired accuracy. The results given in the next section were computed by applying the

Wittrick-Williams algorithm as the customary solution technique.

3. Results and discussions

The first set of results was obtained for a uniform FGB with different boundary conditions

with the letters C, F and S denoting clamped, free and simple-support at each end of the beam.

Four classical boundary conditions are investigated, namely, clamped-free (C-F), simply-

supported (S-S), clamped-simply support (C-S) and clamped-clamped (C-C). The simple

support (S) boundary condition is assumed to be equivalent to a pinned support which prevents

both flexural and axial displacements. A wide range of investigations was carried out by

varying the length to thickness ratio (L/h) and the power law index k of the FGB which

14
controls the material property distribution through thickness. However, the authors have been

highly selective when presenting the results because existing literature already covers a huge

amount of data for natural frequencies and mode shapes with the variations of L/h and k, see

for example Ref.[30]. For presentational purposes, only the results for L/h =10 and k = 0.5

have been used in this paper, but the theory has been extensively validated against published

results.

In order to make the results universal and also to be consistent with the published results, the

following non-dimensional natural frequency parameter is defined

L2 ρb
λi = ωi (45)
h Eb

where ωi is the ith angular natural frequency in rad/s, ρ b and E b are density and Young’s

modulus of the bottom surface of the FGB.

Table. 1 shows the first five natural frequencies of the FGB with L/h = 10 and k = 0.5 for C-

F, S-S, C-S and C-C boundary conditions alongside the results reported in a recently published

paper [30]. The close agreement between the results from the current investigation and the

published ones is clearly evident. The maximum discrepancy between the two sets of results

is less than 4%. The corresponding mode shapes shown in Fig. 4, reveal that for the C-F and

C-C boundary conditions, the first, second, fourth and fifth modes are essentially bending

modes whereas the third one is axial. By contrast, for the S-S and C-S boundary conditions,

the first, second, third and fifth modes are basically bending modes and the fourth one axial.

The next set of results was obtained for a stepped beam (see Fig. 5) made of FGM, for which

some comparative results are available in the literature. The DSM theory developed in this

paper can easily account for such problems with any step location, thickness variation and

boundary conditions. However, for brevity only the results for a cantilever (C-F) stepped FGB

15
with step locations L1 = 0.25L, L1 = 0.5L and L1 = 0.75L (see Fig. 5) and the power law index

k = 0.5 are presented in Table. 2 together with the published results [41]. Note that for

consistency, FGM type – II of Ref. [41] is used so that the results are directly comparable.

Clearly, the results from the current investigation are in close agreement with those of [41].

The final set of results was obtained for a portal frame consisting of three beam members

AB, BC and CD as shown in Fig. 6. The natural frequencies of this frame are available in the

literature [47] when all the three beam members of the frame are made of isotropic material

and the supports at both points A and D are either clamped (built-in) or pinned (simply-

supported). These results are based on exact analytical theory. Using the current theory,

results are obtained for both the boundary conditions C-C and S-S at A and D, respectively

and making (i) all the three members AB, BC and CD isotropic (which is achieved by

substituting the power-law index parameter k to zero and using both the top and bottom surface

material properties to be the same and isotropic), (ii) AB and CD isotropic, but BC made of

FGM, (iii) BC isotropic and AB and CD made of FGM and (iv) AB, BC and CD are all made

of FGM. When computing numerical results, all three members of the portal frame are

assumed to have the same rectangular cross-section and length. The width and depth (height)

of the cross-section are taken to be 0.04m and 0.02m, respectively and length of each member

is set to 1m. When any of the beam members is isotropic, it is considered to be made of steel

with Young’s modulus 200 GPa and density 7500 kg/m3 whereas if it is made of FGM, the

bottom surface is considered to be steel with the above properties and the top surface ceramic

with Young’s modulus 380 GPa and density 3960 kg/m3. The computed natural frequencies

are non-dimensionalised with respect to the metallic properties to give

ρAL4
λi = ωi (46)
EI

16
The results of the investigation when all members of the portal frame (see Fig. 6) are metallic,

are given in Table. 3 showing the first three non-dimensional natural frequencies of the frame

when the points A and D are clamped (C-C) or simply-supported (S-S), together with the

results reported in Ref. [47]. The agreement between the sets of results in Table. 3 is excellent.

Table. 4 shows the three non-dimensional natural frequencies when the vertical members AB

and CD and the horizontal member BC of the frame are made of either isotropic metal or FGM

in turn, as indicated, and the points A and D are clamped (C-C). Similar results are obtained

for the case when the points A and D of Fig. 6 are simply-supported (S-S). The results for the

S-S case are shown in Table. 5. Clearly the results shown in Tables. 4 and 5 when compared

to Table. 3 indicate that significant changes in natural frequencies can occur as a result of

using functionally graded beams. This can have profound influence in the design of fire-

resistant multi-storey and multi-bay building structures.

For illustrative purposes, representative mode shapes for the portal frame of Fig. 6 are

presented in Figs 7 and 8 when the points A and D of the frame are built-in (clamped) and

simply-supported, respectively. The power law index parameter k is set to 0.5 when computing

the mode shapes. Figs 7(a) and 8(a) represent the mode shapes when all three members of the

portal frame are metallic whereas Figs. 7(b) and 8(b) shows the mode shapes when the

columns AB and CD are metallic, but the beam BC is made of FGM. By contrast Figs. 7(c)

and 8(c) show the mode shapes for the case when the beam BC of the portal frame is metallic,

but its columns AB and CD are made of FGM. Finally, Figs. 7(d) and 8(d) show the mode

shapes when all three members of the portal frame are made of FGM. Although the basic

nature of the mode shapes for the portal frame remains the same depending on the order of

the natural frequency on a case to case basis, significant changes in the natural frequencies are

found to occur when using FGM as evident from Figs. 7 and 8. As expected the first mode of

17
the portal frame in each case is a sway mode with the frame oscillating between left and right

with virtually no elastic displacement of the central beam. The second mode shows elastic

deformations of all three members with no nodal point or any point of inflection within any

member. By contrast, the third mode reveals somehow a different picture in that a node with

zero displacement appears towards the top end of each columns whereas a node for the beam

appears near its centre. The mode shapes shown in Figs. 7 and 8 are typical, as expected from

the modal analysis of a portal frame and they are in accord with similar mode shapes reported

by other investigators [47-49].

4. Conclusions

The dynamic stiffness matrix of a functionally graded beam is developed by deriving explicit

expressions for the individual stiffness elements in explicit algebraic form. This is achieved

through the application of symbolic computation. The dynamic stiffness theory is applied by

using the Wittrick-Williams algorithm as solution technique to compute the natural

frequencies and mode shapes of some representative problems of uniform functionally

gradient beams, for which the material properties are considered to vary continuously in the

thickness direction according to a power law distribution. A stepped beam made of

functionally graded material is also investigated for its free vibration characteristics. The

results show good agreement with published results. Importantly, the theory has been applied

to study the free vibration behaviour of a portal frame with its constituent members made of

both isotropic and functionally graded material (FGM). The investigation has shown that

significant changes in the free vibration behaviour are possible when using FGM. The

developed theory can be applied to analyse high-rise building structures made of FGM which

has advantageous mechanical properties of metal and virtuous fire-resistant characteristics of

ceramic and it is in this context, the investigation carried out is expected to be most useful.

18
References

[1] Rasheedat Modupe Mahamood, Esther Titilayo Akinlabi, Functionally Graded Materials,
Springer International Publishing AG, 2017.
[2] Snehashish Chakraverty, Karan Kumar Pradhan, Vibration of Functionally Graded Beams
and Plates, Academic Press, 2016.
[3] Isaac Elishakoff, Demetris Pentaras, Cristina Gentilini, Mechanics of Functionally Graded
Material Structures, World Scientific, Singapore, 2015.
[4] Zheng Zhong, Linzhi Wu, Weiqiu Chen, Mechanics of Functionally Graded Materials and
Structures, Nova Science Publishers Inc, 2012.
[5] A.Y. Tang, J.X. Wu, X.F. Li, K.Y. Lee, Exact frequency equations of free vibration of
exponentially non-uniform functionally graded Timoshenko beams, Int. J. Mech. Sci. 89
(2014) 1–11.
[6] S.R. Li, Z.Q. Wan, J.H. Zhang, Free vibration of functionally graded beams based on both
classical and first-order shear deformation beam theories, Appl. Math. Mech. (English Ed.) 35
(2014) 591–606.
[7] K. Sarkar, R. Ganguli, Closed-form solutions for axially functionally graded Timoshenko
beams having uniform cross-section and fixed-fixed boundary condition, Compos. Part B Eng.
58 (2014) 361–370.
[8] T.P. Vo, H.T. Thai, T.K. Nguyen, F. Inam, Static and vibration analysis of functionally
graded beams using refined shear deformation theory, Meccanica. 49 (2014) 155–168.
[9] T.K. Nguyen, T.P. Vo, H.T. Thai, Static and free vibration of axially loaded functionally
graded beams based on the first-order shear deformation theory, Compos. Part B Eng. 55
(2013) 147–157.
[10] X.F. Li, Y.A. Kang, J.X. Wu, Exact frequency equations of free vibration of exponentially
functionally graded beams, Appl. Acoust. 74 (2013) 413–420.
[11] H.T. Thai, T.P. Vo, Bending and free vibration of functionally graded beams using
various higher-order shear deformation beam theories, Int. J. Mech. Sci. 62 (2012) 57–66.
[12] G. Giunta, D. Crisafulli, S. Belouettar, E. Carrera, Hierarchical theories for the free
vibration analysis of functionally graded beams, Compos. Struct. 94 (2011) 68–74.
[13] J. Murín, M. Aminbaghai, V. Kutiš, Exact solution of the bending vibration problem of
FGM beams with variation of material properties, Eng. Struct. 32 (2010) 1631–1640.

19
[14] M. Şimşek, Fundamental frequency analysis of functionally graded beams by using
different higher-order beam theories, Nucl. Eng. Des. 240 (2010) 697–705.
[15] S.A. Sina, H.M. Navazi, H. Haddadpour, An analytical method for free vibration analysis
of functionally graded beams, Mater. Des. 30 (2009) 741-747.
[16] X.F. Li, A unified approach for analyzing static and dynamic behaviors of functionally
graded Timoshenko and Euler-Bernoulli beams, J. Sound Vib. 318 (2008) 1210-1229.
[17] M. Aydogdu, V. Taskin, Free vibration analysis of functionally graded beams with simply
supported edges, Mater. Des. 28 (2007) 1651-1656.
[18] Z. Zhong, T. Yu, Analytical solution of a cantilever functionally graded beam, Compos.
Sci. Technol. 67 (2007) 481–488.
[19] B. V. Sankar, An elasticity solution for functionally graded beams, Compos. Sci. Technol.
61 (2001) 689-696.
[20] T.P. Vo, H.T. Thai, T.K. Nguyen, A. Maheri, J. Lee, Finite element model for vibration
and buckling of functionally graded sandwich beams based on a refined shear deformation
theory, Eng. Struct. 64 (2014) 12–22.
[21] A.E. Alshorbagy, M.A. Eltaher, F.F. Mahmoud, Free vibration characteristics of a
functionally graded beam by finite element method, Appl. Math. Model. 35 (2011) 412–425.
[22] A. Chakraborty, S. Gopalakrishnan, J.N. Reddy, A new beam finite element for the
analysis of functionally graded materials, Int. J. Mech. Sci. 45 (2003) 519–539.
[23] K.K. Pradhan, S. Chakraverty, Free vibration of Euler and Timoshenko functionally
graded beams by Rayleigh-Ritz method, Compos. Part B Eng. 51 (2013) 175–184.
[24] L.L. Jing, P.J. Ming, W.P. Zhang, L.R. Fu, Y.P. Cao, Static and free vibration analysis of
functionally graded beams by combination Timoshenko theory and finite volume method,
Compos. Struct. 138 (2016) 192-213.
[25] J. Gong, L. Xuan, P. Ming, W. Zhang, Thermoelastic analysis of functionally graded
solids using a staggered finite volume method, Compos. Struct. 104 (2013) 134–143.
[26] J.F. Gong, P.J. Ming, L.K. Xuan, W.P. Zhang, Thermoelastic analysis of three-
dimensional functionally graded rotating disks based on finite volume method, Proc. Inst.
Mech. Eng. Part C J. Mech. Eng. Sci. 228 (2013) 583–598.
[27] S. Rajasekaran, Differential transformation and differential quadrature methods for
centrifugally stiffened axially functionally graded tapered beams, Int. J. Mech. Sci. 74 (2013)
15–31.

20
[28] S. Rajasekaran, Buckling and vibration of axially functionally graded nonuniform beams
using differential transformation based dynamic stiffness approach, Meccanica. 48 (2013)
1053–1070.
[29] Y.S. Al Rjoub, A.G. Hamad, Free vibration of functionally Euler-Bernoulli and
Timoshenko graded porous beams using the transfer matrix method, KSCE J. Civ. Eng. 21
(2017) 792–806.
[30] J.W. Lee, J.Y. Lee, Free vibration analysis of functionally graded Bernoulli-Euler beams
using an exact transfer matrix expression, Int. J. Mech. Sci. 122 (2017) 1-17.
[31] H. Su, J.R. Banerjee, Development of dynamic stiffness method for free vibration of
functionally graded Timoshenko beams, Comput. Struct. 147 (2015) 107–116.
[32] H. Su, J.R. Banerjee, C.W. Cheung, Dynamic stiffness formulation and free vibration
analysis of functionally graded beams, Compos. Struct. 106 (2013) 854-862.
[33] J. Fitch, Solving algebraic problems with REDUCE, J. Symb. Comput. 1 (1985) 211-227.
[34] A.C. Hearn, REDUCE User’s Manual, Version 3.5, RAND Publication, Santa Monica,
California, USA, 1993.
[35] J.R. Banerjee, A.J. Sobey, H. Su, J.P. Fitch, Use of computer algebra in Hamiltonian
calculations, Adv. Eng. Software, 39 (2008) 521-525.

[36] F.W. Williams, W.H. Wittrick, Exact buckling and frequency calculations surveyed, J.
Struct. Eng, ASCE, 109 (1983) 169-187.

[37] F.W. Williams, Review of exact buckling and frequency calculations with optional multi-
level substructuring, Comput. Struct., 48 (1993) 547-552.
[38] J.R. Banerjee, Dynamic stiffness formulation for structural elements: A general approach,
Comput. Struct. 63 (1997) 101-103.
[39] W.H. Wittrick, F.W. Williams, A general algorithm for computing natural frequencies of
elastic structures, Quart. J. Mech. Appl. Math. 24 (1971) 263-284.
[40] D.V. Bambill, C.A. Rossit, D.H. Felix, Free vibrations of stepped axially functionally
graded Timoshenko beams, Meccanica. 50 (2015) 1073-1087.
[41] N. Wattanasakulpong, J. Charoensuk, Vibration characteristics of stepped beams made
of FGM using differential transformation method, Meccanica. 50 (2015) 1089-1101.
[42] M.A. Eltaher, A.E. Alshorbagy, F.F. Mahmoud, Determination of neutral axis position
and its effect on natural frequencies of functionally graded macro/nanobeams, Compos. Struct.
99 (2013) 193-201.

21
[43] K.S. Al-Basyouni, A. Tounsi, S.R. Mahmoud, Size dependent bending and vibration
analysis of functionally graded micro beams based on modified couple stress theory and
neutral surface position, Compos. Struct. 125 (2015) 621-630.
[44] L.A. Pipes, L.R. Harvill, Applied mathematics for engineers and physicists, McGraw-
Hill, 1970.

[45] J.R. Banerjee, Coupled bending-torsional dynamic stiffness matrix for beam elements,
Int. J. Num. Meth. Eng. 28 (1989) 1283-1298.
[46] J.R. Banerjee, Explicit analytical expressions for frequency equation and mode shapes of
composite beams, Int. J. Solids Structs. 38 (2001) 2415-2426.
[47] C.P. Filipich, P.A.A. Laura, In-plane vibrations of portal frames with end supports
elastically restrained against rotation and translation, J. Sound Vib. 117 (1987) 467–473.

[48] R.O. Grossi, C.M. Albarracin, A variational approach to vibrating frames, Proc. Inst.
Mech. Eng., Part K: J. Multibody Dyn. 221 (2007) 247-259.

[49] I. Tatar, Vibration characteristics of portal frames, MSc Thesis, June 2013, Izmir Institute
of Technology, Graduate School of Engineering and Sciences, Izmir, Turkey.

22