7/31/2019 Lu_Y_D_2008
1/139
7/31/2019 Lu_Y_D_2008
2/139
ii
Random Vibration Analysis of Higher-Order Nonlinear Beams and
Composite Plates with Applications of ARMA ModelsYunkai Lu
ABSTRACT
In this work, the random vibration of higher-order nonlinear beams and composite plates
subjected to stochastic loading is studied. The fourth-order nonlinear beam equation is
examined to study the effect of rotary inertia and shear deformation on the root mean
square values of displacement response. A new linearly coupled equivalent linearization
method is proposed and compared with the widely used traditional equivalent
linearization method. The new method is proven to yield closer predictions to the
numerical simulation results of the nonlinear beam vibration. A systematical
investigation of the nonlinear random vibration of composite plates is conducted in which
effects of nonlinearity, choices of different plate theories (the first order shear
deformation plate theory and the classical plate theory), and temperature gradient on the
plate statistical transverse response are addressed. Attention is paid to calculate the
R.M.S. values of stress components since they directly affect the fatigue life of the
structure. A statistical data reconstruction technique named ARMA modeling and its
applications in random vibration data analysis are discussed. The model is applied to the
simulation data of nonlinear beams. It is shown that good estimations of both the
nonlinear frequencies and the power spectral densities are given by the technique.
7/31/2019 Lu_Y_D_2008
3/139
iii
Acknowledgment
I would like to thank my advisor, Professor Thangjitham, for his guidance and support all
through my work.
I would like to thank all my committee members, Professor Case, Professor Hyer,
Professor Lu, and Professor Ragab, for the time they took to attend my exams and
defense as well as their advice on my dissertation.
My thanks also go to Professor. Kraige, Professor. Hendricks, and the ESM department,
for their financial support during my Ph.D. study.
Last but not least, I want to thank my parents, Lu, Zhicheng and Zhao, Guiying. I would
not have been able to make it through all the difficulties and hard times in my life without
their unconditional love and support all the time.
7/31/2019 Lu_Y_D_2008
4/139
iv
Table of Contents
Chapter 1. Introduction ....................................................................................................... 1
Chapter 2. Literature Review .............................................................................................. 4
Chapter 3. Random Vibration of Geometrically Nonlinear Beams .................................. 14
3.1 Solutions to the Nonlinear Random Vibration of Isotropic Beams ........................ 14
3.2 Effect of Inertia of Rotation and Shear Deformation ............................................. 23
3.3 Numerical Results ................................................................................................... 31
Chapter 4. Nonlinear Random Vibration of Composite Plates ......................................... 54
4.1 Governing Equations .............................................................................................. 54
4.2 Stochastic Response of Linear System ................................................................... 604.3 Stochastic Response of Nonlinear System .............................................................. 63
4.4 Temperature Effects on Random Vibrations of Composite Plate ........................... 68
4.5 Comparison between FSDT and CPT ..................................................................... 72
4.6 R.M.S. Stresses Calculation .................................................................................... 73
Chapter 5. ARMA Model and Its Applications in Random Vibration Data Analysis ...... 79
5.1 Introduction ............................................................................................................. 79
5.2 Theoretical Background .......................................................................................... 80
5.3 Applications of ARMA Model in Identifying, Re-generating, and Extending the
Random Vibration Data ................................................................................................ 83
5.4 Comparison between PSD Curve from ARMA Model and Newlands Approach 88
Chapter 6. Future Work .................................................................................................... 96
7/31/2019 Lu_Y_D_2008
5/139
v
6.1 Durability of Structures Subjected to Random Loading ......................................... 96
6.2 Future Work ............................................................................................................ 99
Reference ........................................................................................................................ 101
Appendix A ..................................................................................................................... 116
A.1 Selected eigenfunctions ....................................................................................... 116
2. Solution procedure for a 22 linearly-coupled system ........................................... 117
Appendix B: Derivation of Plate Equations.................................................................... 121
7/31/2019 Lu_Y_D_2008
6/139
vi
List of Figures
Figure 3.1 A beam under pressure .................................................................................... 15Figure 3.2 Correlation between displacement and acceleration for a typical linear beam
(data size: 214
) ........................................................................................................... 27Figure 3.3 Correlation between displacement and acceleration for a typical nonlinear
beam (data size: 214
).................................................................................................. 28Figure 3.4 Two types of loads used in the simulation ...................................................... 33Figure 3.5 A typical stationary Gaussian random process (time domain) ........................ 33Figure 3.6 Histogram of the random process in Figure 3.5 .............................................. 34Figure 3.7 PSDs of the two types of loads used in the simulation.................................... 34Figure 3.8 Displacement R.M.S. of a uniformly loaded F-SS beam vs. different ............ 35Figure 3.9 Mode 1 displacement R.M.S. of a half-uniformly loaded F-SS ...................... 35Figure 3.10 Mode 2 displacement R.M.S. of a half-uniformly loaded F-SS .................... 36Figure 3.11 Displacement R.M.S. (summation of first two modes) of a uniformly loaded
SS-SS beam vs. different random loading PSD levels ............................................. 36Figure 3.12 Coupling effect on mode 1 for a uniformly loaded F-SS beam .................... 38Figure 3.13 Coupling effect on mode 2 for a uniformly loaded F-SS beam .................... 38Figure 3. 14 Coupling effect on mode 1 for a half uniformly loaded F-SS beam ............ 39 Figure 3. 15 Coupling effect on mode 2 for a half uniformly loaded F-SS beam ............ 39 Figure 3.16 Typical mode 1 displacement response (corresponding to data in Table 3.1)
................................................................................................................................... 41Figure 3.17 Typical mode 2 displacement response (corresponding to data in Table 3.1)
................................................................................................................................... 41Figure 3.18 Typical FFT of mode1 displacement response .............................................. 42
Figure 3.19 Typical FFT of mode 2 displacement response ............................................. 42Figure 3.20 Histogram of mode 1 displacement response ................................................ 43Figure 3.21 Histogram of mode 2 displacement response ................................................ 43Figure 3.22 Comparison among different approaches of mode 1 response ...................... 45Figure 3.23 Comparison among different approaches of mode 2 response ...................... 45
7/31/2019 Lu_Y_D_2008
7/139
vii
Figure 3.24 Comparison among different approaches of mode 1 response ...................... 46Figure 3. 25 Comparison among different approaches of mode 2 response ..................... 46Figure 3. 26 Comparison among different approaches of mode 1 response ..................... 47Figure 3. 27 Comparison among different approaches of mode 2 response ..................... 47Figure 3. 28 Comparison among different approaches of mode 1 response ..................... 48Figure 3.29 Comparison among different approaches of mode 2 response ...................... 48Figure 3. 30 Comparison among different approaches of mode 1 and 2 responses ......... 49 Figure 3. 31 Mode 1 R.M.S. response of a SS-SS beam subjected to uniform load ........ 51 Figure 3. 32 Mode 2 R.M.S. response of a SS-SS beam subjected to uniform load ........ 52 Figure 3. 33 Mode 1 R.M.S. response of a F-SS beam subjected to half-uniform load ... 52Figure 3. 34 Mode 2 R.M.S. response of a F-SS beam subjected to half-uniform load ... 53
Figure 4.1 Free body diagram of a rectangular plate element (without bending moments)
................................................................................................................................... 55Figure 4.2 Free body diagram of a rectangular plate element (bending moments only) .. 55Figure 4.3 RMS values vs. square root of power spectral density .................................... 67Figure 4.4 R.M.S. values vs.T (based on FSDT) ......................................................... 72Figure 4.5 Variations of the ratios between FSDT and CPT R.M.S. values .................... 73Figure 4.6 Contour plot of
xx R.M.S. at middle plane of the first layer of a
( 60 / 60 / 60 / 60 ) laminate .............................................................................. 75Figure 4.7 Contour plot ofxx R.M.S. at middle plane of the first layer of a
(30 / 30 / 30 / 30 ) laminate .............................................................................. 76Figure 4.8 Effect of ply angle on the maximal R.M.S. values of stress components .... 77
Figure 5.1 PSD of the mode 1 of the beam displacement simulation data ....................... 85 Figure 5.2 Displacement from simulation of mode 2 of a F-SS beam ............................. 86Figure 5.3 Displacement generated by ARMA(4, 3) model ............................................. 87Figure 5.4 Displacement generated by ARMA(8, 7) model ............................................. 87Figure 5.5 Displacement generated by ARMA(11, 10) model ......................................... 87Figure 5.6 Example spectrum plots of 8 out of the 32 segments ...................................... 91Figure 5.7 Comparison of PSD curves from ARMA(4,3) model ..................................... 92
7/31/2019 Lu_Y_D_2008
8/139
viii
Figure 5.8 PSD of the beam simulation displacement data (Newlands approach) .......... 93Figure 5.9 PSD from ARMA(4,3) model ......................................................................... 94Figure 5.10 PSD from ARMA(8,7) model ....................................................................... 94Figure 5.11 PSD from ARMA(11,10) model ................................................................... 94
7/31/2019 Lu_Y_D_2008
9/139
ix
List of Tables
Table 3.1 Response of a beam (mm) with F-SS boundary condition and subjected to half
uniform load (load PSD = 100000 Pa2/Hz, 2 , 10 1 mh b L h= = = ) ......................... 44
Table 3.2 Response of a beam with F-SS boundary condition and subjected to half
uniform load (load PSD = 100000 Pa2/Hz, 2 , 20 1 mh b L h= = = )......................... 44
Table 3.3 Response of a beam (mm) with F-SS boundary condition and subjected to
uniform load (load PSD = 10000 Pa2/Hz, 2 , 12.5 1 mh b L h= = = ) ..................... 44Tabel 3.4 Response of a beam (mm) with F-Fixed boundary condition and subjected to
half uniform load (load PSD = 100000 Pa2/Hz, 2 , 12.5 1 mh b L h= = = ) ............... 44
Table 3.5 Comparison of R.M.S. response of a 2nd
order beam with that of 4th
order beam
(SS-SS boundary condition with uniform load, 2 , 10 1 mh b L h= = = ) .................. 50Table 3.6 Comparison of R.M.S. response of a 2
ndorder beam with that of 4
thorder beam
(SS-SS boundary condition with half-uniform load, 2 , 10 1 mh b L h= = = ) .......... 50Table 3.7 Comparison of R.M.S. response of mode 1 of 2
ndorder beam with that of 4
th
order beam (F-SS boundary condition with half-uniform load,
2 , 10 1 mh b L h= = = ) ............................................................................................. 50Table 3.8 Comparison of R.M.S. response of mode 2 of 2
ndorder beam with that of 4
th
order beam (F-SS boundary condition with half-uniform load,
2 , 10 1 mh b L h= = = ) ............................................................................................. 51
Table 5.1 Estimated ARMA parameters for different order models ............................... 85Table 5.2 Frequencies predicted by selected ARMA models in Table 5.1....................... 85Table 5.3 Estimated displacement R.M.S. of selected ARMA models ............................ 87
7/31/2019 Lu_Y_D_2008
10/139
1
Chapter 1. Introduction
A great deal of work has been done on the response of beams and plates subjected to
deterministic loading conditions. However, in real life the structure may be subjected to a
stochastic type of loading such as earthquakes, wind turbulence, sea wave, acoustic loads,
etc. These loading conditions are commonly observed on dams, nuclear facilities,
offshore structures, aircraft, etc. The main purpose of this dissertation is to present a
systematical study of the stochastic response of geometrically nonlinear beams/plates
under random excitations.
In this work, a literature review of previous research in the area is given first in Chapter
2. Then, in Chapter 3 random vibration of geometrically nonlinear beams is elaborated
where the detailed procedures to solve the random vibration problem and the traditional
uncoupled equivalent linearization technique are discussed. To study the effect of the
rotary inertia and shear deformation on the root mean square (R.M.S.) of the stochastic
beam response, the fourth-order nonlinear beam equation is examined. The results from
the fourth-order beam equation are compared with those from second-order. A new
7/31/2019 Lu_Y_D_2008
11/139
2
coupled equivalent linearization method is proposed which takes into account the effects
of modal interactions between adjacent modes. Numerical results indicate that the new
method yields closer results to simulation data compared with the uncoupled linearization
approach. In Chapter 4 the focus is moved from one-dimensional beam problem to
rectangular composite plates using both nonlinear classical plate theory (CPT) and first
order shear deformation (FSDT) theory. FSDT takes into account of the transverse shear
strain effect. A nonlinear stress-strain relationship in the von Karman sense is considered
in the formulations of the governing equations. The effects of nonlinearity and
temperature on the R.M.S. values of transverse displacement of the plate and the selected
stress components are examined.
A statistical data characterization and reconstruction technique called ARMA modeling
and its applications in random vibration data analysis are demonstrated in Chapter 5. The
auto-regressive moving averaging (ARMA) model was originally developed as a time-
domain modal analysis method. ARMA model is very efficient in reconstructing the
loading condition directly in the time domain. The model is very concise in its
formulation, requiring very few parameters while preserving the stochastic nature and
spectral information of the original signal history. In an ARMA model, the current value
of the system response is expressed as a linear combination of past values of response
plus a pure white noise. The parameters in the model are determined through a trial and
error procedure in order to minimize the residue variance of the noise. Once the
parameters of the model are known, natural frequencies and damping ratios for all the
modes can be obtained from the autoregressive part. However, the order of an ARMA
7/31/2019 Lu_Y_D_2008
12/139
3
model to fit a certain data set is not unique. In addition, the correlation between the
nonlinear simulation data usually requires a higher order model than the linear case does
in order to accurately represent the spectral properties of the original input, especially the
power spectral density (PSD) plot. It is shown that even though a model can give good
estimations of the frequency values, it may not represent the PSD closely. At the end of
this work, issues regarding the durability of structure under random excitations are
addressed. Future areas of research work are discussed.
7/31/2019 Lu_Y_D_2008
13/139
4
Chapter 2. Literature Review
Stochastic loads such as wind turbulence, sea wave, and acoustical loads are commonly
observed on aerospace, mechanical and civil structures. Consequently, random vibration
analysis is necessary to understand the behavior of these structures under stochastic
loading. For a general review on the random vibration theory, one can refer to the
references (Crandall and Mark 1973, Bolotin 1984, Nigam 1983, Roberts and Spanos
1990, Newland 1993, Solnes 1997, Wirsching et al. 1995, Lin 1976, and Elishakoff
1983). The exact solutions to nonlinear random vibration problems are only available for
certain special cases. Therefore, approximation technique and numerical solutions were
developed to find the probability density functions of the response of the nonlinear
system. For limited cases, the moments of the response can be obtained via solving the
Fokker-Planck equation (Stratanovich 1963, Stratanovich 1967, Risken 1996, and
Gardiner 2004). A set of ordinary differential equations for the moment characteristics of
response can be obtained after applying a closure technique such as the Gaussian closure
method (Iyengar and Dash 1978). Perturbation method (Nayfeh 1993, and Nayfeh and
7/31/2019 Lu_Y_D_2008
14/139
5
Mook 1995) is the most widely used technique dealing with the nonlinear dynamic
response of systems with nonlinearity. It can also be adapted to solve the nonlinear
random vibration of such systems (Crandall 1963). However, by nature the perturbation
method is only applicable when the nonlinearity is small, which greatly limited its usage
in a wide range of problems. Another method called stochastic averaging technique
(Roberts and Spanos 1990, and Socha and Soong 1991) can also be applied to weakly
nonlinear systems. The equivalent linearization (Caughey 1959a, Caughey 1959b,
Caughey, 1963, Spanos 1981, and Roberts et al. 1990) is the most commonly used
approximation method due to its straight-forward formulation and effectiveness. When
the load is white noise, the equivalent linearization yields the same results as Gaussian
closure method (Er 1998).
The random vibrations of beams have been studied since the 1950s (Eringen 1957,
Bogdanoff and Goldberg 1960, Crandall and Yildiz 1962, Elishakoff and Livshits 1984,
and Elishakoff 1987). An exact probability density function of modal displacements was
found by Herbert (1964, 1965). Among all the approaches, the two mostly used are the
perturbation method and the stochastic linearization technique. Eringen (1957),
Elishakoff (1987) and Elishakoff and Livshits (1984) came up with closed-form solutions
for simply supported beams subjected to random loading in the form of infinite modal
summation. Exact solutions by the Fokker-Planck equation method only exist for some
extreme cases. Even if an exact solution exists, a large amount of multiple integrations
are needed to evaluate the root mean square value of the response, which makes it
computationally prohibitive. Fang et al. (1995) and Elishakoff et al. (1995) proposed an
7/31/2019 Lu_Y_D_2008
15/139
6
improved stochastic linearization method by minimizing the potential energy of the beam
under stationary random excitation. They claimed that the new approach improved the
accuracy of the conventional stochastic linearization method. Different variations of the
improved stochastic linearization technique can be found in the literature (Elishakoff and
Zhang 1991, Elishakoff 1991, Zhang et al. 1990, and Fang and Fang 1991).
Since one part of this dissertation studies the random vibration of composite plates using
classical plate theory (CPT) and first order shear deformation plate theory (FSDT), a
comprehensive introduction of composite material as well as different plate theories can
be found in the work of Reddy (1997, 2004). Before the stochastic response of composite
plate is examined, a brief review of some of the work on dynamic response of plates
using different theories is given as follows. Some of the studies investigated the nonlinear
vibrations of composite plates or functionally graded plate (a special type of composite
plate), in which iteration scheme was used similar to the equivalent linearization in the
random vibration analysis.
Kim and Noda (2002) discussed transient displacement of functionally graded composite
plates due to heat flux by a Greens function approach based on the classical plate theory.
Praveen and Reddy (1998) investigated the static and dynamic responses of functionally
graded ceramicmetal plates by using a plate finite element that accounts for the
transverse shear strains, rotary inertia and moderately large rotations in the von Karman
sense. Reddy (2000) analyzed the static behavior of functionally graded rectangular
plates based on the third-order shear deformation plate theory via finite element
7/31/2019 Lu_Y_D_2008
16/139
7
approach. Theoretical formulation along with Naviers solution and finite element model
for the plate were presented. Woo and Meguid (2001) applied the von Karman theory for
large deformation to obtain the analytical solution for the plates and shell under
transverse mechanical loads and a temperature field. Zenkour (2006) presented a general
formulation for functionally graded composite plates using the generalized shear
deformation theory that did not require a shear correction factor. Cheng and Batra
(2000a) presented results for the buckling and steady state vibrations of a simply
supported functionally graded polygonal plate based on Reddys plate theory. Cheng and
Batra (2000b) also related the deflections of a simply supported functionally graded
polygonal plate given by the first-order shear deformation theory and a third-order shear
deformation theory to that of an equivalent homogeneous Kirchhoff plate. Loy et al.
(1999) studied the vibration of functionally graded cylindrical shells using Loves shell
theory and RayleighRitz method. Liew et al. (2001, 2002a, 2002b) used classical plate
theory and the first order shear deformation theory to present the finite element
formulation for the shape and vibration control of functionally graded plates with
integrated piezoelectric sensors and actuators. He et al. (2001) presented the vibration
control of functionally graded plate with integrated piezoelectric sensors and actuators by
a finite element formulation based on CPT. Huang and Shen (2004) solved the nonlinear
vibration and dynamic response of simply supported functionally graded plates subjected
to a steady heat conduction process through an improved perturbation technique. Woo et
al. (2006) provided an analytical solution in terms of mixed Fourier series for the
nonlinear free vibration behavior of composite plates. The nonlinear coupling effects on
the fundamental frequencies were examined. Liew et al (2006) presented the nonlinear
7/31/2019 Lu_Y_D_2008
17/139
8
vibration analysis for layered cylindrical panels subjected to a temperature gradient due
to steady heat conduction along the panel thickness direction. A nonlinear pre-vibration
analysis was conducted to obtain the thermally induced pre-stresses and deformation.
Differential quadrature method with an iteration scheme was employed to find the
nonlinear vibration characteristics of the panel.
Yang and Shen (2001) presented the dynamic response of initially stressed functionally
graded thin plates. Yang and Shen (2002) investigated the free and forced vibration
problems for the shear-deformable functionally graded plate in thermal environment.
Their results indicated that the plates with intermediate material properties did not
necessarily have intermediate dynamic response. Kitipornchai et al. (2004) gave a semi-
analytical solution for the nonlinear vibration of imperfect functionally graded plates
based on higher-order shear deformation theory with temperature dependent material
properties. The sensitivity of the nonlinear vibration characteristics of plates to the initial
geometric imperfection was evaluated. In Yang et al. (2004), a semi-analytical Galerkin-
differential quadrature approach was employed to convert the governing equations into a
linear system of MathieuHill equations. The influences of various parameters such as
material composition and temperature change on the dynamic stability, buckling and
vibration frequencies were demonstrated through parametric studies. The stability of a
functionally graded cylindrical shell subjected to axial harmonic loading was discussed
by Ng et al. (2001). Patel et al. (2005) conducted the finite element analysis for the free
vibration of elliptical composite cylindrical shells based on the high order shear
deformation theory. Sofiyev (2004) and Sofiyev and Schnack (2004) studied the dynamic
7/31/2019 Lu_Y_D_2008
18/139
7/31/2019 Lu_Y_D_2008
19/139
7/31/2019 Lu_Y_D_2008
20/139
11
behavior of nonlinear systems (Rice 1999, Chen and Tomlison 1996, Staszewski 1998,
Boukhrist et al. 1999, and Jaksic and Boltezar 2002). However, when the data consists of
signals from all over the frequency range (white noise or narrow-band white noise),
traditional modeling identification methods have difficulties since extra filtering process
has to be conducted to remove the noise from the data so that the harmonic components
can be exposed. To analyze signal like this, the auto-regressive moving average (ARMA)
model is a very powerful tool. It is also called Box-Jenkins models named after the
people who developed it. A detailed introduction can be found in Box et al. (1994) and
Chatfield (1989). In an ARMA model, the current value of the system response is
expressed as a linear combination of past values plus a white noise. Once the parameters
are determined, natural frequencies, damping ratios (if applicable) can be obtained from
the autoregressive part of the model. The typical procedure for fitting an ARMA models
to a time series involves model identification, model fitting, and model validation. It
should also be pointed out that the applications of ARMA model are not restricted to
engineering field. For instance, it has been used in the analysis of financial data such as
stock market changes and other economical issues (Mills, 1990). Tian and Tan (1987)
used ARMA time series model to study the information of heart sounds of normal human
and patients with cardiovascular disorders. A cardiac functional state which was
determined from the ARMA parameters provided valuable information on the initiation
of heart-failure.
Baek et al. (2006) proposed a modeling method of the mass, the damping coefficient and
the stiffness of a cutting system using an autoregressive moving average (ARMA) model
7/31/2019 Lu_Y_D_2008
21/139
12
and a bisection method. Yoon et al. (2004) compared different algorithms in estimating
the structural dynamic between the endmill and workpiece of a cutting system. Baek et al.
(2006) conducted parameter identification on the experimental data of single-degree-of-
freedom system using ARMA model. Wang et al. (2003) evaluated the nonlinear fluid
force for a freely vibrating cylinder over a wide range of Reynolds numbers, mass and
structural damping ratios. Smail and Thomas (1999) compared the accuracy of three
kinds of ARMA methods (recursive, least-squares output error and corrected covariance
matrix method) in parameter identification of certain simulations and experimental data.
Effects of model orders and sampling frequency were studied. Its found that a good
sampling frequency ranged from three to ten times the maximal frequency of interest.
This information was used while running the simulations for the beam and plate in this
dissertation. Carden and Brownjohn (2007) applied the ARMA modeling technique on
the experimental data from the IASCASCE benchmark four-storey frame structure as
well as two bridge structures. A health-monitoring algorithm was examined that
distinguished a structure in a healthy state from that in an unhealthy state. Mattson and
Pandit (2006) used vector autoregressive (ARV) models to capture the predictable
dynamic properties in the experimental response data. The standard deviation of the
autoregressive residual series provided valuable information on the location of damage in
the structures. Sohn and Farrar (2001) combined auto-regressive and auto-regressive with
exogenous inputs techniques and conducted damage diagnosis of a mass-spring system
with eight degrees of freedom.
7/31/2019 Lu_Y_D_2008
22/139
7/31/2019 Lu_Y_D_2008
23/139
14
Chapter 3. Random Vibration of Geometrically
Nonlinear Beams
In this chapter the fundamentals of random vibration of geometrically nonlinear beams
are elaborated. Issues such as the solution procedures, linearization technique, and effects
of nonlinearity and modal interaction are addressed. A new equivalent coupled
linearization approach is proposed and compared with the traditional equivalent
uncoupled linearization method. The solution to the random vibration of fourth order
beams is obtained with attention paid to the effects of rotary inertia and shear
deformation.
3.1 Solutions to the Nonlinear Random Vibration of Isotropic
Beams
The geometry of a simply supported beam subjected to uniform pressure is shown in
Figure 3.1. The nonlinear equation of motion for the transverse displacement w(x,t) of a
uniform beam (Foda 1999) can be expressed as
7/31/2019 Lu_Y_D_2008
24/139
15
4 2 4 2 4
4 2 2 2 4
2 4 4
2 4 2 2
1
x
w w w E w I wEI c A I
x t t kG x t kG t
w EI w I wN p
x AG x KAG x t
+ + + +
+ =
(3.1)
where , A, E, G, and I represent the density, cross section area, modulus of elasticity,
shear modulus, and moment of inertia of the cross section of the beam, respectively, and
c is the damping factor,
2
0
02
L
x
EA wN N dx
L x
= + represents the axial force, 0N is the
external axial force and assumed to be zero in the following analysis, and is the shear
correction factor.
Figure 3.1 A beam under pressure
The nonlinear equation of motion for the transverse displacement w(x,t) of a isotropic
beam without considering the rotary inertia and shear deformation effects can be
expressed as
4 2 2
4 2 2x
w w w wEI c A N p
x t t x
+ + =
(3.2)
7/31/2019 Lu_Y_D_2008
25/139
16
It is noticed that from Eq. (3.1) to Eq. (3.2), the order of the differential equation drops
from four to two in the time domain.
Following the method of separation of variables, the response of the displacement field
can be expressed as
1
( , ) ( ) ( )
N
n n
n
w x t f x q t
=
= (3.3)
where ( )n
f x represents the n-th eigenfunction which is determined by the boundary
condition and ( )nq t represents the time-dependant part of the n-th modal response.
Selected choices of eigenfunctions for beams with various support conditions are listed in
Appendix A.1.
Substituting Eq. (3.3) into Eq. (3.2) and applying the Galerkins method by left-
multiplying both sides of Eq. (3.3) by ( )nf x and integrating over the span of 0 to L, the
following equation for the n-th mode is obtained after the orthogonality condition is
applied
, ,
1 1 1
( ) ( ) ( ) ( ) ( ) ( ) ( )
N N N
n n n n n n k k i j i j n
k i j
q t q t q t q t q t q t p t = = =
+ + + =
(3.4)
where
7/31/2019 Lu_Y_D_2008
26/139
17
2
0
''''
02
0
''
0,
2
0
' '
,0
02
0
( )
( ) ( )
( )
( ) ( )
2 ( )
( ) ( )
1( ) ( , ) ( )
( )
n L
n
L
n n nL
n
L
k n
n k L
n
L
i j i j
L
n nL
n
c
A f x dx
EIf x f x dx
A f x dx
EA f x f x dx
AL f x dx
f x f x dx
p t p x t f x dx
A f x dx
=
=
=
=
=
(3.5)
Eq. (3.4) can not be solved analytically due to the existence of nonlinear terms. When the
load ( )n
p t is random in nature, the property of interest is the root mean square (R.M.S.)
of the response, which is defined by
[ ]/ 2 22
/ 2
1( )lim
T
x xT
T
x t dtT
= (3.6)
where x(t) represents a stationary process that has a constant mean value ofx
, and T
stands for the period that is under consideration. It can be seen that 2x
is also the
variance of the processx(t).
In the equivalent uncoupled linearization method, a linearized equation in the following
form is sought
,( ) ( ) ( ) ( )n n n n n n nq t q t q t p t + + = (3.7)
where ,andn n n represent the damping factor and stiffness of the equivalent system.
7/31/2019 Lu_Y_D_2008
27/139
18
It can be seen that in Eq. (3.6) different modes of the beam are totally decoupled for each
mode. However, recall that in Eq. (3.4) all the modes are actually coupled through the
nonlinear terms. So it makes more sense if the equivalent equation is written in the linear
coupled format as below
,
1
( ) ( ) ( ) ( )
N
n n n n k n n
k
q t q t q t p t =
+ + = (3.8)
It should be noted that by setting the non-diagonal stiffness terms, , withn k n k , to
zero, Eq. (3.8) is the same as the representation of the traditional equivalent linearized
equation as shown in Eq. (3.7).
The difference between Eq.(3.8) and (3.4) is
,
1
, ,
1 1 1
( ) ( ) ( ) ( )
( ) ( ) ( )
N
n k k n n n n n
k
N N N
n k k i j i j
k i j
q t q t q t
q t q t q t
=
= = =
= +
(3.9)
The goal is to find the optimal values of , andn k n of the equivalent linearized system
so that the square of the difference between linear and nonlinear systems is minimalized
in the statistical sense. This requires that
2
2
,
[ ] 0
[ ] 0
n
n k
E
E
=
=
(3.10)
whereE[] stands for the mathematical expectation.
7/31/2019 Lu_Y_D_2008
28/139
19
To demonstrate this procedure in Eq.(3.9) and Eq.(3.10), we look at a simpler casea
beam that is simply supported at both ends, in which Eq.(3.4) is simplified to the
following
2
,1 ,1
1
( ) ( ) ( ( )) ( ) ( )
N
n n n n n k k n n
k
q t q t q t q t p t +=
+ + + = (3.11)
where
2 2 4
,1 4
4 4
,1 2 4
0
4
2( ) ( , )sin( )
n
n
n k
L
n
c
A
k n E
L
n EG
L
n xp t p x t
L L
+
=
=
=
=
(3.12)
The equivalent linear system in Eq.(3.8) is used which is listed below again for the sake
of convenience
,
1
( ) ( ) ( ) ( )
N
n n n n k k n
k
q t q t q t p t =
+ + =
The difference between Eqs.(3.12) and (3.8) is
2
, ,1 ,1
1 1
( ) ( ( )) ( ) ( ) ( )
N N
n k k n n k k n n n n
k k
q t q t q t q t += =
= + + (3.13)
Therefore,
7/31/2019 Lu_Y_D_2008
29/139
20
[ ] [ ] [ ]
2
,
2
, ,1 ,1
1 1
2
, ,1 ,1
1 1
0 [ ]
( ) ( ) ( ( )) ( ) ( )
( ) ( ) ( ) ( ) 3 ( ) ( ) ( )
n k
N N
k n k k n n k k n k
k k
N N
n k n k n n k n k k n k
k k
E
E q t q t E q t q t q t
E q t q t E q t q t E q t E q t q t
+
= =
+
= =
=
= +
= +
(3.14)
and
2
2
0 [ ]
( ) ( )
n
n n n
E
E q t
=
=
(3.15)
which leads to
2
, ,1 ,1
1
3 ( )
N
n k n n k k
k
n n
E q t
+
=
= +
=
(3.16)
In the above derivation, the following relationship and definitions are used under the
assumption that both the load and response follow zero-mean Gaussian distributions
(Soong 2004):
3[ ] [ ] 0
[ ] 0 ( )
n n n n
k n
E q q E q q
E q q k n
= =
=
(3.17)
and4 2 2
3 2
[ ] 3 [ ]
[ ] 3 [ ] [ ]
n n
k n k n k
E q E q
E q q E q q E q
=
=(3.18)
Another example is given in Appendix A.2 for a beam fixed on one end and simply
supported at the other. In that case, a complete set of quadratic terms maintains and
7/31/2019 Lu_Y_D_2008
30/139
7/31/2019 Lu_Y_D_2008
31/139
22
where ( )PS represents the power spectral density of the original load ( , )p x t =
( ) ( )x P t in Eq. (3.1). By definition, the power spectral density is the Fourier transform
of the autocorrelation function ( )R of load ( , )p x t :
( ) ( ) iPS R e d
= (3.22)
In the traditional equivalent linearization method, the correlation between different
modes is not considered due to the fact that the final linearized equations are decoupled.
However, numerical simulations indicate that under certain boundary conditions, there
are strong correlations between the responses of different modes in the nonlinear
problem. The value of correlation factor, ( )n k n k is calculated from the following
relationship (under the assumption that both the load and response have zero mean
Gaussian distribution)
,
2 2 2
[ ]( )
[ ] [ ]
n kn k
n k
E q qn k
E q E q
= (3.23)
Finally, the displacement R.M.S. of nonlinear random vibration of the beam can be
obtained after an iteration scheme that is similar to that of the uncoupled linearization
method:
(1) Taking the linear part of Eq.(3.8) only and calculate the first estimate of 2[ ]nE q and
[ ]n kE q q via Eq.(3.19) to Eq.(3.21) for each of theNmodes.
7/31/2019 Lu_Y_D_2008
32/139
7/31/2019 Lu_Y_D_2008
33/139
24
4 2 4 2 4
4 2 2 2 4
2 4 4
2 4 2 2
1
x
w w w E w I wEI c A I
x t t kG x t kG t
w EI w I wN p
x KAG x KAG x t
+ + + +
+ =
For a beam simply supported at both ends, (0, ) ( , ) 0w t w L t = = . Assume the solution for
the beam is the summation of the firstNmodes
sin( ) ( )
N
n
n
n xw q t
L
= (3.24)
whereNrepresents the total number of modes considered.
Substituting Eq.(3.24) into Eq.(3.1) and applying the Galerkins method, the following
equation for the n-th mode is obtained after some lengthy manipulation
2
,1 ,1 ,2
1
2
,3 ,3
1
( ) ( ) ( ) ( )
( ( )) ( ) ( )
N
n n n k k n n N n
k
N
n N n N k k n n
k
q t q q t q t
q t q t p t
+ +
=
+ + +
=
+ + +
+ + =
(3.25)
where
2 2
,1 2
2 2 4
, 1 4
,2 2
4 4
,3 2 4
2 2 4 2 2 2
,3 2 6
20
( )
4
( )
4
2( ) ( , )sin( )
n
n k
n N
n N
n N k
L
n
AG n E G
I L
k n E
L
c G
I
n EG
L
k n E n EI AGL
I L
G n xp t p x t
L I L
+
+
+
+ +
+= +
=
=
=
+=
=
(3.26)
7/31/2019 Lu_Y_D_2008
34/139
7/31/2019 Lu_Y_D_2008
35/139
26
Because of the coupling terms in Eq.(3.25), not only the 2[ ]nE q ,2[ ]nE q , and [ ]n nE q q for
each mode need to be estimated, but also the cross terms such as [ ] ( )n k
E q q n k . They
are obtained from the frequency domain analysis as explained in the following. Recall
that for a linear system, the mean square response is obtained by Eq.(3.19)
2 *( ) ( ) ( )n n n pG G S d
=
where ( )nG is the frequency response function and )(*
nG is its complex conjugate.
For a linear system governed by Eq.(3.27), )(nG takes the form
4 2
1( )
n e e e
n n n
Gi
= + +
(3.30)
Furthermore, via random vibration theory, 2[ ]nE q and2[ ]nE q can be calculated from the
following formulas
( ) ( )
( ) ( )
2
2 24 2
42
2 24 2
1[ ] ( )
[ ] ( )
n ne e e
n n n
n ne e e
n n n
E q S d
E q S d
= + +
= + +
(3.31)
where ( )nS represents the power spectral density of the excitation ( )np t .
The result for [ ]n n
E q q , on the other hand, can be obtained from the autocorrelation
betweenn
q andn
q :
7/31/2019 Lu_Y_D_2008
36/139
27
, 2 2 2 2
( ) ( ) ( )
( ) ( ) ( ) ( )n n
n n n nq q
n n n n
E q q E q E q
E q E q E q E q
=
(3.32)
Since the correlation between the displacement and acceleration is unknown, we seek
help from simulation results. After generating 10 series of data with length of 214
,
statistical evaluation of the correlation factor was conducted based on Eq.(3.32).
Numerical simulations were run for different beams with different boundary and loading
conditions. It was found out that the value ofn nq q
fell into a consistent range of -0.88 to
-0.80. For the purpose of simplicity, the value of -0.88 was used in the analytical analysis.
Eventually, the value of [ ]n nE q q is calculated from the following relationship (under the
assumption that both the response and associated acceleration have zero mean)
2 2 2
,[ ] [ ] [ ]n nn n q q n nE q q E q E q= (3.33)
-0.8 -0.4 0 0.4 0.8Displacement, m
-0.2
-0.1
0
0.1
0.2
Acceleration,
106 .
7/31/2019 Lu_Y_D_2008
92/139
83
Another important function associated with an ARMA(p, q) process is the power
spectrum density function. This function, given by Eq.(5.13), is useful in representing the
spectral characteristics of the random signal.
2( 1)2
1
2( 1)
1
...( )
2 ...
qi q i
q
pi p i
p
e ef
e e
=
(5.13)
where is the sampling interval.
General speaking, there are infinite choices of (p, q) to fit the collected data by regression
technique. Therefore, based on the nature of the problem, certain criterions must be
satisfied before an ARMA model can be considered qualified. After this screen
process, the optimal fit is usually taken to be the one that is simplest in format, i.e.,
lowest in order. For detailed procedure as well as adequacy check on the fitted models,
one can refer to Pandit and Wu (1993) and Box et. al (1994).
5.3 Applications of ARMA Model in Identifying, Re-generating,
and Extending the Random Vibration Data
In this section, the applications of ARMA models are demonstrated by some numerical
examples. Suppose some random vibration data of limited length is available from either
experiment or simulation. In this case, the displacement of nonlinear random vibration of
a beam with fixed-simply supported boundary condition (refer to Table 3.1) is used as
7/31/2019 Lu_Y_D_2008
93/139
84
input. The beam is subjected to half uniform load. First, different orders of ARMA
models are fitted to the data and the harmonic components within the data are predicted
using Eq. (5.6) to (5.7). Second, once the ARMA(p, q) parameters 1 , , p , 1 ,, q
are obtained, Eq. (5.1) is used to re-generate the displacement response data and this set
of data is compared with the original signal. Third, sincet is the pure noise that follows
normal distribution, it can be generated via random normal number generator once the
variance is found in the previous steps. Substituting thus-obtainedt
along with the
parameters1 , , p , 1 ,, and q into Eq. (5.1), an analytical express is found that
can be used to generate time series of any desired length depending on the special needs
of the situation.
Selected ARMA(m, n) models are fitted to represent the simulated displacement response
of mode 1 of the beam. The scheme by Pandit and Wu (1993) is used to estimate the
parameters of selected ARMA(p , q) models. The results are listed in Table 5.1 and 5.2.
Figure 5.1 shows the corresponding PSD of the response of the simulation data. It is
observed in Figure 5.1 that the response has a dominant harmonic component at about
900 Hz.
Theoretically, the response of a second-order system could be fitted with a second order
ARMA(2, 1) or even AR(2) model. However, since the sampling interval is so small, the
correlations between adjacent data points must be taken into consideration, which may
require a higher order ARMA model. Another possibility is that there are additional
modes that are introduced into the system due to the stochastic nature of the excitation.
7/31/2019 Lu_Y_D_2008
94/139
85
Table 5.1 Estimated ARMA parameters for different order models
ARMA
parameters
ARMA (p, q) model
(2,1) (3,2) (4,3)
1 1.608 1.465 1.241
2 -1.057 -0.946 -1.160
3 -0.0211 0.668
4 -0.692
1 -0.269 -0.527 -0.591
2 -0.254 -0.680
3 -0.310
Table 5.2 Frequencies predicted by selected ARMA models in Table 5.1
ARMA (p, q) model Predicted frequency, Hz
(2,1) 857
(3,2) 917
(4,3) 875
0 400 800 1200 1600 2000Frequency, Hz
0
10
20
30
40
mm
2/Hz
Figure 5.1PSD of the mode 1 of the beam displacement simulation data
The data in Table 5.2 indicates that the selected ARMA models in Table 5.1 all give very
good predictions of the dominant harmonic frequency component embedded in the signal.
7/31/2019 Lu_Y_D_2008
95/139
86
After the ARMA(p, q) parameters 1 , , p , 1 ,, q are obtained as shown in Table
5.1 and the variance of t is found, Eq. (5.1) is used to re-generate a time series based on
the ARMA model. The re-generated data is compared with the original simulation results.
Although ARMA(2,1) model gives a pretty good prediction of the frequency component
embedded in the signal, a look at the re-generated displacement field by the model
reveals that the simulated displacement quickly blows out of proportion, thus suggesting
an instability in the system. This is part of the problem associated with ARMA
applications. The condition of stability (Box et. al 1994) for a certain ARMA model
requires that the norm of the each of the complex root determined by Eq.(5.6) is less than
one. For the ARMA(2,1) parameters listed in Table 5.1, the norm of the two complex
roots is 1.028 which is just over one. That causes the instability of the re-generated data.
Hence, a higher order model is needed in order to produce a stable time series.
0 1 2 3 4Time, sec
-0.15
0
0.15
Displacement,m
Figure 5.2Displacement from simulation of mode 2 of a F-SS beam
7/31/2019 Lu_Y_D_2008
96/139
87
0 1 2 3 4Time, sec
-0.15
0
0.15
Displacement,m
Figure 5.3Displacement generated by ARMA(4, 3) model
0 1 2 3 4Time, sec
-0.15
0
0.15
Displac
ement,m
Figure 5.4Displacement generated by ARMA(8, 7) model
0 1 2 3 4Time, sec
-0.15
0
0.15
Displace
ment,m
Figure 5.5Displacement generated by ARMA(11, 10) model
Table 5.3 Estimated displacement R.M.S. of selected ARMA models
Displacement R.M.S., mm
simulation ARMA(4,3) ARMA(8,7) ARMA(11,10)
32.1 30.1 34.4 32.3
7/31/2019 Lu_Y_D_2008
97/139
88
Further numerical examples are shown in Figures 5.2 to 5.5 for the mode 2 displacement
of the same beam. The original simulation result is shown in Figure 5.2. In this case
ARMA(3,2) model turns out to be unstable, so the lowest model used is ARMA(4, 3).
During the process, the estimated variance of the pure white noise part of each model is
used to produce random displacement data. Thats why the generated displacement looks
different each case. It is also seen that the generated displacement looks quite different
from the original data. In the original data, it is obvious that there is a significant
harmonic component. While in Figures 5.2 to 5.5, that harmonic component is shadowed
by the random noise. However, the R.M.S. of all the displacement data are very close, as
is shown in Table 5.3. So from the statistical point of view, they can be considered
equivalent signals. This is a very useful application of ARMA model because it can
extend data serial to unlimited length in the time domain while preserving the spectral
information of the original data. It is also observed that as the order of ARMA model
increases, the R.M.S. of the predicted displacement gets closer to that of the original
signal.
5.4 Comparison between PSD Curve from ARMA Model and
Newlands Approach
There are some long-established methods that estimate the power spectral densities of
digital signals. One of the most widely used techniques was summarized by Newland
(1993). The detailed procedures are not repeated here. Although the guideline is the
7/31/2019 Lu_Y_D_2008
98/139
89
same, there are variations in the executions of the technique. In this section, different
trials are conducted in order to get a smoothed PSD curve.
(1) Divide the original data set (usually 218) into segments in the size of 213 to 215, dothe spectral estimation for each segment following Newlands approach (with
smoothing), then average across all the segments.
(2) Same as (1), except that no smoothing is applied during the spectral estimationfor each segment .
(3)
Same as (1), except that during the spectral estimation for each segment,
triangular window was used instead of the default rectangular window.
(4) Do the FFT with each segment, average across all the segments, then do thespectral estimation based on the averaged FFT data (now having the size of one
segment).
It is found out that if there is enough data in each segment (say 213 or more) and enough
segments (say, 32 or more), all the methods yielded almost identical curves. This result
suggests that one can skip the smoothing process and just use the approach in (4) since
this is most efficient one computationally.
As a quick example, the above spectrum estimation technique is applied to the simulated
transverse response data of a 3m by 2m rectangular composite plate with a thickness of
0.06 m. The plate is subjected to a white noise excitation and has a first natural frequency
of 70.2 Hz. The four-layer laminate ( 45 / 45 / 45 / 45 ) has material properties of
7/31/2019 Lu_Y_D_2008
99/139
90
3
1 2 3 1 12 13 12206.84 GPa, 0.1 , 10.34 GPa, 0.25, and 2070 kg/m .E E E E G G = = = = = = =
The size of the simulation data is 218
which is divided into 32 sections. In Figure 5.6
selected 8 out of the 32 segments are displayed. The averaged PSD curve via Newlands
approach is plotted in Figure 5.7.
The PSD plot given by Eq. (5.13) based on an ARMA(4,3) model is plotted in Figure 5.7
along with the estimated PSD curve from simulation data. Although the curves from the
two approaches dont overlap, especially the height of the peaks, the areas under the
curves that represent the variance of the data are very close to each other, which is what
is expected. Generally Newlands approach gives a very conservative estimation for the
PSD curve. It tends to underestimate the peak values.
7/31/2019 Lu_Y_D_2008
100/139
91
0 40 80 120 160Freqency, Hz
0
200
400
100
300
m
m2/Hz
0 40 80 120 160Freqency, Hz
0
200
400
100
300
m
m2/Hz
0 40 80 120 160Freqency, Hz
0
200
400
100
300
m
m2/Hz
0 40 80 120 160Freqency, Hz
0
200
400
100
300
m
m2/Hz
0 40 80 120 160Freqency, Hz
0
200
400
100
300
m
m2/Hz
0 40 80 120 160Freqency, Hz
0
200
400
100
300
m
m2/Hz
0 40 80 120 160Freqency, Hz
0
200
400
100
300
mm
2/Hz
0 40 80 120 160Freqency, Hz
0
200
400
100
300
mm
2/Hz
R.M.S. is 22.30 mmR.M.S. is 21.05 mm
R.M.S. is 20.48 mmR.M.S. is 21.02 mm
R.M.S. is 21.36 mmR.M.S. is 21.30 mm
R.M.S. is 20.83 mmR.M.S. is 21.67 mm
#9 #10
#11 #12
#14 #15
#16 #17
Figure 5.6 Example spectrum plots of 8 out of the 32 segments
(segments #9-#17, each of which is smoothed)
7/31/2019 Lu_Y_D_2008
101/139
92
0 40 80 120 160 200Frequency, Hz
0
50
100
150
200
mm
2/Hz
ARMA(4,3) model
predictionNewland method
Figure 5.7 Comparison of PSD curves from ARMA(4,3) model
and Newlands approach
Now attention is turned to cases when multiple harmonic components exist in the data.
Newlands approach would always work no matter what the data is like. ARMA models
ability to capture those multiple frequency components is demonstrated by the following
example. The simulation data as shown in Figure 5.3 of the mode 2 of a fixed-simply
supported beam is used as the original signal. Figure 5.8 shows the estimated PSD of the
response via Newlands approach. In the previous section, ARMA(4,3) model has been
proven to give an accurate reproduction of the time series. However, the PSD curve of
the ARMA(4,3) model does not represent the spectral characteristic of the original data
well, as is shown in Figure 5.9. First of all, it is not able to pick out the frequency at about
900 Hz. Second, the two frequencies it predicts are 1732 Hz and 2330 Hz, respectively.
Both of them miss the dominant component at around 1800 Hz. This indicates that a
higher order ARMA model is needed if one wants to accurately capture the spectral
7/31/2019 Lu_Y_D_2008
102/139
93
properties of the signal. So in the next steps, ARMA(5,4), ARMA(6,5), ,
ARMA(11,10) models are tested. The details for the intermediate models are not shown
here. When ARMA(11,10) model is selected, it successfully picks out five harmonic
components: 918 Hz, 1733 Hz, 2023 Hz, 2626 Hz, and 3482 Hz. The PSD plot associated
with the ARMA(11,10) model is shown in Figure 5.11. By comparing the curves in
Figures 5.9 and 5.11, it can be found that the PSD curve via ARMA(11,10) model is
almost a perfect match to the one by Newlands approach. Not only does it capture all the
spectral characteristics, but gives a smooth curve with an analytical expression as
compared to the sketchy one estimated from the Newlands approach. Therefore, it can
be concluded that once the order of the ARMA model is high enough, it will accurately
predict the PSD of the signal.
0 1000 2000 3000 4000Frequency, Hz
0
1
2
3
4
mm
2/H
z
Figure 5.8 PSD of the beam simulation displacement data (Newlands approach)
7/31/2019 Lu_Y_D_2008
103/139
94
0 1000 2000 3000 4000Frequency, Hz
0
1
2
3
mm
2/Hz
Figure 5.9 PSD from ARMA(4,3) model
0 1000 2000 3000 4000Frequency, Hz
0
1
2
3
mm
2/Hz
Figure 5.10 PSD from ARMA(8,7) model
0 1000 2000 3000 4000Frequency, Hz
0
1
2
3
mm
2/Hz
Figure 5.11 PSD from ARMA(11,10) model
7/31/2019 Lu_Y_D_2008
104/139
95
In summary, the ability of ARMA models in identifying, re-generating, and extending the
random vibration data is demonstrated and verified in this chapter. If one is only
interested in regenerating the displacement response and using it as a random input for
situations such as structural fatigue analysis, relatively lower ARMA models are efficient
enough to get the job done. However, higher order models are required to accurately
capture the spectral characteristics of the random signal. Overall, ARMA models have
great potentials in the analysis and application of random vibration data.
7/31/2019 Lu_Y_D_2008
105/139
96
Chapter 6. Future Work
6.1 Durability of Structures Subjected to Random Loading
So far in this study the dynamic response of beams and plates under stochastic loading
conditions has been examined. One significant type of reliability concern associated with
oscillating loads of random nature is fatigue. As the magnitude and the direction of the
load vary incessantly with time, the material undergoes fluctuating strains that may
eventually cause the structure to fail at a stress well below the yield stress of the
constituent material. The most straightforward way to evaluate fatigue is the S-N curve.
The S-N curve is obtained through experiments in which a material under study is
subjected to a cyclic stress and the number of cycles to failure is counted. Numerous tests
are run at different stress levels and a smooth curve is drawn with the help of some
knowledge of data processing or linear regression technique. Traditional theories in
fatigue include Miners Rule and Paris equation. Miner (1945) proposed that failure
occurs if the following relationship for a material subjected to kdifferent stress levels is
satisfied
7/31/2019 Lu_Y_D_2008
106/139
97
1
k
i
ii
nC
N=
= (6.1)
where C is a constant determined experimentally and ranges from 0.7 to 2.2, andi
n
represents the number of cycles under the i-th stress level andi
N is the total number of
cycles before failure.
The expected fatigue damage after duration of time Tunder random loading is found to
be (Liou et al. 1999)
[ ] 00
( )( )
( )
d
f m
pE D T v T d
N
+ =
(6.2)
where [ ]E means the mathematical expectation, d is Morrows plastic work interaction
exponent, 0v+ represents the number of stress cycle per unit time for a narrow-band
process, ( )p is the PDF (probability density function) of stress amplitudes, m is the
maximum stress magnitude, and ( )f
N is the number of stress peaks to cause failure if a
constant amplitude is applied and can be obtained from the Basquins equation:
'pfN C = (6.3)
wherep and 'C are material-dependant constants.
In Nigam and Narayanan (1994), the estimated fatigue damage after time T under
stationary narrow-band random vibration is expressed by
(0)[ ( )] ( 2 ) 1
2
bb
q
x
N T qE D T
c
+ = +
(6.4)
7/31/2019 Lu_Y_D_2008
107/139
7/31/2019 Lu_Y_D_2008
108/139
99
2 2 2 2
,
( , ) ( ) ( )mn mnw mn Q Q
m n
W x y S H d
= (6.8)
where (m, n) represents mode number.
Using Eqs. (6.6)-(6.8), the reliability vs. time curve can be obtained for a beam or plate
subjected to a known random excitation. Due to the fact that this is a straightforward
procedure and no material fatigue data is available, no numerical example is given while
the outline of the solution process is summarized here.
6.2 Future Work
The fatigue life analysis mentioned in section 6.1 is one area future work can be devoted.
So far this study has focused on the stationary random process whose statistical
properties do not vary with time. In reality many random excitations are non-stationary
such as earthquake wave. It would be interesting to examine the dynamic response of
structure as time evolves subjected to this type of loading condition. Different approaches
to model the non-stationary process can be found in work of Cederbaum et. al (1992).
On the other hand, although a new linearly coupled linearization method is proposed in
addition to the traditional equivalent linearization to obtain the numerical results in this
study, the limitation of this type of technique is that the load and system response must be
Gaussian. It has been discussed that technically when governing equations of motion of
7/31/2019 Lu_Y_D_2008
109/139
100
the system are nonlinear, the responses can be non-Gaussian even if the loading is.
Hence, in order to pursue more accurate results for multi-degree systems, other
linearization techniques need to be explored for the higher-order beams or plates such as
the Gaussian closure technique, statistical linearization, and statistical non-linearization.
These are all potential fields where future can be conducted.
7/31/2019 Lu_Y_D_2008
110/139
101
Reference
Abdelnaser, A.S. and Singh, M.P. (1993) Random response of antisymmetric angle-ply
composite plates with Levy boundary-conditions, Composite Engineering, 3(9),
817833
Bao, G. and Wang, L. (1995) Multiple cracking in functionally graded ceramic/metal
coatings,International Journal of Solids and Structures, 32, 28532871
Birman, V (1995) Buckling of Functionally Graded Hybrid Composite Plates,
Proceeding of the 10th Conference on Engineering Mechanics, vol. 2, 11992202
Bolotin, V.V. (1984) Random vibrations of elastic systems. The Hague, The Netherlands:
Martinus Nijhoff
Boukhrist, A., Mourot, G. and Ragot, J. (1999) Non-linear dynamic system
identification: a multi-model approach, International Journal of Control, 72(7),
591604
Box, G.E.P., Jenkins, G.M. and Reinsel, G.C. (1994) Time Series Analysis: Forecasting
and Control, 3rd edition, Prentice-Hall, Inc.
Cai, H. and Bao, G. (1998) Crack bridging in functionally graded coatings.International
Journal of Solids and Structures, 35, 701717
Carden, E.P. and Brownjohn, J.M.W. (2008) ARMA modelled time-series classification
for structural health monitoring of civil infrastructure, Mechanical Systems and
Signal Processing, 22, 295314
Caughey, T.K. (1959a) Response of a non-linear string to random loading, Journal of
applied Mechanics, ASME, 26,341-344
Caughey, T.K. (1959b) Response of van der Pol oscillator to random excitation,Journal
of applied Mechanics, ASME, 26,345-348
7/31/2019 Lu_Y_D_2008
111/139
102
Caughey, T. K., Equivalent Linearization Techniques (1963)Journal of Acoustic Society
of America, 35, 1706-1711
Cederbaum, G., Librescu, L. and Elishakoff, I. (1988) Random vibration of laminated
plates modeled within a high-order shear deformation theory, Journal of the
Acoustical Society of America, 84(2), 660-666
Cederbaum, G., Librescu, L. and Elishakoff, I. (1989) Remarks on a dynamic higher-
order theory of laminated plates and its application in random vibration,
International Journal of Solids and Structures, 25(5), 515526
Chatfield, C. (1989) The Analysis of Time Series: An Introduction, Fourth Edition,
Chapman & Hall
Chen, C.S. (2005) Nonlinear vibration of a shear deformable functionally graded plate,
Composite Structures, 68, 295-302
Chen, Q. and Tomlison G.R. (1996) Parametric identification of systems with dry friction
and nonlinear stiffness using a time series model, Journal of Vibration and
Acoustics,118 (2), 252263
Chen, W.Q. and Ding, H.J. (2002) On free vibration of a functionally graded
piezoelectric rectangular plate,Acta Mechanica, 153 (34), 207216
Cheng, Z.Q. and Batra, R.C. (2000a) Deflection relationships between the homogeneous
Kirchhoff plate theory and different functionally graded plate theories, Acta
Mechanica, 52, 143158
Cheng, Z.Q. and Batra, R.C. (2000b) Exact correspondence between eigenvalues of
membranes and functionally graded simply supported polygonal plates, Journal of
Sound and Vibration, 229, 879895
Cheng, Z.Q. and Batra, R.C. (2000c) Three-dimensional thermoelastic deformations of a
functionally graded elliptic plate, Composites: Part B, 31, 97106
Chi, S.H., Chung, Y.L., 2002. Cracking in sigmoid functionally graded coating, Journal
of Mechanics, 18, 4153
Chi, S.H., Chung, Y.L., (2003) Cracking in coating-substrate composites of multi-layered
and sigmoid FGM coatings,Engineering Fracture Mechanics, 70, 12271243
Chi, S.H., Chung, Y.L., (2006a) Mechanical behavior of functionally graded material
7/31/2019 Lu_Y_D_2008
112/139
103
plates under transverse loadPart I: Analysis, International Journal of Solids and
Structures, 43(13), Pages 3657-3674
Chi, S.H., Chung, Y.L., (2006b) Mechanical behavior of functionally graded material
plates under transverse loadPart II: Numerical results, International Journal of
Solids and Structures, 43(13), Pages 3675-3691
Cho, J.R. and Oden, J.T. (2000) Functionally graded material: a parametric study on
thermal-stress characteristics using theCrankNicolsonGalerkin scheme,
Computer Methods in Applied Mechanics and Engineering, 188, 1738
Chung, Y.L., Chi, S.H. (2001) The residual stress of functionally graded materials,
Journal of the Chinese Institute of Civil and Hydraulic Engineering, 13, 19
Chonan, S. (1995) Random vibration of a prestressed, orthotropic, thick rectangular plate
on a generalized flexible foundation,Journal of the Acoustical Society of America,
78(2), 598604
Crandall, S.H. (1963) Perturbation techniques for random vibration of nonlinear systems,
Journal of Acoustical Society of America, 35, 1700-1705
Crandall, S.H., and Mark, W.D. (1973) Random vibration in mechanical systems, New
York: Academic Press
David, J.Y., Jones, S.A., and Giddens, D.P. (1991) Modern spectral analysis techniques
for blood flow velocity and spectral measurements with pulsed Doppler ultrasound,
IEEE Transactions on Biomedical Engineering, 38(6), 589596
Dahlberg, T. (1999) The effect of modal coupling in random vibration analysis, Journal
of Sound and Vibration, 228(1), 157-176
Delale, F., and Erdogan, F. (1983) The crack problem for a nonhomogeneous plane,
ASME Journal of Applied Mechanics, 50, 609614
Drake, J.T., Williamson, R.L., Rabin, B.H., (1993) Finite element analysis of thermal
residual stresses at graded ceramicmetal interfaces, Part II: interface optimization
for residual stress reduction,Journal of Applied Physics, 74, 13211326
Elishakoff, P. (1983) Probabilistic methods in the Theory of Structures, Wiley-
Interscience, New York
7/31/2019 Lu_Y_D_2008
113/139
104
Elishakoff, I., Fang, J. and Caimi, R. (1995) Random vibration of a nonlinearly deformed
beam by a new stochastic linearization technique, International Journal of Solids
and Structures, 32(11), 15711584
Elishakoff, I. (1991). Method of stochastic linearization revised and improved, In
Computational Stochastic Mechanics (Edited by P. D. Spanos and C. A. Brebbia),
pp. 101-I 11. Computational Mechanics Publications and Elsevier Applied Science,
London.
Elishakoff, I. and Livshits, D. (1984). Some closed-form solutions in random vibration of
Bernoulli-Euler beams, International Journal of Engineering Science, 22, 1291-
1302
Elishakoff, I. and Zhang, R. C. (1991) Comparison of the new energy-based versions of
the stochastic linearization technique, In Nonlinear Stochastic Mechanics (Edited
by N. Bellomo and F. Casciati), pp. 201-212. Springer, Berlin.
Elmajid, A., Taya, M. and Hudnut, S. (2001) Analysis of out-of-plane displacement and
stress field in a piezocomposite plate with functionally graded microstructure,
International Journal of Solids and Structures, 38(19), 33773391
Er, G.K. (1998) Multi-Gaussian closure method for randomly excited nonlinear systems,
International Journal of Non-Linear Mechanics, 33(2), 201-214
Erdogan, F., Chen, Y.F. (1998) Interfacial cracking of FGM/metal bonds. In: Kokini, K.
(Ed.), Ceramic Coating, pp. 2937
Erdogan, F., Wu, B.H. (1996) Crack problems in FGM layers under thermal stresses,
Journal of Thermal Stresses, 19, 237265
Eringen, A. C. (1957). Response of beams and plates to random loads, Journal of Applied
Mechanics, 24, 46-52
Fang, J., Elishakoff, I. and Caimi, R. (1995) Nonlinear response of a beam under
stationary random excitation by improved stochastic linearization method, Applied
Mathematical Modelling, 19(2), 106-111
Feldman, E., Aboudi, J. (1997) Buckling analysis of functionally graded plates subjected
to uniaxial loading, Composite Structures, 38, 2936
Foda, M.A., 1999. Influence of shear deformation and rotary inertia on nonlinear free
vibration of a beam with pinned ends, Computers and Structures, 71, 663670
7/31/2019 Lu_Y_D_2008
114/139
105
Ganapathi, M. and Prakash, T. (2006) Thermal buckling of simply supported functionally
graded skew plates, Composite Structures, 74(2), 247250
Gardiner, C.W. (2004) Handbook of Stochastic Methods: for Physics, Chemistry and the
Natural Sciences (Springer Series in Synergetics), 3rd
edition, Springer
Gautier, P.E., Gontier, C., and Smail, M. (1995) Robustness of an ARMA identification
method for modal analysis of mechanical systems in the presence of noise, Journal
of Sound and Vibration, 179(2), 227-242
Gray, C.E., Decha-Umphait K. and Mei, C. (1985) Large Deflection, Large Amplitude
Vibrations and Random Response of Symmetrically Laminated Plates, Journal of
Aircraft, 22(11), 929-930
Gu, P., Asaro, R.J., 1997. Crack deflection in functionally graded materials, International
Journal of Solids and Structures, 34, 30853098
Guler, N.F., Kiymik, M.K., and Guler, I. (1995) Comparison of FFT and AR-based
sonogram outputs of 20 MHz pulsed Doppler data in real time, Computers in
Biology and Medicine, 25(4), 383391
Guler, I., Kara, S., Guler, N.F., and Kiymik, M.K. (1996) Application of autoregressive
and fast Fourier transform spectral analysis to tricuspid and mitral valve stenosis,
Computer Methods and Programs in Biomedicine, 49, 2936
Harichandran, R.S. and Naja, M.K. (1997) Random vibration of laminated composite
plates with material non-linearity, International Journal of Non-Linear Mechanics,
32(4), 707720
He, X.Q., Ng, T.Y., Sivashanker, S., Liew, K.M., 2001. Active control of FGM plates
with integrated piezoelectric sensors and actuators, International Journal of Solids
and Structures, 38, 16411655
He, X.Q., Liew, K.M., Ng, T.Y. and Sivashanker, S. (2002) A FEM model for the active
control of curved FGM shells using piezoelectric sensor/actuator layers,
International Journal for Numerical Methods in Engineering, 54(6), 853870
Herbert, R. E. (1964) Random vibrations of a nonlinear elastic beam, Journal of
Acoustical Society of America, 36, 2090-2094
Herbert, R. E. (1965) On the stresses in a nonlinear beam subject to random excitation,
International Journal of Solids and Structures, 1, 235-242
7/31/2019 Lu_Y_D_2008
115/139
106
Huang, X.L. and Shen, H.S. (2004) Nonlinear vibration and dynamic response of
functionally graded plates in thermal environments, International Journal of Solids
and Structures, 41, 24032427
Ibrahim, R.A. (1987) Structural dynamics with parameter uncertainties, Applied
Mechanics Review, 40(3) (1987) 309328
Ichikawa, K. (2000) Functional Graded Materials in the 21st Century, Kluwer Academic
Publishers, Norwell, MA
Iyengar, R.N. and Dash, P.K. (1978) Study of the random vibration of non-linear systems
by the Gaussian closure technique,Journal of Applied Mechanics, ASME, 45, 393-
399
Jaksic, N. and Boltezar, M. (2002) An approach to parameter identification for a single-
degree-of-freedom dynamical system based on short free acceleration response,
Journal of Sound and Vibration, 250(3), 465483
Javaheri, R.and Eslami, M. R. (2002) Buckling of Functionally Graded Plates under In-
plane Compressive Loading,ZAMM, 82(4), 277283
Javaheri, R. and Eslami, M.R. (2002b) Thermal buckling of functionally graded plates.
AIAA Journal, 40(1), 162169
Javaheri, R. and Eslami, M.R. (2002c) Thermal buckling of functionally graded plates
based on higher order theory,Journal of Thermal Stresses, 25, 603625
Jin, Z.H., Batra, R.C., 1996. Stresses intensity relaxation at the tip of an edge crack in a
functionally graded material subjected to a thermal shock, Journal of Thermal
Stresses, 19, 317339
Jin, Z.H., Noda, N., 1994. Crack tip singular fields in nonhomogeneous materials,ASME
Journal of Applied Mechanics, 61, 738740
Jin, Z.H., Paulino, G.H., 2001. Transient thermal stress analysis of an edge crack in a
functionally graded material,International Journal of Fracture, 107, 7398
Kaluzynski, K. (1987) Analysis of application possibilities of autoregressive modeling to
Doppler blood flow signal spectral analysis, Medical and Biological Engineering
and Computing, 25, 373376
7/31/2019 Lu_Y_D_2008
116/139
107
Kang, J., and Harichandran, R.S. (1999) Random vibration of laminated FRP plates with
material nonlinerity using high-order shear theory, Journal of Engineering
Mechanics, 9, 1081-1088
Kaysser W.A., and Ilschner, B. (1995) FGM research activities in Europe, MRS Bulletin,
20(1), 2226
Kim, K.S., Noda, N. (2002) A Greens function approach to the deflection of a FGM
plate under transient thermal loading,Archive of Applied Mechanics, 72, 127137
Kitipornchai, S., Yang, J., and Liew, K.M. (2004) Semi-analytical solution for nonlinear
vibration of laminated FGM plates with geometric imperfections, International
Journal of Solids and Structures, 41, 22352257
Kitipornchai, S., Yang, J. and Liew, K.M. (2006) Random vibration of the functionally
graded laminates in thermal environments, Computer Methods in Applied
Mechanics and Engineering, 195 (2006) 10751095
Koizumi, M. (1997) FGM activities in Japan, Composites-Part B, 28(1/2), 14
Koizumi M, and Niino, M (1995) Overview of FGM research in Japan, MRS Bulletin,
20(1),1921
Koizumi M. (1993) The concept of FGM, Ceramic Transactions, Functionally Gradient
Materials, 34, 310
Lanhe, W. (2004) Thermal buckling of a simply-supported moderately thick rectangular
FGM plate, Composite Structures, 64, 211218
Lee, Y.D. and Erdogan, F. (1995) Residual/thermal stress in FGM and laminated thermal
barrier coatings,International Journal of Fracture, 69, 145165
Liew, K.M., He, X.Q., Ng, T.Y., and Sivashanker, S. (2001) Active control of FGM
plates subjected to a temperature gradient: modeling via finite element method
based on FSDT, International Journal for Numerical Methods in Engineering, 52,
12531271
Liew, K.M., He, X.Q., Ng, T.Y. and Kitipornchai, S. (2002) Active control of FGM
shells subjected to a temperature gradient via piezoelectric sensor/actuator patches,
International Journal for Numerical Methods in Engineering, 55(6), 653668
7/31/2019 Lu_Y_D_2008
117/139
108
Liew, K.M., Lim, H.K., Tan, M.J. and He, X.Q. (2002) Analysis of laminated composite
beams and plates with piezoelectric patches using the element-free Galerkin
method, Computational Mechanics, 29(6), 486497
Liew, K.M., He, X.Q. and Kitipornchai, S. (2004) Finite element method for the feedback
control of FGM shells in the frequency domain via piezoelectric sensors and
actuators, Computer Methods in Applied Mechanics and Engineering, 193(35),
257273
Liew, K.M., Yang, J., and Kitipornchari, S. (2003) Postbuckling of piezoelectric FGM
plates subject to thermo-electromechanical loading, International Journal of Solids
and Structures, 40, 38693892
Liew, K.M., Yang, J. and Kitipornchai, S. (2004) Thermal post-buckling of laminated
plates comprising functionally graded materials with temperature-dependent
properties,Journal of Applied Mechanics, 71(6), 839850
Liew, K.M, He, X.Q. and Kitipornchai, S. (2004) Finite element method for the feedback
control of FGM shells in the frequency domain via piezoelectric sensors and
actuators, Computer Methods in Applied Mechanics and Engineering, 193(35),
257273
Liew, K.M. Yang J. and Wu, Y.F. (2006) Nonlinear vibration of a coating-FGM-
substrate cylindrical panel subjected to a temperature gradient,Methods in Applied
Mechanics and Engineering, 195(9-12), 1007-1026
Lim, C.W. and He, L.H. (2001) Exact solution of a compositionally graded piezoelectric
layer under uniform stretch, bending and twisting, International Journal of
Mechanical Science, 43(11), 24792492
Lin, Y.K. (1973) Probabilistic theory of structural dynamics, New York: McGraw-Hill
Lin, Y.K. (1976) Probabilistic theory of structural dynamics, Malabar, FL: R.E. Krieger;
Loy, C.T., Lam, K.Y. and Reddy, J.N. (1999) Vibration of functionally graded cylindrical
shells,International Journal of Mechanical Science, 41, 309324
Manohar, C.S. and Ibrahim, R.A. (1999) Progress in structural dynamics with stochastic
parameter variations: 19871998,Applied Mechanics Review, 52(5), 177196
7/31/2019 Lu_Y_D_2008
118/139
109
Mattson, S.G. and Pandit, S.M. (2006) Statistical moments of autoregressive model
residuals for damage localization, Mechanical Systems and Signal Processing,
20(3), 627645
Mills, T.C. (1990) Time Series Techniques for Economists, Cambridge University Press
Miner, M.A. (1945) Cumulative damage in fatigue, Journal ofApplied Mechanics, 12,
159164
Mobarakeh, A.A., Rofooei, F.R., and Ahmadi, G. (2002) Simulation of earthquake
records using time-varying ARMA(2,1) model, Probabilistic Engineering
Mecahnics, 17, 15-34
Najafizadeh, M.M. and Eslami, M.R. (2002a) Buckling analysis of circular plates of
functionally graded materials under uniform radial compression, International
Journal of Mechanical Science, 44, 24792493
Najafizadeh M.M. and Eslami, M.R. (2002b) First-order-theory-based thermoelastic
stability of functionally graded material circular plates, AIAA Journal, 40(7), 1444
1450
Najafizadeh, M.M., Eslami, M.R., 2002b. Thermoelastic stability of circular plates
composed of functionally graded materials under uniform radial compression.
International Journal of Mechanical Science, 44, 24792493
Na, K., and Kim, J. (2004) Three dimensional thermal buckling analysis of functionally
graded materials, Composites, Part B: Engineering, 35 (5), 429437
Nayfeh, A. (1993) Introduction to Perturbation Techniques, Wiley-Interscience
Nayfeh, A. and Mook, D. (1995) Nonlinear Oscillations, Wiley-Interscience
Newland, D.E. (1993) An Introduction to random vibration, spectral and wavelet
analysis, London: Longman, third edition.
Ng, T.Y., Lam, Y.K., Liew, K.M., and Reddy, J.N. (2001). Dynamic stability analysis of
functionally graded cylindrical shell under periodic axial loading, International
Journal of Solids and Structures, 38, 12951309
Ng, T.Y., He, X.Q., and Liew, K.M. (2002) Finite element modeling of active control of
functionally graded shells in frequency domain via piezoelectric sensors and
actuators, Computational Mechanics, 28(1), 19
7/31/2019 Lu_Y_D_2008
119/139
110
Nigam, N.C. (1983) Introduction to Random vibrations, Cambridge, Massachusetts: MIT
Press
Ootao, Y. and Tanigawa, Y. (2000) Three-dimensional transient piezothermoelasticity in
functionally graded rectangular plate bonded to a piezoelectric plate, International
Journal of Solids and Structures, 37(32), 43774401
Pandit, S.M. and Wu, S. (1993) Time series and system analysis with applications,
Malabar, Fla.: Krieger Pub. Co.
Park, J.S. and Kim, J.H. (2005) Thermal postbuckling and vibration analyses of
functionally graded plates,Journal of Sound and Vibration, 289(1-2), 77-93
Patel, B.P., Gupta, S.S., Loknath, M.S. and Kadu, C.P. (2005) Free vibration analysis of
functionally graded elliptical cylindrical shells using higher-order theory,
Composite Structures, 69(3), 259270
Popescu, T.D., and Demetriu, S. (1990) Analysis and Simulation of Strong Earthquake
Ground Motions Using ARMA Models,Automatica, 26(4), 721-737
Pradhan, S.C., Loy, C.T., Lam, K.Y. and Reddy, J.N. (2000) Vibration characteristics of
functionally graded cylindrical shells under various boundary conditions, Applied
Acoustics, 61, 111129
Praveen, G.N. and Reddy, J.N. (1998) Nonlinear transient thermoelastic analysis of
functionally graded ceramicmetal plates, International Journal of Solids and
Structures, 35, 44574476
Reddy, J.N. (1997) Mechanics of laminated composite plates: theory and analysis, Boca
Raton: CRC Press
Reddy, J.N. and Chin, C.D. (1998) Thermomechanical analysis of functionally graded
cylinders and plates,Journal of Thermal Stresses, 21, 593626
Reddy, J.N. (2000) Analysis of functionally graded plates, International Journal for
Numerical Methods in Engineering, 47, 663684
Reddy, J.N. and Cheng, Z.Q. (2001) Three-dimensional solutions of smart functionally
graded plates,Journal of Applied Mechanics, 68(2), 234241
Reddy, J.N. and Cheng, Z.Q. (2002) Frequency correspondence between membranes and
functionally graded spherical shallow shells of polygonal platform, International
Journal of Mechanical Sciences, 44(5), 967985
7/31/2019 Lu_Y_D_2008
120/139
111
Reddy, J.N. (2004) Mechanics of laminated composite plates and shells: theory and
analysis, second edition, Boca Raton: CRC Press
Reiter, T. and Dvorak, G.J. and Tvergaard, V. (1997) Micromechanical models for
graded composite materials, Journal of the Mechanics and Physics of Solids, 45,
12811302
Rice, H.J. (1995) Identification of weakly non-linear systems using equivalent
linearization,Journal of Sound and Vibration, 185 (3), 473481
Risken, H. (1996) The Fokker-Planck Equation: Methods of Solution and Applications,
Springer
Roberts, J.B. and Spanos, P.D. (1990) Random Vibration and Statistical Linearization.
Wiley, New York
Salim, S., Iyengar, N.G.R. and Yadav, D. (1998) Natural frequency characteristics of
composite plates with random properties, Structural Engineering & Mechanics,
6(6), 659671
Shariat, B.A.S., Javaherib, R. and Eslamia, M.R. (2005) Buckling of imperfect
functionally graded plates under in-plane compressive loading, Thin-Walled
Structures, 43, 10201036
Shariat, B.A.S., and Eslami, M.R. (2006