Sunteți pe pagina 1din 17

Mechanical Systems

and
Signal Processing
Mechanical Systems and Signal Processing 23 (2009) 415431
Dynamic condition assessment of a cracked beam with the
composite element model
Z.R. Lu
a
, S.S. Law
b,
a
School of Engineering, Sun Yat-sen University, Guangzhou, Guangdong Province, Peoples Republic of China
b
Civil and Structural Engineering Department, Hong Kong Polytechnic University, Yuk Choi Road, Hunghom, Kowloon, Hong Kong
Received 5 September 2007; received in revised form 16 February 2008; accepted 24 February 2008
Available online 29 February 2008
Abstract
Existing models of local damage in a beam element are usually formulated as a damage in a single element,
and the coupling effect between adjacent damages is simply ignored. This coupling effect is larger in the case of a ne
mesh of nite elements or when there is a high density of damage in the structure. This paper studies such effect from
multiple cracks in a nite element in the dynamic analysis and local damage identication. The nite beam element is
formulated using the composite element method [P. Zeng, Composite element method for vibration analysis of structure,
Journal of Sound and Vibration 218 (1998) 619696] with a one-memberone-element conguration with cracks where the
interaction effect between cracks in the same element is automatically included. The accuracy and convergence speed of
the proposed model in computation are compared with existing models and experimental results. The parameter of the
Christides and Barr [One dimensional theory of cracked BernoulliEuler beams, International Journal of Mechanical
Science 26 (1984) 639648] crack model is found needing adjustment with the use of the proposed model. The response
sensitivity-based approach of damage identication is then applied in the identication of single and multiple crack
damages with both simulated and experimental data. Results obtained are found very accurate even under noisy
environment.
r 2008 Elsevier Ltd. All rights reserved.
Keywords: Composite element; Crack; Response sensitivity; Coupling; Damage; Identication
1. Introduction
The effect of cracks on the dynamic behavior of structures has been a subject of active research in
mechanical, aeronautic and civil engineering for decades. Numerous crack models for a cracked beam
can be found in the literature. The simplest one is a reduced stiffness (or increased exibility) in a nite
element to simulate a small crack in the element [35]. Another simple approach is to divide the cracked
beam into two-beam segments joined by a rotational spring that represents the cracked section [68].
This spring hinge model, combined with fracture mechanics, has been most popular amongst researchers.
An improved version of this model [9] leads to a closed-form solution giving the natural frequencies and
ARTICLE IN PRESS
www.elsevier.com/locate/jnlabr/ymssp
0888-3270/$ - see front matter r 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ymssp.2008.02.009

Corresponding author. Tel.: +852 2766 6062; fax: +852 2334 6389.
E-mail address: cesslaw@polyu.edu.hk (S.S. Law).
mode shapes of the cracked beam directly. Other researchers [1012] solved the differential equations with
compatible boundary conditions satisfying the crack conditions. Also Krawczuk and Ostachowicz [13] and
Lee and Chung [14] have developed the exibility matrix for a beam element with a crack using the energy
method.
The above models do not give a clear description on the crack and its effect on its adjacent material.
They are either localized at a cross-section or averaged over an entire nite element. This problem was
studied by Christides and Barr [2] who developed the one-dimensional vibration theory for the lateral
vibration of a cracked EulerBernoulli beam with one or more pairs of symmetric cracks. The perturbation in
the stress induced by the crack is incorporated through a local function, which has assumed an exponential
decay with distance from the crack, and the parameter involved in the function was determined from
experiments. This cracked beam vibration theory was used for the prediction of the fundamental natural
frequency of a simply supported beam with transverse open cracks in the middle of the beam. The natural
frequency of the cracked beam is obtained by adopting a two-term trial function in the RayleighRitz method.
Shen and Pierre [15] reconsidered the model of Christides and Barr through a Galerkin procedure [16] in which
the deection of the cracked beam was expanded in a series of functions. However, the convergence is very
slow for this type of problem, and a technique is proposed to increase the convergence speed by adding a
supplementary function to a classical set of suitable innitely differentiable co-ordinate functions. Later,
Messina [17] revisited the Galerkin procedure using the global piece-wise smooth functions. It is reported that
the convergence speed of Galerkin procedure can be more than doubled as compared with that based on the
SP technique [15].
However, the coupling effect of adjacent cracks/damages has not been included in the above
studies, and this problem would be very acute with many closely spaced cracks or when the member is
modeled with a ne mesh of nite elements. Since existing nite element conguration cannot accommodate
the interaction zones of individual crack with its discretized element, an open crack model [2] is
included in a composite element [1] in the one-memberone-element nite element model of the structure.
The element shape function is retained to dene displacement at any location of the element. The interaction
effect between cracks in the element is automatically included. Free vibration analysis is conducted
on the cracked beam model and results are compared with those from other existing methods and from
experiments. The dynamic response sensitivities with respect to the crack parameters are used in the
sensitivity-based model updating analysis for the condition assessment. Results from two numerical
simulations and experimental study show that the proposed method is accurate in identifying the cracks even
under noisy condition.
2. Theory
Fig. 1 shows a simple beam with multiple transverse open cracks along its length. The cracks are assumed to
have uniform depth across the width of the beam, and they do not change the mass of the beam. The model on
the crack is briey introduced before it is incorporated into the composite element to obtain the governing
equation for the vibration analysis of the cracked beam.
ARTICLE IN PRESS
1
x
1
x
p
x
j
h
c1
h
cp
h
cj
w
d
x
L
Fig. 1. Beam with multiple cracks.
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 416
2.1. Crack model
According to Christides and Barr [2], the variation of bending stiffness EI along the beam length takes up
the form of
EIx
EI
0
1 m 1 exp2ajx x
c
j=d
(1)
where E is Youngs modulus of the beam, I
0
wd
3
/12 is the second moment of area of the intact beam,
m 1/(1C
r
)
3
, C
r
d
c
/d is the crack depth ratio and d
c
and d are the depth of crack and the beam,
respectively, x
c
is the location of the crack. a is a constant, which governs the rate of decay and it varies with
the crack geometry and material type. Fig. 2 shows the variations of the bending stiffness along the beam
corresponding to different a values. However, Christides and Barr reported from their experiments that this
constant can be taken as 0.667 in general for different crack congurations. This parameter will be studied
with the proposed method for beams with different crack depths and material types.
2.2. Composite element incorporating a crack
Composite element [1] is a relatively new tool for nite element modeling. This method is basically a
combination of the conventional nite element method (FEM) and the highly precise classical theory (CT). In
the composite element method (CEM), the displacement eld is enlarged to the sum of the nite element
displacement and the shape functions from the classical theory. The displacement eld in the FEM satises the
nodal boundary conditions and the analytical functions are obtained from the CT observing also some special
boundary conditions. The enlarged displacement eld can be expressed as
u
CEM
x; t u
FEM
x; t u
CT
x; t (2)
where u
FEM
(x,t) and u
CT
(x,t) are the two parts of the CEM displacement eld with the subscripts dening
those of the FEM and CT, respectively.
Taking a planar beam element as an example, the rst part of the enlarged displacement eld can be
expressed as the product of the shape function vector N(x) and the nodal displacement vector q,
u
FEM
x; t Nxqt (3)
where q(t) [n
1
(t),y
1
(t),n
2
(t),y
2
(t)]
T
and n and y represent the transverse and rotational displacements,
respectively, and
Nx 1 3x=L
2
2x=L
3
; x=L 2x=L
2
x=L
3
; 3x=L
2
2x=L
3
; x=L
3
x=L
2

N
1
x; N
2
x; N
3
x; N
4
x (4)
The second part u
CT
(x,t) is obtained by the multiplication of analytical mode shapes with a vector of
coefcient c ( also called the c degrees-of-freedom or c-coordinates),
u
CT
x; t f
1
xc
1
t f
2
xc
2
t f
N
xc
N
t (5)
ARTICLE IN PRESS
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
0
50
100
150
Location along the beam
E
I
= 0.667
= 1.426
Fig. 2. Variation in EI with different a values.
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 417
where f
r
(r 1,2,y,N) are the analytical vibration shape functions of the beam. The c-coordinates physically
denote the contribution of each mode shape function to the total displacement u(x,t), where c
i
(t) denotes the
contribution of f
i
(x) in the total displacement u(x,t).
Like the FEM, the CEM can be rened using h-renement technique [1] by increasing the number of nite
elements. Moreover, it can also be rened through the c-renement method [1], by increasing the number of
analytical functions in the shape functions.
We can make use of the advantage of the c-renement from the CEM. The beam only needs to be
discretized into one element. This will reduce the total number of degrees of freedom in the nite element
model. The reduction of stiffness due to all the cracks is usually assumed to fall into one single element [18], or
it will be difcult to calculate its distribution along the beam. Such an assumption is not necessary in this work
because the whole beam is represented physically by a single nite element, and the distribution of the bending
stiffness can be obtained from Eq. (1). For a beam with multiple cracks, the bending stiffness distribution
along the beam can be obtained by superposition, and thus including automatically the interaction effect of
adjacent crack damages.
The displacement eld of the CEM for a uniform EulerBernoulli beam element can be written from
Eqs. (24) as
u
CEM
x; t SxQt (6)
where S(x) [N
1
(x),N
2
(x),N
3
(x),N
4
(x),f
1
(x),f
2
(x),y,f
N
(x)] is the generalized shape function of the CEM,
and Q(t) [n
1
(t),y
1
(t),n
2
(t),y
2
(t),c
1
(t),c
2
(t),y,c
N
(t)]
T
is the vector of generalized displacements, and N is the
number of shape functions used from the classical theory.
The elemental stiffness matrix of the cracked beam can be obtained from the following equation:
K
e

Z
L
0
d
2
S
T
dx
2
EIx
d
2
S
dx
2
dx
k
qq
k
qc

k
cq
k
cc

" #
(7)
where the submatrix [k
qq
] corresponds to the element stiffness matrix from the FEM for the cracked beam; the
submatrix [k
qc
] corresponds to the coupling terms of the q-dofs and the c-dofs; submatrix [k
cq
] is a transpose
matrix of [k
qc
], and the submatrix [k
cc
] corresponds to the c-dofs and is a diagonal matrix. If N terms are used
in the shape functions, the dimension of the elemental stiffness matrix is N+4. The stiffness matrix is obtained
from Eq. (7) after obtaining the generalized strain displacement matrix d
2
S/dx
2
from the generalized shape
function S(x). The elements in the submatrices [k
qq
],[k
qc
] and [k
cc
] are shown in Appendix A.
The elemental consistent mass matrix can be expressed as
M
e

Z
L
0
Sx
T
rASx dx
m
qq
m
qc

m
cq
m
cc

" #
(8)
where r is the mass density of the beam, A is the area of the cross-section. The submatrix [m
qq
] corresponds to
the element mass matrix from the FEM for the cracked beam; the submatrix [m
qc
] corresponds to the coupling
terms of the q-dofs and the c-dofs; submatrix [m
cq
] is a transpose matrix of [m
qc
], and the submatrix [m
cc
]
corresponds to the c-dofs and is a diagonal matrix. The dimension of the elemental mass matrix is also N+4.
The mass matrix can be obtained in a similar way as for the stiffness matrix. The elements in the submatrices
[m
qq
],[m
qc
] and [m
cc
] are shown in Appendix B.
After introducing the boundary conditions, the governing equation for free vibration of the cracked beam
can be expressed as
K o
2
MQ 0 (9)
where K and M are system stiffness and mass matrix, respectively, o is the circular frequency, from which and
the natural frequencies and mode shapes of the cracked beam can be obtained. The rst N eigenpairs are the
same as the rst N analytical eigenpairs, while the remaining four sets of eigenpairs are associated with the
unrestrained q-dofs at the two ends of the element.
ARTICLE IN PRESS
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 418
2.3. Convergence property of the CEM
As we know, the accuracy of the solutions of the CEM depends upon two parameters, the number of terms
N in the CT and the bending stiffness decay constant a. First of all, the number of terms N is determined by a
frequency convergence test [15]. The frequency convergence criterion is dened as
max
i1;2;3
jDo
N
i
=o
N
i
j otol (10)
where o
i
N
is the estimation of the ith frequency with N-terms in the CT. Do
i
N
o
i
N
o
i
N1
is the difference of
the ith frequency obtained with the N-terms and (N1)-terms. tol is a small number and it is taken as
2.0 10
5
in this study.
The parameters of the beam for this study are: Youngs modulus E 200 GPa, mass density r
7850 kg/m
3
, width b 0.01 m, depth d 0.01 m and length L 0.2 m. The location of crack is 0.1 m from the
left support. The beam is assumed to be simply supported. Table 1 shows the convergence of the rst three
natural frequencies with different number of terms in CT when the parameter a equals to 0.667 and 1.426. It is
seen from the results that only 10 terms are needed to satisfy the frequency convergence criterion. Calculation
for Table 1 also shows the convergence is little dependent on the value of the parameter a which will be
discussed below. This indicates the robustness of the proposed approach with respect to the parameter a.
Another comparison of the convergence property study is made with the Galerkin procedure [16] from
different number of terms in the mode functions and results are presented in Table 2. Both methods converge
with 10 terms in the shape functions with similar modal frequencies indicating similar accuracy.
A further study is made by comparing the frequency ratio computed for the beam in [17] using the proposed
CEM, the Galerkin method with piece-wise-smooth functions (GPSF) and from Shen and Pierre [15] (SP), and
the results are shown in Table 3. The frequency ratio is dened as the ratio of natural frequency of the cracked
beam to that of the intact beam. The ratios from the CEM method are between those from GPSF and SP
ARTICLE IN PRESS
Table 1
Convergence in the natural frequencies with different number of shape functions
Mode number Number of terms in the shape function
5 7 8 9 10
1 514.7/547.7 (6%) 511.7/546.3 (6.3%) 511.6/546.3 (6.4%) 509.6/544.9 (6.5%) 509.6/544.9 (6.5%)
2 2262.9/2286.4 (1%) 2261.7/2286.2 (1.1%) 2260.7/2285.9 (1.1%) 2260.7/2285.7 (1.1%) 2260.7/2285.9 (1.1%)
3 4785.4/4957.2 (3.5%) 4761.3/4945.6 (3.7%) 4761.3/4945.7 (3.7%) 4745.4/4935.4 (3.8%) 4745.4/4935.4 (3.8%)
Notes: / denotes values (Hz) calculated with a 0.667 and 1.426, respectively.
() denotes the percentage difference relative to results for a 0.667.
Table 2
Comparison on the convergence performance of CEM with Galerkin procedure [16]
Mode
number
Number of terms in the shape function (a 1.426)
5 7 8 9 10
1 547.7/547.0 (0.13%) 546.3/545.7 (0.11%) 546.3/545.5 (0.15%) 544.9/544.1 (0.15%) 544.9/544.1 (0.15%)
2 2286.4/2285.9
(0.02%)
2286.2/2285.8
(0.02%)
2285.9/2285.7
(0.01%)
2285.9/2285.7
(0.01%)
2285.9/2285.6
(0.01%)
3 4957.2/4952.2
(0.10%)
4945.6/4940.1
(0.11%)
4945.7/4940.1
(0.11%)
4935.4/4929.3
(0.12%)
4935.4/4929.3
(0.12%)
Notes: / denotes frequencies (Hz) calculated from CEM and Galerkin procedure, respectively.
() denotes the percentage change relative to CEM results.
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 419
when the same number of terms in the shape function is used, and they are closer to those from GPSF
indicating the correctness of the proposed method.
2.4. Determination of the parameter a
Parameter a has been shown in Fig. 2, governing the distribution of the EI along the beam. The affected
region is large for a small a and vice versa. The parameter a is valid for a crack at any location of the
beam. It is obtained by tting the fundamental natural frequency calculated by CEM to best match the
analytical one from Saez et al. [9]. The parameters of the beam are the same as for the last study and
the crack is assumed to be in the middle span of the beam. The crack depth ratio changes from 0.1 to 0.9 with
increment of 0.1, and the vector of calculated fundamental frequencies are used in the tting using penalty
method [19]. The parameter a is found to be 1.426 after 12 iterations. In the calculation, 10 terms in the
shape functions are used in CEM for the vibration analysis. The drop in the fundamental natural frequency in
terms of crack depth is shown in Fig. 3(a). This gure indicates that the agreement between the two methods is
very good when a equals to 1.426. This value of a is further checked when the crack is at 3L/4 or L/10 of the
beam. The agreement is excellent as shown in Figs. 3(b) and (c). The matching of the two curves for the latter
is excellent for C
r
in the range of 0.10.8, but the difference increases when C
r
is larger than 0.8. The above
results show that the proposed model predicts the dynamic behavior very well when the crack depth ratio is
less than 0.8.
The variations in the ratio of the fundamental natural frequency obtained from a 0.667 [2] are
also shown in the gures, and Christides and Barr is found to underestimate the solutions with the
crack ratios over the range studied. Therefore, a is taken as 1.426 in the proposed composite element crack
model.
A major source of error with the proposed model is the use of analytical mode shapes in Eq. (5), while there
would be some changes in the mode shapes with damage. This error is investigated as follows. Fig. 4 shows the
normalized mode shape at the fundamental frequency of the beam from the CEM and the analytical one from
Saez et al. [9] with the crack locates at mid-span of the beam and with a crack depth ratio of 0.3 and 0.5. The
mode shapes match each other very well with a better matching at the crack location by the proposed model.
Results not shown with the crack at 0.1L of the beam indicate that the mode shape does not change much with
a crack depth ratio of 0.6. These indicate that the use of analytical mode shape with the proposed model could
be acceptable when the crack depth ratio is not exceeding 0.5.
2.5. Forced vibration analysis
The equation of motion of the forced vibration of a cracked beam with n cracks when expressed in terms of
the CET method is
M

Q C
_
Q Kx
L
1
;
d
c
1
; . . . ; x
L
i
;
d
c
i
; . . . x
L
n
;
d
c
n
Q f t (11)
ARTICLE IN PRESS
Table 3
Comparison on the converged frequency ratio
Number of terms in the
shape function
GPSFs SP CEM
1st
mode
2nd
mode
3rd
mode
1st
mode
2nd
mode
3rd
mode
1st
mode
2nd
mode
3rd
mode
9 0.9433 0.9997 0.9493 0.9443 0.9997 0.9497 0.9547 0.9996 0.9588
11 0.9354 0.9997 0.9424 0.9388 0.9997 0.9452 0.9513 0.9996 0.9560
19 0.9165 0.9997 0.9275 0.9239 0.9997 0.9333 0.9370 0.9996 0.9441
29 0.9130 0.9996 0.9248 0.9164 0.9997 0.9275 0.9181 0.9996 0.9290
31 0.9129 0.9996 0.9248 0.9157 0.9997 0.9269 0.9140 0.9996 0.9260
33 0.9129 0.9996 0.9248 0.9152 0.9997 0.9265 0.9139 0.9996 0.9260
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 420
where M and K are the system mass and stiffness matrices, which are the same as those shown in Eq. (9), C is
the damping matrix which represents a Rayleigh damping model in this work as
C a
1
M a
2
Kx
L
1
;
d
c
1
; . . . ; x
L
i
;
d
c
i
; . . . x
L
n
;
d
c
n
(12)
where a
1
and a
2
are constants to be determined from two modal damping ratios. f(t) is the genera-
lized force vector. For an external force F(t) acting at the location x
F
from the left support, f(t) can be
expressed as
f t N
1
x
F
; N
2
x
F
; N
3
x
F
; N
4
x
F
; f
1
x
F
; . . . f
n
x
F

T
Ft (13)
The generalized acceleration Q

, velocity Q

and displacement Q of the cracked beam can be obtained from


Eq. (11) by direct integration method. The physical acceleration u(x,t) is obtained from
ux; t Sx
T

Q (14)
The physical velocity and displacement can be obtained in a similar way.
2.6. Dynamic response sensitivity with respect to the crack parameters
Taking partial derivative of Eq. (11) with respect to the crack location x
L
i
of the ith crack, we have
M
q

Q
qx
L
i
C
q
_
Q
qx
L
i
Kx
L
1
;
d
c
1
; . . . ; x
L
i
;
d
c
i
; . . . x
L
n
;
d
c
n

qQ
qx
L
i
a
2
qKx
L
1
;
d
c
1
; . . . ; x
L
i
;
d
c
i
; . . . x
L
n
;
d
c
n

qx
L
i
_
Q
qKx
L
1
;
d
c
1
; . . . ; x
L
i
;
d
c
i
; . . . x
L
n
;
d
c
n

qx
L
i
Q (15)
ARTICLE IN PRESS
0.1 0.2 0.3
crack at 0.5L
crack at 0.75L
crack at 0.1L
0.4 0.5 0.6 0.7 0.8 0.9
0
0.5
1

c
r
a
c
k
/

0
Ref. [8]
= 1.426
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
0.2
0.4
0.6
0.8
1

c
r
a
c
k
/

0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
0.2
0.4
0.6
0.8
1
crack depth ratio

c
r
a
c
k
/

0
= 0.667
Ref. [8]
= 1.426
= 0.667
Ref. [8]
= 1.426
= 0.667
Fig. 3. Comparison on frequency ratio corresponding to different crack depth ratios.
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 421
where
qKx
L
1
;
d
c
1
; :::; x
L
i
;
d
c
i
; :::x
L
n
;
d
c
n

qx
L
i

Z
L
0
d
2
S
T
dx
2
q
qx
L
i
EI
0
1 1=1 d
c
i
=d
3
1 exp2ajx x
L
i
j=d
" #
d
2
S
dx
2
dx (16)
Since the dynamic response has been obtained from Eq. (11) and the right-hand side of Eq. (16) can be
obtained from numerical integration, the right-hand side of Eq. (15) can be obtained. Note that Eq. (15) is
similar to Eq. (11), and the dynamic response sensitivity (i.e., the generalized acceleration response sensitivity,
velocity response sensitivity and displacement response sensitivity) with respect to the location of the ith crack
can be obtained by direct integration again.
Similarly, the dynamic response sensitivity with respect to the depth of the ith crack can be obtained from
the following equation:
M
q

Q
qd
c
i
C
q
_
Q
qd
c
i
Kx
L
1
;
d
c
1
; . . . ; x
L
i
;
d
c
i
; . . . x
L
n
;
d
c
n

qQ
qd
c
i
a
2
qKx
L
1
;
d
c
1
; . . . ; x
L
i
;
d
c
i
; . . . x
L
n
;
d
c
n

qd
c
i
_
Q

qKx
L
1
;
d
c
1
; . . . ; x
L
i
;
d
c
i
; . . . x
L
n
;
d
c
n

qd
c
i
Q
(17)
where
qKx
L
1
;
d
c
1
; . . . ; x
L
i
;
d
c
i
; . . . x
L
n
;
d
c
n

qd
c
i

Z
L
0
d
2
S
T
dx
2
q
qd
c
i
EI
0
1 1=1 d
c
i
=d
3
1 exp2ajx x
L
i
j=d
" #
d
2
S
dx
2
dx (18)
ARTICLE IN PRESS
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
0
1
2
3
4
N
o
r
m
a
l
i
z
e
d

m
o
d
e

s
h
a
p
e
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
0
1
2
3
4
Location along the beam (m)
N
o
r
m
a
l
i
z
e
d

m
o
d
e

s
h
a
p
e
uncracked
proposed
ref. [8]
crack depth ratio = 0.3
crack depth ratio = 0.5
uncracked
proposed
ref. [8]
Fig. 4. Comparison of the fundamental mode shape.
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 422
2.7. Identication of the crack location and depth
The vector of parameter to be updated is W [X d
c
]
T
, where X x
L
1
; x
L
2
; . . . ; x
L
n
and d
c

d
c
1
; d
c
2
; . . . ; d
c
n
are the vectors of locations and the crack depths of the n cracks, respectively. In the
identication, one or more generalized responses are used. The measured response is simulated by adding
different levels of articial normally distributed random noise into the one calculated from Eq. (11).
The sensitivity-based approach basing on the measured acceleration responses [20] is adopted for the
updating of the vector of parameters W,
d

Q PdW (19)
where dW is the vector of perturbations in the updating parameters, d

Q
^

Q

Q is the differences in the vector
of polluted-generalized acceleration
^

Qand the vector of calculated acceleration Q

. Matrix P consists of the


response sensitivity, which is the rst derivative of the dynamic response with respect to the updating
parameters. These derivatives are calculated from Eqs. (15) and (17). For example, when the ith generalized
acceleration is used, the elements in the sensitivity matrix are shown as
P
q

Q
i
t
1

qx
L
1
q

Q
i
t
1

qx
L
2

q

Q
i
t
1

qx
L
n
q

Q
i
t
1

qd
c
1
q

Q
i
t
1

qd
c
2

q

Q
i
t
1

qd
c
n
q

Q
i
t
2

qx
L
1
q

Q
i
t
2

qx
L
2

q

Q
i
t
2

qx
L
n
q

Q
i
t
2

qd
c
1
q

Q
i
t
2

qd
c
2

q

Q
i
t
2

qd
c
n
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
q

Q
i
t
f

qx
L
1
q

Q
i
t
f

qx
L
2

q

Q
i
t
f

qx
L
n
q

Q
i
t
f

qd
c
1
q

Q
i
t
f

qd
c
2

q

Q
i
t
f

qd
c
n
2
6
6
6
6
6
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
7
7
7
7
7
5
(20)
where t
1
is the beginning of the time history and t
f
is the end of the time history. Matrix P has the dimensions
of t
f
2n. Eq. (19) can be solved with the least-squares method directly as
dW P
T
P
1
P
T
d

Q (21)
Like many other inverse problems, Eq. (19) is an ill-conditioned problem. In order to provide bounds to the
solution, the damped-least-squares method (DLS) [21] is used in the pseudo-inverse calculation. Eq. (19) can
be written in the following form in the DLS method
dW P
T
P lI
1
P
T
d

Q (22)
where l is the non-negative damping coefcient governing the participation of least-squares error in the
solution. When the parameter l approaches zero, the estimated vector dW approaches to the solution obtained
from the simple least-squares method from Eq. (21). L-curve method [22] is used in this paper to obtain the
optimal-regularization parameter l.
The proposed method is considered capable to identify the location of the crack whether to the left or to the
right of the mid-point of the beam, as the transmissibility between a crack and the measurement point will be
different for different locations. The only exception would be when the crack damage is symmetrical to the mid-
point of the beam and the response is obtained at the mid-point when the identied result would not be unique.
3. Applications
3.1. Comparision with existing models and experimental results
Several experimental works in Sinha et al. [18] and Lee and Chung [14] are restudied in the following section
and results are compared with those obtained from the CEM model. They are
Case 1An aluminum cantilever beam with one crack.
Case 2An aluminum freefree supported beam with one crack.
ARTICLE IN PRESS
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 423
Case 3A steel freefree supported beam with one crack.
Case 4An aluminum freefree supported beam with two cracks.
The geometric and material properties and boundary conditions of the beams are given in Table 4. The
beam is discretized into one element and ten shape functions are used in the calculation with the total degrees-
of-freedom in the CEM equals 14, while the total degrees-of-freedom in the nite element model is 34 and 56
[18] for the beam in Cases 1 and 2, respectively. The beam in Case 3 has been discretized [18] into 21 elements
and the total DOFs is 42. The crack depth in the three beams varies in three stages of 4, 8 and 12 mm. The
comparison of the predicted natural frequencies of the beam from the proposed model and those in Lee and
Chung [14] and Sinha et al. [18] and the experimental results are shown in Tables 57. The proposed model, in
general, gives better results than the model in Sinha et al. since the latter crack model is a linear approximation
of the theoretical crack model of Christides and Barr. The results from Lee and Chung are seen to be
comparable with those from the proposed model.
In Case 4, the beam in Case 2 is tested again with a new crack introduced. The rst crack is at 595 mm from
the left end with a xed crack depth of 12 mm; while the second crack is at 800 mm from the left end with the
crack depth varying from 4 to 12 mm in step of 4 mm. Table 8 gives the rst ve natural frequencies of the
beam by CEM method and compares with those from Sinha et al. and the experimental measurement. The
results from CEM are found closer to the experimental prediction than those in Sinha et al. The above
comparisons show that the CEM approach of modeling a beam with crack(s) is accurate for the vibration
analysis. A yet bigger advantage of the model is the much lesser number of DOFs in the resulting nite
element model of the structure.
3.2. Dynamic response and response sensitivity with respect to crack parameters
The parameters of the beam under study are: E 200 GPa, b 25 mm, d 20 mm, and L 2000 mm. The
mass density is r 7850 kg/m
3
. The beam is simply support and with a 3-mm-deep crack at 700 mm from the
ARTICLE IN PRESS
Table 4
Properties of beams [18] in the experimental study
Case 1 Case 2 Case 3
Boundary conditions Cantilever Freefree Freefree
Material Aluminum Aluminum Steel
Youngs modulus 69.79 Gpa 69.79 Gpa 203.91 Gpa
Mass density 2600 kg/m
3
2600 kg/m
3
7800 kg/m
3
Poisson ratio 0.33 0.33 0.33
Beam length 996 mm 1832 mm 1330 mm
Beam width 50 mm 50 mm 25.3 mm
Beam depth 25 mm 25 mm 25.3 mm
Boundary stiffnesses k
t
26.5 MN/m N/A N/A
k
y
150 KNm/rad
Note: k
t
and k
y
are the vertical and rotational stiffness of the support of the beam, respectively.
Table 5
Comparison of natural frequencies (Hz) of the aluminum cantilever beam with one crack (Case 1)
Mode No crack d
c1
4 mm at x
1
275 mm d
c1
8 mm at x
1
275 mm d
c1
12 mm at x
1
275 mm
Exp. Proposed Exp. [18] [14] Proposed Exp. [18] [14] Proposed Exp. [18] [14] Proposed
1 20.000 19.900 20.000 19.640 19.822 19.763 19.750 19.382 19.580
a
19.562 19.000 19.164
a
19.048
a
19.238
2 124.500 124.530 124.250 124.106 124.410 124.302 124.063 123.689 124.008
a
123.971 123.000 123.343
a
123.147
a
123.440
3 342.188 345.202 340.813 340.758
a
343.920 340.978 336.875 336.094 339.263 336.195 326.563 332.383
a
329.937
a
333.473
4 664.375 665.682 662.813 663.020
a
663.539 663.187 662.313 660.584 661.299 662.121 660.313 658.641 656.975 659.060
a
Denotes result better than that from the proposed method.
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 424
left support. A sinusoidal external excitation force acts at mid-span of the beam with a magnitude of 10 N and
at a frequency of 5 Hz. The time step is 0.001 s in calculating the dynamic response. Fig. 5(a) and (b) shows the
response sensitivity of the fth generalized acceleration (c
1
(t)) with respect to the location of the crack and the
depth of the crack. It is noted that the magnitude of the latter sensitivity is almost 50 times of the former,
which means the variation of the acceleration response is much more sensitive to the perturbation in crack
depth than its location.
3.3. Crack identication from response sensitivity-based updating method
Only the fth generalized acceleration is used in the updating and response data in the rst second is used.
Five and ten percent articial random noises are added to the calculated acceleration separately to simulate
the measured response. Initial value for the unknown crack depth is set equal to zero, while the crack
location is initially guessed to be at mid-span. The latter selection has been checked to give acceptable
ARTICLE IN PRESS
Table 6
Comparison of natural frequencies (Hz) of the aluminium free-free beam with one crack (Case 2)
Mode No crack d
c1
4 mm at x
1
595 mm d
c1
8 mm at x
1
595 mm d
c1
12 mm at x
1
595 mm
Exp. Proposed Exp. [18] [14] Proposed Exp. [18] [14] Proposed Exp. [18] [14] Proposed
1 40.000 39.770 39.688 39.379 39.698
a
39.490 39.375 39.094 39.415
a
39.242 39.063 38.857 38.77 38.869
2 109.688 109.340 109.063 108.206 109.311
a
108.633 108.125 107.132 108.200
a
107.670 105.938 106.278
a
105.850
a
106.293
3 215.000 214.795 215.000 214.087 214.927
a
214.230 214.688 213.825 214.654
a
213.986 214.375 213.622 214.085
a
213.631
4 355.000 354.853 354.688 353.107 355.029
a
353.683 353.438 351.872 353.783
a
352.524 350.625 350.881
a
351.136 350.921
5 528.750 529.601 527.188 524.696 529.363 526.540 522.812 520.452 524.684 522.448 513.125 517.219 515.507
a
517.003
a
Denotes result better than that from the proposed method.
Table 7
Comparison of natural frequencies (Hz) of the steel free-free beam with one crack (Case 3)
Mode No crack d
c1
4 mm at x
1
430 mm d
c1
8 mm at x
1
430 mm d
c1
12 mm at x
1
430 mm
Exp. Proposed Exp. [18] [14] Proposed Exp. [18] [14] Proposed Exp. [18] [14] Proposed
1 75.313 75.171 74.688 74.406 74.938 74.777 74. 063 73. 628 74.224
a
74. 225 72. 813 72. 958
a
72.634
a
73. 316
2 207.188 207.217 205.625 204.183 206.262 205.617 202.500 201.283 203.458
a
203.463 197.188 198. 928 197.764 197.348
3 406.250 406.289 405.625 405.368 405.974 405.850 404.688 404.557
a
405.235
a
405.252 403.125 403.916 403.770 403.373
4 667.813 671.687 666.250 668.429
a
670.550 668.938 662.813 665.356 667.615 664.565 655.938 662.874 661.635
a
662.111
a
Denotes result better than that from the proposed method.
Table 8
Comparison of natural frequencies (Hz) of the aluminum freefree beam with two cracks (Case 4)
Mode No crack d
c1
12 mm at x
1
595 mm d
c1
12 mm at x
1
595 mm d
c1
12 mm at x
1
595 mm
d
c2
4 mm at x
2
800 mm d
c2
8 mm at x
2
800 mm d
c2
12 mm at x
2
800 mm
Exp. Proposed Exp. [18] Proposed Exp. [18] Proposed Exp. [18] Proposed
1 40.000 39.770 38.750 38.352 38.607 38.437 37.897 38.246 37.500 37.513
a
37.703
2 109.688 109.340 105.938 105.890
a
106.196 105.938 105.510 106.062 105.625 105.559
a
105.858
3 215.000 214.795 213.750 212.207 212.786 212.813 210.897 211.643 210.000 209.815 209.975
4 355.000 354.853 350.000 348.920 349.843 349.063 347.235 348.410 345.625 345.876 346.374
5 528.750 529.601 512.500 514.575
a
514.735 511.250 512.903
a
513.044 507.500 510.560
a
510.633
a
Denotes result better than that from the proposed method.
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 425
estimates on the crack locations with cracks at different locations on the beam. Convergence of computation is
considered achieved, when the norm of relative difference between two sets of successively identied
parameters equals 1.0 10
5
.
3.3.1. Case 1: identication of single crack
The identied results shown in Table 9 indicate that the parameters of the crack can be identied with very
good accuracy under different noise levels. This indicates the identied results are not sensitive to the articial
noise.
3.3.2. Case 2: identication of two cracks
The cracked beam in the last study is retained with another 5-mm-deep crack introduced at 400 mm from
the left support of the beam. Although the identied results for the two cracks shown in Table 9 are not as
good as those for the single crack, the identied parameters of the two cracks are still very good under
different noise levels with a maximum relative error or around 2.4% in the crack location and 3.7% in the
crack depth.
4. Experimental verication
A freefree steel beam was tested in the laboratory to verify the proposed method. The length, width and
height of the beam are 2.1, 0.025 and 0.019 m, respectively, and the elastic modulus and mass density of the
material are 207 GPa and 7.83 10
3
kg/m
3
, respectively. The beam is suspended at its two ends as shown in
Fig. 6. It was modeled with one Euler-Bernoulli beam element with 10 c-dofs. The rst ve natural frequencies
of the intact beam are computed from the proposed model and compared with those obtained from modal test
ARTICLE IN PRESS
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
Time (s)

Q
5
/
x
L
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
30
20
10
0
10
20
30
Time (s)

Q
5
/
d
c
.
.
.
.
Fig. 5. Acceleration response sensitivity with respect to different crack parameters (a) with respect to crack location and (b) with respect to
crack depth.
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 426
and they are shown in Table 10. A sinusoidal force at the frequency of half of the rst natural frequency of the
beam was applied at the nodal point of the rst vibration mode of the beam 480 mm from the left free end with
an exciter model Ling Dynamic LDS V450. Acceleration in the horizontal direction obtained with a B&K
4370 accelerometer at the middle of the beam was used to identify the crack parameters. Sampling frequency is
2000 Hz. Time history of the input sinusoidal force was also recorded for calculating the numerical response of
the beam. Rayleigh damping is adopted and the experimental modal damping ratios of 0.007 and 0.01 for the
rst two modes are included in the calculation. The convergence tolerance in the identication is same as that
for the simulation study.
It is well-known that modeling error in the initial intact structure has, in general, signicant effect on the
accuracy of the identied results. In most cases, the initial model is updated rst to obtain a good
representation of the intact structure. Table 10 shows that these two sets of analytical and experimental
frequencies match each other very well, indicating that the initial model of the beam is good enough for crack
identication.
Two adjacent cracks are introduced at 1.66 and 1.72 m from the left free end with 3 and 9 mm crack depth,
respectively, and they are created using a machine saw with 1.3-mm-thick cutting blade. The two cracks are
close to each other to make sure the interaction effect between them exists. The rst ve measured natural
frequencies of the damaged beam are found from modal test and they are shown in Table 10. The crack
parameters, namely, crack depth and location are taken as the unknowns to be identied in the inverse
analysis. In the inverse identication, the rst three seconds of the measured acceleration response at mid-span
of the beam is used. Initial guess of the crack depth is taken as zero and the crack location is initially estimated
to be at mid-span. Good prediction on the crack parameters are obtained after 38 iterations. The identied
ARTICLE IN PRESS
19 mm
Cracks
25mm
Accelerometer
Shaker
Fig. 6. Conguration of freefree steel beam for forced vibration test.
Table 9
Identied results
True Identied values
Noise free 5% Noise 10% Noise
Single crack
Crack location x
L
(mm) 700 700.8 705.7 707.3
Crack depth d
c
(mm) 3 2.99 2.97 2.94
No. of iteration required N/A 28 32 35
Two cracks
Crack locations x
L1
/x
L2
(mm) 400/700 400.9/700.4 393.6/709.1 389.6/716.7
Crack depths d
c1
/d
c2
(mm) 5/3 4.97/3.02 4.93/2.94 4.89/2.91
No. of iteration required N/A 31 45 50
Optimal regularization parameter N/A 2.31 10
5
3.42 10
5
8.67 10
5
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 427
locations of the two cracks are 1.641 and 1.702 m, respectively, and the depths are 2.84 and 8.97 mm,
respectively, with a maximum error of 1.15% in crack location and 5.33% in crack depth. This shows the
high accuracy of the proposed crack identication method despite the coupling effect of the two close cracks.
The optimal-regularization parameter is 1.75 10
4
. The analytical results of the updated beam with the
identied crack parameters are found very close to the experimental values as shown in Table 10. The same set
of experimental data has been used in the identication of local damage averaged over an element [23], and the
analytical results from the updated beam are also shown in Table 10 for comparison. It is noted that
the identication with the CEM model on the crack damage gives much higher accuracy than those without
ARTICLE IN PRESS
Table 10
The measured and the analytical Natural frequencies (Hz) and relative error of identication (%) of the experimental beam
Crack scenarios Mode no.
1 2 3 4 5
No crack
Experimental 22.87 62.76 123.05 203.24 303.45
Analytical 22.83/0.18 62.74/0.03 123.04/0.0 203.03/0.12 302.86/0.2
Two cracks
Experimental 22.74 61.77 119.75 198.49 299.50
Analytical 22.72/0.09 61.84/0.11 120.32/0.46 199.85/0.69 301.11/0.54
Analytical [23] 21.71/0.05 62.25/0.78 120.73/0.82 200.55/1.04 301.37/0.62
Note: / denotes the modal frequency/relative error.
0 0.5 1 1.5 2 2.5 3
4
2
0
2
4
Time (second)
F

(
N
)
0 0.5 1 1.5 2 2.5 3
0.4
0.2
0
0.2
0.4
Time (second)
A
c
c
e
l
e
r
a
t
i
o
n

(
m
/
s
2
)
Fig. 7. Time history of experimental and calculated acceleration and applied force (a) applied force and (b) acceleration, __ experiment,
reconstructed.
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 428
a proper model of the multiple cracks. Fig. 7 compares the time history of the calculated acceleration at
mid-span including the identied crack parameters and the corresponding measured acceleration smoothed
with twenty terms orthogonal polynomial function to remove the measurement noise [24]. The time series
match each other very well through out the measured duration.
5. Conclusions
A crack model is incorporated into a composite element for the vibration analysis and for crack
identication. The composite beam element is of one-elementone-member conguration. Modeling with this
type of element would allow the automatic inclusion of interaction effect between adjacent local damages in
the nite element model. The accuracy and convergence of this new composite element has been compared
satisfactory with existing model and experimental results.
Results obtained from a system identication approach based on the response sensitivity show that an
accurate model is essential for an accurate inverse identication, and the formulation of similar composite
elements incorporating different type of damages could be a useful library of damage models which, is more
suitable for structural damage detection than the existing ones.
Acknowledgment
The work described in this paper was supported by a grant from the Hong Kong Research Grant Council
Project No. PolyU 5194/05E.
6. Appendix A. The submatrix of the elemental stiffness matrix
k
qq

R
L
0
d
2
N
1
dx
2
EIx
d
2
N
1
dx
2
dx
R
L
0
d
2
N
1
dx
2
EIx
d
2
N
2
dx
2
dx
R
L
0
d
2
N
1
dx
2
EIx
d
2
N
3
dx
2
dx
R
L
0
d
2
N
1
dx
2
EIx
d
2
N
4
dx
2
dx
R
L
0
d
2
N
2
dx
2
EIx
d
2
N
1
dx
2
dx
R
L
0
d
2
N
2
dx
2
EIx
d
2
N
2
dx
2
dx
R
L
0
d
2
N
2
dx
2
EIx
d
2
N
3
dx
2
dx
R
L
0
d
2
N
2
dx
2
EIx
d
2
N
4
dx
2
dx
R
L
0
d
2
N
3
dx
2
EIx
d
2
N
1
dx
2
dx
R
L
0
d
2
N
3
dx
2
EIx
d
2
N
2
dx
2
dx
R
L
0
d
2
N
3
dx
2
EIx
d
2
N
3
dx
2
dx
R
L
0
d
2
N
3
dx
2
EIx
d
2
N
4
dx
2
dx
R
L
0
d
2
N
4
dx
2
EIx
d
2
N
1
dx
2
dx
R
L
0
d
2
N
4
dx
2
EIx
d
2
N
2
dx
2
dx
R
L
0
d
2
N
4
dx
2
EIx
d
2
N
3
dx
2
dx
R
L
0
d
2
N
4
dx
2
EIx
d
2
N
4
dx
2
dx
2
6
6
6
6
6
6
4
3
7
7
7
7
7
7
5
k
cq

R
L
0
d
2
N
1
dx
2
EIx
d
2
f
1
dx
2
dx
R
L
0
d
2
N
1
dx
2
EIx
d
2
f
2
dx
2
dx . . .
R
L
0
d
2
N
1
dx
2
EIx
d
2
f
n1
dx
2
dx
R
L
0
d
2
N
1
dx
2
EIx
d
2
f
n
dx
2
dx
R
L
0
d
2
N
2
dx
2
EIx
d
2
f
1
dx
2
dx
R
L
0
d
2
N
2
dx
2
EIx
d
2
f
2
dx
2
dx . . .
R
L
0
d
2
N
2
dx
2
EIx
d
2
f
n1
dx
2
dx
R
L
0
d
2
N
2
dx
2
EIx
d
2
f
n
dx
2
dx
R
L
0
d
2
N
3
dx
2
EIx
d
2
f
1
dx
2
dx
R
L
0
d
2
N
3
dx
2
EIx
d
2
f
2
dx
2
dx . . .
R
L
0
d
2
N
3
dx
2
EIx
d
2
f
n1
dx
2
dx
R
L
0
d
2
N
3
dx
2
EIx
d
2
f
n
dx
2
dx
R
L
0
d
2
N
4
dx
2
EIx
d
2
f
1
dx
2
dx
R
L
0
d
2
N
4
dx
2
EIx
d
2
f
2
dx
2
dx . . .
R
L
0
d
2
N
4
dx
2
EIx
d
2
f
n1
dx
2
dx
R
L
0
d
2
N
4
dx
2
EIx
d
2
f
n
dx
2
dx
2
6
6
6
6
6
6
4
3
7
7
7
7
7
7
5
k
cc

R
L
0
d
2
f
1
dx
2
EIx
d
2
f
1
dx
2
dx
R
L
0
d
2
f
1
dx
2
EIx
d
2
f
2
dx
2
dx
R
L
0
d
2
f
1
dx
2
EIx
d
2
f
n1
dx
2
dx
R
L
0
d
2
f
1
dx
2
EIx
d
2
f
n
dx
2
dx
R
L
0
d
2
f
2
dx
2
EIx
d
2
f
1
dx
2
dx
R
L
0
d
2
f
2
dx
2
EIx
d
2
f
2
dx
2
dx
R
L
0
d
2
f
2
dx
2
EIx
d
2
f
n1
dx
2
dx
R
L
0
d
2
f
2
dx
2
EIx
d
2
f
n
dx
2
dx
.
.
.
.
.
.

.
.
.
.
.
.
R
L
0
d
2
f
n1
dx
2
EIx
d
2
f
1
dx
2
dx
R
L
0
d
2
f
n1
dx
2
EIx
d
2
f
2
dx
2
dx
R
L
0
d
2
f
n1
dx
2
EIx
d
2
f
n1
dx
2
dx
R
L
0
d
2
f
n1
dx
2
EIx
d
2
f
n
dx
2
dx
R
L
0
d
2
f
n
dx
2
EIx
d
2
f
1
dx
2
dx
R
L
0
d
2
f
n
dx
2
EIx
d
2
f
2
dx
2
dx
R
L
0
d
2
f
n
dx
2
EIx
d
2
f
n1
dx
2
dx
R
L
0
d
2
f
n
dx
2
EIx
d
2
f
n
dx
2
dx
2
6
6
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
7
7
5
ARTICLE IN PRESS
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 429
7. Appendix B. The submatrix of the elemental mass matrix
m
qq

R
L
0
N
1
rAN
1
dx
R
L
0
N
1
rAN
2
dx
R
L
0
N
1
rAN
3
dx
R
L
0
N
1
rAN
4
dx
R
L
0
N
2
rAN
1
dx
R
L
0
N
2
rAN
2
dx
R
L
0
N
2
rAN
3
dx
R
L
0
N
2
rAN
4
dx
R
L
0
N
3
rAN
1
dx
R
L
0
N
3
rAN
2
dx
R
L
0
N
3
rAN
3
dx
R
L
0
N
3
rAN
4
dx
R
L
0
N
4
rAN
1
dx
R
L
0
N
4
rAN
2
dx
R
L
0
N
4
rAN
3
dx
R
L
0
N
4
rAN
4
dx
2
6
6
6
6
6
4
3
7
7
7
7
7
5
m
cq

R
L
0
N
1
rAf
1
dx
R
L
0
N
1
rAf
2
dx
R
L
0
N
1
rAf
n1
dx
R
L
0
N
1
rAf
n
dx
R
L
0
N
2
rAf
1
dx
R
L
0
N
2
rAf
2
dx
R
L
0
N
2
rAf
n1
dx
R
L
0
N
2
rAf
n
dx
R
L
0
N
3
rAf
1
dx
R
L
0
N
3
rAf
2
dx
R
L
0
N
3
rAf
n1
dx
R
L
0
N
3
rAf
n
dx
R
L
0
N
4
rAf
1
dx
R
L
0
N
4
rAf
2
dx
R
L
0
N
4
rAf
n1
dx
R
L
0
N
4
rAf
n
dx
2
6
6
6
6
6
4
3
7
7
7
7
7
5
m
cc

R
L
0
f
1
rAf
1
dx
R
L
0
f
1
rAf
2
dx
R
L
0
f
1
rAf
n1
dx
R
L
0
f
1
rAf
n
dx
R
L
0
f
2
rAf
1
dx
R
L
0
f
2
rAf
2
dx
R
L
0
f
2
rAf
n1
dx
R
L
0
f
2
rAf
n
dx
.
.
.
.
.
.

.
.
.
.
.
.
R
L
0
f
n1
rAf
1
dx
R
L
0
f
n1
rAf
2
dx
R
L
0
f
n1
rAf
n1
dx
R
L
0
f
n1
rAf
n
dx
R
L
0
f
n
rAf
1
dx
R
L
0
f
n
rAf
2
dx
R
L
0
f
n
rAf
n1
dx
R
L
0
f
n
rAf
n
dx
2
6
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
7
5
References
[1] P. Zeng, Composite element method for vibration analysis of structure, Journal of Sound and Vibration 218 (1998) 619696.
[2] A. Christides, A.D.S. Barr, One-dimensional theory of cracked Bernoulli-Euler beams, International Journal of Mechanical Science
26 (1984) 639648.
[3] W. Ostachowicz, M. Krawczuk, On modelling of structural stiffness loss due to damage, in: Proceedings of Fourth International
Conference on Damage Assessment of Structures, DAMAS, Cardiff, 2001, pp. 185199.
[4] A.K. Pandey, M. Biswas, M.M. Samman, Damage detection from change in curvature mode shapes, Journal of Sound and Vibration
145 (1991) 321332.
[5] A.K. Pandey, M. Biswas, Damage detection in structures using change in exibility, Journal of Sound and Vibration 169 (1994) 317.
[6] M.M.F. Yuen, A numerical study of the eigen parameters of a damaged cantilever beam, Journal of Sound and Vibration 103 (1985)
301310.
[7] P.F. Rizos, N. Aspragathos, A.D. Dimarogonas, Identication of crack location and magnitude in a cantilever beam, Journal of
Sound and Vibration 138 (1990) 381388.
[8] T.D. Chaudhari, S.K. Maiti, A study of vibration geometrically segmented beams with and without crack, International Journal of
Solids and Structures 37 (2000) 761779.
[9] J. Fernandez-Saez, L. Rubio, C. Navarro, Approximate calculation of the fundamental frequency for bending vibrations of cracked
beams, Journal of Sound and Vibration 225 (1999) 345352.
[10] M.W. Ostachowicz, M. Krawczuk, Analysis of the effect of cracks on the natural frequencies of a cantilever beam, Journal of Sound
and Vibration 150 (1991) 191201.
[11] M.H.H. Shen, J.E. Taylor, An identication problem for vibrating cracked beams, Journal of Sound and Vibration 150 (1991)
457484.
[12] E.I. Shifrin, R. Ruotolo, Natural frequencies of a beam with an arbitrary number of cracks, Journal of Sound and Vibration 222
(1999) 409423.
[13] M. Krawczuk, W.M. Ostachowicz, Transverse natural vibration of a cracked beam loaded with a constant axial force, Transactions
of the ASME Journal of Vibration and Acoustics 115 (4) (1993) 524528.
[14] Y.S. Lee, M.J. Chung, A study on crack detection using eigen frequency test data, Computers and Structures 77 (2000) 327342.
[15] M.H.H. Shen, C. Pierre, Natural modes of BernoulliEuler beam with symmetric cracks, Journal of Sound and Vibration 138 (1990)
115134.
[16] L. Meirovitch, Principles and Techniques of Vibrations, PrenticeHall, Upper Saddle River, NJ, 1997.
ARTICLE IN PRESS
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 430
[17] A. Messina, Revisiting Galerkins method through global piece-wise-smooth functions in Christides and Barrs cracked beam theory,
Journal of Sound and Vibration 263 (4) (2003) 937944.
[18] J.K. Sinha, M.I. Friswell, S. Edwards, Simplied models for the location of cracks in beam structures using measured vibration data,
Journal of Sound and Vibration 251 (1) (2002) 1338.
[19] M.I. Friswell, J.E. Mottershead, Finite Element Model Updating in Structural Dynamics, Kluwer Academic Publishers, Dordrecht,
1995.
[20] Z.R. Lu, S.S. Law, Identication of prestress force from measured structural responses, Mechanical System and Signal Processing 20
(8) (2006) 21862199.
[21] A.M. Tikhonov, On the solution of ill-posed problems and the method of regularization, Soviet Mathematics 4 (1963) 10351038.
[22] P.C. Hansen, Regularization tools: a Matlab package for analysis and solution of discrete ill-posed problems, Numerical Algorithms
6 (1994) 135.
[23] Z.R. Lu, S.S. Law, Features of dynamic response sensitivity and its application in damage detection, Journal of Sound and Vibration
303 (1-2) (2007) 305329.
[24] X.Q. Zhu, S.S. Law, Orthogonal function in moving loads identication on a multi-span bridge, Journal of Sound and Vibration 245
(2) (2001) 329345.
ARTICLE IN PRESS
Z.R. Lu, S.S. Law / Mechanical Systems and Signal Processing 23 (2009) 415431 431

S-ar putea să vă placă și