Vous êtes sur la page 1sur 13

52nd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference<BR> 19th

4 - 7 April 2011, Denver, Colorado

AIAA 2011-1938

Application of the Campbell Diagram Concept to


Identification of Fatigue Cracks in Bladed Disk
Assemblies
Oleg V. Shiryayev
Petroleum Institute, Abu Dhabi, United Arab Emirates

Joshua Gaerke, Phillip Cooley and Joseph C. Slater


Wright State University, Dayton, OH 45435

The goal of this research is to further develop a Structural Health Monitoring (SHM)
system for detection of fatigue cracks in blades and bladed disks using harmonic responses
caused by crack nonlinearities. Driving weakly nonlinear systems, such as a cracked structure, results in the generation of harmonics. When these harmonics coincide with a resonance, the resulting information can be used to identify the location of the crack. This
paper presents analytical results obtained for a simple beam model illustrating the phenomenon. The results suggest that superharmonic resonances caused by weak nonlinearity
are a suitable crack detection feature.

I.

Introduction

A means of an automated Structural Health Monitoring (SHM) system for turbomachinery components
would greatly decrease the cost of maintenance inspections and increase the likeliness of structural failure
prevention. Degradation in structures generally involves corrosion, cracking, delamination of composite
layers, and loosened fasteners. In the case of turbine engine disks, blades, and other components, a SHM
system would be greatly advantageous for the purposes of fatigue crack detection and evaluation. Fatigue
cracks are one of the most common causes of failures in turbomachinery (and in many other mechanical
systems) and yet one of the hardest types of damage to detect. Due to the large amount of kinetic energy
stored in the moving parts of turbomachinery, fatigue cracks can cause uncontained failures, which may have
catastrophic consequences. Debris flying out of the engines may damage vital aircraft systems leading to
loss of control and a consequent crash,1 or fatalities on board the aircraft after penetrating the fuselage.2
Therefore, development of reliable techniques for detection of damage is highly critical.
New sensor and actuator capabilities have helped in the development of damage detection. Damage
detection can be performed optically, or using x-ray, ultrasound, and vibration methods. Vibration characteristics of structures, however, has struggled in its development due to focus on behaviors that are sensitive
to effects other than damage, such as thermal effects. Currently, vibration techniques for damage detection
include using time, frequency, and modal concepts,36 most of which do not exploit the inherent nonlinear
behavior of fatigue cracks which would help distinguish damage from other benign phenomenon. In many
instances linear vibration testing for damage detection is not sensitive enough to the very localized behavior that comes from hairline fatigue cracks. Mode shapes and especially natural frequencies can be very
insensitive to fatigue cracks until the damage is too far along.
Unlike many types of damage, fatigue cracks are typically hard to find by inspection due to little if any
material being lost. They do become observable, however, when looking at the nonlinear elastic behavior
that results from the opening and closing of the crack. This nonlinear behavior is very localized and acts
like a bilinear local modulus when strained. This nonlinearity is well known to be true of beams with cracks
Assistant

Professor, Petroleum Institute, AIAA Member


Assistant, Department of Mechanical and Materials Engineering.
Professor, Department of Mechanical and Materials Engineering, joseph.slater@wright.edu. AIAA Associate Fellow
Research

1 of 13
American
of Aeronautics
and Astronautics
Copyright 2011 by Joseph C. Slater and Oleg Shiryayev. Published
byInstitute
the American
Institute of Aeronautics
and Astronautics, Inc., with permission.

also.7, 8 Shiryayev and Slater have exploited this phenomena using changes in statistics of the Randomdec
signatures cause by the onset of nonlinearity due to a crack. This method was developed and then verified
experimentally.911 The SHM community has also investigated other nonlinear response features (see for
example1216 ).
Previous work by Meier, Shiryayev and Slater17, 18 identified candidate features that can be used to
detect fatigue cracks in bladed fan or turbine disks from vibration data. In this document we describe the
extension of the crack detection approach by considering a spectral fingerprint of the structure obtained
by continuously changing the excitation frequency. We describe further investigations of these frequency
response features utilizing a model of a beam. We consider a simple structure instead of a full FE model
of a hypothetical compressor disk18 due to its computational efficiency. This is a necessary step towards
developing effective automated monitoring techniques.

II.

Nonlinear Response Features Resulting From Harmonic Excitations

The bilinear type weak nonlinearity that occurs due to the presence of an opening and closing crack can
be roughly approximated by a system that has a quadratic stiffness term in addition to the linear one as
shown in Equation (1). Higher polynomial terms can also be included to improve the approximation. Only
terms that are even functions will have non-zero coefficients.
y + 20 y + 02 y + y 2 = F (t)

(1)

In this equation 0 is the natural frequency of the linear system, and is small relative to 02 . One of the
differences between responses of linear and nonlinear systems due to harmonic excitations is that responses
of nonlinear systems may contain harmonics other than those at the driving frequencies. Nayfeh and Mook19
provide a detailed analysis of systems with quadratic and cubic nonlinearities under harmonic excitations.
If the system with quadratic nonlinearity is excited by a two term harmonic excitation such as in Equation
(2),
F (t) = F1 (t) + F2 (t) = A1 cos(1 t + 1 ) + A2 cos(2 t + 2 )
(2)
then besides the primary resonance (1 or 2 close to 0 ) several interesting phenomena can be observed
in the response depending on the relationships between driving frequencies 1 and 2 , as well as the linear
natural frequency of the system 0 .
Superharmonic resonance can be excited if 1 or 2 1/20 .
Subharmonic resonance can be excited when 1 or 2 20 .
Combination resonance occurs when 1 + 2 0 , or 1 2 0 .
An alternative approach to account for the nonlinearity introduced by a crack is to consider the forces
resulting from the crack motion directly. When a structure is excited, it experiences alternating states of
tension and compression. Since a crack cannot support a tensile load, it opens as the tension grows in that
section of the structure and closes when tension reverts to compression. As the crack closes, a gradually
increasing compressive force is applied to the structure. When subject to sinusoidal excitation, his non-linear
behavior is best approximated by the half-wave rectified sinusoid shown in Equation 3. The peaks in the first
plot of Figure 1 correspond to compression and the flat regions correspond to tension. The Fourier series
coefficients are shown in the stem plot in Figure 1 and indicate that the presence of a crack will induce a
significant excitation at twice the driving frequency. The zeroth term and fourth term will also amplify the
response but may not be significant enough to be useful.

X
A A
2A
A sin(1 t) + abs (A sin(1 t))
= + sin(1 t)
cos(n1 t)
Fc (t) =
2 1)
2

2
(n
n=2

2 of 13
American Institute of Aeronautics and Astronautics

(3)

Time Domain

Amplitude

0.5

0
0

T/2

2T

3T

Relative Time
Frequency Domain
0.8

Amplitude

0.6
0.4
0.2
0

3
4
5
6
Fundamental Frequency Multiples

10

Figure 1. Half Wave Rectified Sinusoid

This phenomenon will cause a resonant response to occur when a cracked beam is excited at one half
a natural frequency and accounts for the additional peak for the cracked beam in Figure 2. Contrary to
anticipated, but not observed, behavior in previous work by Meier et al., it should be noted in Figure 1
that there is no term corresponding to half of the driving frequency. This implies that there will be no
subharmonic resonance for a breathing crack despite the prediction of such by the quadratic stiffness model.
Further, there is no n = 3 harmonic, but there is a small n = 4 harmonic that is unlikely to be useful.
Linear Beam

ABS fft, dB

0
20
40
60
80

50

100

150
Frequency, Hz

200

250

200

250

Cracked Beam

ABS fft, dB

0
20
40
60
80

50

100

150
Frequency, Hz

Figure 2. Combination resonance excited by two driving functions.

The amplitude of the superharmonic excitation introduced by the crack depends on several factors. The
first is the location of the excitation. Away from resonances, operational deflection shapes (responses to
harmonic excitations) are not close to mode shapes. Instead they are linear combinations of mode shapes.
Further, the shapes are highly dependent on location of excitation, as illustrated by the sensitivity of antiresonances to excitation location. As a result, moving the excitation has a strong direct impact on the strain
amplitude in the vicinity of the crack. As the strain amplitude in the vicinity of the crack is directly related to

3 of 13
American Institute of Aeronautics and Astronautics

the superharmonic excitations that it generates (there is also a zero frequency excitation that will likely not
be observable), knowledge of the excitation location, and the resulting dependence of superharmonic response
amplitude to it, will be posed as an optimization problem to determine the location of the excitation. As
numerous locations (similar to nodal lines) will have identical strains in a given operational deflection shape,
multiple modes will be excited in phases of testing and a means to optimize this process can be determined.
Because the excitation caused by the crack itself is an order of magnitude less (or smaller yet) than the
primary excitation, we will be ignoring, in analysis, the change in response of the system at primary driving
frequency relative to the healthy system to that primary driving frequency. This is a reasonable assumption
since for very small cracks, the energy dissipation is very slight. This slight damping has been used in the past
for damage detection,2024 but our goals here are to focus on where the energy does go, and use resonance
to amplify and find it. Thus we can actually use two separate families of FRF results, one for direct force
input, and one for moment input: the first for use in determining responses (and thus surface strains) to
the primary excitations, and the second to determine the response expected by the induced surface strain
excitation at a resonance. Both can be done on a model basis, and the second is simply a family of modal
cofactors, albeit in mixed direct/strain form.
To account for these resonant responses caused by remote frequency excitations, the sophistication of
current methodology will need to be increased. To achieve this, a much simpler model of a cracked beam
with multiple natural frequencies is being experimented with prior to working on a structural model of a
representative disk. This model represents a beam mounted in clamp on a shaker as shown in Figure 3.

Figure 3. Schematic of a beam mounted on a shaker.

III.

Beam model

The numerical model of the system consists of two linear submodels: one for the case when the crack is
in the closed state, and the other for the crack in the open state. This implies that there are two eigenvalue
problems resulting in two sets of mode shapes and natural frequencies. Switching between the two submodels
occurs when the crack opens and closes, which is checked at every time step based on the curvature of the
beam near the crack location. We proceed by deriving the modal equations for the case when the crack is
closed.
III.A.

Closed crack case

The governing partial differential equation for the continuous Euler-Bernoulli beam is written in Eq. (4),
where y(x, t) is the displacement of any point along the beam.
A

2 y(x, t)
y(x, t)
4 y(x, t)
+
D
+
EI
=0
c
t2
t
x4

Assuming separable solution we can write the total displacement as


X
y(x, t) = w(t) +
Yi (x)qi (t)
i

4 of 13
American Institute of Aeronautics and Astronautics

(4)

(5)

P
where w(t) is the displacement of the clamp and i Yi (x)qi (t) is the sum of modal components representing
displacement with respect to the spring hinge. Substituting Eq. (5) into Eq. (4), dividing by A and
rearranging gives
X

Yi (x)
qi (t) +

Dc
EI X d4 Yi (x)
Dc X
qi (t) = w(t)

Yi (x)qi (t) +
w(t)

A i
A i
dx4
A

(6)

In this derivation we assume the mode shapes are similar to the undamped case and are written as
Yi (x) = a1i cos i x + a2i sin i x + a3i ei x + a4i ei x
One can verify that
X
i

d4 Yi (x)
dx4

(7)

= i4 Yi (x), then:

Yi (x)
qi (t) +

Dc X
EI X 4
Dc
Yi (x)qi (t) +
Yi (x)qi (t) = w(t)

w(t)

A i
A i i
A

(8)

Figure 4. Spring hinge beam configuration.

The eigenvalue problem for the clamped-free beam can be formulated by considering the boundary
conditions (see Fig. 4).
at

x=0:

at

x=L:

d2 Yi (0)
dYi (0)
= Kc
dx2
dx
d3 Yi (0)
d2 Yi (0)
= 0,
=0
dx2
dx3

Yi (0) = 0, EI

(9)
(10)

where Kc is the torsional stiffness of the spring hinge. A set of four equations can be obtained from conditions
in Eqs. (9) and (10). Characteristic equation is obtained by setting the determinant of the resulting matrix
to zero. Solution of the characteristic equation was calculated using Matlab.25 The resulting mode shapes
are expressed as:



 
2i EI
2i EI
Yi (x) = ci (1 + i ) cos i x +
+1+
1 i sin i x
Kc
Kc

+ ei x + i ei x
(11)
where coefficients i are defined as

i = 

2i EI
Kc

2i EI
Kc

(sin i L + cos i L) + 2 sin i L


(12)

( sin i L cos i L) + 2 cos i L + 2ei L

p
Coefficients i are related to the natural frequencies as i = i2 EI/(A). Constant ci is chosen to normalize
the mode shapes
Z L
AYi2 (x)dx = 1
(13)
0

5 of 13
American Institute of Aeronautics and Astronautics

Utilizing orthogonality of the modes, we multiply Eq. (6) by AYi (x) and integrate over the length of the
beam, which results in the set of modal equations shown below.


Z L
Dc
EI 4
Dc
AYi (x)dx w(t)

qi (t) +
qi (t) =
w(t)

(14)
qi (t) +
A
A i
A
0
In a more common form Eq. (14) can be rewritten as
Z L
qi (t) + 2i i qi (t) + i2 qi (t) =
AYi (x)dx (w(t)
2i i w(t))

(15)

where i and i are the corresponding modal damping ratio and natural frequency. Note that the righthand-side of Eq. (15) represents modal forcing, which is dependent on the motion of the clamp.
III.B.

Open crack case

When the crack is open (see Fig. 5), the beam is divided into two continuous segments connected by a
torsional spring as shown in Fig. 6. According to Sundermeyer and Weaver,26 the stiffness of the torsional
spring KT is calculated based on the concepts of linear fracture mechanics and Castiglianos theorem as
KT =

Ebh2
72F1 (ac /h)

(16)

where F1 (ac /h) is the shape factor that depends on geometry and loading. The shape factor is calculated as
F1 (ac /h)

19.6(ac /h)10 40.69(ac /h)9 + 47.04(ac /h)8 32.99(ac /h)7


+ 20.29(ac /h)6 9.975(ac /h)5 + 4.602(ac /h)4
1.047(ac /h)3 + 0.6294(ac /h)2

(17)

Figure 5. Geometry of a transverse edge crack.

Figure 6. Representation of the beam with an open crack.

The mode shapes for the two beam segments in Fig. 6 are written as
Y1i (x1 ) = a1i sin i x1 + a2i cos i x1 + a3i ei x1 + a4i ei x1
Y2i (x2 ) = b1i sin i x2 + b2i cos i x2 + b3i ei x2 + b4i ei x2
6 of 13
American Institute of Aeronautics and Astronautics

(18)
(19)

The eigenvalue problem is formed in order to calculate the mode shapes and natural frequencies by
considering boundary conditions and continuity conditions at the location of the crack. At the clamped end
(x1 = 0) and the free end (x2 = 0):
at

x1 = 0 :

at

x2 = 0 :

dY1 (0)
=0
dx1
d3 Y2 (0)
d2 Y2 (0)
=
0,
=0
dx22
dx32

Y1 (0) = 0,

(20)
(21)

Continuity of displacement, shear force and bending moment is enforced at the crack location (x1 = Xc ,
x2 = L Xc ):
Y1 (Xc ) = Y2 (L Xc )
d3 Y2 (L Xc )
d Y1 (Xc )
=
3
dx1
dx32


d2 Y1 (Xc )
KT dY1 (Xc ) dY2 (L Xc )
=
+
dx21
EI
dx1
dx2


2
d Y2 (L Xc )
KT dY1 (Xc ) dY2 (L Xc )
=
+
dx22
EI
dx1
dx2

(22)

(23)
(24)
(25)

Eqs. (20-25) are written in matrix form and the characteristic equation is formed by equating the determinant
to zero. Solution of characteristic equation is calculated numerically in Matlab. The resulting mode shapes
are expressed as




sin i x + ei x +
Y1i = i
Ai cos i x + 2 Ai 1 2Ki EI
c

i x
(Ai 1) e
, 0 < x < Xc
(26)

Y2i = i i

Bi cos i (L x) (Bi + 2) sin i (L x) + ei (Lx)



(Bi + 1)ei (Lx) , Xc 6 x < L

where coefficients i are expressed as





Ai cos i Xc + 2 Ai 1 2Ki EI
sin i Xc + ei Xc + (Ai 1) ei Xc
c
i =
Bi cos i (L Xc ) (Bi + 2) sin i (L Xc ) + ei (LXc ) (Bi + 1)ei (LXc )

(27)

(28)

Coefficients Ai and Bi are expressed as shown in the appendix, and i are normalization constants such that
Eq. (29) holds.
Z Xc
Z L
AY1i2 (x)dx +
AY2i2 (x)dx = 1
(29)
0

Xc

The parameters for the beam were chosen to replicate the experimental setup that was used in previous
work.17 The beam was made of 2024 aluminum (E = 70 GPa, = 2700 kg/m3 ), with a free length of 0.66 m,
and a rectangular cross-section of 0.0250.0125 m. The distance from the clamp to the crack was Xc = 0.257
m. The stiffness of the torsional spring Kt that represents flexibility of the shaker head assembly has been
varied to achieve natural frequencies that are close to those observed in the past series of experiments.17 The
models use the modal damping ratios obtained from the experimental data. In Table 1, a large difference is
observed between the model and experimental value at the 1st natural frequency. However, higher modes
have frequency values that match experimental data to within 2%. In this work we are mostly concerned
with higher modes because they are the ones that are most affected by the presence of cracks.
Figure 7 illustrates the mode shapes obtained for the the open and closed crack submodels. As expected
the first several modes are mostly unaffected by the presence of the crack. However, modes 4 and 5 have
more noticeable differences between the baseline and the open crack states.
7 of 13
American Institute of Aeronautics and Astronautics

2
baseline
cracked

2.5

baseline
cracked

1.5
1

0.5
0
Y(x)

Y(x)

1.5

0.5

1
1.5

0.5

2
0

2.5

0.5
0

0.1

0.2

0.3

0.4

0.5

0.6

3
0

0.7

0.1

0.2

0.3

X, m

(a) Mode 1.

0.6

0.7

2
baseline
cracked

2.5

baseline
cracked

1.5

1.5

0.5
0
Y(x)

1
0.5

0.5

0.5

1.5

1.5

2.5
0.1

0.2

0.3

0.4

0.5

0.6

3
0

0.7

0.1

0.2

0.3

X, m

0.4

0.5

X, m

(c) Mode 3.

(d) Mode 4.

3
baseline
cracked

2.5
2
1.5
1
Y(x)

Y(x)

0.5

(b) Mode 2.

2
0

0.4
X, m

0.5
0
0.5
1
1.5
2
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

X, m

(e) Mode 5.
Figure 7. Mode shapes of the baseline and cracked beam, a/h = 0.2, Xc = 0.26.

8 of 13
American Institute of Aeronautics and Astronautics

0.6

0.7

Mode
1
2
3
4

Experimental value
(cracked beam)
15.20
139.0
399.9
770.6

Baseline beam
model
22.56
142.2
399.9
786.8

Cracked beam
model, a/h=0.2
22.52
141.8
399.2
786.0

Damping ratio
0.0401
0.0083
0.0018
0.0012

Table 1. Natural frequencies (Hz) and damping ratios of experimental setup and beam models.

The numerical algorithm for the simulation of the cracked beam model utilizes two submodels within
which the beam behaves linearly. It is the switching between two different linear submodels that is responsbile
for the bilinear behavior of the system. The state of the crack is checked at every time step based on the
curvature of the beam near the crack. The curvature of the beam near the crack is calculated based on the
displacements of two points on each side of the crack and the point at the crack location. A coordinate
transformation is performed to maintain continuity of displacement and velocity fields at the intstants when
the state of the crack changes.

IV.

Results

In this work we compare the data obtained from the model of the baseline healthy beam and the model
of the cracked beam. We consider exciting superharmonic resonance near the third mode of the structure at
approximately 400 Hz, hence the driving frequency is varied from 195 Hz to 205 Hz with an increment of 0.5
Hz. Excitation to the structure is provided in terms of acceleration of the shaker head with an amplitude of
40 m/s2 . For each excitation frequency, a 5 seconds long time history displacement data has been obtained
for the point 5 mm away from the free end of the beam. The response reaches the steady state within
the first 3 seconds, hence for further analysis we considered the data in the range of [4, 5] seconds. The
displacement response data was windowed using a Hann window and the FFT was calculated to obtain the
frequency content of the response.
Figure 8 demonstrates the spectral contents of the responses obtained from the models of healthy beam
and cracked beams with the crack depths of 10% and 20% of the cross-section height respectively. The data
obtained from the model of the healthy beam represents typical spectral content of a linear system excitated
away from resonances. There is a signle well-defined peak at the excitation frequency, which could be easily
observed on both the waterfall and contour plots.
The waterfall and contour plots in Figures 8(c)-8(f) suggest that response data obtained from cracked
beam models contains additional harmonics besides the one at the excitation frequency. There are clear
peaks at 2 and 4 the excitation frequency. There is also a spectral line at 3 the excitation frequency,
but it is not nearly as well-defined compared to the ones at even multiples of excitation frequency.
From Figure 8 it is difficult to observe if the superharmonic resonance has been excited when the driving
(3)
frequency passed through the region near 0.5n at approximately 200 Hz. In order to observe excitation of
superharmonic resonances, magnitudes of the harmonics at 2 and 4 the excitation frequency are plotted
in Figure 10 versus the excitation frequency. As the driving frequency passes through the region near 200
Hz, the magnitude of the 2 harmonic is about 10 dB larger than compared to its values when the driving
frequency is 195 Hz or 205 Hz. A similar observation can be made for the magnitude of the 4 harmonic.
To consider this further, another case was evaluated from 95 Hz to 105 Hz to observe the impact on the
4 harmonic when it coincides with a natural frequency. As seen in Figure 9, the 4 harmonic is more
prominent than the 2 under these circumstances and reaches a maximum as the driving frequency passes
through one quarter the natural frequency. These observations suggest that the superharmonic resonances
can be observed in the system with bilinear stiffness characteristic. Also, as the crack becomes larger the
magnitudes of both harmonics increase substantially. The magnitudes of 2 and 4 harmonics for the beam
with a crack 20% deep are about 15 dB higher than those from the beam with a crack that is 10% deep.

9 of 13
American Institute of Aeronautics and Astronautics

Displacement Data

Displacement Data
204
203
Excitation frequency, Hz

FFT magnitude, dB

50
0
50
100
150
200
205

202
201
200
199
198
197

1000
196

200

500
195

Driving frequency, Hz

195
Response frequency, Hz

100

(a) Baseline beam, waterfall plot.

200

300

400 500 600


Frequency, Hz

700

800

(b) Baseline beam, contour plot.


Displacement Data
205

Displacement Data
0

204
203
Excitation frequency, Hz

FFT magnitude, dB

50
100
150
200
250
205

201
200
199
198
197
196

200
Driving
frequency, Hz

202

195

200

800
600
400
Response frequency, Hz

195

1000

(c) Cracked beam, a/h=0.1: waterfall plot.

100

200

300

400 500 600


Frequency, Hz

700

800

(d) Cracked beam, a/h=0.1: contour plot.


Displacement Data
205

204

50

203
Excitation frequency, Hz

FFT magnitude, dB

Displacement Data

100
150
200
250
205

202
201
200
199
198
197

200
Driving
frequency, Hz

196
195

200

800
600
400
Response frequency, Hz

1000

(e) Cracked beam, a/h=0.2: waterfall plot.

195

100

200

300

400 500 600


Frequency, Hz

700

800

(f) Cracked beam, a/h=0.2: contour plot.

Figure 8. Spectral contents of the responses obtained from baseline healthy and cracked beam models.

10 of 13
American Institute of Aeronautics and Astronautics

Figure 9. Contour plot of response driven at 1/4 of natural frequency.

Displacement Data

Displacement Data
30
2 harmonic
4 harmonic

35
40
45
50
55
60
65
70
75
80
195

200
Excitation frequency, Hz

Nonlinear effect magnitude, dB abs(fft)

Nonlinear effect magnitude, dB abs(fft)

30

(a) Cracked beam, a/h =0.1.


Figure 10.
(3)
0.5 n
.

40
45
50
55
60
65
70
75
80
195

205

2 harmonic
4 harmonic

35

200
Excitation frequency, Hz

205

(b) Cracked beam, a/h =0.2.

Magnitudes of the peaks at 2 and 4 the excitation frequency when the excitation frequency is near

11 of 13
American Institute of Aeronautics and Astronautics

V.

Conclusions and Future Work

In this work we considered a spectral fingerprint of the structure for detection of fatigue cracks. This
fingerprint can be obtained by exciting the structure away from any resonances. The fingerprint of a linear
undamaged structure contains a single well-defined harmonic due to excitation. The fingerprints of a cracked
beam contain additional easily observed harmonics at even multiples of excitation frequency. The magnitude
of these harmonics increases by almost an order of magnitude when excitation is near half of a specific natural
frequency, targeting superhamonic resonance condition. In practical setting, such spectral fingerprints can
be easily obtained for the structure of interest.
Future work will focus on experimental implementation of this method. Initially, we will focus on validating analytical results obtained from the beam models using a setup similar to the model described in
this paper: the beam mounted in a clamp on a shaker. Ideally, a non-contact sensor will be used in the
experiment.

VI.

Appendix

Coefficients Ai and Bi in Eq. (26, 27, 28, 29) are expressed as


Ai =

d34 d12 d23 d34 d13 d22 d32 d23 d14 + d32 d24 d13 + d33 d22 d14 d33 d24 d12
d33 d21 d14 d33 d24 d11 d31 d23 d14 + d31 d24 d13 + d34 d11 d23 d34 d13 d21

d31 d24 d12 d31 d22 d14 + d34 d22 d11 d34 d21 d12 d32 d11 d24 + d32 d14 d21
d32 d11 d23 + d32 d13 d21 + d33 d22 d11 d33 d21 d12 d31 d22 d13 + d31 d23 d12
where coefficients dmn for each mode i are given as:


2i EI
d11 = 1
sin(i Xc ) + cos(i Xc ) ei Xc
Kc
Bi =

d31

(30)

(31)

(32)

d12 = 2 sin(i Xc ) + ei Xc ei Xc

(33)

d13 = cos(i (L Xc )) sin(i (L Xc )) ei (LXc )

(34)

d14 = 2 sin(i (L Xc )) + ei (LXc ) ei (LXc )

(35)



2i EI
d21 = sin(i Xc ) 1
cos(i Xc ) ei Xc
Kc

(36)

d22 = 2 cos(i Xc ) ei Xc ei Xc

(37)

d23 = sin(i (L Xc )) cos(i (L Xc )) + ei (LXc )

(38)

d24 = 2 cos(i (L Xc )) ei (LXc ) ei (LXc )

(39)






Kt
2i EI
Kt
Kt
i Xc
sin(i Xc ) 1 +
e
+ 1
cos(i Xc ) sin(i Xc )
= cos(i Xc )
i EI
i EI
Kc
i EI
(40)






Kt
Kt
Kt
d32 = 1
ei Xc 1 +
ei Xc + 2
cos(i Xc ) sin(i Xc )
(41)
i EI
i EI
i EI

Kt  i (LXc )
d33 =
e
sin(i (L Xc )) + cos(i (L Xc ))
(42)
i EI

Kt  i (LXc )
d34 =
e
ei (LXc ) 2 cos(i (L Xc ))
(43)
i EI
12 of 13
American Institute of Aeronautics and Astronautics

References
1 Anonymous, United Airlines Flight 232, McDonnel Douglas DC-10-10, Sioux Gateway Airport, Sioux City, Iowa, July
19, 1989, Tech. Rep. NTSB/AAR-90/06, National Transportation Safety Board, 1990.
2 Anonymous, Uncontained Engine Failure Delta Air Lines Flight 1288 McDonnel Douglas MD-88, N927DA, Pensacola,
Florida, July 6, 1996, Tech. Rep. NTSB/AAR-98/01, National Transportation Safety Board, 1998.
3 Doebling, S. W., Farrar, C. R., Prime, M. B., and Shevitz, D. W., Damage Identification and Health Monitoring
of Structural and Mechanical Systems from Changes in Their Vibration Characteristics: A Literature Review, Tech. Rep.
LA-13070-MS, Los Alamos National Laboratory, Los Alamos, New Mexico, 87545, 1996.
4 Doebling, S. W., Farrar, C. R., and Prime, M. B., A Summary Review of Vibration-Based Damage Identification
Methods, Shock and Vibration Digest, Vol. 30, No. 2, March 1998, pp. 91105.
5 Worden, K. and Dulieu-Barton, J., An Overview of Intelligent Fault Detection in Systems and Structures, Structural
Health Monitoring, Vol. 3, No. 1, 2004, pp. 8598.
6 Carden, E. P. and Fanning, P., Vibration Based Condition Monitoring: A Review, Structural Health Monitoring,
Vol. 3, No. 4, December 2004, pp. 355377.
7 Dimarogonas, A. D., Vibration of Cracked Structures: A State of the Art Review, Engineering Fracture Mechanics,
Vol. 55, No. 5, 1996, pp. 831857.
8 Friswell, M. I. and Penny, J. E. T., Crack Modeling for Structural Health Monitoring, Structural Health Monitoring,
Vol. 1, No. 2, 2002, pp. 139148.
9 Shiryayev, O. V. and Slater, J. C., Improved Structural Health Monitoring Using Random Decrement Signatures:
Application to FEM Data, Structural Control and Health Monitoring, Vol. 15, No. 7, November 2008, pp. 10061020.
10 Shiryayev, O. V. and Slater, J. C., Structural Damage Identification Using Random Decrement Signatures From Experimental Data, Proceedings of the 49th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials
Conference, No. AIAA-2008-2163, April 2008.
11 Shiryayev, O. V. and Slater, J. C., Detection of Fatigue Cracks Using Random Decrement Signatures, Structural Health
Monitoring-An International Journal, Vol. 9, No. 4, July 2010, pp. 347360.
12 Epureanu, B. I. and Yin, S.-H., Identification of Damage in an Aeroelastic System Based on Attractor Deformations,
Computers and Structures, Vol. 82, 2004, pp. 27432751.
13 Hillis, A., Neild, S., Drinkwater, B., and Wilcox, P., Global Crack Detection Using Bispectral Analysis, Proceedins of
the Royal Society A: Mathematical, Physical and Engineering Sciences, Vol. 462, No. 2069, May 2006, pp. 1515 1530.
14 Hunter, N. F., Bilinear System Characterization from Nonlinear Time Series Analysis, Proceedings of the 17th International Modal Analysis Conference, Vol. 2, February 1999, pp. 14881494.
15 Moniz, L., Nichols, J., Nichols, C. J., Seaver, M., Trickey, S. T., Todd, M. D., Pecora, L. M., and Virgin, L. N., A
Multivariate, Attractor-Based Approach to Structural Health Monitoring, Journal of Sound and Vibration, Vol. 283, 2005,
pp. 295310.
16 Nichols, J., Todd, M. D., Seaver, M., and Virgin, L. N., Use of Chaotic Excitation and Attractor Property Analysis in
Structural Health Monitoring, Physical Reviews, Vol. E 67, No. 016209, 2003.
17 Meier, M., Shiryayev, O. V., and Slater, J. C., Investigation of Candidate Features For Crack Detection in Fan and
Turbine Blades and Disks, Proceedings of the 50th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and
Materials Conference, No. AIAA-2009-2671, May 2009.
18 Shiryayev, O. V. and Slater, J. C., Sensitivity Studies of Nonlinear Vibration Features For Detection of Cracks in
Turbomachinery Components, Proceedings of the 51st AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics,
and Materials Conference, No. AIAA-2010-3030, April 2010.
19 Nayfeh, A. and Mook, D., Nonlinear Oscillations, Wiley Interscience, 1995.
20 Li, H. C. H., Weis, M., Herszberg, I., and Mouritz, A. P., Damage Detection in a Fibre Reinforced Composite Beam
Using Random Decrement Signatures, Composite Structures, Vol. 66, 2004, pp. 159167.
21 Yang, J. C. S., Chen, J., and Dagalakis, N. G., Damage Detection in Offshore Structures by the Random Decrement
Technique, ASME Journal of Energy Resources Technology, Vol. 106, March 1984, pp. 3842.
22 Tsai, T., Yang, J. C. S., and Chen, R. Z., Detection of Damages in Structures by the Cross Random Decrement
Technique, Proceedings of the 3rd International Modal Analysis Conference, Orlando, FL, January 1985, pp. 691700.
23 Yang, J. C. S. and Caldwell, D. W., The Measurement of Damping and the Detection of Damage in Structures by the
Random Decrement Technique, 46th Shock and Vibration Bulletin, November 1975, pp. 6768.
24 Zubaydi, A., Haddara, M. R., and Swamidas, A. S. J., Random Decrement Technique for Damage Identification of
Stiffened Plates, 18th International Modal Analysis Conference, San Antonio, TX, USA, 2000, pp. 13991405.
25 The MathWorks Inc., 3 Apple Hill Drive, Natick, MA 01760, M atlab ,
R r2008a ed.
26 Sundermeyer, J. and Weaver, R., On Crack Identification and Characterization in a Beam by Non-linear Vibration
Analysis, Journal of Sound and Vibration, Vol. 183, No. 5, 1995, pp. 857871.

13 of 13
American Institute of Aeronautics and Astronautics

Vous aimerez peut-être aussi