Professional Documents
Culture Documents
David (2009) PDF
David (2009) PDF
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/232389936
CITATIONS READS
34 48
3 authors, including:
David Mba
De Montfort University
290 PUBLICATIONS 2,576 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
European Aviation Safety Agency, EASA. HMGH- Helicopter Main Gearbox Health” View project
All content following this page was uploaded by David Mba on 13 September 2014.
Buckling and Vibration Analysis of Functionally Graded Composite Structures using the Finite
Element Method
ABSTRACT
The authors [1] have previously written a paper on structural integrity of functionally graded composite (FGC) structure
using Mindlin-type finite elements. In this paper, the Mindlin-type element and Reissner-type element have been further
developed for the modelling of FGC plate subjected to buckling and free vibration. The Mindlin-type element formulation is
based on averaging of transverse shear distribution over plate thickness using Lagrangian interpolation. The Reissner-type
element formulation is based on parabolic transverse shear distribution over plate thickness using Lagrangian and Hermitian
interpolation. The composite plate considered in this paper is functionally graded in the longitudinal direction only, but the
FE code developed is capable of analysing composite plates with functional gradation in transverse and radial direction as
well. This study was able to show that the structural integrity enhancement and strength maximisation of composite
structures are achievable through functional gradation of material properties over the structure.
KEY WORDS: Functionally graded material, finite element methods, Mindlin-type element, Reissner-type element, finite
1
List of Symbols Longitudinal direction of the material axis
B Matrix of shape function derivatives Transverse direction of the material axis
Transverse shear strain vector y Transverse direction of the local axis
X Longitudinal strength
Y Transverse strength
1. Introduction
Non-dimensional x and y location Functionally Graded Materials (FGMs) are made such
2
2000 K and a temperature gradient of 1000 K across a In comparison to existing publications, this paper has
cross section <10 mm [2]. In 2000, Reddy [3] presented a been able to give unique contributions to the subject
theoretical formulation and finite element models based matter. These contributions include Mindlin-type element
on third order shear deformation theory for the analysis of formulation, Reissner-type element formulation, finite
through-thickness functionally graded plates. The Navier strain modelling and smooth fibre distribution technique.
solution for simply supported plates based on the linear The Mindlin-type element formulation is based on an
third-order theory and the non-linear static and dynamic assumption of average transverse shear distribution over
finite element results based on the first-order theory were plate thickness using Lagrangian interpolation. The
presented by Reddy in [3]. The results show the effects of Reissner-type element formulation is based on an
volume fractions and modulus ratio of the constituents on assumption of parabolic transverse shear distribution over
deflections and transverse shear stresses. In 2002, plate thickness using Lagrangian and Hermitian
Javaheri et al. [4] derived equilibrium and stability interpolation. Green’s strain-displacement equation was
equations for rectangular simply supported functionally employed in the finite strain modelling. The smooth fibre
graded plates. Javaheri’s derivation was based on the distribution technique is based on the numerical
classical plate theory with the assumption of power law computation of macro-mechanical properties at Gaussian
composition for the material and he studied the buckling quadrature points.
analysis of functionally graded plates under in-plane The authors [1] have previously written a paper on
compression. In 2004, Chen et al. [2] investigated the structural integrity of functionally graded composite
buckling behaviour of FGM rectangular plates subjected structure using Mindlin-type finite elements. In the paper
to non-linearly distributed in-plane edge loads. Chen et [1], two new Mindlin-type plate bending elements were
al. [2] stated that a mesh-free method which approximates derived for the modelling of functionally graded plate
displacements based on scattered nodes (i.e. radial basis subjected to various loading conditions such as tensile
function and polynomial basis) was employed, in-order to loading, in-plane bending and out-of-plane bending.
avoid complicated numerical procedures that arises in the There were two types of non-linearity considered in the
FEM from the use of elements. This FEM complication modelling of the plate, which include finite strain and
was dealt with in this paper. Other useful studies on material degradation. In the Mindlin-type element
functionally graded materials can be found in these formulation, the transverse shear strain is averaged over
references [5 - 11].
3
plate thickness. Its finite element derivation is based on considered include buckling load constraints and natural
In order to model the transverse shear strain more description of this optimisation technique concept.
have been satisfied. The constraints that have been The transverse stiffness of a composite is given below.
4
1 V V
f m E
E 22 E f E m (2) X c 2 X fs Vf Vm m (7)
E1f
The in-plane shear modulus can be obtained from a
The transverse tensile strength of FGMs is given below
similar model to that used for obtaining transverse
X mt rm
stiffness which results in the in-plane shear modulus Yt (8)
kc
expression as given below.
The transverse compressive strength of FGMs is given
1 Vf Vm
below.
12 f m (3)
X mc rm
The Poisson’s ratio of a composite is given below. Yc (9)
kc
12 Vf f Vm m (4)
The shear strength of FGMs is given below.
X ms
S (10)
2.2 Strength Properties of FGMs ks
The longitudinal tensile strength of FGMs is given
below.
2.3 Fibre Distribution Techniques
E This section explains the implementation of fibre
X t X ft Vf Vm m
E1f distribution in the FE code. The equation used for fibre
E (5)
X t X mt Vm Vf 1f distribution is given below.
Em
V f (ξ ) V1 (V2 V1 ) ξ p (11)
The longitudinal compressive strength of FGMs for V f <
where
0.5 is given below.
x x1
ξ for ic 1
E E V x 2 x1
,
X c 2 Vf m 1f f
3 1 f ξ
y y1
, for ic 2
y 2 y1
E E V (6)
X c 2 Vf m 1f f r r1
ξ , for ic 3
3 vm r2 r1
V1 fibre ratio at x1 , y1 or r1
If Vf 05
V 2 fibre ratio at x 2 , y 2 or r2
p power index
5
Manual Fibre Distribution Hence
p 1 p 1
Smooth Fibre Distribution (ξ max ξ min )
V f (ξ ) V1 (V2 V1 )
(ξ max ξ min )(p 1 )
This fibre distribution technique enables the code user
V f ( ) V1 (V2 V1 )Q
to either specify fibre ratios V1 and V2 or the mean fibre
p 1 p 1
(ξ max ξ min )
where Q
ratio V and fibre ratio V1. These fibre ratios are then (ξ max ξ min )(p 1 )
used in computing the fibre ratios at each Gaussian V f (ξ ) V2 Q V1( 1 Q)
quadrature point using equations which will be derived V f V1( 1 Q)
V2
later in this section. Q
(15)
Derivation of mean fibre ratio V for specified fibre ratios The plots of the fibre ratio distribution for all ten
6
fibre distribution in the FGC plate and the traditional
Figure 2: Smooth fibre ratio distribution plot for cases Figure 4: Smooth fibre ratio distribution plot for cases
with P=0 and P=0.5 with P=0 and P=2
7
Figure 7: Average fibre ratio distribution plot for
cases with P=0 and P=2
Figure 5: Average fibre ratio distribution plot for
cases with P=0 and P=0.5
3. Finite Element Equations for Mindlin-type
Element
element theory.
element equation.
8
3.2 Element Mass Matrix Derivation q ( x , y , t ) N ( x , y ) δ b (t ) (23)
(L) (16)
Ni (,) ...
... 0 0
N (x, y) N (,) (26)
Using D’Alembert’s principle, the inertial force vector ... 0 Ni ( ,) 0 ...
vector is:
Equation (18) can be rewritten as:
d F a dxdydz (17)
h d δ t
o N t
o N o δo
Hence, the work done by the inertia force due to an
dW a h dδ tb N tw N w δ b dxdy (28)
infinitesimal virtual displacement field is: x y
h dδ tb N t N δ b
3
12
dW a d q .a dxdydz
volume
(18) Defining the following mass matrices:
d q
t
q dxdydz
M oo h N (ξ, η) N o (ξ, η) dxdy
t
volume o
(29)
x y
Using the expression below
M ww h N (ξ, η) N w (ξ, η) dxdy
t
(30)
q q z q
w
x y
q (x, y, z, t) uv o (19)
qw qw
h3 t
It can be shown that
M 12 N (ξ, η) N (ξ, η) dxdy
x y
(31)
dq t q dq to zdqt q o z q dq tw q w (20) Equation (28) can be rewritten as:
dW a dδ to M oo δ o dδ tb M ww M δ b
Expanding the equation above and integrating with
respect to z gives: dδ t M δ
h/2
t
(32)
w
t t
d q q dz h d q o q d q w q
o
h / 2 (21) Hence it can be shown that the element mass matrix is as
h3
d q t q given below:
12
Defining the following displacement components: M 0
M oo (33)
0 M ww M
q o ( x , y , t ) N o ( x , y ) δ o (t ) (22)
9
N ( x , y)
n
y ( x , y) i y i (37)
4. Finite Element Equations for Reissner-type i 1
equilibrium. The generalised equation of equilibrium is n
w(x, y) Fi (x, y)wi Gi (x, y)wi,x Hi (x, y)wi, y (38)
then linearised in-order to obtain the Reissner-type i1
element equation. Finally, the stiffness matrix and mass (b) Conforming elements
10
These strain components can be obtained from Green’s Using the principle of virtual work, the generalised
strain-displacement equations i.e. Equations (43 – 45). equation of equilibrium can be derived.
They can be divided into two parts which include the dU dW 0 (48)
infinitesimal component derived from the Cauchy’s An approximate solution of this equation of equilibrium
strain-displacement equation, and the additional non- gives the expression for the residual vector.
linear terms in Green’s equation. Linearisation of Equations of Equilibrium and Derivation
u 1 u w
2 2 2
v of Element Equations
x (43)
x 2 x x x In order to restore equilibrium, the residual vector
v 1 u v w
2 2 2 must approach a value of zero. This equilibrium is
y
y 2 y y y
(44)
achieved by employing the expressions below.
δ new δ old δ
u v u u v v w w
xy σ (new
L)
σ old σ
y x x y x y x y
(45) (49)
A new A old A
The total strain vector can be obtained in terms of the
The combination of the above expressions and the
nodal parameters and shape function by substituting the
residual vector expressions results in a final matrix
above displacement and slope components into the
equation, which is given below.
Green’s strain-displacement equations. The total strain
δo R o
K K δ
Ne
vector equation above can then be manipulated to obtain R (50)
e 1 b b
the variation of strain vector in terms of the nodal
The variation of strain energy density at a point inside This derivation follows the same procedure as in
the Lth layer is given below. section 3.2. Hence Eqs.16-18 also applies to this section.
q q z q f q
Generalised Equation of Equilibrium q ( x , y , z , t ) uv o (51)
q w qw
The work done by actual loads can be expressed in
It can be shown that
terms of equivalent nodal loads as given below.
11
d q t q d q to zd q t f d q t
h dδt t
o N o No δ o h dδbNwNw δb
t t
(52)
h3 t t
q o z q f q d q tw q w dδbN N δb
12
dWa 17 h3 dxdy
x y dδt N N δt
t t
Expanding the equation above and integrating with
140
respect to z gives: h3 t t
dδbN N δt dδt N N δb
t t
h/2 10
t
w
(62)
t t
dq q dz h dq q d q q
o w
o
h / 2
Defining the following mass matrices:
h3 17h 3 t
dqt q dq q (53)
12 140 h N (ξ, η) N o (ξ, η) dxdy
t
M oo o
(63)
h3
x y
dqt q dqt q
10
h N (ξ, η) N w (ξ, η) dxdy
t
M ww w
(64)
Defining the following displacement components: x y
q o ( x , y , t ) N o ( x , y ) δ o (t ) (54) h3 t
M
x y 12
N (ξ, η) N (ξ, η) dxdy (65)
q ( x , y , t ) N ( x , y ) δ b (t ) (55)
17 h 3 t
q ( x , y , t ) N ( x , y ) δ t (t ) (56) M 140 N (ξ, η) N (ξ, η) dxdy
x y
(66)
q w ( x , y , t ) N w ( x , y ) δ b (t ) (57)
h3 t
M N (ξ, η) N (ξ, η) dxdy (67)
... Ni (,) 0 ... x y 10
No (x, y) No (,)
Ni (,) ...
(58)
... 0
h3 t
M N (ξ, η) N (ξ, η) dxdy M
t
(68)
... 0 Ni (,) ... x y
10
N (x, y) N (,)
...
(59)
... - Ni (,) 0
Equation (62) can be rewritten as:
v
x ... Fi, x Hi, x Gi, x Pi, x ... dW a dδ tt M δ t
N (x, y) w
(69)
(61)
d δ tb M δ t d δ tt M δ b
... Fi, y Hi, y Gi, y Pi, y ...
y
Equation (18) can be rewritten as: dδ t M δ
given below:
12
M oo 0 0 5.2 Numerical Example of Buckling Case
M 0 M ww M M
(70)
0
Study
M M
Composite Material Data
K λK δ 0
o
(74)
stress
K λK o 0
SKC, Longitudinal stress 1
This gives a characteristic equation and its smallest real concentration factor
roots define the critical buckling load. SKS, Shear stress concentration 1
13
V , Average fibre ratio 0.4 seen that the first and second buckling mode results of the
P, Power term in the fibre 0, 0.5, 1 or 2 Reissner and Mindlin programs are in good agreement.
distribution equation But the third buckling mode results of the Reissner and
A rectangular plate made of a typical FGM with its discrepancy can be attributed to the difference transverse
mid-plane as shown in Figure 8 was considered. The modelling technique employed by both elements. Also
width-to-thickness ratio of the plate is 33. A 72 element the discrepancy could be reduced by using a finer mesh of
mesh was employed for all the case studies. The elements the Mindlin-type element which results in the
used in this exercise include Mindlin-type element and convergence of the Mindlin-type element solution to
Reissner-type element. The boundary condition applied Reissner-type element solution. This is due to the fact
in the case studies is that edge x=0 is a clamped edge. A that the Reissner-type element (based on parabolic
load of 0.1kN was applied as an equivalent nodal loading transverse shear modelling through the plate thickness)
at edge x=2 for the buckling case study. gives a more accurate result in comparison with the
Table 2, shows the comparison of the buckling desirable effect of 20% increment in critical buckling load
eigenvalue results for all ten fibre ratio distribution cases. (relative to the traditional composite case) for the first and
Maximisation of the buckling eigenvalue for most second buckling mode is achieved by using functionally
buckling modes is usually the desired effect required for graded composite.
14
Table 2: Buckling Eigenvalue 6. Dynamic Analysis: Procedure and
Buckling Eigenvalue, min = Fcritical / Fo
Numerical Example
st nd rd
P V1 1 Mode 2 Mode 3 Mode
6.1. Dynamic Analysis Solution
(Mindlin) (Mindlin) (Mindlin)
Consider an ideal structure with no damping forces,
0 Vav=0.4 8.130 75.021 145.28
the dynamic finite element matrix equation can be
0.5 0.5 8.504 75.482 150.91
reduced to the expression below.
0.5 0.55 8.665 75.452 184.14
0.5 0.6 8.808 75.256 181.92 M δ t K δ t F t (79)
1 0.5 8.752 75.045 183.72 In a natural mode, each point of a structure executes
1 0.55 9.006 74.364 169.74 harmonic motion about the position of static equilibrium
1 0.6 9.221 73.170 153.49 at the same frequency. Hence it can be assumed that, at a
δ t δ cos (ω t )
~
2 0.55 9.439 71.192 130.65 (80)
1 0.5 8.766 73.231 184.33 On comparing natural frequency problem and buckling
1 0.55 9.024 72.586 169.26 problem, it can be shown that both problems are
2 0.6 9.767 65.177 94.007 The matrix equation of the dynamic eigenvalue problem
15
techniques which include the simple iteration and 0.5 0.55 318.12 1276.0 1893.4
subspace iteration techniques [15 - 17]. 0.5 0.6 323.60 1281.7 1900.4
The FE model used in this section is as described in 2 0.5 325.83 1300.4 1913.3
Section 5.2. Table 3, shows the comparison of the natural 2 0.55 337.79 1314.4 1927.2
vibration frequency results for all ten fibre ratio 2 0.6 349.19 1317.0 1930.6
distribution cases. Maximisation of the natural frequency P V1 1st Mode 2nd Mode 3rd Mode
for most free vibration modes is usually the desired effect (Reissner) (Reissner) (Reissner)
required for design purposes. On looking at the table 0 Vav=0.4 300.24 1257.3 1832.3
below, it can be seen that the results of the Reissner and 0.5 0.5 312.86 1275.2 1855.8
Mindlin programs are in good agreement for all vibration 0.5 0.55 318.67 1282.2 1864.9
modes. Also the natural frequency increases as the 0.5 0.6 No No No
exponent P values increases. This is expected because the solution solution solution
stiffness at the root region of the composite plate
1 0.5 319.64 1287.9 1865.7
increases with increase in exponent P value. Finally, in
1 0.55 328.53 1298.7 1876.9
this vibration analysis, an average desirable effect of 15%
1 0.6 336.96 1306.4 1884.6
increment in natural frequency (relative to the traditional
2 0.5 326.34 1305.9 1884.8
composite case) for all the vibration mode is achieved by
2 0.55 338.40 1319.6 1900.2
using this functionally graded composite.
2 0.6 No No No
Table 3: Natural vibration frequencies solution solution solution
Natural Vibration Frequencies (Hz)
16
Maximum critical buckling load (i.e. maximum performing a buckling and vibration analysis of a
Hence, using the definition of design criterion above, Mindlin-type element formulation, Reissner-type element
the results in the previous sections have been summarised formulation, finite strain modelling and smooth fibre
below and the optimum design was determined from the distribution technique. A methodical approach was used
vibration = 1.13
REFERENCES
17
[3] Reddy J. N., 2000; Analysis of functionally graded means of the interaction integral method.
plates. International Journal of Numerical Methods in Engineering Fracture Mechanics, 71, 1907-1950.
Engineering; 47: 663-684 [11] Yang Y.Y., 1998; Stress analysis in a joint with a
[4] Javaheri R. and Eslami M.R. 2002; Buckling of functionally graded material under a thermal loading
functionally graded plates under in-plane by using the Mellin transform method. Int. J. Solids
compressive loading. Z. Angew. Math. Mech. 82, 4, Structures, Vol. 35, No. 12 pp 1261-1287.
277 – 283 [12] Agarwal B.D. and Broutman L.J., 1990; Analysis and
[5] Shi J. and Askew C., 1997; Thermal stress in performance of fiber composites. 2nd edition.
cylinders made of composite materials. Key [13] El-Zafrany A. and Cookson R. A., 1986; Derivation
Engineering Materials, Vols 127-131, pp. 1109-1116 of Lagrangian and Hermitian shape functions for
[6] Silvia E.C.N. and Paulino G.H., 2004; Topology triangular elements. International Journal of
optimisation applied to the design of functionally Numerical Engineering, Vol. 23, pp. 275-285.
graded material (FGM) structures; XXI ICTAM, [14] El-Zafrany A. and Cookson R. A., 1986; Derivation
[7] Durodola J.F. and Adlington J.E., 1997; Functionally quadrilateral elements. International Journal of
graded material properties for disks and rotors; Key Numerical Methods in Engineering, Vol. 23, pp.
[8] Nogata F., 1995; Intelligent functionally graded [15] Clint M. and Jennings A., 1970; The evaluation of
[9] Pompe W., Worch H., Epple M., Friess W., Gelinsky [16] Corr R. B. and Jennings A., 1976; A simultaneous
M., Greil P., Hempel U., Scharnweber D. and Schulte iteration algorithm for symmetric eigenvalue
K., 2003; Functionally graded materials for problem. International Journal of Numerical Methods
biomedical applications. Materials Science and in Engineering, Vol. 10, pp. 647-663.
Engineering, A362, 40-60. [17] Bathe K. J. and Wilson E. L., 1976; Numerical
[10] Paulino G.H. and Kim J., 2004; A new approach to methods in finite element analysis. Prentice Hall
18