International Journal of Mechanical Sciences 57 (2012) 19–33

Contents lists available at SciVerse ScienceDirect

International Journal of Mechanical Sciences

journal homepage: www.elsevier.com/locate/ijmecsci

A transfer matrix method for free vibration analysis and crack

identification of stepped beams with multiple edge cracks and
different boundary conditions
Mostafa Attar n
School of Mechanical Engineering, The University of Western Australia, 35 Stirling Highway, Crawley WA 6009, Australia

a r t i c l e i n f o abstract

Article history: This paper illustrates an analytical approach to investigating natural frequencies and mode shapes of a
Received 31 October 2010 stepped beam with an arbitrary number of transverse cracks and general form of boundary conditions.
Received in revised form A new method to solve the inverse problem of determining the location and depth of multiple cracks is
19 September 2011
also presented. Based on the Euler–Bernoulli beam theory, the stepped cracked beam is modeled as an
Accepted 17 January 2012
Available online 28 January 2012
assembly of uniform sub-segments connected by massless rotational springs representing local
flexibility induced by the non-propagating edge cracks. A simple transfer matrix method is utilized
Keywords: to obtain the general form of characteristic equation for the cracked beam, which is a function of
Cracked beam frequency, the locations and sizes of the cracks, boundary conditions, geometrical and physical
Natural frequency
parameters of the beam. The proposed method is then used to form a system of 2N equations in order
Transfer matrix
to identify N cracks exploiting 2N measured natural frequencies of the damaged beam. Various
Characteristic equation
Crack identification numerical examples for both direct and inverse problem are provided to validate the present approach.
The results are in good agreement with those obtained by finite element and experimental methods.
& 2012 Elsevier Ltd. All rights reserved.

1. Introduction which can be derived from the stress intensity factors in the
theory of fracture mechanics. Under the most general loading, the
Dynamic behavior of machine components represents one of local flexibility can be represented by a matrix of dimension 6  6
the main problems in solid mechanics and must be controlled to [4]. Local flexibility coefficients depend on the size of the crack
ensure the safety and reliability of structures against collapse or and crack plane’s geometry. In the case of transverse vibration of
to assess their residual load carrying capacity. It has a crucial beams under pure bending, the cracked section may be replaced
importance, especially for aerospace, mechanical, civil and ocean by a single rotational spring representing local flexibility of the
engineering. Mechanical members like beams and columns, crack [5]. Using the local flexibility for investigating free vibration
which are widely used in high speed machinery or aircraft of edge-cracked beams includes two aspects; the first one is the
structures, may contain imperfections such as cracks. The cracks effects of the cracks on the structural dynamic characteristics like
may develop from flaws due to applied cyclic loads, mechanical natural frequencies and mode shape of damaged beams as a
vibrations, aerodynamic loads, etc. and it is obvious that they ‘‘direct problem’’ and the second one is how to predict the
cause a lower structural integrity and change dynamic properties location and size of the cracks from the measured information
such as natural frequencies and mode shapes of the components, of the damaged beam system as an ‘‘inverse problem’’. The direct
so should be certainly taken into account in the vibration analysis analysis of vibrating beams in the presence of edge cracks is
of the structures. necessary to solve the inverse problem.
Investigating the dynamic behavior of cracked beams has Euler–Bernoulli theory has been used in many previous studies
received a great deal of attention over the recent years. Dimar- on the cracked beams with various boundary conditions [6–9].
ogonas [1] and Gasch [2] presented comprehensive reviews of Narkis and Elmalah [10] analyzed vibration of the simply sup-
various methods in tackling a cracked structural member. Dimar- ported cracked beam by using a massless rotational spring and
ogonas and Paipetis [3] suggested an attractive method for dividing the beam into two sub-beams. Zheng and Fan studied
modeling the open edge crack in a beam as a local flexibility beams with hollow rectangular and circular sections [11]. Lele
and Maiti [12] provided a new method based on the Timoshenko
theory for crack identification in beams and Loya et al. [13]
n
Tel.: þ61 422 658830. studied the effect of cracks on the natural frequencies of a simply
E-mail address: mostafa@mech.uwa.edu.au supported Timoshenko beam. In general, there are two main

0020-7403/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijmecsci.2012.01.010
20 M. Attar / International Journal of Mechanical Sciences 57 (2012) 19–33

categories to study dynamic behavior of a cracked beam; con- may be solved by an appropriate numerical method to yield the
tinuous method and discrete method. In the continuous method, crack parameters.
a beam is divided into a number of sub-beams (sub-segments) Various detailed numerical examples are also given to demon-
connected by rotational springs and differential equations are strate the effectiveness of the proposed procedure. The results are
solved for each beam individually considering boundary condi- in good agreement with those obtained by finite element and
tions [5,14] and discrete technique involves the finite element experimental methods. The calculated frequency equation and
method [15]. In the field of inverse problems, greater attention in corresponding mode shapes can be useful to evaluate the influ-
recent years has been devoted to detecting, locating and quanti- ence of parameters like boundary conditions, crack depth, crack
fying the extent of the crack, based on changes of fundamental location, number of cracks and structural constants, reducing
frequencies, mode shapes or dynamic flexibility [12,16–21]. Liang significantly computational time in comparison with a detailed
and co-workers [16] utilized frequency contour plot method finite element analysis. Moreover, it can provide a wide range of
exploiting the first three natural frequencies to detect a crack in data sets which are essential for most of the crack detection
a beam. Lee [17] used a combination of the finite element method procedures.
and the Newton–Raphson procedure to identify multiple cracks.
Studies on the bending vibration of edge-cracked beams are
mostly presented for uniform beams with only one or at most two 2. Local flexibility due to a crack
cracks and there are a few studies on the vibration of non-uniform
beams with cracks. Jang and Bert [22] analyzed free vibration of An open crack on an elastic structure can be considered as a
stepped beams with different boundary conditions and Nagules- source of local flexibility due to the strain energy concentration at
waran [23] studied the vibration of a stepped beam with axial the surrounding area of the crack tip. The idea of replacing a crack
force. Nandwana and Maiti [7] presented a method for crack by a massless spring is presented to establish the relation
detection in a stepped cantilever beam. Shifrin and Ruotolo [14] between the strain energy concentration and applied loads. The
proposed a technique which can be used to analyze the vibratory flexibility coefficients are expressed in terms of stress intensity
characteristics of a beam with multiple cracks. However, only factors, utilizing Castigliano’s theorem. Generalized loading con-
uniform beams can be solved using their approach. ditions for a beam element of circular or rectangular cross-section
The transfer matrix method (TMM) is a prevalent and efficient with a transverse surface crack are shown in Fig. 1. The crack has
tool for free vibration analysis of beams with non-uniform a tip line parallel to z-axis and the bar is loaded with axial load, P1,
mechanical properties. This method, first introduced by Pestel shear loads, P2 and P3, bending moments, P4 and P5, and torsional
and Leckie [24], has been the subject of several research papers torque, P6. According to the Castigliano’s theorem, the additional
[25–27]. Modified transfer matrix methods are also developed to displacement caused by the crack is given as
study the dynamic behavior of beams with various attachments Z 
@U @
[28,29]. ui ¼ ¼ JdA ð1Þ
@P i @P i Ac
The objective of the present paper is to present an analytical
method to investigate free vibration of a stepped beam having an where U is strain energy due to the crack, Ac is crack section, ui is
arbitrary number of transverse open cracks with general form of the additional displacement in the direction of the loading Pi and
boundary conditions and provide an efficient approach to solving J is strain energy density function given by Tada et al. [30] as
the inverse problem of detecting multiple open cracks in a beam. 2 !2 !2 !2 3
 
1 4 X 6 X6 X6
Both ends of the beam carry lumped masses (m1 and m2) and are J¼ 0 K Ii þ K IIi þð1 þuÞ K IIIi 5 ð2Þ
supported by linear (K1 and K2) and rotational (KR1 and KR2) E i¼1 i¼1 i¼1
springs. Using the elastic end conditions for the beam, one may
where n is Poisson’s ratio, E is Young’s modulus, E0 ¼ E for plane
easily model all kinds of two end supports by choosing appro-
stress and E0 ¼ E=ð1n2 Þ for plane strain and K ni ðn ¼ I,II,IIIÞ is the
priate values for stiffness of the springs. The present model is
crack stress intensity factor of mode n corresponding to the
based on the continuous method and the stepped cracked beam is
generalized loading P i . The SIF in the Eq. (2) is determined as
modeled as an assembly of uniform sub-segments connected by
pffiffiffiffiffiffi
massless rotational springs representing local flexibility induced K ni ¼ si paF ni ðn ¼ I,II,IIIÞ, ði ¼ 1,2,. . .,6Þ ð3Þ
by open non-propagating edge cracks. The flexibilities of these
In which, si is the stress at the crack cross-section due to ith
springs are calculated using fracture mechanics theory [11].
independent force, a is the crack depth and Fni denotes a geometry
Based on the Euler–Bernoulli beam theory, differential equa-
dependant non-dimensional crack configuration factor. Now, the
tions for free vibrations are derived for each segment. Four
flexibility coefficient, by definition, is
unknown coefficients appear in the solution of deflection function
Z 
for each sub-segment of cracked beam. To determine these @u @2
C ij ¼ i ¼ JdA ð4Þ
constants the transfer matrix method is employed to satisfy the @Pj @P i @P j Ac
conditions at all boundary points of the sub-segments, which
leads to a general frequency equation for the damaged beam. This According to Eqs. (2)–(4), the elements of the local flexibility
equation is expressed in terms of the elements of the overall matrix depend only on the degrees of freedom being considered
transfer matrix. In addition, the mode shapes of the damaged for the moments and forces applied on the crack section. The full
beam play a crucial role in providing the local and whole
information of the structure. In addition to natural frequencies,
equivalent mode shapes can be derived using the present TMM. b a
To solve the inverse problem, this study presents an effective a P6 P6
scheme based on the transfer matrix method and the Newton– P5 x D P5 P1
x P1
h P2 P4
Raphson iteration procedure to identify N cracks exploiting 2N z P2
P4
z
measured natural frequencies of the damaged beam. In this y P3
y P3
technique, the proposed transfer matrix method is utilized to
form a system of 2N equations, where the locations and sizes of Fig. 1. Beam with an open edge crack under generalized loading condition:
the cracks are unknown parameters. This system of equations (a) rectangular cross-section and (b) circular cross-section.
 M. Attar / International Journal of Mechanical Sciences 57 (2012) 19–33 21

compliance matrix is 6  6, while in this study the flexural terms of f and c as [11]
vibration of the beam in x–y plane is investigated and P5 is the qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
only load in Eq. (4) that the beam could take, so C55 is the only 2f þ 14c 1
l¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð7dÞ
component in the flexibility matrix. The SIF in Eq. (3) is then 2
2 14c
reduced to the form
pffiffiffiffiffiffi P 5 :h
K I5 ¼ s5 pzF I5 , s5 ¼ ð5a; bÞ 3. Direct problem formulation
2I

where P5 is bending moment and I is the second moment of area Consider an intact beam with uniform cross-section area A,
of the cross-section. z and Z are spatial variables to measure the Young’s modulus E, moment of area I and density r. The
crack length and width along the crack plane (crack plane depth is differential equation that governs transverse motion of an elastic
h for rectangular and D for circular cross-section), as shown in Euler–Bernoulli beam can be expressed as
Fig. 2. The origin of the zZ axis in Fig. 2 is different from the
@4 yðx,tÞ @2 yðx,tÞ
origin of the yz axis in Fig. 1. Now C55 is obtained as follows: EI þ rA ¼0 ð8Þ
@x4 @t 2
2 3
Z ! "Z Z ! #
@2 6 K 2I5 @2 a b=2
K 2I5 where yðx,tÞ is transverse deflection function of the beam at axial
7
C 55 ¼ 4 0 dA5 ¼ 0 d Zdz ð6Þ coordinate x and time t. The free vibration solution can be found
@P25 E 2
@P5 0 b=2 E
Ac using the method of separation of variables as
yðx,tÞ ¼ YðxÞUTðtÞ ð9Þ
Finally, using Eq. (5a) and (5b) into Eq. (6) and integrating
along the crack edge, as shown in Fig. 2, the local flexibility for Using Eq. (9) in Eq. (8) and rearranging yields
rectangular-sectional beam derived as [11] 4
d YðxÞ 4
"Z # b YðxÞ ¼ 0 ð10aÞ
6ph a=h  2 dx4
C¼ 0 f HðlÞ df ð7aÞ
E Ir 0 2
d TðtÞ
þ o2 TðtÞ ¼ 0 ð10bÞ
Following the same procedure for circular cross-section, the dt 2
4
local flexibility obtained as [11] where b ¼ o2 ðrA=EIÞ and o is the natural frequency of the entire
2 8 pffiffiffiffiffiffiffiffiffiffi 9 3 uniform beam. The solution of the Eq. (10a) can be expressed as
Z Z ff2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
16D 4 a=D < 2 2
=
YðxÞ ¼ B1 cosðbxÞ þ B2 sinðbxÞ þ B3 coshðbxÞ þ B4 sinhðbxÞ ð11aÞ
pffiffiffiffiffiffiffiffiffi2ffi ½14c ½2f þ 14c 1½HðlÞ dc df5
2
C¼ 0
E Ic 0 :  ff ;
In which, B1  B4 are unknown constants. The expressions for
ð7bÞ slope yðxÞ, bending moment MðxÞ and shear force VðxÞ are
obtained as
where h is rectangular section depth, Ir is bh3/12, D is circular
@YðxÞ  
section diameter, Ic is pD4/64 and C55 is replaced by C. For yðxÞ ¼ ¼ bU B1 sinðbxÞ þ B2 cosðbxÞ þ B3 sinhðbxÞ þ B4 coshðbxÞ
rectangular cross-section f ¼ z=h and for circular cross-section @x
f ¼ z=D and c ¼ Z=D denote normalized spatial variables, and H is ð11bÞ
a function of non-dimensional parameter l as [11]
@Y 2 ðxÞ 2 
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi MðxÞ ¼ EI ¼ EIb U B1 cosðbxÞB2 sinðbxÞ
ð2=plÞtanðpl=2Þð0:923 þ 0:199ð1sinðpl=2ÞÞ4 Þ @x2
HðlÞ ¼ ð7cÞ þ B3 coshðbxÞ þ B4 sinhðbxÞ ð11cÞ
cosðpl=2Þ

In which, for rectangular cross-section l is expressed in terms @Y 3 ðxÞ 3 

VðxÞ ¼ EI ¼ EIb U B1 sinðbxÞB2 cosðbxÞ þ B3 sinhðbxÞ
of f (i.e., l ¼ f) and for circular cross-section l is expressed in @x3
þ B4 coshðbxÞ ð11dÞ
Sectional crack
As shown in Fig. 3, a section of uniform beam with an open
b edge crack is divided into two beam segments and the crack
S
position is assumed as the separation point. Eqs. (11a)–(11d) can
a a be written in the form of matrix for both segments i and iþ1,
h

S x xc
S-S 2( xc )
Sectional crack

C x
S
1(xc) i i 1
a a

h
a
S y
S-S

Fig. 2. A transverse open crack: (a) in a rectangular cross-sectional beam and Fig. 3. Equivalent massless rotational spring and discontinuity of slope at crack
(b) in a circular cross-sectional beam. position.
22 M. Attar / International Journal of Mechanical Sciences 57 (2012) 19–33

xn , N
xn , 2
xn , 1
xi , N
xi , 2
xi , 1
x2 , N
x2 , 2
x2 , 1
x1, N
x1, 2
x1,1

KR1 KR 2
... ...
m1 ... C2,1 C 2, 2 C2, N
... m2
e1 ... Ci,1 Ci , 2 Ci ,N ... C e2
J1 I1 C1,1 C1, 2 C1,N I n n ,1 Cn, 2 Cn,N J2 x
I2 Ii
K1 K2
L1 L2 Li Ln
y L

Fig. 4. A multi-stepped free–free beam with N (N ¼N1 þN2 þ ?þ Nn) edge open cracks composed of n uniform beam segments and carrying lumped masses, linear and
rotational springs at both ends.

which can be treated separately, as Table 1

8 9 8 9 Common boundary conditions modeled by appropriate value of stiffness of
> Y i ðxÞ > > B1,i >
>
> >
> >
> >
> springs.
> yi ðxÞ =
< > >
< >
B2,i =
M ðxÞ ¼ ½SðxÞU B3,i > ð12aÞ Boundary conditions Linear and rotational stiffness of
>
> i >
> >
> >
>
> >
> >
> >
> springs (m1 ¼m2 ¼ 0, e1 ¼ e2 ¼ 0, J1 ¼J2 ¼ 0)
: V i ðxÞ ; : B4,i ;
K1 KR1 K2 KR2
8 9 8 9
> Y i þ 1 ðxÞ > > B1,i þ 1 > Clamped–free 0 0
>
> >
> >
> >
>
N N
>
< yi þ 1 ðxÞ >
= >
< B2,i þ 1 >
= Pinned–pinned N 0 N 0
¼ ½SðxÞU B3,i þ 1 ð12bÞ Clamped–pinned N N N 0
> M i þ 1 ðxÞ > > >
>
> >
> >
> >
> Clamped–clamped N N N N
>
: V >
; >
: >
;
ðxÞ B4,i þ 1 Free–free 0 0 0 0
iþ1

where
2 3
cosðbxÞ sinðbxÞ coshðbxÞ sinhðbxÞ These cracks are located at points xi,1 ,xi,2 ,. . .,x1,Ni with sizes
6 b sinðbxÞ b cosðbxÞ b sinhðbxÞ b coshðbxÞ 7 ai,1 ,ai,2 ,. . .,a1,Ni . To model a general form of boundary conditions,
6 7
½SðxÞ ¼ 6
6 EIb2 cosðbxÞ 2 2 2
7
the ends of the beam carry lumped masses (ms with eccentricity es
4 EIb sinðbxÞ EIb coshðbxÞ EIb sinhðbxÞ 7
5
3 3 3 3 and rotary inertia Js, s¼1,2) and are supported by linear springs (K1
EIb sinðbxÞ EIb cosðbxÞ EIb sinhðbxÞ EIb coshðbxÞ
and K2) and rotational springs (KR1 and KR2) as shown in Fig. 4.
ð12cÞ Choosing the appropriate value for lumped masses and stiffness of
There is continuity among the deflection, bending moment and springs, one may model all kinds of two end supports. For instance,
shear force at the crack position for segments i and iþ1, but there springs stiffness coefficients for five common end conditions are
is a discontinuity between slopes which is revealed by the bending listed in Table 1. The geometrical parameters indicated in Fig. 4 are
moment and rotation of the spring representing the edge crack. listed in Table 2, where Ci,j is the local flexibility coefficient of jth crack
These boundary conditions at crack position, xc, may be written as at the ith step section and can be computed by Eqs. (7a) and (7b).
Eq. (12a) can be rewritten for a segment, xi,j  1 ox oxi,j, as the
Y i ðxc Þ ¼ Y i þ 1 ðxc Þ ð13aÞ
following general form:
M i ðxc Þ ¼ M i þ 1 ðxc Þ ð13bÞ fQ i,j ðxÞg ¼ ½Si ðxÞUfBi,j g ð15aÞ

V i ðxc Þ ¼ V i þ 1 ðxc Þ ð13cÞ where

8 9
yi þ 1 ðxc Þyi ðxc Þ ¼ CUMi ðxc Þ ð13dÞ >
> Y i,j ðxÞ >
>
>
> >
>
>
< yi,j ðxÞ >
=
Using Eqs. (13a)–(13d) yields fQ i,j ðxÞg ¼ ð15bÞ
8 9 8 9 > M i,j ðxÞ >
Y i þ 1 ðxc Þ > 2 1 0 0 0 3 > Y i ðxc Þ >
> >
>
>
> > > >
> >
> >
>
>
> > > > : V i,j ðxÞ ;
>
< yi þ 1 ðxc Þ = 6 0 1 C 0 7 > < yi ðxc Þ >
=
6 7
¼ 6 7U
M i þ 1 ðxc Þ > 4 0 0 1 0 5 > Mi ðxc Þ > ð14Þ
>
> > > > 8 1 9
>
> >
> >
> >
> > Bi,j >
: V ðx Þ ; 0 0 0 1 : V ðx Þ ; >
> >
>
iþ1 c i c >
> >
>
>
> 2 >
>
< Bi,j =
Now consider a multi-stepped beam with an arbitrary number of fBi,j g ¼ ð15cÞ
open non-propagating edge cracks as shown in Fig. 4. It consists of n >
> B3i,j >
>
>
> >
>
uniform beam steps and the length of ith step is Li, the second >
> >
>
>
: B4i,j >
;
moment of area is Ii and the number of cracks in the ith step is Ni.
 M. Attar / International Journal of Mechanical Sciences 57 (2012) 19–33 23

Table 2
Geometrical parameters for stepped beam considered in this study.

No. of segment Li Ii Ni xi,0 xi,j xi,Ni þ 1 Ci,j

1 L1 I1 N1 x1,0 ¼0 x1,1,x1,2, y ,x1,N1 x1, N1 þ 1 ¼ L1 C1,1,C 1,2, y ,C 1,N1

2 L2 I2 N2 x2,0 ¼L1 x2,1,x2,2, y ,x2,N2 x2, N2 þ 1 ¼ L1 þ L2 C2,1,C 2,2, y ,C 2,N2
3 L3 I3 N3 x3,0 ¼L1 þL2 x3,1,x3,2, y ,x3,N3 x3, N3 þ 1 ¼ L1 þ L2 þ L3 C3,1,C3,2, y ,C3,N3
^ ^ ^ ^ ^ ^ ^ ^
i Li Ii Ni xi,0 ¼ L1 þ L2 þ    þ Li  1 xi,1,xi,2, y ,xi,Ni xi,Ni þ 1 ¼L1 þL2 þ    þ Li Ci,1,Ci,2, y ,Ci,Ni
^ ^ ^ ^ ^ ^
n Ln In Nn xn,0 ¼L1 þL2 þ    þ Ln  1 xn,1,xn,2, y ,xn,Nn xn,Nn þ 1 ¼L1 þ L2 þ    þ Ln Cn,1,Cn,2, y ,Cn,Nn

2 3
cosðbi xÞ sinðbi xÞ coshðbi xÞ sinhðbi xÞ ½Tci,j  ¼ ½C i,j U½Si ðxi,j ÞU½Si ðxi,j1 Þ1 ð23bÞ
6 b sinðbxÞ bi cosðbi xÞ bi sinhðbi xÞ bi coshðbi xÞ 7
6 i 7
½Si ðxÞ ¼ 6
6 EIi b2 cosðbi xÞ 2 2 2 7 Using Eqs. (22) and (23b) leads to
4 i EIi bi sinðbi xÞ EIi bi coshðbi xÞ EIi bi sinhðbi xÞ 7
5
3 3 3 3
EIi bi sinðbi xÞ EIi bi cosðbi xÞ EIi bi sinhðbi xÞ EIi bi coshðbi xÞ fQ 1,2 ðxÞg ¼ ½S1 ðxÞU½S1 ðx1,1 Þ1 U½Tc1,1 UfQ 1,1 ð0Þg ð24Þ
ð15dÞ Similar equation can be obtained for the (N1 þ1)th segment of
4
In which, bi ¼ o2 ðrAi =EIi Þ and Ai represents area on the ith the first step (x1,N1 oxoL1) as
step. The boundary condition for a crack positioned at x ¼xi,j can fQ 1,N1 þ 1 ðxÞg ¼ ½S1 ðxÞU½S1 ðx1,N1 Þ1 U½Tc1,N1 U½Tc1,N1 1     ½Tc1,2 U½Tc1,1 fQ 1,1 ð0Þg
be written, similar to Eq. (14), as
ð25Þ
fQ i,j þ 1 ðxi,j Þg ¼ ½C i,j UfQ i,j ðxi,j ÞgO ð16aÞ
According to Eq. (25), a relation between the deflections,
Where slopes, bending moments and shear forces of the ends of the first
2 3 step may be expressed as
1 0 0 0
60 1 C i,j 07 fQ 1,N1 þ 1 ðL1 Þg ¼ ½T 1,N1 þ 1 U½Tc1,N1 U½Tc1,N1 1     ½Tc1,2 U½Tc1,1 fQ 1,1 ð0Þg
6 7
½C i,j  ¼ 6 7 ð16bÞ
40 0 1 05 ð26Þ
0 0 0 1
Let
Using an equilibrium approach, the boundary conditions for ½Ti,j ¼ ½T i,j þ 1 U½Tci,j U½Tci,j1     ½Tci,2 U½Tci,1  ð27Þ
the stepped beam can be derived as
Then Eq. (26) can be rewritten as
Mð0Þ ¼ e1 m1 o2 Yð0Þ þ K R1 yð0ÞðJ 1 þm1 e21 Þo2 yð0Þ ð17aÞ
fQ 1,N1 þ 1 ðL1 Þg ¼ ½T1,N1 UfQ 1,1 ð0Þg ð28Þ
Vð0Þ ¼ K 1 Yð0Þ þm1 o2 Yð0Þ þ e1 m1 o2 yð0Þ ð17bÞ
Correspondingly, the relationships between the ends of the
MðLÞ ¼ e2 m2 o2 YðLÞK R2 yðLÞ þ ðJ2 þ m2 e22 Þo2 yðLÞ ð17cÞ other steps may be derived as
fQ 2,N2 þ 1 ðL1 þ L2 Þg ¼ ½T2,N2 UfQ 2,1 ðL1 Þg ð29aÞ
VðLÞ ¼ K 2 YðLÞm2 o2 YðLÞe2 m2 o2 yðLÞ ð17dÞ
General formulas expressed above can be used to model the fQ 3,N3 þ 1 ðL1 þ L2 þL3 Þg ¼ ½T3,N3 UfQ 2,1 ðL1 þ L2 Þg
cracked stepped beam as an assembly of uniform sub-segments ^ ð29bÞ
connected by massless rotational springs in order to derive
characteristic equation and calculate the natural frequencies fQ i,Ni þ 1 ðL1 þ L2 þ    þ Li1 þ Li Þg
and mode shapes of the beam. Eq. (15a) can be obtained for a ¼ ½Ti,Ni UfQ i,1 ðL1 þ L2 þ    þLi2 þ Li1 Þg ð29cÞ
segment in which 0 oxox1,1 as
fQ 1,1 ðxÞg ¼ ½S1 ðxÞUfB1,1 g ð18Þ ^
fQ n,Nn þ 1 ðL1 þ L2 þ    þ Ln1 þ Ln Þg
Eq. (18) may be rewritten as
¼ ½Tn,Nn UfQ n,1 ðL1 þL2 þ    þ Ln2 þ Ln1 Þg ð29dÞ
fQ 1,1 ðxÞg ¼ ½S1 ðxÞU½S1 ð0Þ1 UfQ 1,1 ð0Þg ð19Þ
Considering the continuity of the deflection, slope, bending
Using Eqs. (16a) and (19) for a crack positioned at x ¼x1,1
moment and shear force at the common interface of two neigh-
becomes
boring steps, the relationship between the left end of the first step
fQ 1,2 ðx1,1 Þg ¼ ½C 1,1 U½S1 ðx1,1 ÞU½S1 ð0Þ1 UfQ 1,1 ð0Þg ð20Þ (x ¼0) and the right end of the nth step can be expressed as the
following form:
Similar equation for a segment in which x1,1 oxox1,2 is
obtained as fQ n,Nn þ 1 ðLÞg ¼ ½TUfQ 1,1 ð0Þg ð30Þ

fQ 1,2 ðxÞg ¼ ½S1 ðxÞUfB1,2 g ¼ ½S1 ðxÞU½S1 ðx1,1 Þ 1

UfQ 1,2 ðx1,1 Þg ð21Þ where ½T  is the overall transfer matrix and can be expressed as

Substituting Eq. (20) into Eq. (21) yields ½T ¼ ½Tn,Nn U½Tn1,Nn1    ½T 2,N2 U½T1,N1 ð31Þ

fQ 1,2 ðxÞg ¼ ½S1 ðxÞU½S1 ðx1,1 Þ1 U½C 1,1 U½S1 ðx1,1 ÞU½S1 ð0Þ1 UfQ 1,1 ð0Þg Matrix ½T has a general form as below:
ð22Þ 8 9 2 3 8 9
> Y n,Nn þ 1 ðLÞ > T T 12 T 13 T 14 > Y 1,1 ð0Þ >
>
> >
> 6 11 > >
< yn,Nn þ 1 ðLÞ >
> = 6 T 21 T 22 T 23 7 >> y1,1 ð0Þ >>
Two matrixes, in general forms, are defined as ½T i,j  and ½Tci,j  T 24 7 < =
¼6 7U ð32Þ
where > M n,Nn þ 1 ðLÞ >
> >
6 T 31 T 32 T 33
4
7
T 34 5 >>
M 1,1 ð0Þ >
>
>
> >
> >
> >
>
1 : ; T 41 T 42 T 43 T 44 : ;
½T i,j  ¼ ½Si ðxi,j ÞU½Si ðxi,j1 Þ ð23aÞ V n,Nn þ 1 ðLÞ V 1,1 ð0Þ
24 M. Attar / International Journal of Mechanical Sciences 57 (2012) 19–33

Substituting Eqs. (17a)–(17d) into Eq. (32) gives problem, natural frequencies of the beam with an arbitrary
8 9 number of steps and surface cracks and any kind of boundary
>
> Y n,Nn þ 1 ðLÞ >
>
>
> >
> conditions may be obtained easily by applying a suitable numer-
>
< yn,Nn þ 1 ðLÞ >
=
ical technique to solve Eq. (35f). The characteristic equation for
> e2 m2 o2 Y n,Nn þ 1 ðLÞK R2 yn,Nn þ 1 ðLÞ þðJ 2 þm2 e22 Þo2 yn,Nn þ 1 ðLÞ > the cracked beam could also be derived in explicit form by the
>
> >
>
>
> >
>
: K 2 Y n,Nn þ 1 ðLÞm2 o2 Y n,Nn þ 1 ðLÞe2 m2 o2 yn,Nn þ 1 ðLÞÞ ; proposed method. Neglecting rotary inertia and eccentricity for
8 9 the lumped masses at both ends of the beam, frequency equation
>
> Y 1,1 ð0Þ >
> may be written by the simplified form as
>
> >
>
>
< y1,1 ð0Þ >
=
¼ ½TU e1 m1 o2 Y 1,1 ð0Þ þ K R1 y1,1 ð0Þ-ðJ 1 þ m1 e2 Þo2 y1,1 ð0Þ ½ðK 1 m1 o2 ÞðK R2 T 24 þ T 34 ÞT 31 K R2 T 21 U½K R1 ðK 2 T 13 T 43 Þ
>
> 1 >
>
>
> >
> T 42 þK 2 T 12 m2 o2 ðK R1 T 13 þ T 12 Þ½ðK 1 m1 o2 ÞðK 2 T 14 T 44 Þ
>
: >
;
K 1 Y 1,1 ð0Þ þm1 o2 Y 1,1 ð0Þ þe1 m1 o2 y1,1 ð0Þ
þ T 41 K 2 T 11 m2 o2 ðK 1 T 14 T 11 Þ
ð33Þ
þ m1 m2 o4 T 14 U½K R1 ðK R2 T 23 þ T 33 Þ þ T 32 þ K R2 T 22  ¼ 0 ð36Þ
From Eq. (33), one obtains
( ) " # ( )
Y n,Nn þ 1 ðLÞ T 11 e1 m1 o2 T 13 þ ðm1 o2 K 1 ÞT 14 T 12 þðK R1 ðJ 1 þm1 e21 Þo2 ÞT 13 þ e1 m1 o2 T 14 Y 1,1 ð0Þ
¼ U ð34aÞ
yn,Nn þ 1 ðLÞ T 21 e1 m1 o2 T 23 þ ðm1 o2 K 1 ÞT 24 T 22 þðK R1 ðJ 1 þm1 e21 Þo2 ÞT 23 þ e1 m1 o2 T 24 y1,1 ð0Þ
" # ( ) " # ( )
e2 m 2 o 2 K R2 þ ðJ2 þ m2 e22 Þo2 Y n,Nn þ 1 ðLÞ T 31 e1 m1 o2 T 33 þ ðm1 o2 K 1 ÞT 34 T 32 þ ðK R1 ðJ 1 þ m1 e21 Þo2 ÞT 33 þ e1 m1 o2 T 34 Y 1,1 ð0Þ
U ¼ U
K 2 m2 o2 e2 m2 o 2 yn,Nn þ 1 ðLÞ T 41 e1 m1 o2 T 43 þ ðm1 o2 K 1 ÞT 44 T 42 þ ðK R1 ðJ 1 þ m1 e21 Þo2 ÞT 43 þ e1 m1 o2 T 44 y1,1 ð0Þ
ð34bÞ

Substituting Eq. (34a) into Eq. (34b) yields the general form of Corresponding to each natural frequency, calculated from
frequency equation (characteristic equation) for cracked beam, Eq. (35a), one may determine the unknown constants, fBi,j g, for
shown in Fig. 4, as the following form: each sub-segment of the cracked beam, in order to obtain the
F ¼ f 1 Uf 4 f 2 Uf 3 ¼ 0 ð35aÞ mode shape of the entire beam. At first, the unknown coefficients
for the first segment, fB1,1 g, are determined. Then, the unknown
where coefficients for the other segments may be obtained, using the
f 1 ¼ e2 m2 o2 T 11 e1 e2 m1 m2 o4 T 13 e2 m2 o2 K 1 T 14 þ e2 m1 m2 o4 T 14 following procedure.
For the first segment, substituting Eqs. (15d), (17a) and (17b)
K R2 T 21 þ ðJ2 þ m2 e22 Þo2 T 21 þ e1 m1 o2 K R2 T 23 e1 m1 ðJ2 þ m2 e22 Þo4 T 23 into Eq. (18) leads to
þ K 1 K R2 T 24 m1 o2 K R2 T 24 K 1 ðJ2 þ m2 e22 Þo2 T 24 þ m1 ðJ 2 þm2 e22 Þ 8 9
>
> Y 1,1 ð0Þ >
>
4 2
o T 24 T 31 þe1 m1 o T 33 þK 1 T 34 m1 o T 34 2
ð35bÞ >
> >
>
>
< y1,1 ð0Þ >
=
> e1 m1 o2 Y 1,1 ð0Þ þK R1 y1,1 ð0ÞðJ 1 þm1 e21 Þo2 y1,1 ð0Þ >
f 2 ¼ e2 m2 o2 T 12 e2 m2 o2 K R1 T 13 e2 m2 ðJ 1 þm1 e21 Þo4 K 1 T 13 >
> >
>
>
> >
>
e1 e2 m1 m2 o4 T 14 K R2 T 22 þ ðJ 2 þ m2 e22 Þo2 T 22 K R1 K R2 T 23 : K 1 Y 1,1 ð0Þ þm1 o2 Y 1,1 ð0Þ þe1 m1 o2 y1,1 ð0Þ ;
þ K R2 ðJ1 þ m1 e21 Þo2 T 23 þ ðJ 2 þ m2 e22 Þo2 K R1 T 23
8 9
ðJ 1 þm1 e21 ÞðJ2 þ m2 e22 Þo4 T 23 e1 m1 o2 K R2 T 24 2 3 >> B11,1 >
>
1 0 1 0 >
> >
>
þ e1 m1 ðJ 2 þ m2 e22 Þo4 T 24 T 32 K R1 T 33 þ ðJ1 þ m1 e21 Þo2 T 33 >
> >
>
6 b1 7 > 2 >
6 0 b1 0 7 < B1,1 =
e1 m1 o2 T 34 ð35cÞ ¼6 7U ð37aÞ
6 EI1 b21
4 0 EI1 b1
2
0 7 5 >> B31,1 >
>
>
> >
>
0
3
EI1 b1 0
3
EI1 b1 > > >
>
f 3 ¼ K 2 T 11 m2 o2 T 11 e1 m1 o2 K 2 T 13 þ e1 m1 m2 o4 T 13 >
: B41,1 >
;
2 4 2
þ m1 o K 2 T 14 K 1 K 2 T 14 m1 m2 o T 14 þm2 o K 1 T 14
e2 m2 o2 T 21 þe1 e2 m1 m2 o4 T 23 þ e2 m2 o2 K 1 T 24 From Eq. (37a) one obtains
e2 m1 m2 o4 T 24 T 41 þ e1 m1 o2 T 43 m1 o2 T 44 þ K 1 T 44 ð35dÞ
( ) " #
Y 1,1 ð0Þ 1 0 1 0
f 4 ¼ K 2 T 12 m2 o2 T 12 þK R1 K 2 T 13 ðJ 1 þm1 e21 Þo2 K 2 T 13 ¼ UfB1,1 g ð37bÞ
y1,1 ð0Þ 0 b1 0 b1
m2 o2 K R1 T 13 þm2 ðJ1 þ m1 e21 Þo4 T 13 þ e1 m1 o2 K 2 T 14
e1 m1 m2 o4 T 14 e2 m2 o2 T 22 e2 m2 o2 K R1 T 23
þ e2 m2 ðJ 1 þ m1 e21 Þo4 T 23 e1 e2 m1 m2 o4 T 24 T 42 " # ( )
K R1 T 43 þ ðJ 1 þ m1 e21 Þo2 T 43 e1 m1 o2 T 44 ð35eÞ e1 m1 o2 K R1 ðJ1 þ m1 e21 Þo2 Y 1,1 ð0Þ
U
K 1 þ m1 o2 e1 m1 o2 y1,1 ð0Þ
Therefore, free vibration analysis of the cracked multi-stepped
beam leads to F¼0, where F for a specific damaged beam with 2 3
2 2
known geometrical and physical parameters and boundary con- EI1 b1 0 EI1 b1 0
ditions, is a function of natural frequencies, the positions and ¼4 3 3
5UfB1,1 g ð37cÞ
0 EI1 b1 0 EI1 b1
depths of the cracks, as the following form:
Fðoi ,a1,1 ,a1,2 ,. . .,a1,N1 ,. . .,an,1 ,an,2 ,. . ., Substitution of Eq. (37b) in Eq. (37c) yields
an,Nn ,x1,1 ,x1,2 ,. . .,x1,N1 ,. . .,xn,1 ,xn,2 ,. . .,xn,Nn Þ ¼ 0 ð35fÞ " #
Q 11 Q 12 Q 13 Q 14
In which the crack depths are used to compute the local UfB1,1 g ¼ 0 ð38aÞ
Q 21 Q 22 Q 23 Q 24
flexibilities by Eqs. (7a) and (7b). Using this approach in the direct
 M. Attar / International Journal of Mechanical Sciences 57 (2012) 19–33 25

where For xi,j1 o xo xi,j , j ¼ 1 : fBi,1 g ¼ ½Si ðxi,0 Þ1 U½Si1 ðxi1,Ni1 þ 1 ÞUfBi1,Ni1 þ 1 g
2 ð42bÞ
Q 11 ¼ e1 m1 o2 þ EI1 b1 Q 21 ¼ K 1 þm1 o2
Q 12 ¼ K R1 b1 ðJ 1 þm1 e21 Þo2 b1 Q 22 ¼ e1 m1 o 2
b1 þ EI1 b31
2 2 ð38bÞ
Q 13 ¼ e1 m1 o EI1 b1 Q 23 ¼ K 1 þm1 o2
3
Q 14 ¼ K R1 b1 ðJ 1 þm1 e21 Þo2 b1 Q 24 ¼ e1 m1 o2 b1 EI1 b1 4. Results and discussion for direct problem

In the same manner, substituting Eqs. (15d), (17c), (17d) and In order to illustrate the proposed method in a direct problem,
(18) into Eq. (32) one can obtain free vibration analyses for various cracked beams are presented in
( )
Y n,Nn þ 1 ðLÞ this section. Results obtained according to this procedure for
yn,Nn þ 1 ðLÞ some combinations of crack and beam dimensions are compared
2 3 with available results.
2
T 11 EI1 b1 T 13 b1 ðT 12 EI1 b21 T 14 Þ T 11 þEI1 b21 T 13 b1 ðT 12 þEI1 b21 T 14 Þ
¼4 2
5
T 21 EI1 b1 T 23 b1 ðT 22 EI1 b21 T 24 Þ T 21 þEI1 b21 T 23 b1 ðT 22 þEI1 b21 T 24 Þ Example 1. Free vibration of a square-sectional cantilever beam
having two cracks.
UfB1,1 g ð39aÞ
" # ( )
Fig. 5 shows a square-sectional cantilever beam with two open
e2 m2 o2 K R2 þ ðJ 2 þm2 e22 Þo2 Y n,Nn þ 1 ðLÞ edge cracks. The material parameters for the beam are
U
K 2 m2 o2 e2 m2 o2 yn,Nn þ 1 ðLÞ E¼181 Gpa, r ¼ 7860 kg/m3, Poisson’s ratio n ¼ 0.29 and its
2 3 dimensions are 0.02 m  0.02 m  0.8 m. The normalized crack
2 2 2
T 31 EI1 b1 T 33 b1 ðT 32 EI1 b1 T 34 Þ T 31 þEI1 b1 T 33 b1 ðT 32 þEI1 b21 T 34 Þ
¼4 2 2 2
5 sizes and locations are a1,1/h¼0.2, a1,2/h¼ 0.3, x1,1/L¼0.3182 and
T 41 EI1 b1 T 43 b1 ðT 42 EI1 b1 T 44 Þ T 41 þEI1 b1 T 43 b1 ðT 42 þEI1 b21 T 44 Þ
x1,2/L¼0.6812. Using Eqs. (27), (31) and (32) for this beam one
UfB1,1 g ð39bÞ obtains
fQ 1,3 ðLÞg ¼ ½TUfQ 1,1 ð0Þg ð43aÞ
Using Eq. (39a) in Eq. (39b) leads to
" #
Q 31 Q 32 Q 33 Q 34 ½T ¼ ½T1,2 ¼ ½T 1,3 U½Tc1,2 U½Tc1,1  ð43bÞ
UfB1,1 g ¼ 0 ð40aÞ
Q 41 Q 42 Q 43 Q 44 In which
where ½T 1,3  ¼ ½S1 ðx1,3 ÞU½S1 ðx1,2 Þ1 ð43cÞ
2
Q 31 ¼ e2 m2 ðT 11 EI1 b1 T 13 Þ þððJ2 þ m2 e22 Þ
o 2 2
o K R2 Þ
2 2
½Tc1,2  ¼ ½C 1,2 U½S1 ðx1,2 ÞU½S1 ðx1,1 Þ1 ð43dÞ
UðT 21 EI1 b1 T 23 ÞT 31 þ EI1 b1 T 33
2
Q 32 ¼ e2 m2 o2 b1 ðT 12 EI1 b1 T 14 Þ þ b1 ððJ 2 þm2 e22 Þo2 K R2 Þ ½Tc1,1  ¼ ½C 1,1 U½S1 ðx1,1 ÞU½S1 ðx1,0 Þ1 ð43eÞ
2 2
UðT 22 EI1 b1 T 24 Þb1 ðT 32 EI1 b1 T 14 Þ where x1,0 ¼0 and x1,3 ¼L. For this beam frequency equation can
2
o 2
Q 33 ¼ e2 m2 ðT 11 þEI1 b1 T 13 Þ þ ððJ 2 þm2 e22 Þ o2 K R2 Þ be obtained by use of Eq. (35a) and setting K1 ¼N, KR1 ¼N, K2 ¼0,
2 2 KR2 ¼ 0, m1 ¼m2 ¼ 0, e1 ¼e2 ¼0, J1 ¼J2 ¼0 as
UðT 21 þEI1 b1 T 23 ÞT 31 EI1 b1 T 33
2 F ¼ T 34 UT 43 T 33 UT 44 ¼ 0 ð44Þ
Q 34 ¼ e2 m2 o2 b1 ðT 12 þEI1 b1 T 14 Þ þ b1 ððJ 2 þ m2 e22 Þo2 K R2 Þ
2 2 By solving Eq. (44) natural frequencies of the beam, shown in
UðT 22 þEI1 b1 T 24 Þb1 ðT 32 þ EI1 b1 T 14 Þ
Fig. 5, can be easily obtained. It is obvious that by setting C1,1 ¼0,
2
2
o
Q 41 ¼ ðK 2 m2 ÞUðT 11 EI1 b1 T 13 Þe2 m2 o2 ðT 21 EI1 b21 T 23 Þ C1,2 ¼0 in the above equations, one obtains the natural frequencies of
2 the corresponding intact beam. The lowest three natural frequencies
T 41 þ EI1 b1 T 43
2 2 of the beam, obtained from experiment [19], two dimensional ANSYS
Q 42 ¼ b1 ðK 2 m2 o2 ÞUðT 12 EI1 b1 T 14 Þe2 m2 o2 b1 ðT 22 EI1 b1 T 24 Þ
model [20], finite element method [20] and present study are listed in
2
b1 ðT 42 EI1 b1 T 44 Þ Table 3, indicating that the results calculated by the procedure
2
Q 43 ¼ ðK 2 m2 o2 ÞUðT 11 þ EI1 b1 T 13 Þe2 m2 o2 ðT 21 þ EI1 b1 T 23 Þ
2 proposed in this article agree well with those obtained from experi-
2 ment. Fig. 6 represents a plot of characteristic equation (Eq. (44))
T 41 EI1 b1 T 43
versus o for this beam, with and without cracks.
2 2
Q 44 ¼ b1 ðK 2 m2 o2 ÞUðT 12 þ EI1 b1 T 14 Þe2 m2 o2 b1 ðT 22 þ EI1 b1 T 24 Þ The reduction of first three natural frequencies of a beam with
b1 ðT 42 þEI1 b1 T 44 Þ
2
ð40bÞ same dimensions and properties as in Example 1, but with only
one edge crack and for various boundary conditions are shown
Combining Eqs. (38a) and (40a), one obtains in Fig. 7, where the vertical axis stands for the frequency ratio of
8 9
3 > 1 the natural frequencies of the cracked beam to the natural
2 > B1,1 >> 8 9
Q 11 Q 12 Q 13 Q 14 > >
>
>
> > 0 >
>
6 7 >> 2 >
> >
> >
>
6 Q 21 Q 22 Q 23 Q 24 7 < B1,1 = < 0 =
6 7U 3 ¼ ð41Þ
6 Q 31 Q 32 Q 33 Q 34 7 > B1,1 > > 0 > b
4 5 >> > >
> > >
>
> > : ;
Q 41 Q 42 Q 43 Q 44 > >
>
>
> 0
: B41,1 >
; h
x1, 2
Unknown coefficients for the first segment may be calculated x1 ,1
from Eq. (41). Following the procedure of TMM, unknown con-
stants for the other segments may be derived by using the
I a1,1 a1 , 2 x
constants of the previous segment as
y L
For xi,j1 o x oxi,j , ja1 : fBi,j g ¼ ½Si ðxi,j1 Þ 1
U½C i,j1 U½Si ðxi,j1 ÞUfBi,j1 g
ð42aÞ Fig. 5. Uniform cantilever beam with rectangular cross-section and double cracks.
26 M. Attar / International Journal of Mechanical Sciences 57 (2012) 19–33

Table 3 the former example and the total length of the beam is L¼0.3 m.
Natural frequencies of intact and cracked beam (with double cracks). In order to illustrate the efficiency of the present method, the
lowest four natural frequencies of the stepped beam for various
Methods Natural frequencies, os (rad/s)
crack depths, crack locations and boundary conditions are com-
o1 o2 o3 puted and given in Table 5. A finite element analysis is also
carried out using ANSYS software to verify the results obtained for
Experimental measurements [19] Intact 151.896 955.691 2666.93 the stepped beam. A two dimensional model of the cracked beam
Cracked 151.073 937.878 2571.63
is built with 8-node PLANE183 elements and a fine mesh of
2D ANSYS model [20] Cracked 151.475 936.76 2568.13 quarter-point singular elements is used at the vicinity of the
Finite element method [20] Cracked 151.707 943.232 2607.65 crack, as shown in Fig. 9b. The dependence of the results on the
Present study Intact 152.208 953.873 2670.87 mesh density has been checked by convergence test. A reasonable
Cracked 151.058 937.405 2583.64 agreement between TMM results and those achieved by finite
element analysis is observed for all cases of the stepped beam in
L¼ 0.8 m, h ¼b ¼0.02 m, a1,1/h ¼0.2, a1,2/h ¼0.3, x1,1/L¼ 0.3182, x1,2/L¼ 0.6812, Table 5, but the relative error increases for deeper cracks.
E¼ 181 Gpa, r ¼ 7860 kg/m3.
On the other hand, for theoretical analysis, the comparison of
mode shapes of a damaged and intact beam can help detect the
crack location and its size. The vibration modes of the beam,
6 shown in Fig. 9, may be found using Eqs. (41), (42a) and (42b).
Intact Beam Fig. 10a–g shows the first four normalized mode shapes of the
Cracked Beam damaged beam for different crack depths and boundary condi-
4 tions. Exploring these figures, it can be seen that the vibration
mode shapes of the cracked beam depend on the number, depth
and position of cracks, number of steps and boundary conditions.
It is obvious that when the crack is located at a larger curvature
2
ln l T34.T43-T33.T44l

position, the deviations of mode shapes of intact and damaged

beams are more noticeable and for a specific crack location, the
higher crack depth results in the increasing of the distortion at
0
crack section.

-2 5. Inverse problem

As discussed above, the presence of cracks changes the

-4 characteristic equation and has influences on the mode shapes
and natural frequencies of the beam. In an inverse problem, crack
positions and corresponding depths are unknown parameters
-6 and natural frequencies of the beam can easily be determined
0 500 1000 1500 2000 2500 experimentally through various testing techniques. When there
(rad/s) are N open edge cracks in the damaged beam (N ¼N1 þN2 þ
?þNn), according to Eq. (35f) the locations and sizes of the
Fig. 6. Variation of characteristic equation with o for intact and cracked cracks may be obtained by 2N measured natural frequencies. In
cantilever beam (Example 1).
this way the crack detection procedure leads to a system of 2N
equations as
frequencies of the same but intact beam. As shown in these F i ¼ Fðoi ,a1,1 ,a1,2 ,. . .,a1,N1 ,. . .,an,1 ,an,2 ,. . .,
figures and concluded in many previous researches, the reduction an,Nn ,x1,1 ,x1,2 ,. . .,x1,N1 ,. . .,xn,1 ,xn,2 ,. . .,xn,Nn Þ ¼ 0 i ¼ 1,2,. . .,2N
of natural frequencies of a beam depends on the crack depth,
ð45Þ
crack position, boundary conditions of the beam and mode of
vibration. There are several techniques to solve nonlinear equations
systems. The commonest methods are the Newton-type techni-
Example 2. Free vibration of a square-sectional cantilever beam ques [31]. Some other techniques are also available, such as
having three cracks. Broyden [32], secant [33] and trust-region [34] method. To solve
this system of equations an appropriate numerical method can be
As the second example, a cantilever beam with three open
used. In this study the iterative Newton–Raphson algorithm
edge cracks, Fig. 8, is analyzed. The material properties have been
[17,35] is utilized as follows:
taken for the beam as E ¼210 Gpa, r ¼7860 kg/m3, n ¼0.29 and
Newton–Raphson algorithm:
the geometrical parameters for the beam are h¼0.02 m,
b¼0.02 m, L¼0.5 m. The normalized crack sizes for three cracks
 Calculate the overall transfer matrix from Eq. (31).
are the same as a1,1/h ¼a1,2/h¼a1,3/h¼0.1 and the lowest six
 Set k¼0.
natural frequencies of the beam for various crack locations are
 Choose appropriate initial assumptions of N crack depths
given in Table 4. Results obtained in Ref. [17] are also listed in
(a1,1,a1,2,y,a1,N1,y,an,1,an,2,y,an,Nn)0 and N crack locations
Table 4, demonstrating a good agreement with those obtained by
(x1,1,x1,2,y,x1,N1,y,xn,1,xn,2,y,xn,Nn)0.
using the procedure suggested in this article, like Example 1.
Repeat
Example 3. Free vibration of a stepped beam having three cracks.

Third example is a stepped beam with triple cracks, Fig. 9a.  Compute {F}k ¼{F1 F2?F2N}T from Eqs. (35a) and (45).
The material properties have been taken for the beam identical to  Evaluate the Jacobian matrix [J]k from Eq. (48).
 M. Attar / International Journal of Mechanical Sciences 57 (2012) 19–33 27

Fig. 7. Variation of the lowest three normalized natural frequencies versus normalized crack depth (a/h) and normalized crack position (xc/L) for a beam with single crack.
(a) Clamped–free, (b) pinned–pinned, (c) clamped–pinned, (d) clamped–clamped, and (e) free–free beam.

ðxs,t Þk þ 1 ¼ ðxs,t Þk þ pUdxs,t s ¼ 1,2,. . .,n t ¼ 1,2,. . .,Ns

 Solve the linear system
k ¼ kþ 1 ð47Þ
½Jk UfDg ¼ fFgk ð46Þ
where p is an underrelaxation parameter and can be used to avoid
where {D}¼{da1,1 da1,2? da1,N1?dan,1 dan,2? dan,Nn dx1,1
overshoots and increase reliability of Newton–Raphson method
dx1,2?dx1,N1?dxn,1 dxn,2 ?dxn,Nn}T.
especially in early steps of iterations.
 Set
Until convergence criteria are satisfied and elements of fDg
ðas,t Þk þ 1 ¼ ðas,t Þk þ pUdas,t s ¼ 1,2,. . .,n t ¼ 1,2,. . .,N s become adequately small.
28 M. Attar / International Journal of Mechanical Sciences 57 (2012) 19–33

The index k is an iteration index and the Jacobian matrix [J]k in crack size and location with the actual values. It is observed that the
each iteration can be computed as crack identification results are within 0.4% error for normalized crack

2 @F 1 @F 1 @F 1 @F 1 @F 1 @F 1 @F 1 @F 1 @F 1 @F 1 @F 1 @F 1 3
@a1,1 @a1,2  @a1,N1  @an,1 @an,2  @an,Nn @x1,1 @x1,2  @x1,N1  @xn,1 @xn,2  @xn,Nn
6 @F 7
 66
2 @F 2
 @F 2
 @F 2 @F 2
 @F 2 @F 2 @F 2
 @F 2
 @F 2 @F 2
 @F 2 7
7
jK 6 @a1,1 @a1,2 @a1,N1 @an,1 @an,2 @an,Nn @x1,1 @x1,2 @x1,N1 @xn,1 @xn,2 @xn,Nn
7 ð48Þ
6^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ 7
4 5
@F 2N @F 2N @F 2N @F 2N @F 2N @F 2N @F 2N @F 2N @F 2N @F 2N @F 2N @F 2N
@a1,1 @a1,2  @a1,N1  @an,2 @an,2  @an,Nn @x1,1 @x1,2  @x1,N1  @xn,1 @xn,2  @xn,Nn

where partial derivatives of F1,F2,y,F2N and the elements of the location and 5.09% for normalized crack depth, which are satisfactory
Jacobian matrix can be computed numerically as when compared to actual crack parameters. For cases 1, 2, 9 and 14,

@F i Fðoi ,a1,1 ,. . .,as,t þ e,. . .,an,Nn ,x1,1 ,. . .,xn,Nn ÞFðoi ,a1,1 ,. . .,as,t ,. . .,an,Nn ,x1,1 ,. . .,xn,Nn Þ
¼ 9e9 5 1 ð49aÞ
@as,t e
@F i Fðoi ,a1,1 ,. . .,an,Nn ,x1,1 ,. . .,xs,t þ e,. . .,xn,Nn ÞFðoi ,a1,1 ,. . .,an,Nn ,x1,1 ,. . .,xs,t ,. . .,xn,Nn Þ
¼ 9e9 51 ð49bÞ
@xs,t e

To verify the validity and accuracy of the strategy outlined the errors in the iterative detection procedure of crack size and
above, three cases are considered in which natural frequencies location are plotted in Fig. 11b and c. The rapid convergence of crack
computed by FEM are employed as measured input data.

Example 4. A simply supported beam with single crack. x3,1

x2 , 1
In this example a simply supported beam with two types of x1,1
cross-sections (circular and rectangular) and a crack is consid-
ered, as shown in Fig. 11a. The inverse problem is solved for KR1 KR2
fourteen cases by employing the first two natural frequencies m m
e I1 a 2 ,1 e
a 1 ,1 I 2 I3 a 3 ,1 x
calculated from finite element analysis as inputs. Case 1 has the J J
circular cross-section in which the input natural frequencies are K1 K2
taken from Ref. [36] and cases 2–14 concern the rectangular L1 L2 L3
cross-sectional beam and input data of them are obtained from y
Ref. [37]. The geometrical and material properties of the beam are L
D¼ 0.03 m, L¼ 1 m, E¼206 Gpa, r ¼ 7800 kg/m3, n ¼0.29 for case
1 and h¼0.02 m, b¼0.012 m, L¼0.5 m, E¼210 Gpa, r ¼7860 kg/m3,
n ¼0.3 for cases 2–14. Table 6 shows a comparison of the predicted

b
h
x 1, 3
x1, 2
x1 ,1

I a1,1 a1, 2 a1, 3 x

y L Fig. 9. (a) Stepped beam with circular cross-section and triple cracks. (b) Two
dimensional ANSYS model with singular elements at crack tip for finite element
Fig. 8. Uniform cantilever beam with rectangular cross-section and triple cracks. analysis.

Table 4
Natural frequencies of a cantilever beam with triple cracks.

Case Crack locations Methods Natural frequencies, os (rad/s)

x1,1/L x1,2/L x1,3/L o1 o2 o3 o4 o5 o6

1 0.2 0.4 0.6 FEM [17] 416.8933 2612.065 7323.879 14356.68 23589.91 35603.94
Present study 416.9159 2612.213 7324.210 14357.28 23592.02 35604.06

2 0.2 0.4 0.8 FEM [17] 417.0652 2620.375 7318.436 14299.97 23600.29 35573.62
Present study 417.0864 2620.455 7318.811 14301.02 23602.31 35574.00

3 0.2 0.6 0.8 FEM [17] 417.6291 2617.683 7315.436 14300.48 23601.47 35573.74
Present study 417.6464 2617.786 7315.833 14301.53 23603.48 35574.12

4 0.4 0.6 0.8 FEM [17] 418.7431 2610.199 7310.806 14337.70 23575.09 35597.99
Present study 418.7517 2610.361 7311.243 14338.46 23577.32 35598.16

L¼ 0.5 m, h ¼0.02 m, b ¼0.02 m, a1,1/h ¼ a1,2/h ¼a1,3/h ¼ 0.1, E ¼210 Gpa, r ¼7860 kg/m3.
 M. Attar / International Journal of Mechanical Sciences 57 (2012) 19–33 29

Table 5
Natural frequencies of intact and cracked stepped beam for various crack sizes and boundary conditions.

Case Boundary conditions Normalized Natural frequencies (rad/s)

crack sizes

Present method Finite element model (ANSYS) Error (%)

a1/h1 a2/h2 a3/h3 o1 o2 o3 o4 o1 o2 o3 o4 o1 o2 o3 o4

1 Clamped–free 0 0 0 446.497 3075.83 8004.68 16216.3 445.805 3054.13 7903.62 15951.1 0.155 0.711 1.28 1.66
2 0.2 0 0 441.353 3075.61 7973.79 16088.1 440.571 3053.88 7873.46 15824.8 0.177 0.712 1.27 1.66
3 0.2 0.2 0 440.698 3042.17 7973.54 15874.5 439.911 3020.52 7872.83 15615.0 0.179 0.717 1.28 1.66
4 0.2 0.2 0.2 440.664 3034.30 7904.25 15704.0 439.880 3012.60 7804.34 15446.6 0.178 0.720 1.28 1.67
5 0.4 0.4 0.4 421.910 2902.82 7587.65 14257.3 420.414 2878.26 7483.90 13993.3 0.356 0.853 1.39 1.89
6 0.6 0.6 0.6 375.719 2588.12 6839.14 11613.9 373.604 2562.91 6739.97 11380.1 0.566 0.984 1.47 2.05

7 Pinned–pinned 0 0 0 1367.19 5051.15 12057.0 21215.2 1360.12 5012.35 11924.9 20732.6 0.520 0.774 1.11 2.58
8 0.2 0 0 1360.51 5001.22 11963.9 21164.3 1353.40 4962.65 11832.5 20682.2 0.525 0.777 1.11 2.33
9 0.2 0.2 0 1347.96 5001.21 11805.6 21164.3 1341.02 4962.59 11674.8 20681.7 0.517 0.778 1.12 2.33
10 0.2 0.2 0.2 1341.57 4952.02 11713.0 21112.5 1334.67 4913.14 11582.4 20629.0 0.517 0.791 1.12 2.34
11 0.4 0.4 0.4 1262.08 4645.14 10697.6 20811.4 1260.72 4599.61 10555.8 20312.3 0.108 0.990 1.34 2.46
12 0.6 0.6 0.6 1082.22 3956.31 8655.43 20811.4 1080.21 3909.59 8547.02 19666.4 0.187 1.19 1.27 5.82

13 Clamped–pinned 0 0 0 2024.98 6446.03 14041.4 23980.8 2014.26 6378.69 13832.4 23339.5 0.532 1.06 1.51 2.75
14 0.2 0 0 2024.29 6428.30 13939.8 23814.1 2013.57 6361.10 13733.8 23179.9 0.532 1.06 1.50 2.74
15 0.2 0.2 0 2009.49 6418.56 13777.2 23770.4 1998.74 6352.30 13571.7 23137.8 0.538 1.04 1.51 2.73
16 0.2 0.2 0.2 1996.55 6356.35 13685.1 23738.4 1985.93 6290.10 13479.9 23105.2 0.535 1.05 1.52 2.74
17 0.4 0.4 0.4 1906.53 6074.91 12648.1 23075.8 1898.15 6004.78 12438.8 22437.9 0.441 1.17 1.68 2.84
18 0.6 0.6 0.6 1691.43 5413.83 10666.0 21941.8 1679.31 5347.18 10503.6 21244.1 0.722 1.25 1.55 3.28

19 Clamped–clamped 0 0 0 2906.00 7937.52 16254.5 27041.4 2887.00 7826.96 15956.8 26165.7 0.658 1.41 1.87 3.35
20 0.2 0 0 2906.00 7905.81 16125.9 26877.1 2887.00 7796.24 15832.4 26011.8 0.658 1.40 1.85 3.33
21 0.2 0.2 0 2881.45 7905.81 15909.5 26876.9 2862.68 7796.18 15618.7 26011.1 0.656 1.41 1.86 3.33
22 0.2 0.2 0.2 2881.45 7874.67 15785.5 26711.0 2862.62 7765.39 15498.7 25852.8 0.658 1.41 1.85 3.32
23 0.4 0.4 0.4 2802.14 7682.34 14466.7 25781.3 2784.33 7572.49 14188.7 24928.5 0.640 1.45 1.96 3.42
24 0.6 0.6 0.6 2600.52 7267.50 12077.8 24129.4 2598.79 7163.46 11873.3 23181.2 0.067 1.45 1.72 4.09

25 Free-free 0 0 0 3321.35 8060.94 16178.6 26992.6 3298.36 7968.34 15945.5 26267.5 0.697 1.16 1.46 2.76
26 0.2 0 0 3311.40 7992.30 16003.3 26795.5 3288.37 7900.60 15772.1 26075.2 0.700 1.16 1.47 2.76
27 0.2 0.2 0 3270.18 7992.27 15792.8 26795.3 3247.02 7900.48 15563.5 26074.6 0.713 1.16 1.47 2.76
28 0.2 0.2 0.2 3260.76 7925.22 15622.9 26595.6 3237.6 7833.88 15395.7 25877.9 0.716 1.17 1.48 2.77
29 0.4 0.4 0.4 3072.02 7494.02 14044.9 25487.2 3044.27 7395.94 13802.9 24758.9 0.912 1.33 1.75 2.94
30 0.6 0.6 0.6 2641.32 6473.28 11089.5 23559.1 2612.8 6378.69 10877.5 22811.7 1.09 1.48 1.95 3.28

31 Clamped–free with a 0 0 0 205.286 2158.62 6560.36 14163.2 204.712 2140.75 6472.26 13905.8 0.280 0.835 1.36 1.85
32 mass m¼ rA1L attached 0.2 0 0 202.994 2158.21 6541.87 14060.4 202.423 2140.25 6450.26 13802.1 0.282 0.839 1.42 1.87
33 at the free end 0.2 0.2 0 202.496 2141.22 6533.69 13894.5 201.921 2123.39 6441.60 13636.6 0.285 0.840 1.43 1.89
34 0.2 0.2 0.2 202.393 2128.13 6469.15 13796.8 201.817 2110.37 6377.66 13538.5 0.285 0.841 1.43 1.91
35 0.4 0.4 0.4 193.154 2031.64 6182.81 12733.6 192.601 2014.61 6095.26 12493.2 0.287 0.845 1.44 1.92
36 0.6 0.6 0.6 170.773 1801.52 5509.96 10712.0 170.285 1786.47 5409.07 10414.6 0.287 0.842 1.86 2.85

37 Free-free with masses, 0 0 0 60.8960 289.970 2066.32 5782.59 60.6347 287.169 2042.49 5637.22 0.431 0.975 1.17 2.58
38 linear and rotational 0.2 0 0 60.8959 289.722 2060.31 5731.39 60.6342 286.961 2035.05 5581.63 0.432 0.962 1.24 2.68
39 springs attached at 0.2 0.2 0 60.8956 289.722 2042.04 5731.37 60.6197 286.939 2014.46 5581.14 0.455 0.970 1.37 2.69
40 the free ends (m1 ¼ m2 ¼ 0.2 0.2 0.2 60.8954 289.473 2036.36 5681.07 60.6121 286.645 2008.45 5528.11 0.467 0.986 1.39 2.77
41 0.5rA1L, K1 ¼ K2 ¼ EI1/L3, 0.4 0.4 0.4 60.8934 287.755 1944.05 5367.07 60.5791 284.329 1913.4 5205.97 0.519 1.20 1.60 3.09
42 KR1 ¼ KR2 ¼ EI1/L) 0.6 0.6 0.6 60.8878 282.579 1739.40 4666.52 60.5193 278.288 1706.61 4512.42 0.609 1.54 1.92 3.41

L¼ 0.3 m, L1 ¼ L2 ¼ L3 ¼ 0.1 m, h1 ¼ h3 ¼0.0075 m, h2 ¼0.009375 m, x1,1/L¼ 0.2, x2,1/L¼ 0.5, x3,1/L ¼0.8, E¼ 210 Gpa, r ¼7860 kg/m3.

parameters in the first 50 iterations shows the capability of the errors in the computed values are greater than errors in Example 4,
proposed method. Note that for cases involving symmetry, the but estimated depths and positions are adequately reliable.
developed method cannot estimate a unique value for crack location.
In other words, if the initial guess for normalized crack location in Example 6. A cantilever beam with triple cracks.
case 2 of Table 6 is less than 0.5, the algorithm converges to 0.1. But if
the initial guess is between 0.5 and 1, the estimated crack location The cantilever beam with three open edge cracks is considered
will be 0.9. identical to Example 2. The values of first six natural frequencies
obtained in Ref. [17] were depicted in Table 4 for various four
Example 5. A stepped beam with double cracks. cases. Present crack detection approach is again used (considering
In order to verify and demonstrate the accuracy and applicability p¼0.1) and the identification of crack parameters is shown in
of the present technique for prediction of crack parameters in stepped Table 8. The calculated normalized crack position and depth are in
beams, the inverse problem is solved for four various cases presented good agreement with actual crack details, which is promising and
in Table 5. The crack detection procedure is utilized for cases of shows the accuracy of the proposed approach in identification of
clamped–free, pinned–pinned, clamped– pinned, and clamped– multiple cracks.
clamped stepped beam. The results obtained from finite element It is worth remarking that suitable selection of initial guesses
model of the beam are used as input natural frequencies. The actual for crack parameters in the inverse procedure plays an important
and predicted crack parameters calculated from identification algo- role in convergence of the scheme. When there are multiple
rithm are listed in Table 7 for different boundary conditions. The cracks in the beam, the convergence of the solution is more
30 M. Attar / International Journal of Mechanical Sciences 57 (2012) 19–33

Fig. 10. First four normalized mode shapes of the stepped cracked beam (L¼0.3 m, L1 ¼L2 ¼L3 ¼0.1 m, D1 ¼D3 ¼0.015 m, D2 ¼ 0.03 m) with triple cracks located at x1,1/
L¼ 0.2, x2,1/L¼0.5, x3,1/L¼ 0.8 as the crack depths change. —; ai/Di ¼ 0, —  —  —; ai/Di ¼ 0.2, — — —; ai/Di ¼ 0.4, —  —  —; ai/Di ¼0.6, i¼ 1, 2, 3. (a) Clamped–free,
(b) pinned–pinned, (c) clamped–pinned, (d) clamped–clamped, (e) free–free, (f) clamped–free with a mass m¼ rA1L attached at the free end, and (g) free–free with masses
m1 ¼ m2 ¼0.5rA1L, linear springs K1 ¼K2 ¼EI1/L3 and rotational springs KR1 ¼ KR2 ¼ EI1/L attached at the free ends.

dependent upon the appropriate choice of the initial guesses, contaminated input data on the estimated crack parameters,
especially for crack locations. normally distributed random noises are generated and introduced
In practice, measured frequencies of the beam, which are used into the natural frequencies of Table 4 as follows:
as input data in the inverse problem, are always influenced by
measurement noises. In order to study the effect of noise- oci ¼ oi ð1 þ s  randnÞi ¼ 1,2,    ,6 ð50Þ
 M. Attar / International Journal of Mechanical Sciences 57 (2012) 19–33 31

Case 1 Cases 2 14
b

Cross section : D h

xc

a
I
x

y L

200 150

Error in xc/L (%)

Error in a/h (%)

Case 1 Case 1
150 Case 2 100 Case 2
Case 9 Case 9
100 Case 14 Case 14
50
50
0 0
0 10 20 30 40 50 0 10 20 30 40 50
Iterations Iterations

Fig. 11. (a) Uniform simply supported beam with a transverse open crack (Example 4). (b) Variation of the relative error in the iterative detection procedure of crack
extent. (c) Variation of the relative error in the iterative detection procedure of crack location.

Table 6
Crack identification of a simply supported beam with a transverse open crack.

Case Input data (rad/s) Normalized crack location (xc/L) Normalized crack depth (a/D or a/h)

o1 o2 Actual Initial guess Predicted Error (%) Actual Initial guess Predicted Error (%)

1 351.356 1521.66 0.5 0.25 0.50149 0.298 0.5 0.3 0.49012 1.976
2 1177.66 4705.604 0.1 0.25 0.10040 0.4 0.1 0.3 0.10005 0.05
3 1176.402 4694.446 0.2 0.45 0.20018 0.09 0.1 0.3 0.10039 0.39
4 1162.672 4557.815 0.2 0.45 0.20002 0.01 0.3 0.25 0.30344 1.15
5 1165.559 4644.925 0.3 0.45 0.30001 0.003 0.2 0.25 0.20158 0.79
6 1149.363 4563.174 0.3 0.45 0.30001 0.003 0.3 0.25 0.30344 1.15
7 1160.876 4686.538 0.4 0.45 0.40002 0.005 0.2 0.25 0.20159 0.8
8 1105.78 4608.053 0.4 0.45 0.40001 0.0025 0.4 0.25 0.40210 0.525
9 1159.108 4712.566 0.5 0.35 0.49946 0.788 0.2 0.45 0.20158 0.79
10 1099.58 4712.566 0.5 0.35 0.50018 0.036 0.4 0.15 0.40121 0.303
11 1003.482 4479.635 0.6 0.75 0.59999 0.002 0.6 0.35 0.56940 5.1
12 1124.204 4446.84 0.7 0.85 0.7 0 0.4 0.35 0.40235 0.588
13 1042.566 4140.357 0.7 0.55 0.7 0 0.6 0.45 0.56944 5.093
14 1100.028 4060.228 0.8 0.55 0.79999 0.0013 0.6 0.45 0.56944 5.093

D ¼ 0.03 m, L ¼1 m, E¼ 206 Gpa, r ¼ 7800 kg/m3, n ¼ 0.29 for circular cross-section (case 1) and h ¼ 0.02 m, b¼ 0.012 m, L¼0.5 m, E ¼210 Gpa, r ¼7860 kg/m3, n ¼0.3 for
rectangular cross-section (cases 2–14).

here oci is the noise-contaminated natural frequency and randn is matrix involving matrices with dimension of 4  4, which is much
the random noise generator function in MATLAB with a zero mean easier than handling big matrices in finite element or other forms
and a standard deviation of s. The noise level is controlled by the of analysis. Then, this overall transfer matrix is used in the inverse
standard deviation and inverse problem is solved for three procedure to estimate the location and depth of multiple cracks,
different values of s. The generated random numbers are added not needing the direct problem to be solved many times.
to natural frequencies of case 3 and 4 in Table 4 and crack Furthermore, a non-uniform beam can be investigated by using
identification results are given in Table 9. According to this table, the proposed method and approximated with a multi-stepped
present method provides a satisfactory estimate with a maximum beam. The present method can also be extended to tackle direct
relative error of 9.08% and 4.88% for normalized crack size and and inverse problems for short beams when the transfer matrices
location, respectively. It is also observed that the reliability of the are modified based on the Timoshenko beam theory.
present method is influenced by the level of noise pollution.
Consequently, it is worth mentioning that the iterative procedure
might not converge properly for a high level of s. 6. Conclusions
The main feature of the present damage identification
approach is reducing required computation time by employing In this study, a transfer matrix method is presented to
the transfer matrix method combined with the Newton–Raphson investigate the free transverse vibrations of stepped beams with
iteration procedure. To solve the problem of detection of the multiple edge cracks and solve the inverse problem of the cracks
cracked beam, one just needs to compute the overall transfer identification. The Euler–Bernoulli beam theory is employed to
32 M. Attar / International Journal of Mechanical Sciences 57 (2012) 19–33

Table 7
Crack identification of the beam with double cracks presented in Table 5.

Case no. in Table 5 Boundary conditions Normalized crack location (xc/L) Normalized crack depth (a/h)

Actual Initial guess Predicted Error (%) Actual Initial guess Predicted Error (%)

3 Clamped–free 0.2 0.1 0.20437 2.19 0.2 0.35 0.20766 3.83

0.5 0.7 0.51439 2.88 0.2 0.35 0.21163 5.82

9 Pinned–pinned 0.2 0.3 0.21156 5.78 0.2 0.35 0.21808 9.04

0.5 0.3 0.52239 4.48 0.2 0.3 0.20043 0.22

15 Clamped–pinned 0.2 0.1 0.20911 4.58 0.2 0.35 0.20692 3.46

0.5 0.7 0.51895 3.79 0.2 0.3 0.21507 7.53

21 Clamped–clamped 0.2 0.1 0.21733 8.66 0.2 0.35 0.21320 6.60

0.5 0.3 0.53199 6.40 0.2 0.35 0.20882 4.41

Table 8
Crack identification of a cantilever beam with triple open cracks.

Case Normalized crack location (xc/L) Normalized crack depth (a/h)

Actual Initial guess Predicted Error (%) Actual Initial guess Predicted Error (%)

1 0.2 0.25 0.20006 0.030 0.1 0.05 0.10037 0.370

0.4 0.5 0.39990 0.025 0.1 0.3 0.10055 0.550
0.6 0.7 0.59987 0.022 0.1 0.2 0.10035 0.350

2 0.2 0.25 0.20016 0.080 0.1 0.3 0.10044 0.440

0.4 0.5 0.40001 0.003 0.1 0.3 0.10048 0.480
0.8 0.75 0.80008 0.010 0.1 0.3 0.10037 0.370

3 0.2 0.1 0.20011 0.055 0.1 0.3 0.10046 0.460

0.6 0.7 0.59998 0.003 0.1 0.3 0.10046 0.460
0.8 0.9 0.80005 0.006 0.1 0.3 0.10037 0.370

4 0.4 0.25 0.39959 0.103 0.1 0.3 0.10025 0.250

0.6 0.5 0.59960 0.066 0.1 0.3 0.10066 0.660
0.8 0.75 0.79997 0.040 0.1 0.3 0.10036 0.360

L¼ 0.5 m, h ¼0.02 m, b ¼0.02 m, a1/h ¼a2/h ¼ a3/h ¼ 0.1, E ¼210 Gpa, r ¼7860 kg/m3.

Table 9
Crack identification of two cases presented in Table 4 with noise-contaminated input data.

Case no. in Table 4 Noise level (s) Normalized crack location (xc/L) Normalized crack depth (a/h)

Actual Initial guess Predicted Error (%) Actual Initial guess Predicted Error (%)

3 0.0001 0.2 0.1 0.19497 2.51 0.1 0.3 0.09835 1.65

0.6 0.7 0.59955 0.074 0.1 0.3 0.09984 0.157
0.8 0.9 0.79722 0.347 0.1 0.3 0.10251 2.51

3 0.0002 0.2 0.1 0.19025 4.87 0.1 0.3 0.10090 0.903

0.6 0.7 0.59958 0.070 0.1 0.3 0.09742 2.58
0.8 0.9 0.78809 1.49 0.1 0.3 0.10507 5.07

3 0.0003 0.2 0.1 0.19777 1.11 0.1 0.3 0.10496 4.96

0.6 0.7 0.60034 0.057 0.1 0.3 0.09575 4.25
0.8 0.9 0.79537 0.579 0.1 0.3 0.10049 0.487

4 0.0001 0.4 0.25 0.40170 0.426 0.1 0.3 0.09835 1.65

0.6 0.5 0.59960 0.066 0.1 0.3 0.10136 1.36
0.8 0.75 0.80031 0.038 0.1 0.3 0.10185 1.85

4 0.0002 0.4 0.25 0.39123 2.19 0.1 0.3 0.09526 4.74

0.6 0.5 0.59429 0.952 0.1 0.3 0.10389 3.89
0.8 0.75 0.79711 0.361 0.1 0.3 0.10251 2.51

4 0.0003 0.4 0.25 0.38139 4.88 0.1 0.3 0.09092 9.08

0.6 0.5 0.59314 1.14 0.1 0.3 0.10775 7.75
0.8 0.75 0.79934 0.082 0.1 0.3 0.10287 2.87

L¼ 0.5 m, h ¼0.02 m, b ¼0.02 m, a1/h ¼a2/h ¼ a3/h ¼ 0.1, E ¼210 Gpa, r ¼7860 kg/m3.
 M. Attar / International Journal of Mechanical Sciences 57 (2012) 19–33 33

model the stepped cracked beam as an assembly of uniform sub- [10] Narkis Y, Elmalah E. Crack identification in a cantilever beam under uncertain
segments connected by massless rotational springs. The general end conditions. Int J Mech Sci 1996;38(6):499–507.
[11] Zheng DY, Fan SC. Vibration and stability of cracked hollow-sectional beams.
form of characteristic equation for the cracked beam is obtained,
J Sound Vib 2003;267(4):933–54.
which is a function of natural frequencies, the locations and sizes [12] Lele SP, Maiti SK. Modeling of transverse vibration of short beams for crack
of the cracks, boundary conditions, geometrical and physical detection and measurement of crack extension. J Sound Vib 2002;257:
parameters of the beam. Numerical examples are provided to 559–83.
[13] Loya JA, Rubio L, Fernandez-Saez J. Natural frequencies for bending vibrations
confirm the direct problem solution for various cracked beams.
of Timoshenko cracked beams. J Sound Vib 2006;29:640–53.
The predicted natural frequencies by the proposed approach [14] Shifrin EI, Ruotolo R. Natural frequencies of a beam with an arbitrary number
agree well with those obtained from experiment and FEM. Results of cracks. J Sound Vib 1999;222:409–23.
also demonstrate that vibration mode shapes and reduction of [15] Morassi A. Crack-induced changes in eigenfrequeices of beam structures.
J Eng Mech 1993;119:1768–803.
natural frequencies of the cracked beam depend on the number, [16] Liang RY, Choy FK, Hu J. Detection of cracks in beam structures using
depth and position of the cracks, number of steps and boundary measurements of natural frequencies. J Franklin Inst 1991;328(4):505–18.
conditions. The transfer matrix method is then utilized to provide [17] Lee J. Identification of multiple cracks in a beam using natural frequencies.
a simple and efficient technique to identify the locations and sizes J Sound Vib 2009;320:482–90.
[18] Ratcliffe CP. Damage detection using a modified Laplacian operator on mode
of N cracks by exploiting 2N measured natural frequencies as shape data. J Sound Vib 1997;204:505–17.
input data. A system of 2N equations is solved by an appropriate [19] Ruotolo R, Surace C. Damage assessment of multiple cracked beams:
numerical method to identify 2N unknown parameters of the numerical results and experimental validation. J Sound Vib 1997;206(4):
cracks. The main advantage of the present damage identification 567–88.
[20] Lee J. Identification of multiple cracks in a beam using vibration amplitudes.
procedure is reducing computation time in comparison to other J Sound Vib 2009;326:205–12.
methods. The predicted crack depths and positions in numerical [21] Gladwell GML. Inverse problems in vibration. Dordrecht, The Netherlands:
examples show that the present approach is theoretically capable Kluwer Academic Publishers; 2004.
of estimating crack parameters with high performance. [22] Jang SK, Bert CW. Free vibration of stepped beams: higher mode frequencies
and effects of steps on frequency. J Sound Vib 1989;132:164–8.
[23] Naguleswaran S. Vibration and stability of an Euler–Bernoulli beam with up
to three-step changes in cross-section and in axial force. Int J Mech Sci
Acknowledgments 2003;45:1563–79.
[24] Pestel EC, Leckie FA. Matrix methods in elastomechanics. London: McGraw-
The author gratefully acknowledges Professor Elena Pasternak, Hill; 1983.
[25] Takahashi I. Vibration and stability of non-uniform cracked Timoshenko
School of Mechanical Engineering, The University of Western
beam subjected to follower force. Comput Struct 1999;71:585–91.
Australia, for her helpful comments and suggestions. [26] Ruotolo R, Surace C. Natural frequencies of a bar with multiple cracks.
J Sound Vib 2004;272:301–16.
[27] Bilello C, Bergman LA. Vibration of damaged beams under a moving mass:
References
theory and experimental validation. J Sound Vib 2004;274:567–82.
[28] Liu WH, Huang CC. Vibrations of a constrained beam carrying a heavy tip
[1] Dimarogonas AD. Vibration of cracked structures: a state of the art review. body. J Sound Vib 1988;123(1):15–29.
Eng Fract Mech 1996;55(5):831–57. [29] Aleyaasin M, Ebrahimi M, Whalley R. Vibration analysis of distributed-
[2] Gasch R. A survey of the dynamic behaviour of a simple rotating shaft with a lumped rotor systems. Comput Methods Appl Mech Eng 2000;189:545–58.
transverse crack. J Sound Vib 1993;160(2):313–32. [30] Tada H, Paris PC, Irwin GR. The stress analysis of cracks handbook. New York,
[3] Dimarogonas AD, Paipetis SA. Analytical methods in rotor dynamics. London: USA: ASME Press; 2000.
Applied Science Publisher; 1983. [31] Press WH, Teukolsky SA, Vetterling WT, Flannery BP. Numerical recipes in
[4] Papadopoulos CA, Dimarogonas AD. Coupled longitudinal and bending fortran. 2nd ed. Cambridge: Cambridge University Press; 1992.
vibrations of a rotating shaft with an open crack. J Sound Vib 1987;117(1): [32] Broyden CG. A class of methods for solving nonlinear simultaneous equa-
81–93.
tions. Math Comput 1965;19(92):577–93.
[5] Rizos PF, Aspragathos N, Dimarogonas AD. Identification of crack location and
[33] Denis JE, Wolkowicz H. Least change secant methods, sizing, and shifting.
magnitude in a cantilever beam from the vibration modes. J Sound Vib
SIAM J Numer Anal 1993;30:1291–314.
1990;138:381–8.
[34] Conn AR, Gould NIM, Toint PL. Trust-region methods. Philadelphia, PA: SIAM;
[6] Shen MHH, Pierre C. Natural modes of Bernoulli-Euler beams with symmetric
2000.
cracks. J Sound Vib 1990;138:115–34.
[7] Nandwana BP, Maiti SK. Detection of the location and size of a crack in [35] Grosan C, Abraham A. A new approach for solving nonlinear equations
stepped cantilever beams based on measurements of natural frequencies. systems, IEEE Transactions on systems. Man Cybern—A Syst Hum
J Sound Vib 1997;203:435–46. 2008;38(3):698–714.
[8] Fernandez-Saez J, Rubio L, Navarro C. Approximate calculation of the [36] Zheng DY, Kessissoglou NJ. Free vibration analysis of a cracked beam by finite
fundamental frequency for bending vibrations of cracked beams. J Sound element method. J Sound Vib 2004;273:457–75.
Vib 1999;225:345–52. [37] Xiang JW, Chen XF, Li B, He YM, He ZJ. Identification of a crack in a beam
[9] Fernandez-Saez J, Navarro C. Fundamental frequency of cracked beams in based on the finite element method of a B-spline wavelet on the interval.
bending vibrations: an analytical approach. J Sound Vib 2002;256:17–31. J Sound Vib 2006;296:1046–52.