See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/268357894

Vibration-based analysis of damage in composites

Article  in  Materials Forum · January 2008

CITATIONS READS

5 364

4 authors, including:

Mohammad Mehdizadeh
RAMPF COMPOSITE SOLUTIONS, ONTARIO, CANADA
9 PUBLICATIONS   10 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Mohammad Mehdizadeh on 04 May 2016.

The user has requested enhancement of the downloaded file.

MATERIALS FORUM VOLUME 33 - 2009
Edited by Dr Steve Galea, Associate Professor Wingkong and Professor Akira Mita
© Institute of Materials Engineering Australasia Ltd

VIBRATION-BASED ANALYSIS OF DAMAGE IN COMPOSITES

K. Oruganti 1, M. Mehdizadeh 1, S. John 1, I. Herszberg 2
1
RMIT University, School of Aerospace, Mechanical & Manufacturing Engineering, Bundoora Campus,
Bundoora, VIC 3083, Australia
2
Cooperative Research Centre for Advanced Composite Structures (CRC-ACS),
506 Lorimer Street, Fishermans Bend, VIC 3207, Australia
Corresponding Author: S John. Email: [email protected]

ABSTRACT
In recent times, composite materials are supplanting conventional metals in aerospace, automotive, civil and
marine industries. These material types are however prone to delamination thus necessitating the early detection
of the flaw before it initiates into a serious defect. Constant health monitoring of such structures is highly
essential to prevent sudden failure. An offline approach was commonly employed where the parts were removed
from service. This not only consumed a lot of time but risked damage to the part during operation and handling.
Hence, the demand for a robust Real Time Health Monitoring System which not only detects the presence of the
damage but also estimates the remaining life of the composition. A detailed understanding of the various proven
methods and techniques and their applicability in the analysis of vibration signatures obtained from damaged
structures under dynamic conditions is essential to develop a reliable Structural Health Monitoring System
(SHMS). This paper highlights the utilisation of Vibration based testing on Carbon/Epoxy composite beams for
damage detection. Such composites are commonly used in the aerospace and marine industry. This paper reports
the processing of the vibration signatures from healthy and damaged composite beams upon excitation and
analysis of the mode shapes acquired. The study comprises of testing carbon/epoxy composite beams with
various embedded delaminations with a mechanical actuator and a scanning laser vibrometer (SLV) as a sensor
for recording the frequency response and analysing the acquired signatures based on Displacement and
Curvature Mode Shapes. The paper also includes Finite Element Analysis (FEA) and C-scan techniques used on
the composite beams to validate the results. In recent times, there has been a lot of focus on vibration based
damage detection as it has proved to be more reliable and as it can be incorporated into an online health
monitoring system for damage identification. Thus, this paper also attempts to summarize the recent trends and
progress within this area of SHM.

KEY WORD

Damage Detection, Dynamic response, Displacement mode shape, Curvature Mode shape

1 INTRODUCTION
There are a number of methods for the detection
Composite materials are replacing conventional and localisation of the damage in a composite
metals in aerospace, automotive, civil and marine material. In general, structural damage detection
industries in modern times. This is mainly due to can be classified as global-damage detection and
their high strength and light weight characteristics. local-damage detection [10]. Local-damage
But with all the advantages they have, they are detection largely refers to techniques referring to
prone to delamination. Thus it is highly essential non-destructive testing (NDT) such as CT scanning
that the initial crack be detected before it initiates and ultrasonic methods. It is primarily used to
into a serious flaw [1,2,3]. Constant health detect local damage in structures-damage existence
monitoring of these structures is required to prevent and its location. The main advantage of local
sudden failure. An offline approach was often used damage detection techniques are data and
where in the parts were removed from theoretical models of the undamaged structure are
service/operation and examined [5]. This consumes not needed. However, for large and more complex
a lot of time. Such an approach also risks damage structures, it is very difficult to detect damage
to the parts. Hence a real time health monitoring using any of the local damage detection techniques.
system is needed to reduce these risks and save Thus, local damage detection techniques can only
time. be used to detect damages in certain parts of the

496
 

structure. For detecting damage throughout the acquired structural response signals. Therefore, it is
whole structure, especially some large and complex important to develop and extract a more reliable
structures, a methodology called global vibration- and sensitive damage feature index with the aid of
based structural damage detection [2,3] has been advanced methods by combination of artificial
proposed. In brief, it is based on the principle intelligence and signal processing. The traditional
where the structure is assumed as a dynamic system type of methods are basically 1)method based on
with mass, stiffness, and damping. Upon damage the change of natural frequency 2) method based on
within the structure, there would be a change in the the change of structural flexibility or stiffness 3)
structural parameters along with the frequency- method based on statistical information 4) method
response function (FRF) and modal parameters of based on power flow.  Added to the TTDD
the structural system. Hence, the change of the (Traditional type damage detection) methods,
structural modal parameters can be treated as a several concepts of Modern-type vibration-based
signal of early damage occurrence within the structural damage detection methods have also
structural system [9]. been investigated. Briefly,

Vibration-based structural damage detection is a a) Wavelet Analysis Method (WA):

relatively new and emerging area of research
within SHM and its development can be divided The idea behind using this method is based on
into traditional and modern type. The traditional- splitting the dynamic signal of a structural response
types of damage detection methods utilises into a series of local basic function called wavelets
mechanical characteristics of structures, such as [1,10], so as to detect the special characteristics of
natural frequencies, modal damping, modal shapes, the structure using scaling and transformation
etc. However, this kind of method generally property of wavelets. The spectrum graph obtained
requires experimental modal analysis or transfer using wavelet transform can indicate the damage
function measure, and is not very existence directly.
convenient for online detection of structures. The
so called ‘‘traditional-type” damage detection b) Neural Network Method (NN):
techniques have three main disadvantages: (1)
More experimental dependence is time-consuming Within a well trained NN, one can input the
and proves expensive (2) Using traditional-type experimentally measured real structural damage
method for damage detection, it is very difficult to feature index into the trained NN, and the output of
establish a universal methodology for various the trained NN will be able to give the location and
structures as the properties shall vary according to severity of the structural damage. The prerequisite
size, shape construction etc. (3) The traditional- to realize this intention is only to obtain the
type methods generally are not very sensitive to dynamic response data of damaged structure by
initial tiny damage within the structures. The online measurement and some known information
“modern-type” such as techniques incorporating to train the NN. The dynamics-based damage
vibration signatures for analysis refer to the detection is an effective method due to its ability of
damage detection methods based on response acquiring both the global and the local information
signals acquired from excitation of structures. Its of the structure.
advantages can be summarised as: (1) Compared to
the traditional type techniques, it is less dependant c) Curvature mode shape assessment:
on experiments. Vibration responses at few points
on the structure are sufficient for damage detection. Lestari and Hanagud [5] used few lower order
(2) Using the more modern-type techniques, curvature modes to identify damage location based
smaller structural damage can be detected by the on the difference between the acquired data of
construction and extraction of better characteristic healthy and damaged structures. The mathematical
information from structural dynamic response relationship of both the measured data and dynamic
signals. Among the modern-type methods for parameters was used to estimate the severity of the
structural damage detection, the representative ones damage. The detection technique was extended
include Wavelet analysis, Genetic algorithm (GA) using a Heaviside step function. The amplitude
and Artificial Neural Network (ANN). difference of curvature mode shapes was employed
to estimate the local stiffness loss due to damage.
The principal challenges in structural damage The first three curvature modes were used in the
detection are how to ascertain emergence, location damage assessment as the curvature nodal points of
and extent of structural damage using the obtained damaged structures at higher modes may shift
structural dynamic responses. While identifying significantly from the original undamaged case.
initial damage within structures, the influence on This could generate misleading results. Hamey et
structural response caused by a small damage al. [7] evaluated several damage detection
within the structure is very minute, and it is algorithms in carbon/epoxy composite beams with
difficult to identify the small damage from the several possible damage configurations. The results

497
 

proved that all the methods of the curvature modes which might include changes in the second moment
measured by the piezoelectric sensors could be of inertia, material loss due to damage, friction or
used as a potential tool in damage detection contact effects between delaminated part and
methods and the damage index method (DIM) primary part for delamination case. By assuming
detected and isolated the damage better than other is the stiffness loss parameter at the location of
ones reviewed. Saying that, in all the algorithms damage, , the effect of damage on the
used, the information of healthy structures was
needed in the implementation. These algorithms stiffness distribution of the beam was expressed as
could possibly be a vital part of the neural network.
Lestari and Qiao [2] also adopted a similar
approach of using the curvature mode shapes to (2)
detect debonding and core crushing in composite where H is a Heaviside step function; EI0 is the
honeycomb sandwich beams, and both the location bending stiffness of the undamaged structures; x1
and relative size of damage were evaluated. In their and x2 are the beginning and the end of damage
study, the delamination detection in composite area, respectively as shown in Fig 1.
plates was carried out using a scanning laser
vibrometer (SLV) with a mechanical actuator, from
which the displacement mode shapes were
measured directly. The curvature mode shapes
were then implemented with a few relatively new
damage detection algorithms to evaluate the
location and size of the delamination.
Experimentally measured data of dynamic response
are sometimes not directly useful for damage Figure 1. Details of damage region
detection. They are often coupled with damage
detection algorithms, as part of data reduction The derivation of mathematical relationship of
procedures to extract the useful information. highly localized or notched type of damage could
be found in the previous work by Lestari [5].
2 THEORETICAL BACKGROUND Currently, extension of this analysis taking the
thickness (h) of the panel into consideration is in
progress.
Equation of Motion
The analytical relationship is developed based on
Euler–Bernoulli beam theory, and the governing 3 EXPERIMENTAL VALIDATION
equation for beam in vibration can be expressed as
In this study, experiments on the Vibrometer (SLV)
were conducted on carbon/epoxy composite beams
with simulated defects to evaluate system setup
parameters.  Various beams (Fig 2) were
(1) manufactured by the vacuum bag resin infusion
where EI(x) is the bending stiffness distribution of technique (VBRI) and hand lay-up technique [14,
the beam, m(x) is the mass distribution of the beam 15, 16]. A total of 9 specimens were manufactured
per unit length along the x-axis, and W(x,t) is the within which 7 samples had artificial
displacement of the beam as a function of the delaminations, 1 specimen with no delamination
location (x) and time (t). Lestari and Hanagud [5] (healthy specimen) and 1 specimen for material
have extended this relation onto a thin composite testing. Carbon fibre reinforced epoxy
plate using a Heaviside step function to address the unidirectional (UD) prepregs each 0.125mm thick
extent of damage in the panel. The authors found were used to make the composite beams along with
the effect of the neutral axis changes due to the Teflon impregnated glass film (~60µm) inserted as
damage and assumed it to be negligible. The artificial delaminations. A total of 32 plies were cut
distribution of the bending stiffness along the beam and arranged into a 550mm X 450mm rectangle
is a function of location of damage, and the loss of and stacked using the hand layup technique
stiffness due to the damage was expressed as a non- [14,15,16] using a hand roller onto a steel plate in a
dimensional parameter. The stiffness loss factor Quasi Isotropic & Symmetric layup [(+45 0 -45
was introduced in the model as a measure of the 90)s]n ; n=6 (Figure 2).
extent of the damage presence in the structures.
This factor represents equivalent bending stiffness
loss in the damaged area, which can be caused by a
combination of several factors. Few parameters

498
 

Figure 2. Various carbon/epoxy beams manufactured with defects for testing

After the over-laminate was fully cured, the In addition, the natural frequencies are very
composite was ready to be machined to make the sensitive to interference, especially at the higher
required beam specimens for testing. order modes.

Vibrations were induced in the samples using a

mechanical shaker while being fixed to the shaker
using thin double sided tape to minimize damping.
An experimental setup was prepared using the
Laser Vibrometer for testing of these composites as
shown in Fig 3. Upon excitation, the responses
were recorded using the Laser Vibrometer software
for analysis. A database consisting of responses
from all the carbon epoxy samples was recorded for
further evaluation. The effect of damage on
frequency response primarily provides global
information about the condition of structures.

Figure 3. Experimental Setup

499
 

Few examples of interference include fibre The difference in the displacement and curvature
misalignment during manufacture, introduction of mode shapes between the intact and the damaged
non-negligible mass by sensors, and simulated specimen is utilized to detect the location of the
approximate boundary conditions that prompt the defect. The difference in the signal between the
largest error in the frequency measurement [17]. healthy and defect sample would potentially
The location of the defect can be estimated with the indicate the presence and location of the damage. A
percentage of change in the natural frequencies of difference algorithm was used to estimate the
intact and damaged structures which depend on the change in displacement between healthy and defect
position of the defect for a particular mode of sample to the original value. The resultant value (δ)
vibration.  From the responses, the first 4 modes described below highlighted the change in
were selected and displacement mode shapes for signatures. 
the selected modes were generated. When the
responses of the healthy sample to the defect were
compared against each other, there were noticeable
differences in the peaks as well as difference in the (3)
displacement mode shapes. This difference
potentially indicates the presence of damage where, - Magnitude of displacement of a point
within the sample compared against the healthy on the reference specimen acquired. -
one. The response signatures obtained after
comparing sample #1 composite beam with the Magnitude of displacement of the same point over
response signature of the healthy specimen specimen under test (defect sample). - Ratio of
(assumed as the reference) is illustrated in Figure 4 absolute difference of displacement magnitudes
and 5. between Healthy and Defect specimen to the
reference displacement magnitude from a healthy
4 MODE SHAPE ANALYSIS specimen. was computed for modes 1, 2 and 3
respectively. Lestari et al [5], used few lower order
A crack or flaw within a structure reduces the modes to identify damage location based on the
stiffness and increases the damping in the structure. difference between the acquired data of healthy and
Such a reduction in stiffness is allied with damaged structures. Lower order modes were
decreases in the natural frequencies and alteration preferred in the damage assessment as the nodal
of the modes of vibration of the structure. Several points of damaged structures at high modes may
researchers have used one or more of the above shift significantly from the original undamaged
characteristics to detect and locate a crack. case. This could generate misleading results.
Systematic change in the first mode shape with The resultant curve could be used to estimate the
respect to the damage location within a cantilever approximate damage size and location within the
beam was observed during the analysis of various specimen under test. Only results from mode 1
mode shapes [19]. They based their findings using were applicable for damage assessment as higher
finite element analysis to obtain the natural modes produced misleading peaks. Numerous
frequencies and the mode shapes of a damaged researchers have studied the systematic change in
structure. From changes in frequency they could the first mode shape with respect to damage
determine the presence of crack or damage in a identification [19].
structure. Numerous researchers have probed into
the utility of displacement mode shapes and
curvature mode shapes in damage detection [18].

500
 

a  b 

Figure 4. Comparison of Full range FRF signal between Sample 1 and a healthy Sample

a 

Healthy Sample 
b 

Figure 5. Close up plots for portions (a) & (b) in Figure 4.

501
 

The actual location of the delamination is validated magnitudes. The resultant curvature plots indicate
using the C-Scan result as shown in Figure 6. the presence and location of damage. As observed
from the curvature plots for individual samples
shown in Fig 7, only Mode 1 was analysed to
indicate the presence of damage within the
specimen as higher modes produce results which
were less convincing. From the resultant curve (Fig
8) using the subtracting algorithm, the sudden spike
in magnitudes extends between points 11 and 15.
As the figure illustrates, the sudden spike in
magnitudes extends between points 11 and 15.
Figure 6. Actual location of delamination within sample #1 as Thus the damage zone is approximately 38.68 mm
determined from a C-Scan similar to the estimation from displacement mode
shapes.
Numerous researchers have probed into the utility
of curvature mode shapes in damage detection Sample 1‐ Curvature Mode Shape‐ Mode 1
[18].The difference in the curvature mode shapes
3.00E‐04
between the intact and the damaged specimen is

Magnitude 
2.00E‐04
utilized to detect the location of the defect. 1.00E‐04
0.00E+00
 The Curvature ( at a point is given by ‐1.00E‐04
1 6 11 16 L5
(4)
21 26 31 36 41 46
51 56
Where, Sample Point
The curvature at a section,
M - The bending moment at the section,
E - The modulus of elasticity Healthy Sample ‐ Curvature Mode 
I - The second moment of the cross-sectional 3.00E‐04 Shape‐ Mode 1
area.
Magnitude 

2.00E‐04
1.00E‐04
Curvature mode shape can be obtained by using
piezoelectric sensors mounted on the sample [20] 0.00E+00
or by the application of a numerical algorithm to ‐1.00E‐04
1 6 11 16 L5
convert displacement mode shapes to curvature 21 26 31
36 41 46 51
mode shapes. Curvature mode shape is the second 56
Sample Point
derivative of a displacement mode shape. By using
the Central Difference Approximation (CDA)
technique over displacement mode shapes; Figure 7. Curvature mode shape for sample 1 and healthy
curvature mode shapes can be generated. Equation sample for the first mode of vibration
below estimates the curvature at a point using the
central difference approximation method.  
 

(5)
where,
The curvature at a section or nodal point
– Magnitudes of displacements at nodal
points 1, 2 and 3 respectively
h- Distance between 2 consecutive nodal points.

Using the central difference approximation method

the curvatures for the samples were calculated from
the displacement data obtained from the PSV-400 Figure 8. Resultant curvature plot between defect sample #1 and
Healthy sample
Laser Vibrometer System. A subtracting algorithm
was employed to compute the absolute difference
between the healthy and defect curvature

502
 

1 FINITE ELEMENT ANALYSIS The mode shapes acquired from FEA were very
similar and nodal displacements were very close to
the experimental results. Fig 10 shows the resultant
Finite element analysis (FEA) has wide curvature plots for mode 1 acquired by both FEA
applications and enjoys extensive utilization in the and LSV system as an example.
structural, thermal and fluid analysis areas [19].
The advantages of FEA are numerous and a
important. In this study, FEA was initiated to
complement and validate the damage detection
algorithms. Initial FE models were created by
modelling the composite with the solid element
(8node82) in ANSYS. A load was applied load (in
direction) at 400mm from the left end of the
beam, node numbering was used to get x = 0.4 m, y
= 0.0 m as shown in Figure 9.A bilinear element
(Link10) in ANSYS was used to model the  
delamination as shown in Figure 9. A delamination   Excitation pt.
was simulated using Link10 which is a 3-D spar   Delamination

element designed as compression-only element,  

 
which means that the stiffness is zero if the element
b
is under tension.  

Delamination (0.25 mm thick)

100 40

150
Delamination Excitation pt.
1-2 kHz
550 Figure 10. Comparison of Mode 1 Resultant curvature plots
acquired from (a) ANSYS and (b) LSV system

Figure9. FE model describing damage. The discrepancy between the frequency values
shown in Table 1 is probably due to the boundary
The initial results from the FE analysis that were condition differences between the 2D model and
obtained for various natural frequencies and the actual experimental test coupon, where only a
displacement mode shapes were very encouraging. fraction of the edge was constrained.
Table 1 describes the comparison between natural
frequencies obtained experimentally and using
FEA. CONCLUSION

Table 1. Comparison between Natural Frequencies obtained The damage within the Carbon/Epoxy beams could
from FE model and LSV system. be identified successfully by both the displacement
mode shape analysis as well as curvature mode
Mode FEA Laser Vibrometer shape analysis. The displacement mode shape
provides a global snapshot through the frequency
(Hz) (Hz) domain range, is a valid method for a globalized
technique of damage identification. The curvature
1 136.795 111.3 mode shape being the double derivative of the
displacement mode shape is more of a localised
2 336.526 278.1 technique. The size and location of the
delamination in one place was successfully
3 468.86 540 identified by both these techniques and validated
using the C-scan results. A denser scanning mesh is
4 1207 1005 employed by the use of the Laser Vibrometer
system as compared to a system using a PZT-

503
  

PVDF system. It can measure the displacement 8 Pandey, A.K., Biswas, M. and Samman, M.M.
response at individual scan points, thus generating 1991. ‘‘Damage Detection from Changes in
the displacement mode shapes. The present study Curvature Mode Shapes,’’ Journal of Sound
focuses on the identification of the presence, and Vibration, 145(2):321–332.
location, and size of delamination in the laminated 9 Kessler, S.S., Spearing, S.M., Atalla, M.J.,
composite plates by extracting the modal Cesnik, C.E.S. and Soutis, C. 2002. ‘‘Damage
parameters obtained from the SLV measurement Detection in Composite Materials using
system. The experimental program comprises of Frequency Response Methods,’’ Composites
testing Carbon/epoxy composite laminated plates Part B: Engineering, 33(1):87–95.
with embedded delamination using the SLV 10 Y.J. Yan, L.H. Yam, Online detection of crack
actuator–sensor measurement system. Finite damage in composite plates using embedded
Element Analysis (FEA) has been initiated to piezoelectric actuators/sensors and wavelet
complement the damage detection and to analysis, Composite Structures 58 (1) (2002)
demonstrate the validity of the damage detection 29–38.
algorithms. Future work involves improving the 11 C.Y. Kao, S.L. Hung, Detection of structural
FE model to increase the overall accuracy. The damage via free vibration responses generated
observation of the results indicates that the by approximating artificial neural networks,
experimental approach using the SLV measurement Computers & Structures 81 (28–29) (2003)
systems is somewhat successful in extracting the 2631–2644.
modal parameters and thus helps in determining 12 J.H. Chou, J. Ghaboussi, Genetic algorithm in
the existence, location and size of delamination. structural damage detection, Computers &
Structures 79 (14) (2001) 1335–1353.
References 13 C.J. Lu, Y.T. Hsu, Vibration analysis of an
inhomogeneous string for damage detection by
wavelet transform, International Journal of
1 Y.J. Yan, L. Cheng, Z.Y. Wu L.H. Yam., Mechanical Sciences 44 (4) (2002) 745–754.
Development in vibration-based structural 14 Mallick, P., Fibre-Reinforced Composites:
damage detection technique. Mechanical Materials, Manufacturing and Design. 1 ed.
Systems and Signal Processing 21 (2007) 1988, New York: Marcel and Dekker Inc.
2198–2211 15 Barbero, E., Introduction to composite
2 Lestari W, Qiao PZ. Damage detection of materials design. 1 ed. 1999, Philadelphia:
fiber-reinforced polymer honeycomb sandwich Taylor and Francis Inc.
beams. Composite Structures 2005;67(3):365– 16 Schwartz, M., Composite materials volume II:
73. processing, fabrication and application. 1 ed.
3 Ratcliffe CP, Bagaria WJ. Vibration technique 1996, New Jersey: Prentice Hall PTR
for locating delamination in a composite beam. 17 Kessler, S.S., Spearing, S.M., Atalla, M.J.,
AIAA J 1998; 36:1074–7. Cesnik, C.E.S. and Soutis, C, Damage
4 M.K. Yoon, D. Heider, J.W. Gillespie, C.P. Detection in Composite Materials using
Ratcliffe, R.M. Crane, Local damage detection Frequency Response Methods. Composites
using the two-dimensional gapped smoothing Part B: Engineering, 2002. 33(1): p. 87–95.
method, Journal of Sound and Vibration 279 18 Hamey, C.S., Lestari, W., Qiao, P., Song, G,
(1–2) (2005) 119–139. Experimental Damage Identification of
5 Lestari, W. and Hanagud, S. 2001. ‘‘Detection Carbon/Epoxy Composite Beams using
of an Edge Notch Defect by using a Single Curvature Mode Shapes. International Journal
Mode Based Methods,’’ In: Proceedings of 3rd of Structural Health Monitoring, 2004. 3(4): p.
International Workshop on SHM, pp. 1343– 333–353.
1355. 19 Yan, Y.J., L. H. Yam, FEM modeling method
6 Adams, R.D., Cawley, P., Pye, C.J. and Stone, of damage structures for structural damage
B.J. 1978. ‘‘A Vibration Technique for Non- detection. Composite Structures, 2006. 72(2):
destructively Assessing the Integrity of p. 193-199.
Structures,’’ Journal of Mechanical 20 Lestari W, Q.P., Damage detection of fiber-
Engineering Science, 20(2):93–100. reinforced polymer honeycomb sandwich
7 Hamey, C.S., Lestari, W., Qiao, P. and Song, beams. Composite Structures, 2005. 67(3): p.
G. 2004. ‘‘Experimental Damage Identification 365–7. 
of Carbon/Epoxy Composite Beams using
Curvature Mode Shapes,’’ International
Journal of Structural Health Monitoring,
3(4):333–353.

504

View publication stats