Journal of Sound and Vibration 387 (2017) 5365

Contents lists available at ScienceDirect

Journal of Sound and Vibration

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

Multi-cracks identification method for cantilever beam

structure with variable cross-sections based on measured
natural frequency changes
Kai Zhang a, Xiaojun Yan a,b,c,n
a
School of Energy and Power Engineering, Beihang University, Beijing 100191, China
b
Collaborative Innovation Center of Advanced Aero-Engine, Beijing 100191, China
c
National Key Laboratory of Science and Technology on Aero-Engine Aero-Thermodynamics, Beijing 100191, China

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

Article history: Cantilever beam's crack identification can provide critical information which is helpful to
Received 5 June 2015 determine whether the structure be healthy or not. Among all crack identification
Received in revised form methods, the methods based on measured structure's natural frequency changes own
24 August 2016
advantages of simplicity and easy for operation in practical engineering. To accurately
Accepted 20 September 2016
identify multi-cracks characteristics for cantilever beam structure with variable cross-
Handling Editor: I. Trendafilova
Available online 21 October 2016 sections, a mathematical model, which is based on the concept of modal strain energy, is
established in this investigation. And to obtain cantilever beam's natural frequency result
Keywords: with higher resolution, a signal processing method based on Hilbert-Huang Transform
Cantilever beam
(HHT) is also proposed, which can overcome the disadvantage of fast Fourier transform
Variable cross-sections
(FFT) in the aspect of frequency resolution and incapability of handling nonlinear vibra-
Multi-cracks identification
Frequency changes tion caused by crack breathing phenomenon. Based on above mathematical model and
Time-frequency analysis signal processing method, the method of identifying multi-cracks on cantilever beam with
variable cross-sections is presented. To verify the accuracy of this multi-cracks identifi-
cation method, experimental examples are conducted, and the results show that the
method proposed in this investigation can accurately identify the cracks characteristics,
including their locations and relative depths.
& 2016 Elsevier Ltd. All rights reserved.

1. Introduction

For cantilever beam structures in practical engineering, existed cracks will lower their loading capacity, and may cause
fractures and further serious faults. Crack identification is considered to be an efficient way to avoid such faults. It can
provide critical information to help determining whether the structure should be maintained or replaced. For crack iden-
tification, the vibration based method is usually adopted because of its advantage of identifying crack with no requirement
to access to regions near the crack location, and consuming less time and cost [1]. Since the structure to be tested does not
need to be disassembled from the whole system in most of the cases, the vibration based crack identification is very suitable
for on-line monitoring [14]. For example, Xingwu Zhang et al. [5] presented a method based on the measured natural
frequencies to identify one crack on a beam. M. El-Gebeilya and Y.A. Khulief [6] used the measured vibration signal and

n
Corresponding author at: School of Energy and Power Engineering, Beihang University, Beijing 100191, China.
E-mail addresses: yanxiaojun@buaa.edu.cn, buaa405@sina.com (X. Yan).

http://dx.doi.org/10.1016/j.jsv.2016.09.028
0022-460X/& 2016 Elsevier Ltd. All rights reserved.
54 K. Zhang, X. Yan / Journal of Sound and Vibration 387 (2017) 5365

wavelet transform to identify crack on the inner surface of pipe. Zhixiong Li et al. [7] measured the vibration of complex gear
transmission systems in wind turbines by multi-channel sensors, and presented a method to identify gear's crack by ana-
lyzing the measured vibration signal.
To accurately identify the existed cracks characteristics, including their location and depth etc., the primary work for
vibration based crack identification is to establish a mathematical model, which describes the relationship of cracks
characteristics and cantilever beam's vibration related parameters, such as dynamic response, modal shape or natural
frequencies [810]. For uniform cross-section cantilever beam, this mathematical model can be easy derived from its vi-
bration equation [1113], which is a linear differential equation with constant coefficient. But for variable cross-section
cantilever beam, its variable cross-sectional area and moment inertia make it difficult to establish such mathematical model
based on its vibration equation. Another important factor that should be considered when establishing the model for multi-
cracks identification is that different combinations of existing multi-cracks may cause same changes of vibration related
parameter [14,15]. So usually more than one vibration related parameters are adopted for identifying multi-cracks [8,9], and
to obtain these parameters, a more complex measuring system is usually adopted in present researches [16,17]. For easy of
on-line monitoring application, this investigation only takes natural frequency as a measuring parameter, which is in-
dependent of selection of measuring point and can be obtained by one measuring point [18,19]. By introducing the concept
of modal strain energy, the relationship between variable cross-section cantilever beam's natural frequency and its modal
strain energy is derived, then a mathematical model describing variable cross-section cantilever beam cracks characteristics
and natural frequency changes is established.
After above mentioned mathematical model is established, another key issue for identify variable cross-section canti-
lever beam's multi-cracks is how to measure its natural frequency changes. In general, structure's natural frequency can be
obtained by processing its vibration signal [20]. However, the nonlinear vibration caused by crack breathing phenomenon
[21] makes this single processing be difficult [22]. Thus, this investigation proposes a signal processing method based on
Hilbert-Huang Transform (HHT), to obtain cantilever beam's natural frequency result with higher resolution, since HHT has
capability of processing nonlinear vibration signal [23,24]. The method can decompose cantilever beam's vibration signal
into several signal components, which have different frequency band and are respectively related to cantilever beam's
natural frequencies in different vibration modes. And applying HHT on those decomposed signal components, the cantilever
beam's accurate result of natural frequency is obtained.
To verify the correctness of the multi-cracks identification method, experimental verifications are also conducted in this
investigation. Experimental results show that the method presented in this investigation can accurately identify multi-
cracks of cantilever beam with variable cross-sections, just by taking natural frequency changes as measuring parameters.

2. Principle of multi-cracks identification

To identify multi-cracks on variable cross-section cantilever beam based on the measured natural frequency changes,
following issues should be considered. Firstly, a mathematical model which describes relationship between cantilever
beam's natural frequency changes and cracks characteristics is established. And secondly, cantilever beam's natural fre-
quency changes are required to be accurately measured. Thirdly, by substituting the measured natural frequency changes
into the mathematical model, the characteristics of cantilever beam's cracks can be obtained.

Fig. 1. A cantilever beam's model.

 K. Zhang, X. Yan / Journal of Sound and Vibration 387 (2017) 5365 55

2.1. Mathematical model describing relationship between cantilever beam's natural frequency change and crack characteristics

Sketch of a typical cantilever beam structure with variable cross-section is shown in Fig. 1. It has natural frequency i
and modal shape function Yi(x) in the ith vibration mode. Since modal shape function Yi(x) is usually normalized by mass, as
shown in Eq. (1), the ith natural frequency of cantilever beam i can be expressed as Eq. (2) [25].
l
A ( x) Yi ( x) dx = 1
2
0 (1)
l
i2 = 0 EJ ( x) Y i ( x) dx
2
(2)

In Eqs. (1) and (2), l, and E respectively denote the cantilever beam's length, density and Young's modulus; A(x) and J(x)
denote the cross-sectional area and moment of inertia at the location x. The product of E and J(x) reflects the cantilever
beam's stiffness at the location of x.
The crack existing in a cantilever beam is usually described as its location lc, depth dc and width wc [11], as shown in
Fig. 1. It will affect cantilever beam's stiffness near the crack area, but not change mass, since the parameter and A(x) is
constant. As a result, the ith natural frequency and modal shape function change from h,i and Yh,i(x) to be c,i and Yc,i(x)
(Note that, the subscript h and c respectively stand for healthy and cracked structure). And the relationship of cantilever
beam's modal parameter and crack's characteristics will be further discussed by introducing the concept of modal strain
energy.

2.1.1. Natural frequency of healthy structure

The structure's modal strain energy can be used to indicate the relationship of structure's strain energy and modal shape,
in the status of free vibration. For a healthy cantilever beam, its modal strain energy Wh,i in the ith vibration mode is
expressed as Eq. (3) [26].
1 l
EJ ( x) Yh, i ( x) dx
2
Wh, i =
2
0 (3)

Considering the expression of cantilever beam's natural frequency in Eq. (2), the relationship of cantilever beam's natural
frequency and modal strain energy can be expressed as Eq. (4).

h2, i
Wh, i =
2 (4)

2.1.2. Natural frequency of cracked structure

For a cracked cantilever beam as shown in Fig. 1, its modal strain energy Wc,i can be considered as the difference value
between healthy cantilever beam's modal strain energy Wh,i and the released strain energy Wi yielded by crack surface, as
expressed in Eq. (5);

2c, i
Wc, i = Wh, i Wi =
2 (5)

where Wi can be calculated by referring to strain energy release rate G defined by Griffith [27], as expressed in Eq. (6).

( 1 2) KI2 d
Wi = GdA = E (6)

In Eq. (6), is the Poisson's ratio of cantilever beam material, is the area of crack surface, and KI denote the stress
intensity factor. For an edge cracked beam shown in Fig. 2, the value of KI is expressed as Eq. (7).
d
KI = d c F c ( y )
bc (7)

In Eq. (7), the term F(dc/bc) is a geometrical factor depending on relative crack depth dc/bc. It is expressed as Eq. (8)
[2830].
d d d 2 d 3 d 4
F c = 1.122 0.231 c + 10.550 c 21.710 c + 30.382 c
bc bc bc bc bc (8)

And s(y) in Eq. (7) denotes the nominal stress at different layers of cracked location. It is expressed as Eq. (9);
2 m y
( y) =
h (9)

where sm is the nominal stress at the top layer. According to the cantilever beam's vibration equation [26], sm can be
56 K. Zhang, X. Yan / Journal of Sound and Vibration 387 (2017) 5365

Fig. 2. Cracked beam and its stress intensity s and stress intensity factor KI.

expressed by modal shape function Yi(x), as shown in Eq. (10).

Eh
m = Yi ( lc )
2 (10)

Thus, the released strain energy Wi can be written as Eq. (11).

d d 2
1 2 ) c F c bc EJ ( lc ) Yi ( lc )
2
Wi =
2
( bc bc (11)

2.1.3. Relationship between natural frequency changes and crack's characteristics

According to the expression of healthy and cracked cantilever beam's natural frequency, the change ratio of the ith
natural frequency ri and released strain energy Wi has following relationship.
i h, i c, i 2Wi
ri = = =
h, i h, i h, i (12)

While it also can be written as a product of a constant /h2, i , and two functions f(dc/bc) and Si(lc), as expressed in Eq. (13);

(1 2 2
) dc F dc b EJ
ri =
bc bc
c ( lc ) Yi ( lc ) 2
h2, i
d
= f c Si ( lc )
h, i bc
2
(13)

where function f is related to relative crack depth dc/bc and function Si is related to crack's location lc. In Eq. (13), , f and Si
are respectively expressed as Eq. (14) to Eq. (16):

= ( 1 2 ) (14)
d d d 2
f c = c F c
bc bc bc (15)
Si ( lc ) = bc EJ ( lc ) Yi ( lc )
2
(16)

If a cantilever beam has more than one crack, its total released modal strain energy is considered to be the summation of
the released modal strain energy caused by every signal crack [31]. Thus, the natural frequency change ratio in the ith
vibration mode can be expressed as Eq. (17):
m ( j)
d
ri = 2
f c Si lc( )
h, i j = 1 bc( j)
j
( ) (17)
 K. Zhang, X. Yan / Journal of Sound and Vibration 387 (2017) 5365 57

Fig. 3. Hammer test method for natural frequency measurement.

2.2. Measurement of cantilever beam's natural frequency

After obtaining the mathematical model of Eq. (17), The next issue for identifying cantilever beam's crack is to measure
cantilever beam's natural frequencies respectively in healthy status and cracked status, and further obtain the natural
frequency change ratio. Usually, cantilever beam's natural frequency can be measured by hammer test, as shown in Fig. 3. A
hammer applies a pulsed excitation to cantilever beam, and make it vibrate with its natural frequencies. Then, measuring
the vibration signal in time domain by a sensor, and processing this signal by spectral analysis [11], the natural frequencies
of the cantilever beam can be obtained. Traditional method of spectral analysis is FFT [20]. Though it has good performance
of analyzing linear vibration's frequency characteristic. Its resolution of frequency analysis is limited by original vibration
signal's sampling frequency and signal size. However, cantilever beam's crack identification requires natural frequency
results with higher resolution. And, the vibration of cracked cantilever beam is usually nonlinear because of crack's
breathing phenomenon [21], so FFT is no longer feasible to process nonlinear vibration signal [22].
The vibration signal excited by hammer test has several major signal components, which are respectively related to
cantilever beam's natural frequencies. For purpose of obtaining accuracy result of cantilever beam's natural frequencies,
first, the original vibration signal is decomposed into several signal components by utilizing empirical mode decomposition
(EMD) [32]. While each signal component's frequency is supposed to be equal to cantilever beam's natural frequency in
different modes. Then, to make the decomposed signal components have purer frequency spectrum, band pass filters [33]
are applied on each signal components. At last, by utilizing HHT [34], the signal components transient frequency varying
with time is obtained. And the average value of each signal components transient frequency is considered to be the can-
tilever beam's natural frequency in different vibration modes.

2.3. Detection and identification of multi-cracks on a cantilever beam

According to the discussion in Section 2.1, identifying multi-cracks in a cantilever beam with variable cross-sections
claims to detect the existence of cracks, identify the cracks location and relative depth.
The existence of cracks is detected by comparing the measured natural frequencies of the cantilever beam in working
status and initial healthy status. If these two sets of measured natural frequencies are equal, that means the beam in the
working status is still healthy. In reverse, the difference between these two sets of natural frequencies indicates that can-
tilever beam has cracks.
After the existence of cracks is detected, cracks characteristics are supposed to be identified. For m cracks on cantilever
beam, 2 m characteristics are needed to be determined, because one crack is described by two characteristics: crack location
lc and relative crack depth dc/bc. By substituting the measured natural frequency change ratios in 2m vibration modes into
the mathematical model established in Section 2.1, an equation set with 2m equations and 2m unknown parameters is
formed, as expressed in Eq. (18). And solving this equation set, all of the 2m unknown parameters are worked out, which
means that all of the m cracks location lc and relative depth dc/bc are identified.
m ( j)
r = f d c S l ( j)
1 2 ( )
h,1 j = 1 bc( j)
1 c

m ( j)
r = f d c S l ( j)
2 h2,2
j = 1 bc
( )
( j) 2 c


m ( j)
d


h,2m j = 1 b
c
( j) ( )
r2m = 2 f c S2m lc( )
j

(18)

It should be pointed out that, Eq. (18) is a nonlinear equation set. To solve this set of nonlinear equation, the Newton
iteration method [35,36] is applied in this investigation. First, a vector x is defined, which is expressed as Eq. (19).
58 K. Zhang, X. Yan / Journal of Sound and Vibration 387 (2017) 5365

d( )
1 ( 2) ( m)
x= c l ( 1) dc l ( 2) ... dc l ( m)


c c ( m) c
(
bc
1) (
bc
2)
bc
(19)

And rewriting Eq. (18) as Eq. (20).

m ( j)
r f d c S l ( j)
1
h2,1 j = 1 bc( j)
1 c ( )

F1 m ( j) 0

F2 r2 f dc S2 lc( j)
F= = h,2 j = 1 bc( j)
2 ( ) = 0



F2m
0
m ( j)
d



r2m 2 f c S2m lc( )
h,2m j = 1 b
( j)
j

c
( )

(20)

Thus, the iterative equations, which is used to solve Eq. (18), can be expressed as Eq. (21).
1
( ) F ( x ( ))
x ( k + 1) = x ( k) F x ( k) k
(21)

Considering the value of dc/bc belongs to a range [0, 1], and lc is greater than 0 and smaller than the whole length of
cantilever beam l, the convergence criterion is defined as Eq. (22);

( k
)
i = 1 x2( i 1 ) x2( i ) 1 +
m k+1 1
l
m
( ) x( ) )
i = 1 x2( i
k+1 k
2i
<
m
i = 1 (x ) +
(k)
2i 1
1
l
m
i = 1 ( x( ) )
k
2i (22)

where is a convergence factor, which should be chosen as a much small number. Iteratively calculating Eq. (21) until the
vector x(k) and x(k 1) satisfies Eq. (22), the detailed characteristics of cracks on cantilever beam can be obtained.
According to above analysis of cracks identification, to identify all cracks characteristics, the number of measured natural
frequencies in different vibration modes should be no less than twice of cracks number. But in fact, the number of actual
cracks is unknown before the identification. So the natural frequencies in different vibration modes are needed to be
measured as more as possible in this investigation. And because practical cantilever beam only has a few cracks, it usually
can be guaranteed that the number of measured natural frequencies is more than twice of number of cracks. If a cantilever
beam has m cracks, and its natural frequency change ratios are measured in 2 m (m 4m) vibration modes, all of the 2m
natural frequency change ratios should be substituted into Eq. (18), and there will be m identified cracks. Among all these m
identified cracks, there are m actual cracks and m-m false cracks. The actual cracks relative depths are significantly greater
than zero and smaller than one. While the false cracks relative depth are identified to approach to zero.

3. Procedures of multi-cracks identification

According to the principle of multi-cracks identification in Section 2, the procedures of crack identification can be
summarized as shown in Fig. 4.
The first step is measurement of natural frequencies. In this step, the cantilever beam's natural frequencies in 2m vi-
bration modes are respectively measured in the initial healthy status and working status by utilizing hammer test and
spectral analysis method proposed in previous.
The second step is detection of cracks. In this step, the measured cantilever beam's natural frequencies in healthy status
and working status are compared. And the existence of cracks is detected based on the difference of these two sets of
natural frequencies. The natural frequency change ratios in 2m vibration modes are further calculated.
The third step is establishment of a mathematical model, which describes natural frequency change ratio and cracks
characteristics. In this step, according to the cantilever beam's detailed geometry and material attributes, the constant , and
two functions f(dc/bc) and Si(lc) in the mathematical model are respectively calculated, as shown in Eq. (14) to Eq. (16). And
the mathematical model is further established, as a form of equation set Eq. (18).
The final step is to locate and evaluate the cracks. In this step, the measured natural frequency change ratios in 2m
vibration modes are substituted into Eq. (18), and utilizing Newton iteration method, m cracks are identified (including m
actual cracks and m-m false cracks). The actual cracks relative depth is identified to be significantly greater than zero and
smaller than one, while the false cracks relative depth is identified to approach to zero.
 K. Zhang, X. Yan / Journal of Sound and Vibration 387 (2017) 5365 59

Fig. 4. Summarized procedures of multi-cracks identification.

4. Example of multi-cracks identification

To prove the accuracy of the multi-cracks identification method presented in this investigation, experimental example is
conducted. In this example, two cantilever beams with variable cross-sections are adopted. These two cantilever beam have
the same geometric parameters and material attributes, as listed in Table 1, but different crack cases, as shown in Fig. 5. The
first beam has one crack, its relative location is lc/l 32%, and the crack's relative depth is dc/bc 10%. The second beam has
two cracks: one locates at lc/l 3%, and has relative depth of dc/bc 10%; another locates at lc/l 50%, and has relative depth
dc/bc 25%. All these cracks are machined by wire-electrode cutting in the edge of beam.
According to the summarized procedure of the multi-cracks identification method presented in previous, the following
steps are needed for cracks identification: (1) measurement of natural frequencies; (2) detection of cracks; (3) establishment
of mathematical model describing natural frequency change ratios and crack's information; (4) localization and evaluation
of cracks.

4.1. Measurement of natural frequencies

For the two adopted cantilever beams in the example, their natural frequencies in initial healthy status are measured at
first. And then, different cases of cracks are machined in these cantilever beams. At last, the cracked cantilever beams
natural frequencies are further measured.

Table 1
Geometric parameters and material properties of the cantilever beam.

Description Symbol Value

Length l 130 mm
Thickness h 1.8 mm
Width in the fixed end b1 12 mm
Width in the tip end b2 22 mm
Mass density 8900 kg/m3
Young's modulus E 97  109 N/m2
Poisson's ratio 0.27
60 K. Zhang, X. Yan / Journal of Sound and Vibration 387 (2017) 5365

Fig. 5. Crack cases for identification examples.

In the measurement of natural frequencies, hammer test is utilized to make cantilever beam vibrate with its natural
frequencies. The acquired vibration acceleration signal of cantilever beam in healthy status is shown in Fig. 6. This vibration
signal is decomposed into four signal components, which are respectively related to cantilever beam's natural frequencies in
the first four vibration modes. And band pass filters are applied on these signal components to make their frequency
spectrum be purer. After that, these decomposed and filtered signal components are processed by HHT to acquire their
transient frequency varying with time. The average value of each signal components transient frequency is considered to be
the cantilever beam's natural frequency in corresponding vibration mode.
For the cantilever beam in the healthy status, the four decomposed and filtered signal components, and their corre-
sponding transient frequency varying with time are shown in Fig. 7. From Fig. 7, the cantilever beam's natural frequencies in
the first four vibration modes can be observed, which are 46.625 Hz, 336.030 Hz, 992.345 Hz and 1866.505 Hz. Applying the
same method, the cracked cantilever beams natural frequencies in the first four vibration modes are also measured and
listed in Table 2.

4.2. Detection of cracks

From the measured natural frequencies of the cantilever beam in healthy status and two cracked statuses, it can be
observed that, both of the two cracked beams natural frequencies in the first four vibration modes are different compared
with their natural frequencies in healthy status. That indicates cracks existence in cantilever beam. The calculated natural
frequency change ratios of the cantilever beams for two crack cases are listed in Table 3.

4.3. Establishment of mathematical model

After cracks existence is detected, a mathematical model is supposed to be established for help of identifying cracks
characteristics. The general expression of this mathematical model is established in Section 2.1. Substituting the detailed
geometry parameters and material attributes of cantilever beam into the mathematical model's general expression, the
constant is calculated to be 2.799, and the function f(dc/bc) and Si(lc) are also calculated and shown in Fig. 8.
To verify the validity of the mathematical model, numerical simulation is conducted by using finite element method
(FEM). The healthy and cracked cantilever beams are respectively modeled as shown in Fig. 9. According to the established
mathematical model and FEM result, the cantilever beam's natural frequency change ratio varying with single crack's lo-
cation and relative depth is calculated and plotted in Fig. 10. In Fig. 10, the curves denote the cantilever beam's natural
frequency change ratio results, which are calculated by the established mathematical model, and the points are the FEM

Fig. 6. Vibration acceleration signal of the cantilever beam.

 K. Zhang, X. Yan / Journal of Sound and Vibration 387 (2017) 5365 61

Fig. 7. Decomposed and filtered signal components and their transient frequency varying with time. (a) The1st signal component and its transient fre-
quency varying with time. (b) The 2nd signal component and its transient frequency varying with time. (c) The 3rd signal component and its transient
frequency varying with time. (d) The 4th signal component and its transient frequency varying with time.

results of cantilever beam's natural frequency change ratios. From Fig. 10, it can be observed that, the results obtained by
mathematical model and FEM are almost the same, which verifies that the mathematical model proposed in this in-
vestigation is correct.
62 K. Zhang, X. Yan / Journal of Sound and Vibration 387 (2017) 5365

Table 2
Measured natural frequency results for healthy and cracked beam structure.

Vibration mode 1 Vibration mode 2 Vibration mode 3 Vibration mode 4

Healthy status 46.625 Hz 336.030 Hz 992.345 Hz 1866.505 Hz

Crack case 1 46.600 Hz 336.015 Hz 992.040 Hz 1865.475 Hz
Crack case 2 46.505 Hz 335.000 Hz 990.850 Hz 1862.900 Hz

Table 3
Calculated natural frequency change ratios.

Vibration 1 2 3 4
mode

Crack case 1 1.072E  03 8.928E  05 6.146E  04 1.103E  03

Crack case 2 5.141E  03 6.121E  03 3.011E  03 3.859E  03

Fig. 8. Calculated curves of function f and Si.

Fig. 9. Healthy and cracked cantilever beam's model. (a) Healthy cantilever beam. (b) Cracked cantilever beam.

4.4. Localization and evaluation of cracks

For each crack case, to locate and evaluate the cracks information, the natural frequency change ratios in four vibration
modes, which are measured in Section 4.1, are required to be substituted in the established mathematical model. And an
equation set with four unknown parameters and four equations is formed. Utilizing Newton iteration method, this equation
set is worked out, and two cracks are identified.
In Newton iteration method, the initial vector x(0) is chosen as {0.5, 0.3 l, 0.5, 0.6 l} for each crack case. That means the
(1) (1) (1) (2) (2) (2)
initial cracks characteristics are assumed to be lc 0.3 l, dc /bc 0.5 and lc 0.6 l, dc /bc 0.5. For each crack case,
iteratively calculating Eq. (21), the results of vector x and convergence criterion in each step are obtained and listed in
Table 4.
From Table 4, it can be seen that, the convergence criterion of case 1 is smaller than 1E-4 after five iterative calculations;
and for case 2, after nine iterative calculations, the convergence criterion is smaller than 1E-4. That means the cracks
identification iterative calculations for all crack cases are convergent. And the detailed identification results of multi-cracks
on cantilever beam are shown in Fig. 11.
 K. Zhang, X. Yan / Journal of Sound and Vibration 387 (2017) 5365 63

Natural frequency change ratio Natural frequency change ratio

dc/bc=10% Model FEM dc/bc=10 Model FEM

dc/bc=20% Model FEM dc/bc=20 Model FEM
dc/bc=30% Model FEM dc/bc=30 Model FEM

Relative crack location [%] Relative crack location [%]

(a) The1st vibration mode (b) The 2nd vibra tion mode
Natural frequency change ratio Natural frequency change ratio

dc/bc=10% Model FEM dc/bc=10% Model FEM

dc/bc=20% Model FEM dc/bc=20% Model FEM
dc/bc=30% Model FEM dc/bc=30% Model FEM

Relative crack location [%] Relative crack location [%]

(c) The 3rd vibration mode (d) The 4th vibration mode
Fig. 10. Curves of natural frequency change ratio varying with crack location and relative depth.

Table 4
Results of iterative calculations for each crack case.

Iterations ( 1) (k) ( 1) ( 2) (k) ( 2) Convergence criterion

(k) d x2 l (k) d x4 l
x1 c x3 c

c c
bc( ) bc( )
1 2

Crack case 1 1 0.44762 0.31445l 0.37963 0.41887l 0.17865

2 0.21365 0.32412l 0.12774 0.10436l 0.50667
3 0.124463 0.320463l 0.00712 0.01324l 0.39563
4 0.120475 0.320226l 0.00401 0.01139l 0.01974
5 0.120474 0.32022l 0.00401 0.01138l 2.6E  05
Crack case 2 1 0.32147 0.12479l 0.37963 0.52449l 0.28927
2 0.21365 0.06746l 0.33125 0.51108l 0.16806
3 0.11796 0.03994l 0.27462 0.50012l 0.16984
4 0.11469 0.03211l 0.28946 0.50014l 0.004032
5 0.11032 0.03146l 0.28466 0.50012l 0.01051
6 0.11047 0.03112l 0.28334 0.50010l 0.00165
7 0.11044 0.03124l 0.28230 0.50012l 0.00101
8 0.11043 0.03112l 0.28231 0.50011l 0.00014
9 0.11043 0.03111l 0.28230 0.50011l 2.2E  05

For case 1, the two identified cracks respectively locate at relative location of lc/l 1% and lc/l 32%. However, the first
identified crack has a relative depth approach to zero, which indicate that this identified crack can be ignored. The
relative depth of the actual relative crack locating at lc/l 32% is identified to be dc/bc 12%. For case 2, the two iden-
tified cracks locate at relative location of lc/l 3% and lc/l 50%, and the relative depth of these two cracks are re-
spectively evaluated to be dc/bc 11% and dc/bc 28%. The identification result show that the multi-cracks identification
method presented in this investigation can accurately identify the cracks locations, and the result of cracks relative
depth are also acceptable.
64 K. Zhang, X. Yan / Journal of Sound and Vibration 387 (2017) 5365

Fig. 11. Multi-cracks identification result for different crack case.

5. Conclusion

To identify multi-cracks characteristics in a cantilever beam with variable cross-sections, this investigation establishes a
mathematical model, which can describe cantilever beam's natural frequency change ratios and its cracks characteristics.
And a processing method based on Hilbert-Huang Transform (HHT) is also proposed, which can obtain cantilever beam's
natural frequency result with higher resolution. Based on above, a multi-cracks identification method is presented. To verify
the correctness of the multi-cracks identification method proposed in this investigation, experimental examples of multi-
cracks identification are conducted. And the identification result shows that the method presented in this investigation can
accurately identify multi-cracks location and relative depth.

Acknowledgements

This study is supported by National Natural Science Foundation of China (Grant No. 11272025), and Defense Industrial
Technology Development Program (Grant No. B2120132006).

References

[1] F.P. Kopsaftopoulos, S.D. Fassois, Vibration based health monitoring for a lightweight truss structure: experimental assessment of several statistical
time series methods, Mech. Syst. Signal Process. 24 (2010) 19771997.
[2] W.L. Bayissa, N. Haritos, S. Thelandersson, Vibration-based structural damage identification using wavelet transform, Mech. Syst. Signal Process. 22
(2008) 11941215.
[3] S. El-Ouafi Bahlous, H. Smaoui, S. El-Borgi, Experimental validation of an ambient vibration-based multiple damage identification method using
statistical modal filtering, J. Sound Vib. 325 (2009) 4968.
[4] A. Katunin, Vibration-based spatial damage identification in honeycomb-core sandwich composite structures using wavelet analysis, Compos. Struct.
118 (2014) 385391.
[5] X. Zhang, R.X. Gao, R. Yan, X. Chen, C. Sun, Z. Yang, Multivariable wavelet finite element-based vibration model for quantitative crack identification by
using particle swarm optimization, J. Sound Vib. 375 (2016) 200216.
[6] M. El-Gebeily, Y.A. Khulief, Identification of wall-thinning and cracks in pipes utilizing vibration modes and wavelets, Appl. Math. Model. 40 (2016)
53355348.
[7] Z. Li, X. Yan, X. Wang, Z. Peng, Detection of gear cracks in a complex gearbox of wind turbines using supervised bounded component analysis of
vibration signals collected from multi-channel sensors, J. Sound Vib. 371 (2016) 406433.
[8] F. Khoshnoudian, A. Esfandiari, Structural damage diagnosis using modal data, Sci. Iran. 18 (2011) 853860.
[9] X. Wang, N. Hu, H. Fukunaga, Z.H. Yao, Structural damage identification using static test data and changes in frequencies, Eng. Struct. 23 (2001)
610621.
[10] M. Link, M. Weiland, Damage identification by multi-model updating in the modal and in the time domain, Mech.Syst. Signal Process. 23 (2009)
17341746.
[11] G.-R. Gillich, Z.-I. Praisach, Modal identification and damage detection in beam-like structures using the power spectrum and timefrequency analysis,
Signal Process. 96 (Part A) (2014) 2944.
[12] D.-G. Kim, S.-B. Lee, Structural damage identification of a cantilever beam using excitation force level control, Mech. Syst Signal Process 24 (2010)
18141830.
[13] M. Cao, W. Xu, W. Ostachowicz, Z. Su, Damage identification for beams in noisy conditions based on Teager energy operator-wavelet transform modal
curvature, J Sound Vib 333 (2014) 15431553.
[14] K. Mazanoglu, M. Sabuncu, A frequency based algorithm for identification of single and double cracked beams via a statistical approach used in
experiment, Mech Syst Signal Process 30 (2012) 168185.
[15] H.Y. Guo, Z.L. Li, Structural damage identification based on Bayesian theory and improved immune genetic algorithm, Exp. Syst. Appl. 39 (2012)
 K. Zhang, X. Yan / Journal of Sound and Vibration 387 (2017) 5365 65

64266434.
[16] O.S. Salawu, Detection of structural damage through changes in frequency: a review, Eng. Struct. 19 (1997) 718723.
[17] P. Cawley, R.D. Adams, The location of defects in structures from measurements of natural frequencies, J. Strain Anal. Eng. Des. 14 (1979) 4957.
[18] A. Esfandiari, F. Bakhtiari-Nejad, A. Rahai, Theoretical and experimental structural damage diagnosis method using natural frequencies through an
improved sensitivity equation, Int. J. Mech. Sci. 70 (2013) 7989.
[19] A. Majumdar, D.K. Maiti, D. Maity, Damage assessment of truss structures from changes in natural frequencies using ant colony optimization, Appl.
Math. Comput. 218 (2012) 97599772.
[20] M. Nishiyama, K. Watanabe, Frequency characteristics of hetero-core fiber optics sensor for mechanical vibration, Sens. Actuators A: Phys. 209 (2014)
154160.
[21] T.H. Patel, A.K. Darpe, Influence of crack breathing model on nonlinear dynamics of a cracked rotor, J. Sound Vib. 311 (2008) 953972.
[22] M. Silani, S. Ziaei-Rad, H. Talebi, Vibration analysis of rotating systems with open and breathing cracks, Appl. Math. Model. 37 (2013) 99079921.
[23] A. Baccigalupi, A. Liccardo, The Huang Hilbert transform for evaluating the instantaneous frequency evolution of transient signals in non-linear
systems, Measurement 86 (2016) 113.
[24] Q. Zhang, Y. Wang, Y. Sun, L. Gao, Y. Yue, HilbertHuang Transform based method for monitoring the crack of concrete arch by using FBG sensors, Optik
- Int. J. Light Electron Opt. 127 (2016) 34173422.
[25] G. Genta, Vibration of Structures and Machines: Practical Aspects, Springer, New York, 1999.
[26] J.R. Barber, Intermediate Mechanics of Materials, Springer, Dordrecht, Netherlands, 2011.
[27] N. Perez, Fracture Mechanics, Kluwer Academic Publishers, Boston, 2004.
[28] P.K. Jena, D.N. Thatoi, J. Nanda, D.R.K. Parhi, Effect of damage parameters on vibration signatures of a cantilever beam, Procedia Eng. 38 (2012)
33183330.
[29] P. Gudmundson, Eigen frequency Changes of structures due to cracks, notches or other geometrical changes, J. Mech. Phys. Solids 30 (1982) 339353.
[30] H. Tada, P. Paris, G. Irwin, The Analysis of Cracks Handbook, ASME Press, New York, 2000.
[31] M. Kisa, Free vibration analysis of a cantilever composite beam with multiple cracks, Compos. Sci. Technol. 64 (2004) 13911402.
[32] J. Wang, P. Shang, J. Xia, W. Shi, EMD based refined composite multiscale entropy analysis of complex signals, Phys. A: Stat. Mech. Appl. 421 (2015)
583593.
[33] Y. Qin, J. Wang, B. Tang, Y. Mao, Higher density wavelet frames with symmetric low-pass and band-pass filters, Signal Process. 90 (2010) 32193231.
[34] D. Guo, Z.K. Peng, Vibration analysis of a cracked rotor using HilbertHuang transform, Mech. Syst. Signal Process. 21 (2007) 30303041.
[35] X. Wang, W.-W. Li, L. Dai, On inexact Newton methods based on doubling iteration scheme for symmetric algebraic Riccati equations, J. Comput. Appl.
Math. 260 (2014) 364374.
[36] S. Zhao, An efficient implementation scheme of the simplified Newton iteration for block systems of nonlinear equations, Appl. Numer. Math. 39 (2001)
225237.