Thesis T Ooijevaar PDF
Thesis T Ooijevaar PDF
Thesis T Ooijevaar PDF
Ted Ooijevaar
VIBRATION BASED STRUCTURAL HEALTH
MONITORING OF COMPOSITE SKIN-STIFFENER
STRUCTURES
Ted Ooijevaar
De promotiecommissie is als volgt samengesteld:
Voorzitter en secretaris:
prof.dr. G.P.M.R. Dewulf Universiteit Twente
Promotor:
prof.dr.ir. R. Akkerman Universiteit Twente
prof.dr.ir. T. Tinga Universiteit Twente
This research project was financially supported by the Eco-Design ITD within the
Clean Sky framework (grant agreement number CSJU-GAM-ED-2008-001).
ISBN 978-90-365-3624-0
DOI 10.3990/1.9789036536240
Copyright © 2014 by T.H. Ooijevaar, Enschede, The Netherlands
Printed by Ipskamp Drukkers B.V., Enschede, The Netherlands
Cover: photograph of the composite skin-stiffener structures used in the present research.
The curved lines represent the steady state velocity responses measured at the damaged area
of a pristine (upper curve) and impact damaged (lower curve) skin-stiffener structure.
VIBRATION BASED STRUCTURAL HEALTH
MONITORING OF COMPOSITE SKIN-STIFFENER
STRUCTURES
PROEFSCHRIFT
door
te Wognum
Dit proefschrift is goedgekeurd door de promotoren:
prof.dr.ir. R. Akkerman
prof.dr.ir. T. Tinga
Summary
i
ii
are attributed to the interaction between the skin and stiffener that occurs during
opening and closing motion of the damage. It is shown that linear damage
identification methods (e.g. modal domain methods) are feasible for low excitation
amplitudes, but the presence of nonlinear dynamic effects cannot remain silent for
higher amplitudes. The nonlinear dynamic effects can act as strong indicator of
damage, but can also be useful for characterization purposes.
The nonlinear dynamic effects introduced by the skin-stiffener damage urges
the development of nonlinear damage identification methods. A study on the
understanding and feasibility of using nonlinear vibro-acoustic modulations for
the detection, localization and characterization of impact damage in a composite T-
beam is presented. A time domain analysis at multiple spatial locations is used
to detect and localize impact damage in a skin-stiffener connection, based on
locally increased amplitude modulation effects. Analysis of the characteristics of
the nonlinear modulations opens the ability to characterize the nonlinear dynamic
behavior introduced by the damage at the skin-stiffener interface.
The work presented in this thesis showed that the relations between the
characteristics of the structure, the potential damage scenarios and the damage
identification method together define the performance of the vibration based
damage identification strategy. Therefore, it is concluded that the design of
a vibration based damage identification strategy is made-to-measure work and
requires a thorough physical understanding of the potential failure mechanisms,
the critical damage locations and their effect on the dynamic behavior. To aid in
this process, a scenario based procedure for the design of a damage identification
strategy is proposed. All findings presented in this thesis contribute to the
development of a design tool for research engineers, to assist the implementation
of structural health monitoring technology in safety-critical composite structures.
Samenvatting
Composiet materialen combineren een hoge sterkte en stijfheid met een relatief
lage dichtheid. Deze materialen kunnen echter complexe schadevormen vertonen,
zoals transversale scheuren en delaminaties. Deze schades kunnen de structurele
eigenschappen van een component aanzienlijk beïnvloeden. Periodieke inspecties
zijn noodzakelijk om de integriteit van een component gedurende zijn levensduur te
kunnen garanderen. De huidige inspectietechnieken zijn vaak tijdrovend, kostbaar
en vereisen dat de componenten gemakkelijk toegankelijk zijn. Het monitoren van
de structurele integriteit op basis van trillingen is een veelbelovend alternatief en
omvat het continue monitoren van een component middels een geïntegreerd sensor-
systeem. Deze technieken zijn gebaseerd op het concept dat het dynamische gedrag
van een component kan veranderen indien schade optreedt.
Hoewel er veel schade-identificatiemethodes zijn beschreven in de literatuur, zijn
er nog tal van moeilijkheden bij de praktische toepassing van deze methodes,
vooral voor complexe structuren. De prestaties van een schade-identificatiemethode
gebaseerd op trillingen zijn sterk afhankelijk van het ontwerp van een structuur
en de schade die wordt beschouwd. Dit proefschrift richt zich op de identificatie
van de schade in verstijfde composiet structuren. Het hoofddoel is om
ontwerprichtlijnen te ontwikkelen voor de detectie, lokalisatie en karakterisatie
van schade in verstijfde composiet structuren gebaseerd op veranderingen in het
dynamische gedrag.
Een literatuurstudie ondersteund door een analytisch model toonde aan dat de
kromming van trilvormen in combinatie met een modale schadeindex (MSE-DI)
algoritme een potentieel krachtige parameter en methode zijn voor de identificatie
van schade in verschillende verstijfde composiet structuren. Een experimentele
opstelling, inclusief een shaker en laser-vibrometer, is gebruikt om het dynamisch
gedrag te meten. Een lineair dynamische systeembeschrijving is verkregen
door het toepassen van een experimentele modaal analyse. De dynamische
metingen toonden aan dat het MSE-DI-algoritme in staat is om nauwelijks zichtbare
impactschade (BVID) in verstijfde composiet structuren te detecteren, te lokaliseren
en een schatting van de omvang te geven. Geconcludeerd is dat de methode
bijzonder effectief is voor het monitoren van de integriteit van de verbinding tussen
de huid en de verstijver. De methode bleef onbeslist in het geval dat de schade zich
puur in de huid van de structuur bevindt.
Experimenten toonden aan dat schade in de verbinding tussen de huid en
iii
iv
Summary i
Samenvatting iii
Nomenclature ix
1 Introduction 1
1.1 Background and motivation . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Composite structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Fiber reinforced plastics . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.2 Damage types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.3 Damage classification . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Structural health monitoring . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.2 Classifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.3 Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.4 Major technology gaps . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Objective and scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
v
vi Contents
7 Discussion 135
7.1 Design of an SHM strategy . . . . . . . . . . . . . . . . . . . . . . . . . 135
7.1.1 Scenario based design procedure . . . . . . . . . . . . . . . . . . 136
7.1.2 Combination of approaches . . . . . . . . . . . . . . . . . . . . . 140
7.2 Application of vibration based SHM . . . . . . . . . . . . . . . . . . . . 141
7.2.1 Integrated sensing . . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.2.2 Operational and environmental effects . . . . . . . . . . . . . . 145
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
Dankwoord 165
Publications 169
Nomenclature
The symbols used in this thesis are classified into a Greek or a Roman symbol group.
Although some symbols can represent multiple quantities, its intended meaning
follows from the textual context. An overview of the most important abbreviations,
symbols and terminology used in the present thesis:
Greek symbols
α Extent of bilinearity [-]
βi Damage index of beam element i [-]
β ij Damage index of plate element ij [-]
(n)
γi Geometrical function of beam element i and mode n [1/m3 ]
(n)
γij Geometrical function of plate element ij and mode n [1/m2 ]
ǫ Control factor of the nonlinear function [-]
ζ (n) , ζ n Viscous damping of mode n [%]
θ Rotation angle [rad]
λn , λ∗n Poles (complex conjugate pair) of mode n [rad/s]
µ Mean value of the damage index β ij [-]
ν Poisson’s ratio [-]
ρ Density [kg/m3 ]
σn Damping factor of mode n [rad/s]
σ Standard deviation of the damage index β ij [-]
τ Shear stress [MPa]
φ Instantaneous phase [rad]
φp Phase of the pump excitation signal [rad]
φc Phase of the carrier excitation signal [rad]
χd Damage parameter (e.g. crack length, loss of stiffness)
χe Influence of the environmental and the operational
conditions (e.g. temperature, humidity)
ψn Mode shape vector of mode n [-]
ψmA , ψB Mode shape vector of case A and mode m, case B and mode n [-]
n
ω Frequency [rad/s]
ωdn Damped natural frequency of mode n [rad/s]
ωn Undamped natural frequency of mode n [rad/s]
ix
x Contents
Roman symbols
Ac Carrier amplitude [m]
Asb1 Left sideband amplitude [m]
Asb2 right sideband amplitude [m]
Ainst Instantaneous amplitude (signal envelope) [m/s]
a Length of intact beam segment 1 [m]
ac Spanwise delamination location [m]
a Dimensionless length intact beam segment 1 [-]
ac Dimensionless spanwise delamination location [-]
b Length of delaminated beam segment [m]
b Dimensionless delamination length [-]
c Viscous damping coefficient [Ns/m]
c Length of intact beam segment 4 [m]
c Dimensionless length intact beam segment 4 [-]
D Flexural rigidity of a plate-like structure [Nm]
Dij Weighted average flexural rigidity of element ij of a plate-like [Nm]
structure
E Young’s modulus [N/m2 ]
Fop Operational load vector [N]
Ftest Test load vector [N]
Fa Excitation force [N]
Fp Pump excitation amplitude [N]
Fc Carrier excitation amplitude [N]
(n)
FB,i Local fractional modal strain energy of mode n and element i [-]
(n)
Fij Local fractional modal strain energy of mode n and element ij [-]
Fspring Spring force [N]
f Frequency [Hz]
fp Pump excitation frequency [Hz]
fc Carrier excitation frequency [Hz]
fa Harmonic excitation frequency [Hz]
f (n) , f n Natural frequency of the nth mode [Hz]
(n) (n)
fB , fT Natural frequency of the nth bending and torsion mode [Hz]
f inst Instantaneous frequency [Hz]
G Shear modulus [N/m2 ]
Gp , Gc Constants [-]
g Force vector of elastic forces, damping forces, etc. [N]
Contents xi
Operators
∗ Complex conjugate
˜ Damaged case
T Transpose
Abbreviations
ADL Allowable damage limit
AE Acoustic emission
AU Acousto-ultrasonics
BVID Barely visible impact damage
CBM Condition based maintenance
CMIF Complex mode indicator function
EMA Experimental modal analysis
Contents xiii
Introduction
Composite materials combine a high strength and stiffness with a relatively low
density. This makes them extremely useful in applications where weight plays an
important role such as aircraft and wind turbines. These materials can, however,
exhibit unconventional and complex types of damage, like transverse cracks and
delaminations. These damage scenarios are often invisible, but can severely
influence the structural performance of a component, and hence tremendously
decrease its service life. Periodic inspections are required to ensure the integrity of a
component during its life. The current scheduled visual inspections are often time-
consuming, costly and require the components to be readily accessible, as shown in
Figure 1.1. Structural health monitoring (SHM) technologies propose a promising
alternative and involve the continuous monitoring of a structure by employing
a nondestructive testing (NDT) approach based on an integrated sensor system.
The output of this process is information regarding the ability of the structure to
perform its intended function in consideration of the applied loadings, aging and
degradation resulting from the operational environments.
Another motivation for the development of SHM technologies is that they
potentially can reduce the maintenance costs and increase the operational
availability of a system. Firstly, by making maintenance routines shorter and
more effective, given the actual physical condition of the component. Secondly, by
optimizing the maintenance intervals utilizing condition based maintenance (CBM)
routines rather than relying on the conventional time-based inspection intervals [1].
The use of structural health monitoring technologies will not only provide safety
benefits or enable new possibilities for maintenance concepts, but can also have a
significant influence on the design concepts. The change in design of lightweight
structures from a safe life to a damage tolerant design supported by a monitoring
system is, although far from reality, also considered as a potential weight-saving
benefit.
1
2 Chapter 1. Introduction
(a) Wind turbine inspection (b) Aircraft fuselage inspection (c) Bridge inspection
(photo courtesy of Spiegel (photo courtesy of Sandia (photo courtesy of Aspen
Online). National Laboratories, USA). Aerials, USA).
Figure 1.1 An overview of potential application fields for structural health monitoring techniques.
The research presented in this thesis was performed within the European research
program Clean Sky [2], which focuses on the development of breakthrough
technologies to significantly improve the environmental performances of airplanes
and air transport. Clean Sky comprises six Integrated Technology Demonstrators
(ITDs). The work presented here was part of the Eco-Design ITD, which
concentrates on green design and production, withdrawal and recycling of aircraft
by optimal use of raw materials and energy. The development of structural health
monitoring technologies is one of the objectives of this research program. Structural
components are usually replaced prior to their end of life, which is costly and
resource inefficient. A monitoring system can allow for an increase in the service
life of aircraft components and hence reduces the costs and long-term ecological
impact of these components.
To summarize, the development of structural health monitoring technologies for
composite structures is aiming to provide safety, cost saving (maintenance and
weight) as well as environmental benefits. However, the number of successful
practical applications of structural health monitoring technologies is still limited.
This is mainly due to the complexity of the composite components, the variety
of potential damage scenarios and the high performance demands of the damage
identification method. The work described in this thesis primarily concerns the
relation between those three aspects in order to achieve a higher level of maturity
of the structural health monitoring technologies. Background information about
composite structures and the typical damage scenarios is presented in the next
section, followed by a general introduction to structural health monitoring. The
1.2. Composite structures 3
Structure level
Meso
z Micro
x y
3
1 2
and the ability to reduce manufacturing costs by lowering the number of individual
components in an assembly. These strengths are why research in composites has
flourished over the past decades and why composites are more and more employed
over a variety of applications in, for example, the aerospace, automotive and marine
industry.
Transverse crack
Delamination
3
2
1
Figure 1.3 Schematic representation of transverse crack and delamination in a [0/90/0]s laminate.
1.2. Composite structures 5
Figure 1.4 Cross section of a glass-reinforced epoxy laminate with transverse cracks and delaminations
caused by an impact test.
are embedded within the composite structure. This makes them barely detectable
during, for example, visual inspections. Figure 1.4 shows a micrograph of a cross
section of a laminate with both transverse cracks and delaminations originating
from an impact test. Although delaminations do not lead to complete fracture, they
can seriously affect the thermal and mechanical properties of the laminate.
The type of damage that significantly decreases the load carrying capacity of a
laminate is fiber failure. Fiber related failure in a laminate is mostly accompanied
by matrix related damage like transverse cracks and delaminations. Typical failure
modes can involve local buckling of fibers, fiber breakage and fiber pull-out.
1.3.1 General
Structural health monitoring is the multidisciplinary process of implementing a
strategy for damage identification in a way that nondestructive testing becomes
an integral part of the structure. This process involves the definition of potential
damage scenarios for a structure, the observation of the structure over a period of
time using periodically spaced measurements, the extraction of damage sensitive
parameters (features) from these measurements and the analysis of these features
to determine the current state of health of the structure (classification). The output
of this process is periodically updated information regarding the ability of the
structure to perform its intended function in consideration of the applied loadings,
aging and degradation resulting from the operational environments.
In contrast to conventional nondestructive testing techniques that are operated off-
line during maintenance, structural health monitoring techniques can be operated
off-line as well as on-line. On-line refers, in this case, to the monitoring during
operation of the system or structure. The structural health monitoring technique is
part of the on-board systems. Sensors are permanently attached (surface sensors)
or embedded (integrated sensors) in the structure. As a result, information on the
structural state is available at arbitrary times.
As stated by Farrar and Doebling [6], the process of structural health monitoring is
fundamentally one of statistical pattern recognition. A statistical pattern recognition
process aims to classify data (patterns) based on either a priori knowledge or on
information extracted from the patterns. The patterns to be classified are usually
groups of measurements or observations. A pattern recognition process covers all
stages of an investigation from problem formulation up to the interpretation of the
results. Figure 1.5 shows a simplified process, which consists of a diagnostic and a
prognostic part. The diagnostic analyses are used to estimate the current state of the
structure. The prognostic analysis evaluates the damage evolution and estimates
the residual service life [7]. The present thesis focuses on the diagnostic part of the
structural health monitoring process, which can be divided into a four-step process:
Diagnostics Prognostics
Representation Current Damage
Actuation Response pattern state evolution Decision
Feature pattern
Repair,
Sensor Feature Failure
Structure Classifier replacement,
system extractor probability residual life?
High
Level 1, 2, 3 Level 4 Low
Damage
1. Operational evaluation
Operational evaluation answers questions regarding the implementation of
the structural health monitoring system, such as possible failure modes,
operational and environmental conditions and data acquisition related
limitations.
2. Data acquisition
This step defines the data acquisition in terms of the quantities to be
measured, the type and quantity of sensors to be used, the locations where
these sensors are to be placed and the hardware to be used. Moreover, it
defines the data fusion and cleansing, which is the determination of which
data is necessary and useful in the feature extraction process.
3. Feature extraction
This step in the structural health monitoring process receives the most
attention in the literature. Feature extraction is the process of identifying
damage sensitive parameters from measured data. These damage features are
defined in the time, frequency or modal domain. Information reduction and
condensation is also of concern for a large quantity of data, particularly if
comparisons of many measurements over the service life of the structure are
required.
4. Classification
The last step is concerned with the implementation of algorithms (e.g. neural
networks) that operate on the extracted features to distinguish between the
damaged and the undamaged structural state and to quantify the damage
state of the structure. Statistical methods are used to establish the feature’s
sensitivity to damage and to prevent false damage identification.
1.3.2 Classifications
Damage identification methods can be classified in different ways. This section
summarizes the most important classifications used in this thesis.
identification:
Local and global methods Damage identification techniques are usually classified
as local or global [13]. This classification is based on the relative size of the area that
can be inspected at once by the method with respect to the overall dimensions of
the structure. The local methods concentrate on a part of the structure and are
usually considered to be more sensitive than the global methods. They are capable
of detecting small damages such as cracks, but their application requires a prior
knowledge of the location of the damaged area. The global methods can analyze
a relatively large area at once, but the resolution is, however, rather limited. As a
consequence, only relatively severe damage cases can be identified.
Table 1.1 An overview of the most commonly used nondestructive testing (NDT) techniques.
1.3.3 Techniques
A wide range of nondestructive testing techniques can be employed for damage
identification purposes. An overview of the most commonly used nondestructive
testing techniques and their characteristics is presented in Table 1.1. The majority
of these techniques can only be applied when the structure is not in operation
(‘off-line’) and readily accessible. Consequently, only a few of these techniques
are suitable to be applied in a health monitoring environment. As part of this
10 Chapter 1. Introduction
Table 1.2 An overview of the dynamics based nondestructive testing (NDT) technologies.
The basic principles of the dynamics based techniques are described in Appendix A.
Each technique comprises a large amount of literature and is usually treated
as a different field of research. Table 1.2 provides a more detailed comparison
of their performances. The low frequency structural vibration (SV) and electro-
mechanical impedance (EMI) techniques primarily rely on standing wave patterns,
while the higher frequency acoustic emission (AE), acousto-ultrasonics (AU) and
ultrasonic testing (UT) utilize traveling wave characteristics. The former group of
methods provide data that is relatively easy to interpret. More complex structures
can be analyzed with these methods and a relatively large area can be explored at
once. The frequency range, and hence the resolution, is however limited [13]. As a
consequence, only relatively severe damage such as delaminations can be identified.
The latter group of methods are usually considered to be more sensitive. They
are capable of detecting small damage such as cracks [29]. For that reason, these
wave propagation based technologies are increasingly being explored for aircraft
applications [33, 34]. The downside is the more complex interpretation of the data,
in particular in case of non-flat or complex (composite) structures [35]. The rating
for the sensitivity is linked to the operational frequency range [11], while the other
aspects are ranked according to the available literature. It should be noted that these
ratings are rather subjective. The intention here is, however, to give an impression
of the relative strengths and weaknesses rather than to condemn techniques.
1.3. Structural health monitoring 11
Although many structural health monitoring techniques have been proposed in the
literature, there are still numerous difficulties in the practical application of these
approaches. The most important technical issues that need to be resolved before
structural health monitoring technologies can make the transition from a research
topic to actual practice are summarized below.
In addition to the technical issues described above, there are other nontechnical
issues that must be addressed before structural health monitoring technologies
can make the transition to actual application. These issues include, for example,
convincing operators, engineers and authorities of the potentials of the technology
as well as the certification of the technologies. More detailed discussions on this
topic are provided by Boller [36] and Farrar et al. [37].
Damage
Structure identification
method
Strategy for
structural
health
monitoring
(Actuation)
Damage
& sensing
scenario
technology
Figure 1.6 The multidisciplinary framework for the design of a structural health monitoring system.
skin and stiffener. Impacts near these connections can lead to local skin-stringer
separation. This is a safety-critical failure mode, because it can significantly affect
the structural performance of the component while remaining invisible from the
outer surface. Skin-stiffener structures are therefore considered as a good candidate
for health monitoring.
The structural vibration based health monitoring approaches are considered in the
present work. These methods are based on the concept that the dynamic behavior
of a structure can change if damage occurs. The motivation is twofold: firstly,
because they do not require the structure to be readily accessible. Secondly, because
these low frequency methods provide data that is relatively easy to interpret.
This provides opportunities to analyze more complex structures, such as the skin-
stiffener structures. A drawback of the vibration based methods is the limited
sensitivity. The identification of barely visible impact damage, as discussed in
Section 1.2.3, sets the lower bound for the capabilities of the approach.
This work will contribute to the development of a design tool for research engineers,
to assist the implementation of structural health monitoring technology in safety-
critical composite structures.
14 Chapter 1. Introduction
Numerical
Loendersloot, Loendersloot,
et al. [38] et al. [39]
Linear
dynamic Experimental
behavior
2 & 3 stiffener
T-beam
structure
Chapter 3 Chapter 4
Vibration
Introduction Discussion Conclusions
methods
Chapter 1 Chapter 2 Chapter 7 Chapter 8
Nonlinear Vibro-acoustic
dynamics modulations
Nonlinear Chapter 5 Chapter 6
dynamic
behavior
1.5 Outline
The outline of the thesis is schematically illustrated in Figure 1.7. The core
comprises five chapters. Chapters 3 to 6 are reproduced from research papers. As a
consequence, some of the essential details are repeated in the different chapters. The
author apologizes for any inconvenience caused by the chosen presentation. From
a more positive point of view, however, the reader is able to study any individual
chapter without having to miss out on any essential details.
Chapter 2 provides an overview of the vibration based damage identification
methods. The basic concept of these methods is explained based on a generalized
description of a damaged system. A literature study supported by an analytical
beam model are used to select a potentially powerful approach for the detection,
localization and characterization of damage in the skin-stiffener structure. Two
approaches are considered. The first approach utilizes mode shape curvatures and
assumes a linear dynamic behavior, while the second approach is focused on the
nonlinear dynamic effects that are introduced by the damage.
Chapter 3 is focused on the experimental feasibility of a vibration based damage
identification method to identify an artificial delamination at the skin-stiffener
connection of a composite T-beam. A force-vibration set-up, including a laser-
vibrometer system, is used to measure the dynamic behavior of the T-beam
experimentally. Both bending and torsion modes are considered in the analysis.
Special attention is paid to the effect of the number of measurement points.
Chapter 4 extends the work described in Chapter 3 to larger and more complex
skin-stiffener structures (i.e. a 2 stiffener structure and a 3 stiffener structure with
non-uniform skin thickness). Impact induced damage scenarios are considered.
References 15
The relation between the damage location, the structural design and the dynamic
behavior is investigated in order to extract recommendations for the effective
application of the methodology. Readers are referred to the book chapters of
Loendersloot et al. [38, 39] for numerical studies utilizing finite element models
of the structures that were used in Chapters 3 and 4.
The vibration based method used in Chapters 3 and 4 assumes a linear dynamic
behavior. The observation of potential nonlinear dynamic effects caused by
damage can also be a strong indicator of the damage. Chapter 5 describes a
study on the interaction of a low frequency vibration with skin-stiffener damage.
This interaction can yield dynamic phenomena that exhibit complicated nonlinear
behavior. Different phenomena are linked to measured waveforms with the help of
phase portraits.
The nonlinear effects introduced by skin-stiffener damage urges the development
of nonlinear damage identification methods. Chapter 6 concerns a study on the
understanding and feasibility of using nonlinear vibro-acoustic modulations for
the detection, localization and characterization of impact damage in a composite
T-beam. A time domain analysis of the vibro-acoustic modulation phenomena
is presented at multiple spatial locations. From a broader perspective, this work
intends to contribute to the development of enhanced methods for the identification
and characterization of damage in advanced composite structures.
The complete work is put into a broader perspective in Chapter 7. Additional
design recommendations and guidelines are extracted based on the work presented
in this thesis. The practical application of the methods is discussed. Finally,
Chapter 8 presents the important conclusions and provides the recommendations
for further research.
References
[1] T. Tinga. Application of physical failure models to enable usage and load based
maintenance. Reliability Engineering & System Safety, 95(10):1061–1075, 2010.
[2] Clean Sky website: http://www.cleansky.eu, visited on June 28th , 2013.
[3] R.L. Foye. Finite element analysis of the stiffness of fabric reinforced composites.
Technical report, NASA Contractor Report, 1992.
[4] Volume 3. Polymer matrix composites: material usage, design and analysis. In
Composite materials handbook CMH-17-3G, Chapter 12, page 952. SAE International, 2012.
[5] J. Baaran. Visual inspection of composite structures. Technical report, Institute of
Composite Structures and Adaptive Systems, DLR Braunschweig, 2009.
[6] C.R. Farrar, S.W. Doebling, and D.A. Nix. Vibration-based structural damage
identification. Philosophical Transactions of the Royal Society A: Mathematical, Physical and
Engineering Sciences, 359(1778):131–149, 2001.
[7] C.R. Farrar, H. Sohn, F.M. Hemez, M.C. Anderson, M.T. Bement, P.J. Cornwell, S.W.
Doebling, J.F. Schultze, N. Lieven, and A.N. Robertson. Damage prognosis: current
16 Chapter 1. Introduction
status and future needs. Technical report, Los Alamos National Laboratory, Los
Alamos, NM, 2001.
[8] S.W. Doebling, C.R. Farrar, M.B. Prime, and D.W. Shevitz. Damage identification and
health monitoring of structural and mechanical systems from changes in their
vibration characteristics: a literature review. Technical report, Los Alamos National
Laboratory, NM, USA, 1996.
[9] A. Rytter. Vibration based inspection of civil engineering structures. Ph.D. thesis, Aalborg
University, 1993.
[10] H. Sohn, C.R. Farrar, F.M. Hemez, D.D. Shunk, D. Stinemans, and B.R. Nadler. A
review of structural health monitoring literature: 1996-2001. Technical report, Los
Alamos National Laboratory, Los Alamos, NM, 2003.
[11] K. Worden, C.R. Farrar, G. Manson, and G. Park. The fundamental axioms of structural
health monitoring. Proceedings of the Royal Society A: Mathematical, Physical and
Engineering Sciences, 463(2082):1639–1664, 2007.
[12] K. Worden and J.M. Dulieu-Barton. An overview of intelligent fault detection in
systems and structures. Structural Health Monitoring, 3(1):85, 2004.
[13] C.-P. Fritzen and P. Kraemer. Self-diagnosis of smart structures based on dynamical
properties. Mechanical Systems and Signal Processing, 23(6):1830–1845, 2009.
[14] E.S. Sazonov, P. Klinkhachorn, U.B. Halabe, and H. V.S. GangaRao. Non-baseline
detection of small damages from changes in strain energy mode shapes. Nondestructive
Testing And Evaluation, 18(3-4):91–107, 2003.
[15] S. Park, C. Lee, and H. Sohn. Reference-free crack detection using transfer impedances.
Journal of Sound and Vibration, 329(12):2337–2348, 2010.
[16] J.C. Abry, S. Bochard, and A. Chateauminois. In situ detection of damage in CFRP
laminates by electrical resistance measurements. Composites Science and Technology,
59:925–935, 1999.
[17] R. Schueler, S.P. Joshi, and K. Schulte. Damage detection in CFRP by electrical
conductivity mapping. Composites Science and Technology, 61(6):921–930, 2001.
[18] C. Hellier. Handbook of nondestructive evaluation. McGraw-Hill, 2003.
[19] ISO 9934-1 Non-destructive testing – Magnetic particle testing – Part 1: general
principles.
[20] B.A. Auld and J.C. Moulder. Review of advances in quantitative eddy current
nondestructive evaluation. Journal of Nondestructive Evaluation, 18(1), 1999.
[21] P.J. Schilling, B.R. Karedla, A.K. Tatiparthi, M.A. Verges, and P.D. Herrington. X-ray
computed microtomography of internal damage in fiber reinforced polymer matrix
composites. Composites Science and Technology, 65(14):2071–2078, 2005.
[22] X.P.V. Maldague. Introduction to NDT by active infrared thermography. Materials
Evaluation, 60(9):1060–1073, 2002.
[23] A. Kesavan, S. John, and I. Herszberg. Strain-based structural health monitoring of
complex composite structures. Structural Health Monitoring, 7(3):203–213, 2008.
[24] E.P. Carden and P. Fanning. Vibration based condition monitoring: a review. Structural
Health Monitoring, 3(4):355–377, 2004.
[25] D. Montalvão, N.M.M. Maia, and A.M.R. Ribeiro. A review of vibration-based
structural health monitoring with special emphasis on composite materials. Shock and
Vibration Digest, 38(4):295–326, 2006.
[26] G. Park, H. Sohn, C.R. Farrar, and D.J. Inman. Overview of piezoelectric
impedance-based health monitoring and path forward. The Shock and Vibration Digest,
References 17
35(6):451–463, 2003.
[27] V. Gopal, M. Annamdas, and C.K. Soh. Application of electromechanical impedance
technique for engineering structures: review and future issues. Journal of Intelligent
Material Systems and Structures, 21(1):41–59, 2009.
[28] T. Kundu, S. Das, and K.V. Jata. Detection of the point of impact on a stiffened plate by
the acoustic emission technique. Smart Materials and Structures, 18(3):035006, 2009.
[29] A. Raghavan and C.E.S. Cesnik. Review of guided-wave structural health monitoring.
The Shock and Vibration Digest, 39(2):91–114, 2007.
[30] Z. Su, L. Ye, and Y. Lu. Guided lamb waves for identification of damage in composite
structures: a review. Journal of Sound and Vibration, 295(3-5):753–780, 2006.
[31] Y.Y. Hung and H.P. Ho. Shearography: an optical measurement technique and
applications. Materials Science and Engineering: R: Reports, 49(3):61–87, 2005.
[32] Y.Y. Hung. Applications of digital shearography for testing of composite structures.
Composites Part B: Engineering, 30(7):765–773, 1999.
[33] W.J. Staszewski, S. Mahzan, and R. Traynor. Health monitoring of aerospace composite
structures – Active and passive approach. Composites Science and Technology,
69(11-12):1678–1685, 2009.
[34] K. Diamanti and C. Soutis. Structural health monitoring techniques for aircraft
composite structures. Progress in Aerospace Sciences, 46(8):342–352, 2010.
[35] R.P. Dalton, P. Cawley, and M.J.S. Lowe. The potential of guided waves for monitoring
large areas of metallic aircraft fuselage structure. Journal of Nondestructive Evaluation,
20(1), 2001.
[36] C. Boller. Structural health monitoring – Its association and use. In W. Ostachowicz
and J.A. Güemes, editors, New Trends in Structural Health Monitoring, volume 542 of
CISM International Centre for Mechanical Sciences, Chapter 1, pages 1–79. Springer
Vienna, 2013.
[37] C.R. Farrar and K. Worden. An introduction to structural health monitoring.
Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering
Sciences, 365(1851):303–15, 2007.
[38] R. Loendersloot, T.H. Ooijevaar, L.L. Warnet, A. de Boer, and R. Akkerman. Vibration
based structural health monitoring and the modal strain energy damage index
algorithm applied to a composite T-beam. In C.M.A. Vasques and J.D. Rodrigues,
editors, Vibration and Structural Acoustics Analysis: Current Research and Related
Technologies, Chapter 6, pages 121–150. Springer, 2011.
[39] R. Loendersloot, T.H. Ooijevaar, A. de Boer, and R. Akkerman. Development of a
damage quantification model for composite skin-stiffener structures. In C. Boller and
H. Janocha, editors, New Trends in Smart Technologies, pages 99–108. Fraunhofer Verlag,
2013.
Chapter 2
Abstract
An overview of the vibration based damage identification methods
is presented in this chapter. The basic concept of these methods is
explained based on a generalized description of a damaged system.
One of the challenges in the development of a successful approach
is the selection of a damage identification method. A literature
study supported by an analytical model showed that mode shape
curvatures combined with the modal strain energy damage index
(MSE-DI) algorithm are a potentially powerful damage feature and
classifier. However, when severe nonlinear dynamic effects are
introduced by the damage, the modal domain based methods are
suffering under their linear system assumption. Alternative methods
in, for example, the time or frequency domain are in that case
required to obtain an adequate description of the nonlinear behavior
introduced by the damage.
19
20 Chapter 2. Overview of vibration based damage identification methods
2.1 Introduction
The basic concept of the vibration based damage identification methods is that
the dynamic behavior of a structure can change if damage occurs [1]. Damage in
a structure can alter the structural integrity and therefore the physical properties
like stiffness, mass and/or damping. The dynamic behavior of a structure is a
function of these physical properties and will therefore directly be affected by the
damage. The dynamic behavior can be described by time, frequency and modal
domain parameters. The changes in these parameters or properties derived from
these parameters are used as indicators of damage.
The structural vibration based approaches generally allow for a relatively easy
interpretation of the measured responses, have the ability to analyze complex
structures and do not require the structure to be readily accessible in order to
be able to identify damage [2]. Drawbacks are, however, the limited sensitivity
compared to higher frequency approaches and the number of required sensors in
case the standing wave patterns need to be described [3]. As a consequence, the
application of structural vibration based methods has to be tailored to the structure
and the expected damage cases to fully utilize the potential of this technology.
This chapter has two objectives. The first objective is to provide an overview of
the structural vibration based damage identification methods. For this purpose,
a fundamental description of the structural vibration based damage identification
problem is given, followed by a short literature overview of the damage features
and (statistical) classifiers that are commonly addressed. The second objective is to
select a promising damage identification method for the detection, localization and
characterization of skin-stiffener damage. The selection of an appropriate method
is often all but straightforward and finally a matter of compromise. To aid in
this process, two basic principles are discussed, namely the effect of the potential
damage case on the dynamic behavior and the consequences involved with the
information reduction in the signal processing. The former determines whether the
parameter is sensitive to the damage and is analyzed by considering an analytical
model of a delaminated composite beam. This model is a simplified representation
of the damage that is expected in skin-stiffener connections. The latter is explained
and demonstrated based on a mass-spring-damper system with a bilinear stiffness,
representing the potential opening and closing behavior of defects in a skin-stiffener
connection.
where M is the mass matrix, g the force vector of elastic forces, damping forces,
etc. depending on the displacements q, the velocities q̇, the time t, a damage
parameter χd and a parameter χe that indicates the influence of the environmental
and operational conditions (e.g. temperature, humidity). The damage parameter
χd characterizes the damage scenario in terms of e.g. the crack length or the loss
of stiffness. Equation (2.1) describes the fundamentals of the damage diagnosis
process, whereas Equation (2.2) provides a starting point for a prognostic analysis
(e.g. damage evolution and residual service life estimation) [4]. The damage
evolution is usually considered as a rather slow process compared to the dynamic
behavior of the structure. The damage parameter χd is, therefore, assumed constant
during the short time span of a dynamic measurement. The forced excitation of the
system consists of the operational loads Fop (t) and the test loads Ftest (t). The latter
is absent in the case of a passive method. For the active approach, the test loads are
generated by an actuator (e.g. piezoelectric element, shaker). The transformation
of the input signal u(t) of the actuator to the forces and moments applied to the
structure can be described by:
A similar transfer function can be defined for the response side of the system:
The transfer function hout relates the state variables to the actually measured system
response y(t) (e.g. strains or voltage). The damage parameter χd in Equations (2.3)
and (2.4) can also account for any failure or debonding of the actuator or sensor.
Although the damage parameter χd provides the most direct description of the
damage, this parameter can usually not be measured directly. Alternatively, the
changes in the dynamic response related to this damage parameter is considered
in the damage identification process. For this purpose, the system response y(t) is
usually compared with an earlier obtained baseline response y0 . The associated
forward problem relates the deviations in the response ∆y(t) with the damage
characteristics χd and the environmental and operational conditions χe :
∆y = y − y0 = f (χd , χe ) . (2.5)
The key challenge for damage diagnosis is, however, to solve the inverse problem,
22 Chapter 2. Overview of vibration based damage identification methods
Selection of damage
Damage type feature & classifier
Environmental
Inverse problem influences
Transverse -1 χe Time
cracking
χ d = f ( Δy , χ e )
Δy(t)
Damage Changes in Changes in
Frequency
Delamination χd structural dynamic
ΔY(ω)
properties behavior
Fiber Forward problem Modal
failure Δy= f ( χ d , χ e ) ΔY(ωn)
Figure 2.1 Schematic illustration of the forward and inverse problem involved with vibration based damage
identification of composite structures.
χd = f −1 (∆y, χe ) . (2.6)
The complexity involved with this problem comes from the non-uniqueness of the
solution and the usually incomplete response data due to a limited number of
measurement locations.
Figure 2.1 schematically illustrates the forward and inverse problem. They connect
the potential damage scenarios in composite materials (left side) with the wide
range of damage features (right side). Damage identification, first, aims to
maximize the correlation between the damage parameter and the selected damage
feature. Second, it is practically also desired to condense the information to
a lower dimensional parameter space. The deviations in the measured time
domain parameters ∆y(t) are therefore often replaced by parameters from the
frequency ∆Y (ω ) or modal ∆Y (ωn ) domain. For both purposes, the selection
of an appropriate damage feature is an important step in the vibration based
damage identification approach. Many additional aspects (e.g. sensing system,
operational conditions, computational costs, etc.) should be taken into account in
this process. The selection is also significantly complicated by the wide range of
potential damage scenarios and the large variety of damage features. None of the
approaches is, at the moment, able to solve all problems in all structures. The
selection of an appropriate approach is, therefore, often a matter of compromise.
Figure 2.2 A concise overview of the wide range of damage identification approaches. The vibration based
methods are categorized according to their damage sensitive feature and (statistical) classifier.
can also be found in the books of Boller et al. [10] and Balageas et al. [11]. More
specifically, Montalvão et al. [1] and Zou et al. [2] provide reviews that are dedicated
to vibration based health monitoring of composite materials. The review by Li [12]
is specifically focused on strain based methods. Reviews only concentrating on
nonlinear damage identification methods are presented by Worden et al. [13] and
Jhang [14].
The vibration based damage identification methods can be categorized according
to their damage sensitive feature and (statistical) classifier, as schematically shown
in Figure 2.2. The earliest studies on this subject report on the relation between
natural frequency shift or modal damping changes and structural damage. Adams
and Cawley [15–18] were among the first to actively research this subject back
in the seventies. These modal domain methods comprise a large part of the
available literature. Nowadays, research is however also directed towards statistical
time series methods [19] (e.g. cross-correlations and auto-regressive moving
average representations) and advanced signal processing techniques (e.g. Wavelet
transforms [20] and Hilbert-Huang transforms [21]) to extract damage features in
the time and frequency domain. A comprehensive list of the damage features that
are used in the literature is presented in Table 2.1. Appendix B provides an overview
of the wide variety of classifiers associated with these features.
The comparison and assessment of the performance of the different methods is
rather difficult. Each method can have individual advantages and disadvantages
regarding the specific application, structure and damage case under investigation.
The review studies primarily list the available literature, but do not allow for
a detailed comparison of the performance of each method. It lies beyond the
scope of this chapter to discuss each method individually. Several experimental
and numerical comparison studies by different authors can be found in the
24 Chapter 2. Overview of vibration based damage identification methods
Table 2.1 An overview of the vibration based damage features categorized according to the time, frequency
and modal domain. An elaborate overview of the classifiers associated with these features is presented in
Appendix B.
literature [9, 12, 22–27]. Although these studies are limited to a selection of
methods, structures and damage cases, they provide a general perception of the
strengths and weaknesses of the evaluated methods. In summary, the comparative
studies concluded that mode shape curvatures (damage feature) combined with
the modal strain energy damage index (MSE-DI) algorithm [28, 29] (classifier)
outperforms other frequently applied damage identification methods based on
natural frequencies [16, 30], mode shapes [31, 32], dynamic flexibility [33–35] and
other mode shape derivative based classifiers [36, 37]. Mode shape curvatures tend
to be a relatively sensitive damage feature [9]. Humar et al. [25] stated that, among
the evaluated methods, the MSE-DI formulation is most successful in detecting
and localizing damage, but is lacking in the quantification of the damage. These
observations are supported by Alvandi and Cremona [24]. They also concluded
that this approach showed a good stability regarding noisy signals. Drawbacks of
using mode shape curvatures are the number of required data points to accurately
describe the mode shapes and the numerical errors introduced with the calculation
of the second derivative [9, 27, 37].
Several dynamic models have been developed for the analysis of the effect of
delaminations on dynamic behavior of beam structures. To overcome difficulties
associated with potential nonlinear dynamic effects introduced by a delamination,
many modeling attempts assume that the dynamic behavior of the structure
remains linear, i.e. the effect of the damage on the material parameters (e.g.
stiffness) is independent of the deformation. Della and Shu [38] categorize the linear
dynamic models of delaminated structures into two classes: the ‘free mode’ and the
‘constraint mode’ models. The ‘free mode’ models assume that the delaminated
layers deformed ‘freely’ without interacting with each other. On the other hand
the ‘constraint mode’ models assume that the delaminated layers are ‘constrained’
and will have identical transverse deformations. In addition, the delaminated
layers are assumed to be free to slide over each other in the axial direction. Both
representations show discrepancies with respect to what is expected in practice.
Nevertheless, the predicted frequencies based on the ’constrained model’ were
closer to the experimental values [39].
26 Chapter 2. Overview of vibration based damage identification methods
1 2 4
3 Free side
a−c
The first ‘constraint mode’ model was presented by Mujumdar and Suryanarayan [39].
This model determines the natural frequencies and mode shapes of an isotropic
beam containing a single through-width delamination in flexural vibration. The
delaminated beam model consists of three spanwise regions, a delaminated section
and two integral sections (see Figure 2.3). The delaminated section is composed of
two separate beams above and below the delamination plane and they are joined at
their ends to the integral beam segment. It is assumed that no gap exists between
the delaminated components and that they are free to slide, i.e. there is no friction
between the components. Each beam is treated as a Bernoulli-Euler beam. The
equations of motion are formulated for each beam segment. The dynamic solution
for the complete beam is obtained by imposing appropriate boundary conditions at
the ends of the two integral segments and continuity conditions at the delamination
junctions.
The basic theory is based on variations in the distribution of shear stress. Both
normal and shear stresses are developed in a beam that exhibits non-uniform
bending. In a beam without a delamination, the shear stress is distributed as
shown in Figure 2.4a. However, in case a delamination is present, the layers are
free to slide and the shear stress on the delaminated surface therefore equals zero
by definition. The shear stress distribution for this case is shown in Figure 2.4b.
This discontinuity in shear stress results in a decrease in the flexural rigidity of the
delaminated segment, which will affect the dynamic properties of the beam.
The two separate beams in the delaminated segment also give rise to coupling
between the axial and flexural deformation. Therefore, continuity of axial forces
2.4. Damage feature selection 27
τ τ N (a) N
N N
τ
(a) (b) (b)
Figure 2.4 Distribution of shear Figure 2.5 Delaminated layers without (a) and with (b)
stress τ in (a) a pristine and (b) a continuity conditions at the junctions between the delaminated
delaminated beam segment. and integral sections.
and displacements at the junctions between the integral and delaminated segments
is required. A normal force N is introduced, as indicated in Figure 2.5. These
normal forces compress one segment and stretch the other such that their ends lie
in the same plane. This internal loading results in an internal bending moment,
which partly compensates for the loss of flexural rigidity due to the new shear
stress distribution.
Grouve et al. [40] extended the analytical model of Mujumdar and Suryanarayan [39]
for laminated beams. Experimental validation showed an averaged difference
between theoretical and experimental resonance frequencies for the delaminated
beams equal to 3.3%. The differences between experimental and calculated
resonance are partly caused by the model assumptions made. The delaminated
segments are modeled to stay in contact during the vibration and slide without
friction. This is a requirement which cannot be satisfied in practice. Opening of the
delamination during vibration will result in a decrease of stiffness, while friction
between the segments will increase the flexural rigidity.
Analytical results
The model proposed by Grouve et al. [40] is used to analyze the effect of a
delamination on the dynamic behavior. A cantilever composite beam structure
(dimensions 300×20×2.0 mm) built from 16 unidirectional plies of carbon fiber
AS4D is considered in the analysis. The thermoplastic matrix is PEKK. The laminate
has a [0/90]4,s lay-up with the material properties as in Table 2.2. The dimensionless
delamination parameters (ac , b and h2 ) as defined in Figure 2.3 are used in the
investigation.
Table 2.2 The material properties of a unidirectional layer of carbon AS4D/PEKK. The material properties
are obtained from data provided by Fokker Aerostructures B.V., Hoogeveen, The Netherlands.
E1 E2 G12 ν12 ρ
[GPa] [GPa] [GPa] [-] [kg/m3 ]
133 10.5 5.56 0.3 1590
28 Chapter 2. Overview of vibration based damage identification methods
The effect of the delamination on the dynamic behavior of the composite beam
is a combination of two effects: one is the discontinuity in shear stress over the
beam thickness, the other is the stretching and compressing of the delaminated
beam segments. The first effect results in a reduction of flexural rigidity, while the
second effect tends to compensate for this. The weakening effect of a delamination
is only noticed in case a shear force is acting in the beam. This means that the
effect of a delamination vanishes if the delamination is located in regions where
the shear force approaches zero, but is the largest at locations where the shear
force is maximum. The non-uniform distribution of shear stress across the length
of the beam causes the effect to be dependent on the axial location ac of the
delamination. As the shear stress distribution differs for every vibration mode,
the delamination will also affect each mode differently. These effects are clearly
demonstrated in Figure 2.6, showing the second and third vibration mode shape
and its derivatives for a through-width delamination (ac = 0.5, b = 0.2, h2 = 0.5) in a
composite beam. Note that the shear force is a function of the third derivative of the
displacement. The delamination is located near a point where the third derivative
of the second mode approaches zero, but is large for the third mode. Consequently,
the delamination has a negligible effect on the second mode, while the third mode
is clearly affected. Figure 2.7 shows that a similar but opposite effect on the second
and third mode is observed when the delamination moves to the axial location
ac = 0.3.
The effect of the local delamination is more pronounced in the mode shape
curvatures than observed in the mode shapes themselves. The delamination
causes clear discontinuities in the mode shape curvature at the boundaries of the
delaminated segment. These discontinuities are different for each vibration mode
and reveal the presence, the location, as well as the size of the delamination.
The size of the discontinuity is described by the continuity equation for bending
moments at the junction between the integral and delaminated segment [39]. The
discontinuities are the largest when the delamination is located at a nodal point in
the mode shape and at the neutral plane. They vanish for locations near the surface
and at an antinode. The ’constrained mode’ model, however, does not incorporate
the potential opening of the delaminated segment in the results. The opening of the
delamination allows for additional curvature changes for delamination locations
where the shear related curvature effects are minimal.
Besides the literature studies discussed in Section 2.3, the analytical results also
showed that modal curvatures are an appropriate damage feature in the case of
delamination damage. The discontinuities caused at the delamination boundaries
are significantly more pronounced than, for example, deviations in mode shapes.
This makes curvatures an interesting candidate for damage identification. The
direct relation between curvature and strain for a beam in bending also opens
possibilities for integrated sensing technologies based on, for example, optical fiber
Bragg gratings [40] or piezoelectric strain sensors [41].
2.4. Damage feature selection 29
x delamination x delamination
z z
1 1
Mode 2 Mode 2
Norm. u(z) [−]
Mode 3
Mode 3
−1 −1
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Position z/L [−] Position z/L [−]
(b) Mode shape. (b) Mode shape.
1 1
Norm. du/dz [−]
2 Mode 3
1
2
Mode 2
0 0
2
Mode 3 −1
−2 Mode 2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Position z/L [−] Position z/L [−]
(d) Mode shape, second derivative. (d) Mode shape, second derivative.
4
Norm. d u/dz [−]
2
3
Mode 2
2 0
Mode 3
3
0 −2 Mode 2 Mode 3
−2 −4
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Position z/L [−] Position z/L [−]
(e) Mode shape, third derivative. (e) Mode shape, third derivative.
Figure 2.6 The second and the third vibration Figure 2.7 The second and the third vibration
mode shape u(z) and the derivatives for a pristine mode shape u(z) and the derivatives for a pristine
(blue) and delaminated (red) composite beam (blue) and delaminated (red) composite beam
(ac = 0.5, b = 0.2, h2 = 0.5). (ac = 0.3, b = 0.2, h2 = 0.5).
30 Chapter 2. Overview of vibration based damage identification methods
Information
condensation
Modal domain Y(ωn)
Damage
temperature Frequency domain Y(ω) features
T
Time domain y(t)
displacement impedance
x Z
damping Damage parameter χd
Δζ
stiffness mass
force Δm
Δk
F damage
scenario strain
non- thermal
ε
voltage linearities ΔT velocity
U dimensions dx/dt
l
acceleration electrical
2
d x/dt
2 resistance
R
Figure 2.8 Damage features classified according to the amount of information condensation. The outer
circles represent a higher level of information condensation.
Fspring(t)
αk x(t) αk
Ftest(t)
m
c x(t)
k
(1-α)k
Figure 2.9 A simplified representation of the opening and closing behavior of damage by a mass-spring-
damper system with a bilinear stiffness subjected to a dynamic loading Ftest (t). The force exerted by the
spring is given by: Fspring (t) = k( x ) x (t).
often scaled vectors without a physical quantity. Finally, for damage identification
purposes typically a selection of the most sensitive modes is considered to obtain
an indicator. The compressed subset of parameters must be able to describe the
damage scenario. Information condensation can result in a robust indicator that has
a strong correlation with the damage scenario. Each transformation is, however,
also accompanied by the loss of potentially valuable information caused by the
intermediate assumptions. These assumptions can inherently limit the applicability
of the selected damage features depending on the desired performance level, as
defined in Section 1.3.2.
The consequences involved with the information condensation process for modal
parameters can easily be demonstrated by considering a mass-spring-damper
system. Damage can open and close under an appropriate dynamic loading. The
dynamic effects induced by the opening and closing motion are represented in a
simplified way by a mass-spring-damper system with a bilinear stiffness, depicted
in Figure 2.9. The governing dynamic equation reads:
with:
(
αk x≥0
k( x) = (2.8)
k x<0
and where m represents the mass, c the damping, k ( x ) the stiffness, x (t) the
time-dependent displacement and Ftest (t) the (harmonic) force. The parameter α
describes the extent of the bilinearity. The system properties are chosen arbitrary
and equal: m = 1 kg, k = 105 N m-1 , c = 5 Ns m-1 , α = 0.5. The time response
32 Chapter 2. Overview of vibration based damage identification methods
−5
x 10
α=0.5 (bilinear) f 2f α=0.5 (bilinear)
2 a a
0
α=1.0 (linear) 10 α=1.0 (linear)
1
|X| [m]
3f
a 4f
x [m]
0 −2
a 5f
10 a
−1
−4
−2 10
10 10.1 10.2 10.3 10.4 0 50 100 150
Time [s] Frequency [Hz]
(a) Time responses x (t). (b) Magnitude of the Fourier spectra | X ( f )|.
−4
x 10
5
α=0.5 (bilinear) α=0.5 (bilinear)
α=1.0 (linear) 10
0
α=1.0 (linear)
|X| [m]
x [m]
0
−2
10
−5 10
−4
0 5 10 15 20 0 50 100 150
Time [s] Frequency [Hz]
(c) Time responses x (t). (d) Magnitude of the Fourier spectra | X ( f )|.
Figure 2.10 The time responses and the magnitudes of their Fourier transformed signals obtained for a (a,
b) single tone ( f a = 25 Hz) and a (c, d) linear swept harmonic ( f a = 1–150 Hz) excitation of a mass-spring-
damper system with a linear (α = 1.0) and a bilinear (α = 0.5) stiffness.
the total response signal. The transformation of the time domain information to the
frequency domain significantly complicates, in this case, the interpretation of the
nonlinear response.
Transformation to the modal domain will further condense the system information.
Modal analysis essentially assumes a time invariant and a linear system. Several
modal analysis methods are available in the literature [43]. The method used in the
present work extracts the modal parameters by applying an experimental modal
analysis (EMA) procedure to the frequency response functions (FRFs) between the
excitation and the response location. These frequency response functions HFx (ω )
can be calculated in different ways. By assuming that mainly the response signal is
subject to noise, a least square estimate is given by [44]:
S Fx (ω )
HFx (ω ) = , (2.9)
S FF (ω )
with S FF (ω ) the auto-power spectral density of the excitation signal and S Fx (ω ) the
cross-power spectral density between the excitation and response signal.
When the frequency response functions have been calculated, the modal parameters
can be estimated. The experimental modal analysis procedure used in the present
work consists of two steps. In the first step, the number of modes, the damped
natural frequencies and the associated damping factors are determined from
the complete set of frequency response functions with the help of a singular
value decomposition. A detailed description of this process is presented in [45].
The second step concerns the estimation of the modal parameters by curve
fitting a mathematical modal-parameter model to all measured frequency response
functions. The modal model in partial fraction form for a multi-degree of freedom
system is given by [46, 47]:
Nmodes
Rn R∗n
H (ω ) = ∑ + , (2.10)
n =1
jω − λn jω − λ∗n
−60
α=0.5
Modal fit
|H | [dB]
−80
α=1.0
Modal fit
Fx
−100
−120
0 50 100 150
Frequency [Hz]
Figure 2.11 The curve fitted modal model and the magnitude of the original frequency response functions
between the harmonic sweep ( f a = 1–150 Hz) excitation and the displacement response of a mass-spring-
damper system with a linear (α = 1.0) and a bilinear (α = 0.5) stiffness.
where σn is the damping factor, ωdn is the damped natural frequency, ωn is the
undamped natural frequency and ζ n the relative damping of mode n. The residue
matrix Rn describes the mode strength and is related to the mode shape vector [49]:
Rn = Qn ψn ψnT , (2.12)
where Qn is a scaling constant and ψn is the mode shape vector for the nth mode.
The curve fitted mathematical model and the original frequency response function
for the linear (α = 1.0) and bilinear (α = 0.5) mass-spring-damper system are
presented in Figure 2.11. The frequency responses are reduced to a set of modal
parameters as shown in Table 2.3. The modal vectors are not presented here
because a single degree of freedom system is considered. An exact description
is obtained for the linear system. The modal representation of the bilinear system,
however, largely ignores the nonlinear effects and only provides a linear description
of the nonlinear system. The estimated natural frequency corresponds to a linear
system with a stiffness of k = 0.68 · 105 N m-1 (32% degradation). This stiffness is
essentially the result of a variable stiffness (k = 1.0 · 105 and αk = 0.5 · 105 N m-1 )
within each oscillation and therefore provides an incorrect characterization of the
damage. These results show that the modal parameter changes can be used to
indicate the presence of structural changes, but that the information is reduced
Table 2.3 Modal parameters estimated by curve fitting the modal model to the frequency response functions
of the linear (α = 1.0) and the nonlinear (α = 0.5) system.
beyond the point where it allows for an adequate description of the damaged
system. This is primarily the case when damage causes severe nonlinear dynamic
effects. An adequate description is obtained if weakly nonlinear dynamic effects are
present.
References
[1] D. Montalvão, N.M.M. Maia, and A.M.R. Ribeiro. A review of vibration-based
structural health monitoring with special emphasis on composite materials. Shock and
Vibration Digest, 38(4):295–326, 2006.
[2] Y. Zou, L. Tong, and G.P. Steven. Vibration-based model-dependent damage
(delamination) identification and health monitoring for composite structures – A
review. Journal of Sound and Vibration, 230(2):357–378, 2000.
[3] C.-P. Fritzen and P. Kraemer. Self-diagnosis of smart structures based on dynamical
properties. Mechanical Systems and Signal Processing, 23(6):1830–1845, 2009.
[4] C.R. Farrar, H. Sohn, F.M. Hemez, M.C. Anderson, M.T. Bement, P.J. Cornwell, S.W.
Doebling, J.F. Schultze, N. Lieven, and A.N. Robertson. Damage prognosis: current
status and future needs. Technical report, Los Alamos National Laboratory, Los
Alamos, NM, 2001.
36 Chapter 2. Overview of vibration based damage identification methods
[5] S.W. Doebling, C.R. Farrar, M.B. Prime, and D.W. Shevitz. Damage identification and
health monitoring of structural and mechanical systems from changes in their
vibration characteristics: a literature review. Technical report, Los Alamos National
Laboratory, NM, USA, 1996.
[6] S.W. Doebling, C.R. Farrar, and M.B. Prime. A summary review of vibration-based
damage identification methods. The Shock and Vibration Digest, 30(2):91–105, 1998.
[7] H. Sohn, C.R. Farrar, F.M. Hemez, D.D. Shunk, D. Stinemans, and B.R. Nadler. A
review of structural health monitoring literature: 1996-2001. Technical report, Los
Alamos National Laboratory, Los Alamos, NM, 2003.
[8] E.P. Carden and P. Fanning. Vibration based condition monitoring: a review. Structural
Health Monitoring, 3(4):355–377, 2004.
[9] W. Fan and P. Qiao. Vibration-based damage identification methods: a review and
comparative study. Structural Health Monitoring, 10(1):83–111, 2011.
[10] C. Boller, F.-K. Chang, and Y. Fujino. Encyclopedia of structural health monitoring. John
Wiley & Sons, Ltd, Chichester, UK, 2009.
[11] D. Balageas, C.-P. Fritzen, and A. Güemes. Structural health monitoring. ISTE Ltd, 2006.
[12] Y.Y. Li. Hypersensitivity of strain-based indicators for structural damage identification:
a review. Mechanical Systems and Signal Processing, 24(3):653–664, 2010.
[13] K. Worden, C.R. Farrar, J. Haywood, and M. Todd. A review of nonlinear dynamics
applications to structural health monitoring. Structural Control and Health Monitoring,
15:540–567, 2008.
[14] K.-Y. Jhang. Nonlinear ultrasonic techniques for nondestructive assessment of micro
damage in material: a review. Precision Engineering and Manufacturing, 10(1):123–135,
2009.
[15] R.D. Adams, D. Walton, J.E. Flitcroft, and D. Short. Vibration testing as a
nondestructive test tool for composite materials. ASTM Special Technical Publication 580,
Composite Reliability, pages 159–175, 1975.
[16] P. Cawley and R.D. Adams. The location of defects in structures from measurements of
natural frequencies. Journal of Strain Analysis, 14(2):49–57, 1979.
[17] P. Cawley and R.D. Adams. A vibration technique for non-destructive testing of fibre
composite structures. Journal of Composite Materials, 13(2):161–175, 1979.
[18] P. Cawley and R.D. Adams. Improved frequency resolution from transient tests with
short record lengths. Journal of Sound and Vibration, 64(1):123–132, 1979.
[19] S.D. Fassois and J.S. Sakellariou. Time-series methods for fault detection and
identification in vibrating structures. Philosophical Transactions of the Royal Society A:
Mathematical, Physical and Engineering Sciences, 365(1851):411–48, 2007.
[20] M. Rucka and K. Wilde. Application of continuous wavelet transform in vibration
based damage detection method for beams and plates. Journal of Sound and Vibration,
297(3-5):536–550, 2006.
[21] H.F. Hu, W.J. Staszewski, N.Q. Hu, R.B. Jenal, and G.J. Qin. Crack detection using
nonlinear acoustics and piezoceramic transducers – Instantaneous amplitude and
frequency analysis. Smart Materials and Structures, 19(6):065017, 2010.
[22] C.R. Farrar and D.A. Jauregui. Comparative study of damage identification algorithms
applied to a bridge: I. Experiment. Smart Materials and Structures, 7:704–719, 1998.
[23] C.R. Farrar and D.A. Jauregui. Comparative study of damage identification algorithms
applied to a bridge: II. Numerical study. Smart Materials and Structures, 7:720–731, 1998.
[24] A. Alvandi and C. Cremona. Assessment of vibration-based damage identification
References 37
Abstract
A vibration based damage identification method is investigated
experimentally for a 2.5-dimensional composite structure. The
dynamic response of an intact and a locally delaminated 16-layer
unidirectional carbon fiber PEKK reinforced T-beam is considered.
A force-vibration set-up, including a laser vibrometer system, is
employed to measure the dynamic behavior of the T-beam. The
modal strain energy damage index (MSE-DI) algorithm is applied
using the bending and torsion modes. Special attention is paid to the
effect of the number of measurement points required to detect and
localize the delamination accurately.
1 Reproduced from: T.H. Ooijevaar, R. Loendersloot, L.L. Warnet, A. Boer, R. Akkerman. Vibration
based structural health monitoring of a composite T-beam. Composite Structures, 92(9):2007-2015,
2010.
39
40 Chapter 3. Vibration based structural health monitoring of a composite T-beam
3.1 Introduction
Development of structural health monitoring (SHM) technologies for composite
based structural components for aircraft is one of the objectives of the European
research program Clean Sky, Eco-Design ITD. An increase of the part’s service life
reduces its cost and long-term ecological impact.
One of the key issues in composite structures is the early detection and localization
of damage. Often service induced damage does not involve visible plastic
deformation, but internal matrix related damage, like transverse cracks and
delaminations. Their detection imposes costly maintenance techniques. Vibration
based damage identification methods are promising as an alternative for the time-
consuming and costly nondestructive testing (NDT) methods currently available.
These methods also offer the potential to be used as a real-time health monitoring
system. The change of the dynamic properties is employed to identify damage such
as delaminations.
Grouve et al. [1] used a vibration based structural health monitoring technique
by monitoring shifts in natural frequencies to detect a delamination in a simple
composite beam structure. However, it is often complicated to localize damage
uniquely. Stubbs et al. [2] proposed a technique which is capable of detecting and
localizing damage in beam-like one-dimensional structures. They were the first
to introduce a damage identification method based on the observation that local
changes in the modal strain energy of the vibration modes of a structure are a
sensitive indicator of damage. Cornwell et al. [3] extended the method for plate-like,
hence two-dimensional, structures. Only recently, the step to three-dimensional
frame structures was made [4, 5].
Until now, the modal strain energy damage index (MSE-DI) algorithm is hardly
used on anisotropic, composite structures. Moreover, this methodology is mainly
applied to one-dimensional beam and two-dimensional plate-like structures with
only flexural vibrations. The question is whether the methodology is applicable for
more complex and larger structures with a minimum amount of sensors. A step in
this direction is to consider a 2.5-dimensional composite beam structure with a more
general deformation pattern, including bending and torsion, in order to extract
a maximum of damage related information from the measured vibration modes.
Duffey et al. [6] are the only ones who included torsion modes and vibrations
in their research. Torsion vibrations can become rather complex and tend to be
more difficult to measure. However, beside bending modes, torsion modes could
also provide damage related information. This could become advantageous, in
particular for larger structures, if the number of measurable vibration modes is
limited and the number of sensors is to be reduced. Therefore it is also important
to investigate the effect of the number of measurement points on the damage
identification results.
This chapter mainly presents the experimental work performed at the University
3.2. The modal strain energy damage index algorithm 41
The strain energy of a vibration mode is referred to as the modal strain energy of
that mode. Consequently, the total modal strain energy is the sum of the modal
strain energy contributions of all modes considered. The modal strain energy is
calculated by linking the deformation of a structure to the strain. A distinction can
be made between axial, flexural and torsional deformation-strain relations [6]. Only
the bending and the torsional deformations are considered in the present work. The
mechanical relations read (see Figure 3.1 for the coordinate system used):
∂2 u x/y My/x ∂θ xy Tz
= , = , (3.1)
∂z2 EIy/x ∂z GJxy
with u the displacement, θ the rotation, M the bending moment, T the torque, EI
and GJ respectively the bending and torsion rigidity of the beam. The subscript
‘(x/y)’ refers either to x or y. The rotation angle for small rotations is defined
as: θ xy (z) = ∂uy ( x, z) ∂x. The strain energies for the bending and torsional
deformations are obtained by integrating the squared strains over the length l of
the structure [6]:
Z l
!2 Z l 2
1 ∂2 u x/y 1 ∂θ xy
UB = EIy/x dz , UT = GJxy dz . (3.2)
2 0 ∂z2 2 0 ∂z
Consider the structure to vibrate in the nth bending mode. The modal strain energy
of this bending mode is given by:
(n)
2
(n) 1
Z l ∂2 u x/y
UB = EIy/x dz , (3.3)
2 0 ∂z2
42 Chapter 3. Vibration based structural health monitoring of a composite T-beam
(n)
with u x/y (z) the normalized amplitude of the mode shape. Subsequently, the
structure is discretized in N elements in axial direction (i.e. z-coordinate). The
(n)
strain energy UB,i for the nth mode and associated with the ith element is then
given by:
(n)
2
(n)
Z
1 zi ∂2 u x/y (n)
N
(n)
UB,i = EIy/x dz with: UB = ∑ UB,i , (3.4)
2 z i −1 ∂z2 i =1
(n) (n)
(n) UB,i (n) ŨB,i
FB,i = (n)
, F̃B,i = (n)
, (3.5)
UB ŨB
for the intact and damaged structure respectively. The fractional strain energy
remains relatively constant in the undamaged elements, under the assumption that
the damage is small and primarily located at element k [3, 7]:
(n) (n)
F̃B,i ≈ FB,i for i 6= k . (3.6)
This equation is rearranged and after substituting Equations (3.3), (3.4) and (3.5) it
can be derived that:
R zi . 2 . R . 2
f y/x ∂2 ũ(n) ∂z2 dz
EI
lf
EI y/x ∂ 2 (n)
ũ ∂z 2
dz
z i −1 x/y 0 x/y
1≈ R . 2 . R . 2 for i 6= k ,
zi 2 (n) 2 l 2 (n) 2
z EI y/x ∂ u x/y ∂z dz 0 EIy/x ∂ u x/y ∂z dz
i −1
(3.7)
The mean value theorem for integrals [8] is applied to the flexural rigidities EIy/x (z)
f y/x (z) in Equation (3.7) to obtain an average rigidity estimate for each
and EI
integral. For the integral over element i of the pristine case:
Z z . 2 Z z . 2
i (n) i (n)
EIy/x (z) ∂2 u x/y ∂z 2
dz = EIy/x (z0 ) ∂2 u x/y ∂z2 dz , (3.8)
z i −1 z i −1
where z0 is a point within the integration interval and EIy/x (z0 ) = ( EIy/x )i is the
weighted average flexural rigidity of element i. Similar estimates can be obtained for
the other integrals in Equation (3.7). Subsequently, a ratio of the flexural rigidities
is obtained by assuming that the change in the average value of the flexural rigidity
3.3. Composite T-beam structure 43
for the integrals over the entire length l of the beam is negligible for small damage:
R zi (n)
. 2 .R
(n)
. 2
(n)
.
∂2 ũ x/y ∂z2 ∂2 ũ x/y ∂z2
l
( EIy/x )i z i −1 dz 0 dz γ̃i γ̃(n)
≈ R . 2 .R . 2 = . for i 6= k ,
f y/x )i
( EI (n) (n) (n)
∂2 u x/y ∂z2 ∂2 u x/y ∂z2
zi l γi γ(n)
z i −1 dz 0 dz
(3.9)
(n)
where γi is the integral over element i and γ(n) the integral over the entire length
l. Stubbs et al. [2] and Cornwell et al. [3] used the right-hand side of Equation (3.9)
(n)
to define a local damage index β i . This indicator is approximately equal to 1
for the undamaged elements, i.e. i 6= k, but can change at the damaged element,
i.e. i = k. Finally, the information of each mode shape is combined into a single
damage index β i by using the definition proposed by Cornwell et al. [3], which is a
summation over the number of modes Nmodes considered:
,
Nmodes h . i Nmodes h . i
(n) (n) (n) (n)
β i = ∑ γ̃i γ̃ ∑ i γ γ . (3.10)
n =1 n =1
The derivation of the damage indicator for the torsional deformations follows
the same route. The starting point is the strain energy formulation for torsional
deformations as presented in Equation (3.2).
Clamping area
2.2 mm 3
1
2 x
m
40 mm 3 60 m
Shear 1 2
center mm
1000
50 mm z
100 mm
Figure 3.1 The composite T-beam with a [0/90]4,s laminate lay-up. The global and material coordinate
systems are indicated.
Side view
y Clamping area 100 mm Delaminated region
z
450 mm
60 1000 mm
Figure 3.2 A side view of the composite T-beam. The location and dimensions of the artificial delamination
at the T-joint are indicated.
3.4. Experimental analysis of a sub-structure 45
point of excitation and the measuring points along the T-beam are determined
using a laser vibrometer. The modal parameters, i.e. the natural frequencies, the
damping values and the mode shapes, are obtained from these frequency response
functions by using experimental modal analysis (EMA) [11]. Firstly, the effect of the
delamination on the natural frequencies and mode shapes is investigated. Secondly,
the modal strain energy damage index algorithm is evaluated using the mode shape
data from the intact and delaminated T-beam.
The complete dynamic set-up and data acquisition scheme used for the experimental
validation are shown in Figure 3.3. The T-beam is horizontally and vertically
clamped at one side employing the hydraulic clamps of an Instron 8516 Fatigue
system. For all tests a 6 kN vertical clamping force is combined with a clamping
pressure of 200 bar applied to the horizontal clamps. These clamping conditions
are obtained by analyzing the robustness of a frequency response plot to changes
in the clamping conditions. The repeatability of the experimental set-up has been
checked in Section 3.4.3. The T-beam was excited by a shaker with a stinger and
force transducer connected to a fixed point at the beam and a spring connected to
its support. A chirp excitation force has been applied to the structure. The laser
vibrometer sensor head is mounted on a traverse system, which has a horizontal
scanning range restricted to 590 mm. The laser vibrometer is used to measure the
velocities at an equally spaced measuring grid containing 3×30 points (L1/M1/R1
until L30/M30/R30, see Figure 3.4) within the range of the traverse system. These
grid lines were selected to be able to distinguish between torsion and bending
modes and to approximate the torsion angle. Two accelerometers, one at the point
of excitation and one at grid point R1, were used for validation of the responses
measured by the laser vibrometer. The frequency response functions between the
excitation force at the fixed excitation point and the velocities at all laser vibrometer
grid points are recorded by a Siglab system. A frequency range of 25–1025 Hz, with
a resolution of 0.313 Hz was selected. A measurement at each grid point consists of
20 averages, without overlap.
The modal parameters (i.e. natural frequency, mode shapes and damping values)
were obtained from the FRFs by using experimental modal analysis [11, 12]. A
Complex Mode Indicator Function (CMIF) together with a global curve fitting
procedure, based on a Rational Fraction Polynomial (RFP) function [13], was used to
estimate the poles and residue mode shape vectors. Slightly complex modal vectors
were considered to be real normal, by neglecting non-proportional damping effects.
The mode shapes from both T-beams were used for damage identification by the
modal strain energy damage index algorithm.
46 Chapter 3. Vibration based structural health monitoring of a composite T-beam
40
30
13.
Magnitude [dB]
20
10
0
-10
-20
0 200 400 600 800 1000 1200
Frequency [Hz]
12.
11. 10. 9.
2.
4.
1.
6.
5.
8. 7.
Figure 3.3 Dynamic set-up and data acquisition systems used for the experimental investigation.
3.4.3 Results
Bottom view x
1000 mm 60
50 590 mm scanning range laser
R1 20 mm R30 204 mm
50
41 41
z
M1 M30
37
L1 L2 L30
3x30 measurement points Excitation point
Figure 3.4 A bottom view of the composite T-beam. The location of the excitation point and the 3×30
measurement points is indicated.
First of all the repeatability on the frequency response plot was tested. Three
tests were performed on each T-beam with the laser vibrometer positioned at grid
point R1. Figure 3.5 shows the frequency response plots of the intact T-beam. The
different curves compare well. The average standard deviation over all measured
peak values was 0.56%. Overall, the peak values of the torsion modes show a higher
standard deviation (average: 0.75%) than for the bending modes (average: 0.31%).
Even smaller standard deviations over the measured peak values (average 0.33%)
were found for the delaminated T-beam. The largest standard deviation in this
case was 0.61% for the 5th bending mode with an average frequency of 814.0 Hz.
The natural frequencies used for comparison of the intact and delaminated T-beam
are obtained from curve fitted data of the complete set of 3×30 FRF measurements
instead of a single FRF measurement. The standard deviations are of the same order
of magnitude.
The repeatability on the mode shapes was assessed in terms of the modal assurance
criterion (MAC). This criterion provides a measure for the degree of consistency
between two modal vectors. A value close to one indicates a good correspondence
between the modal vectors, whereas a value close to zero indicates that the modal
40
|HFv| [dB]
20
Case I
0 Case II
Case III
0 100 200 300 400 500 600 700 800 900 1000
Frequency [Hz]
Figure 3.5 The repeatability of the frequency response function based on three measurements performed at
grid point R1 of the pristine T-beam.
48 Chapter 3. Vibration based structural health monitoring of a composite T-beam
1 1
Case A Case A
0.5 Case B 0.5 Case B
uy (z) [−]
uy (z) [−]
Left Left
0 Center 0 Center
(n )
(n )
Right Right
−0.5 −0.5
−1 −1
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
z−coordinate [m] z−coordinate [m]
(a) 4th torsion mode shape (MAC = 0.999). (b) 7th bending mode shape (MAC = 0.995).
Figure 3.6 The repeatability of the (a) 4th torsion and (b) 7th bending mode shape of the delaminated
T-beam measured at the 3×30 grid points according to the left, center and right grid line.
with ψm A the modal vector of mode m obtained at the first measurement (case A),
B
ψn the modal vector of mode n obtained at the second measurement (case B),
∗ the complex conjugate and T the transpose. Figures 3.6(a) and 3.6(b) show
the comparison of two mode shape measurements for the 4th torsion mode and
the 7th bending mode of the delaminated T-beam. The mode shapes of the two
measurements correspond well. Overall, this holds also for the other measured
bending and torsion mode shapes with, respectively, average MAC values of 0.960
and 0.975. Only a few mode shapes show slightly larger deviations. The reason for
these deviations can be various. Possible explanations are the effects caused by the
clamping, the spring tension of the shaker support, the positioning and alignment
of the set-up and several other noise-sensitive aspects.
Since all average standard deviations do not exceed the 1% and the average
MAC values of the delaminated T-beam exceed the 0.950, it is concluded that the
natural frequencies and mode shapes obtained at the experimental set-up satisfy
the demands in terms of repeatability. Results on the fabrication repeatability of
the intact and delaminated T-beams are not available yet. Equal process conditions,
materials and tools are used. The weight of the intact (518 grams) and delaminated
(514 grams) T-beam correspond well.
3.4. Experimental analysis of a sub-structure 49
60 Intact
Damaged
|H | [dB] 40
20
Fv
−20
0 100 200 300 400 500 600 700 800 900 1000
Frequency [Hz]
Figure 3.7 FRF comparison of the intact and delaminated T-beam at grid point R1.
First of all the effect of the delamination on the natural frequencies and mode
shapes is investigated. In Figure 3.7 the frequency response plots of the intact
and delaminated T-beam at grid point R1 are compared. This figure indicates that
the natural frequencies of the delaminated T-beams are lower than for the intact
T-beam. Qualitatively this corresponds well with the literature [15, 16]. It can also
be observed that the difference in natural frequency is larger for higher modes.
Table 3.1 provides an overview of the natural frequencies f (n) and damping values
ζ (n) obtained from the curve fitted FRF data of the intact and delaminated T-beam.
The frequencies are grouped based on their dominating type of mode shape. The
bending modes in the yz-plane and torsion modes about the z-axis are respectively
referred by Byz and Tz . Coupling between some successive modes was found.
Some modes are experimentally not measured due to various reasons. The natural
frequency of the first torsion mode is smaller than the lowest frequency of 25 Hz
of the excitation signal. Other modes, like the 6th torsion mode, are not measured
because of the low response. In that case, the point of excitation almost coincides
with a point of zero amplitude in the mode shape. This overview also confirms
the shift in natural frequencies to lower values of the delaminated T-beam in the
case of the bending modes. Note that the natural frequencies of the torsion modes
hardly change. The delamination underneath the T-joint mainly affects the natural
frequencies of the bending modes and has less influence on the natural frequencies
of the torsion modes. Numerical results presented in [10] confirm this conclusion.
This observation can be explained by the fact that the delamination is located close
to the shear center of the cross section of the T-beam. The (viscous) damping
values ζ (n) are also determined (see Table 3.1) but no solid conclusions could be
drawn from these values. The statistic variations exceed the changes caused by the
delamination.
Each mode of the intact and delaminated T-beam, mentioned in Table 3.1, is related
to a mode shape. A selection of experimentally obtained mode shapes is shown
in Figures 3.8, 3.9 and 3.10. The mode shapes are determined for a section of the
50 Chapter 3. Vibration based structural health monitoring of a composite T-beam
Table 3.1 Experimentally obtained natural frequencies f (n) and damping values ζ (n) of mode n.
T-beam due to the limited scanning range of the laser vibrometer. In most cases it
is possible to extrapolate the mode shapes, supported by the numerical model [10].
The mode shapes from the intact and delaminated T-beam are used for damage
identification by the modal strain energy damage index algorithm, presented in
Section 3.2. One of the advantages of this algorithm is that damage related
information of separate modes is combined to obtain a damage index distribution,
using Equation (3.10). The required derivatives of the mode shapes are obtained
after applying a cubic spline interpolation (piecewise function) to the data points of
3.4. Experimental analysis of a sub-structure 51
1 0 1 0
uy (z) [−]
u y
−1 −1
−0.04 0.04 1 z−coordinate [m] −0.04 0.04 1 z−coordinate [m]
u y
−1 −1
1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0
z−coordinate [m] z−coordinate [m]
(5) (5)
(a) Intact ( f B = 840 Hz). (b) Delaminated ( f˜B = 816 Hz).
Figure 3.8 Experimentally obtained 5th bending mode shapes (MAC = 0.852).
1 0 1 0
uy (z) [−]
0 0
0.5 0.5
(n)
u y
−1 −1
−0.04 0.04 1 z−coordinate [m] −0.04 0.04 z−coordinate [m]
1
x−coordinate [m] x−coordinate [m]
uy (z) [−]
1 1
0 0
(n)
u y
−1 −1
1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0
z−coordinate [m] z−coordinate [m]
(7) (7)
(a) Intact ( f B = 934 Hz). (b) Delaminated ( f˜B = 920 Hz).
Figure 3.9 Experimentally obtained 7th bending mode shapes (MAC = 0.833).
52 Chapter 3. Vibration based structural health monitoring of a composite T-beam
1 0 1 0
uy (z) [−]
u y
−1 −1
−0.04 0.04 1 z−coordinate [m] −0.04 0.04 1 z−coordinate [m]
u y
−1 −1
1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0
z−coordinate [m] z−coordinate [m]
(9) (9)
(a) Intact ( f T = 616 Hz). (b) Delaminated ( f˜T = 601 Hz).
Figure 3.10 Experimentally obtained 9th torsion mode shapes (MAC = 0.971).
the mode shapes. The interpolation is used to improve the calculation of the mode
shape derivatives at discrete points in the z-direction. In the analysis, the scanning
area is subdivided into 40 elements.
The results obtained by considering only the 4th , 5th , 7th and 8th bending dominated
mode shapes and the full set of 3×30 grid points, are shown in Figure 3.11. The
mode shapes of the lowest bending modes are not taken into account. These modes
show some coupling or other kind of distortions and contain less information about
the presence and location of the delamination. The data obtained at the center grid
line clearly identifies the presence, location and size of the delamination. This is
mainly due to the fact that the measuring points are close to the delamination. The
left and right grid line also show higher damage indices at the delaminated region,
apart from some fitting related artifacts. The ratio between the damage index of
the damaged area and the intact area is lower, but the location is still predicted
correctly. It was shown that the value of the damage index is a measure for the
sensitivity to identify damage at a certain distance from the measurement points.
The same analysis is performed by using less data points. Figure 3.12 shows the
results obtained by using 3×30, 3×15, 3×10, 3×8, 3×6 and 3×5 equally spaced
measurement points at each grid line to define the mode shape of the T-beam.
The bending modes incorporated into the analysis remains unchanged. The ratio
between the damage indices of the damaged area and the intact area is lower for a
decreasing number of data points. It is shown that the number of measurement
points affects the sensitivity to identify damage at a certain distance from the
3.4. Experimental analysis of a sub-structure 53
10 10
Damage Index β [−]
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
z−coordinate [m] z−coordinate [m]
(a) Left grid line. (b) Right grid line.
10
Damage Index β [−]
0
0 0.2 0.4 0.6 0.8 1
z−coordinate [m]
(c) Center grid line.
Figure 3.11 Damage index β i distributions obtained for the bending dominated modes (4th , 5th , 7th and
8th ) defined by 3×30 grid points and using 40 MSE-DI elements.
measured points. The damage indices obtained from the center grid line of the
T-beam clearly indicate the presence, location and size of the delamination using 8
equally spaced data points within the scanning range. The left and right grid line
show higher damage indices at the delaminated region in case 10 equally spaced
data points were used. The location of the delamination is predicted correctly,
despite the lower absolute values of the damage index compared to the damage
indices at the center grid line. The difference between the results obtained from the
left and right grid line can be allocated to differences between mode shapes, and
therefore derivatives, at the left and right grid line.
5 5
0 0
0 0.2 0 0.2
0.4 0.4
0.6 0.6
0.8 0.8
1 z−coordinate [m] 1 z−coordinate [m]
Data
Damage Index β [−]
points
10
30
15
10
5 8
6
0 5
0 0.2
0.4
0.6
0.8
1 z−coordinate [m]
Figure 3.12 Damage index β i distributions obtained for the bending dominated modes defined by 3×30,
3×15, 3×10, 3×8, 3×6 and 3×5 equally spaced grid points.
grid points of the 2nd , 3rd , 4th , 7th , 9th , 10th and 12th torsion mode shape. The
damage index based on torsion dominated mode shapes gives no indication of
presence, location and size of the delamination. This conclusion matches the
conclusions drawn based on the modal parameters. The explanation for this is
the small distance between delamination and shear center at the cross section of the
T-beam.
3.5 Conclusions
The results of the vibration experiments show that the modal strain energy damage
index algorithm can be used to identify a delamination in a composite T-beam
structure in the case of bending vibrations. Although measurements of changes
in natural frequencies of the bending modes only indicate the presence of the
delamination easily, the 4th and higher bending mode shapes and the damage
3.5. Conclusions 55
Damage Index β
4
0
0 0.2 0.4 0.6 0.8 1
z−coordinate [m]
Figure 3.13 Damage index β i distribution obtained for the torsion dominated modes (2nd , 3rd , 4th , 7th ,
9th , 10th and 12th ) defined by 30 grid points and using 40 MSE-DI elements.
index algorithm also predict the location and size of the defect. The damage
index distributions provide a more pronounced identification of the delamination
compared to the single mode shapes. The combination of the different mode shape
derivatives in the damage index algorithm makes the methodology sensitive, even
with a low number of measurement points.
The small distance between the location of the delamination and the shear center
of the cross section of the structure causes the torsion modes to be hardly affected
for this structure. Only a few mode shapes indicate the presence and location of
the delamination. The natural frequencies and damage index remain inconclusive,
as was shown by the measurements. No solid conclusions can be drawn in this
particular case. Perhaps there are other cases where torsional deformations will
have an influence on the damage identification results.
It was shown that measurements at a certain distance from the local delamination,
like the left and right grid line, can be used to identify the delamination. The
damage index is affected by the distance between the measured points and the
delamination.
A reduction in the number of measurement points affects the sensitivity to identify
damage at a certain distance from the measured points. The ratio between the
damage index of the damaged area and the intact area is lower, but the location is
still predicted correctly by using a minimum of eight equally spaced measurement
points within the limited scanning range at the center grid line.
A next step towards the practical application of vibration based structural health
monitoring is to use integrated sensor systems for the mode shape measurements
used in the damage identification process. For this purpose, the numerical model,
as proposed in [10] can be used to optimally place the different sensors. Moreover,
larger structures with multiple stiffeners should be investigated to further study the
capabilities of the vibration based damage identification approach.
56 Chapter 3. Vibration based structural health monitoring of a composite T-beam
Acknowledgements
The authors would like to acknowledge the support of Fokker Aerostructures B.V.,
Hoogeveen, The Netherlands, for the manufacturing of the composite T-beam
structures on which the experiments were performed. This work was carried out
in the framework of the European project Clean Sky, Eco-Design ITD (grant
agreement number CSJU-GAM-ED-2008-001).
References
[1] W.J.B. Grouve, L.L. Warnet, A. de Boer, R. Akkerman, and J. Vlekken. Delamination
detection with fibre Bragg gratings based on dynamic behaviour. Composites Science
and Technology, 68(12):2418–2424, 2008.
[2] N. Stubbs and C.R. Farrar. Field verification of a nondestructive damage localization
and severity estimation algorithm. In Proceedings of the 13th International Modal Analysis
Conference (IMAC XIII), pages 210–218, 1995.
[3] P.J. Cornwell, S.W. Doebling, and C.R. Farrar. Application of the strain energy damage
detection method to plate-like structures. Journal of Sound and Vibration, 224(2):359–374,
1999.
[4] H. Li, H. Yang, and S.-L.J. Hu. Modal strain energy decomposition method for damage
localization in 3D frame structures. Journal of Engineering Mechanics, 132(9), 2006.
[5] H. Li, H. Fang, and S.-L.J. Hu. Damage localization and severity estimate for
three-dimensional frame structures. Journal of Sound and Vibration, 301(3-5):481–494,
2007.
[6] T.A. Duffey, S.W. Doebling, C.R. Farrar, W.E. Baker, and W.H. Rhee. Vibration-based
damage identification in structures exhibiting axial and torsional response. Journal of
Vibration and Acoustics, 123(1):84, 2001.
[7] A. Alvandi and C. Cremona. Assessment of vibration-based damage identification
techniques. Journal of Sound and Vibration, 292(1-2):179–202, 2006.
[8] J. Stewart. Calculus: early transcendentals. 5th edition. Brooks/Cole - Thomson
Learning, 2003.
[9] A. Offringa, J. List, J. Teunissen, and H. Wiersma. Fiber reinforced thermoplastic butt
joint development. In Proceedings of the International SAMPE Conference, pages 1–16,
2008.
[10] R. Loendersloot, T.H. Ooijevaar, L.L. Warnet, A. de Boer, and R. Akkerman. Vibration
based structural health monitoring and the modal strain energy damage index
algorithm applied to a composite T-beam. In C.M.A. Vasques and J.D. Rodrigues,
editors, Vibration and Structural Acoustics Analysis: Current Research and Related
Technologies, Chapter 6, pages 121–150. Springer, 2011.
[11] B.J. Schwarz and M.H. Richardson. Experimental modal analysis. In CSI Reliability
Week, Orlando, FL, pages 1–12, 1999.
[12] M.H. Richardson and B.J. Schwarz. Modal parameter estimation from operating data.
Sound and Vibration, 37(1):28–39, 2003.
[13] M.H. Richardson and D.L. Formenti. Global curve fitting of frequency response
measurements using the rational fraction polynomial method. In 3rd International
Modal Analysis Conference, pages 1–8, 1985.
References 57
[14] R.J. Allemang. The modal assurance criterion – Twenty years of use and abuse. Journal
of Sound and Vibration, 1:14–21, 2003.
[15] P.M. Mujumdar and S. Suryanarayan. Flexural vibrations of beams with delaminations.
Journal of Sound and Vibration, 125(3):441–461, 1988.
[16] J. Lee. Free vibration analysis of delaminated composite Beams. Computers and
Structures, 74(2):121–129, 2000.
Chapter 4
Abstract
The feasibility of a vibration based damage identification approach
for impact damage in two advanced composite skin-stiffener structures
is investigated in this chapter. Mode shape curvatures combined
with the modal strain energy damage index (MSE-DI) algorithm
are utilized to identify the damage. Special attention is paid to
the effective application of this vibration based methodology by
investigating the relation between the damage location, the structural
design and the dynamic behavior. The performance of a 1D and
2D formulation of the MSE-DI algorithm is compared for several
damage scenarios. Experiments demonstrated the capabilities of the
MSE-DI algorithm to detect, localize and roughly quantify the size
of barely visible impact damage in advanced composite structures.
It is concluded that the method is particularly effective for health
monitoring of skin-stiffener connections. The most effective results
were obtained by considering the 1D formulation in the direction of
the stiffeners for the stiffener mid-section and perpendicular to the
stiffeners for the stiffener run-out. The method remained inconclusive
in the case of pure skin related damage. The results obtained in this
study contribute to the development of guidelines for vibration based
structural health monitoring of advanced composite skin-stiffener
structures.
1 Reproduced from: T.H. Ooijevaar, L.L. Warnet, R. Loendersloot, R. Akkerman and T. Tinga,
Impact damage identification in composite skin-stiffener structures based on modal curvatures.
In preparation for: Structural Control and Health Monitoring, 2014.
59
60 Chapter 4. Damage identification in skin-stiffener structures based on curvatures
4.1 Introduction
One of the key issues in composite structures is the early identification of damage.
Often service induced damage does not involve visible plastic deformation, but
internal matrix related damage, like transverse cracks and delaminations. Their
identification often imposes time-consuming and costly nondestructive inspection
techniques. Structural health monitoring (SHM) technologies propose to be a
promising alternative.
A wide range of technologies is currently employed for health monitoring
purposes [1]. The majority of these technologies utilizes local or global dynamic
characteristics [2, 3]. This classification is based on the relative size of the wave
length with respect to the overall structural dimensions. Global methods, typically
structural vibration based technologies, provide data that is relatively easy to
interpret, but the frequency range and hence the resolution are limited [3]. As a
consequence, only relatively severe damage can be identified. Typical applications
are found in civil and offshore structures [4–8]. The local methods, comprising
higher frequency wave propagation technologies, are capable of detecting small
damage such as cracks [9–13], but it is often more complicated to interpret the data,
in particular in the case of non-flat or complex structures. The selection of the best
technology for a SHM system is often a matter of compromise. None of the local or
global approaches is, at the moment, able to solve all problems in all structures [14].
The number of successful practical applications of the SHM technologies on
composite aircraft structures is limited compared to applications in civil and
offshore industry [14–17]. This is mainly due to the complexity of the components,
the variety of potential damage scenarios and the high demands on safety and
reliability of the SHM system. As a consequence, the unique identification
(presence, location and severity) of damage in realistic composite structures by
methodologies operating under field conditions is considered as one of the key
challenges for these technologies.
This chapter addresses an experimental feasibility study on the identification
of impact damage in two advanced (i.e. multiple stiffeners, non-uniform skin
thickness, large dimensions) composite skin-stiffener structures, representing
typical aircraft structures. The present study is an extension of earlier performed
research by the authors on a 1 m composite T-beam specimen with an artificial
delamination [18, 19]. A global vibration based method is considered. This
approach allows for a relatively easy and physical interpretation of the measured
responses, which is desired for the investigation of complex composite structures.
Mode shape curvatures combined with the modal strain energy damage index
(MSE-DI) algorithm [20, 21] are respectively utilized as damage feature and
classifier to detect and localize the naturally originated defects. Studies by
different authors [4, 22–26] showed that this mode shape curvature based approach
outperforms other frequently applied modal domain damage identification methods
4.2. Composite skin-stiffener structures 61
based on natural frequencies [27–29], mode shapes [30, 31], dynamic flexibility [32–
34] and other mode shape derivative based features [35, 36]. The MSE-DI algorithm
essentially utilizes the curvature changes that locally arise under dynamic loading
of damaged structures. The performance of a 1D and 2D formulation of the
MSE-DI algorithm is compared for several local damage scenarios. Up to now,
the feasibility of the method is mainly demonstrated on concrete and metallic
structures [4, 6, 24, 37]. Application to composite structures is limited to relatively
simple composite beams and plates [38–40] with mainly severe and well-defined
artificial damage, such as a saw cut. The present work aims to extend this to more
advanced composite structures and impact induced damage scenarios.
Although the low frequency technologies allow for a relatively easy interpretation
of the responses, their sensitivity to identify small and local damage is rather
limited [16, 23, 41]. Special attention is therefore paid to the effective application of
the vibration based methodology. The effectiveness of the methodology is directly
related to aspects such as the frequency level, the location, orientation, type and
size of the damage, the design of the structure, the location and number of sensors,
etc [18, 22]. This chapter addresses the relation between the damage location, the
structural design and the dynamic behavior in order to extract recommendations
for the effective application of the methodology. The results obtained in this
study contribute to the development of guidelines for vibration based structural
health monitoring of advanced composite skin-stiffener structures. Numerical
studies performed on the same skin-stiffener structures were presented in two book
chapters [42, 43].
2.16mm
Excitation 2.16mm
1 3
point 2
3
1 2
x Impact location: case II
y Damaged area: case II
z
Bottom view Side view
400 mm Damaged area: case II
38 Impact location: case II
70mm
#89 #99
31 31
~120 mm
Top-ply fibre
142mm
282mm
orientation
Excitation
y
point
10mm #01 x #11
8mm Measurement point 40mm
Figure 4.1 Three dimensional and bottom view of the composite skin-stiffener specimen with 2 butt-joint
stiffeners. The dimensions, the measurement points (dots) and the impact locations are indicated.
The second structure is a larger and more advanced three stiffener plate structure
(1304×456 mm) with a non-uniform skin thickness distribution, as shown in
Figure 4.2. It represents a more realistic aircraft structure. The plate consists of
a 16-layer quasi-isotropic mid-section and 44-layer and 30-layer end-sections. The
stiffeners are made of a 15-layer quasi-isotropic stack. The structure contains a Z-
shaped preform based stiffener run-out with a titanium insert. Offringa et al. [45]
stated that this run-out design allows the structure to withstand a 50 J impact on the
stiffener from the outside, satisfying airliner and business jet impact requirements.
Only limited damage should be visible at this impact level.
The skin-stiffener structures are vulnerable for impact damage. The connection
between skin and stiffener is considered as the location where damage due to
impact would have the largest effect on the integrity of the structure. A small
delamination at this connection can significantly reduce the local bending stiffness
and hence the performance of the skin-stiffener structure. This direct structural
effect along with the potential long-term delamination growth make the skin-
stiffener connection a good candidate for health monitoring.
4.2. Composite skin-stiffener structures 63
4 mm
t = 4.1
1 mm
t = 2.2
8 mm
t = 6.0
92mm
67.5 67.5 67
#203
228mm
153 mm 150mm
456mm
50
Excitation
y
point #001
19mm x 25mm
26mm A Impact location: #029 26mm 35mm
Measurement point
case III Damaged area:
case III
Figure 4.2 Three dimensional and bottom view of the composite aircraft structure with three stiffeners,
a non-uniform skin thickness and the impact resistant stringer run-out design. The dimensions, the
measurement points (dots) and the impact location are indicated. The dotted vertical lines indicate the edges
of the transition zones between the sections with different thicknesses.
Three impact induced damage scenarios, summarized in Table 4.1, are considered
in this study. Naturally originated defects are obtained by applying local impact at
several locations from the outside (i.e. the side of the skin without the stiffeners)
of the structure with the help of a Dynatup 8250 falling weight impact machine. A
repeated impact of 10 J was used for the two stiffener structure, while a single 50 J
impact was applied in the case of the three stiffener structure. Visual and ultrasonic
inspections revealed the damage mechanisms occurring.
Figure 4.1 depicts the impact location and damaged area of damage scenario I
and II in the structure with two stiffeners. The first damage scenario, shown in
Figure 4.3(a), consists of local transverse cracks and a small delamination at the
skin in between both stiffeners. More severe damage is obtained for the second
damage case at the skin-stiffener connection of the same structure. This damage
consists of first-ply transverse cracks and a delamination between the first two
64 Chapter 4. Damage identification in skin-stiffener structures based on curvatures
layers of the skin, as shown in Figure 4.3(b). The interface between the injection
molded filler and skin also failed over a significant length. The structure with three
stiffeners was impacted at a stiffener run-out, shown in Figure 4.2. This part of
the skin-stiffener connection is most vulnerable for failure under impact [45]. The
impact resulted in failure of the connection between the skin and the Z-shaped
stiffener run-out. Cracks are indicated in Figure 4.3(c). An ultrasonic inspection
also revealed delamination damage between the titanium insert and the skin.
The three damage scenarios considered show typical fiber reinforced material
related failure mechanisms. They were, except for a small dent, invisible from the
outside of the structure. The damage scenarios are referred to as barely visible
impact damage (BVID), according to the criteria used in aircraft industry [46].
Aircraft structures have historically relied on visual methods to identify BVID.
Presently, the dent depth is used as the damage metric to define BVID. The
largest dent depth measured for the two composite structures equaled 0.16 mm,
which is smaller than the 0.3 mm BVID threshold as stated in [46]. Structures
containing BVID must sustain ultimate load (UL) for the life of the aircraft structure.
Crack A-A A
Delamination and Skin Crack B-B
transverse cracks delamination A B B
(a) Damage case I. (b) Damage case II. (c) Damage case III.
Figure 4.3 Visible part of the impact damage introduced in the (a, b) two stiffener composite structure and
(c) three stiffener composite structure.
4.3. Damage identification procedure 65
This criterion sets the lower bounds for the identification capabilities of an SHM
approach.
Figure 4.4 A schematic representation of the damage identification procedure. The Fa (t) is the excitation
force, v(t) the velocity response, HFv (ω ) the frequency response function between an excitation and
(n)
measurement point, uz ( x, y) is the normalized mode shape of mode n, the tilde represents the damaged
case.
66 Chapter 4. Damage identification in skin-stiffener structures based on curvatures
y
(1,Ny)
ly (Nx,Ny)
yj
(i,j)
yj-1
z
(Nx,1)
(1,1)
0 x
0 xi-1 xi lx
Figure 4.5 A plate structure divided into Nx × Ny elements in x- and y-direction respectively.
The strain energy of a vibration mode is referred to as the modal strain energy
of that mode. The modal strain energy can be obtained by linking the flexural,
axial and/or shear modal deformation of a structure to the strain [47]. The present
work concentrates on the flexural deformations of the skin-stiffener structures. This
type of deformation is expected to be dominant for the vibration modes that are
excited by the shaker. The strain energy U, based on the normalized flexural mode
(n)
shape uz ( x, y) in z-direction, of each individual mode n of an isotropic vibrating
plate-like structure is represented by:
Z ly Z l x !
(n) 2
!
(n) 2
(n) 1 ∂2 u z ∂2 u z
U = D ( x, y) +
2 0 0 ∂x2 ∂y2
(n)
! (n)
! (n)
!2
∂2 u z ∂2 u z ∂2 u z
+ 2ν + 2 (1 − ν ) dxdy , (4.1)
∂x2 ∂y2 ∂x∂y
with D ( x, y) the flexural rigidity of the plate, ν the Poisson’s ratio, lx and ly the
dimensions of the plate structure in x- and y-direction respectively. The total modal
strain energy is approximated by the sum of Equation (4.1) over a limited set of
Nmodes modes, using the theory of modal superposition. The plate-like structure
can be discretized in Nx × Ny elements in x- and y-direction respectively, as shown
4.3. Damage identification procedure 67
in Figure 4.5. Each element ij contains a modal strain energy described by:
Z y Z x (n)
!2 (n)
!2
(n) 1 j i ∂2 u z ∂2 u z
Uij = D ( x, y) +
2 y j −1 x i −1 ∂x2 ∂y2
(n)
! (n)
! (n)
!2
∂2 u z ∂2 u z ∂2 u z
+ 2ν + 2 (1 − ν ) dxdy , (4.2)
∂x2 ∂y2 ∂x∂y
with:
Ny Nx
(n)
U (n) = ∑ ∑ Uij , (4.3)
j =1 i =1
and xi−1 , xi and y j−1 , y j the boundaries of element ij of the discretized structure
in x- and y-direction respectively. Similar quantities can be defined for a damaged
(n)
structure, using the normalized mode shapes ũz ( x, y) of the damaged structure.
Subsequently, the local fractional strain energies, as defined by Cornwell et al. [21],
are:
(n) (n)
(n)
Uij (n)
Ũij
Fij = , F̃ij = , (4.4)
U (n) Ũ (n)
for the pristine and damaged structure respectively. The fractional strain energy
remains relatively constant in the undamaged elements, under the assumption that
the damage is small and primarily located at element kl [21, 22]:
(n) (n)
F̃ij ≈ Fij for i, j 6= k, l . (4.5)
This equation is rearranged and by substituting Equations (4.1), (4.2) and (4.4) it
can be derived that:
R y j R xi .R l R
e (n) dxdy y lx e (n)
y j−1 xi−1 D w̃ 0 0 D w̃ dxdy
1≈ R . for i, j 6= k, l , (4.6)
y j R xi (n) dxdy
R ly R l x
(n) dxdy
y x D w 0 0 D w
j −1 i −1
where w(n) ( x, y) represents all terms between the square brackets in the integrand
of Equation (4.2), representing the information of the mode shape derivatives. The
mean value theorem for integrals [48] is applied to the stiffness terms D ( x, y) and
De ( x, y) in Equation (4.6) to obtain an average stiffness estimate for each double
integral. For the integral over element ij of the pristine case:
Z y Z x Z y Z x
j i j i
D ( x, y) w(n) ( x, y)dxdy = D ( x0 , y0 ) w(n) ( x, y)dxdy , (4.7)
y j −1 x i −1 y j −1 x i −1
68 Chapter 4. Damage identification in skin-stiffener structures based on curvatures
(n)
where γij is the integral of w(n) ( x, y) over element ij and γ(n) the integral of
w(n) ( x, y) over the entire dimensions l x and ly . Stubbs et al. [20], Cornwell et al. [21]
and Alvandi [22] used the right-hand side of Equation (4.8) to define a local damage
(n)
index β ij . This indicator is approximately equal to 1 for the undamaged elements,
i.e. i, j 6= k, l, but can change at the damaged element, i.e. i, j = k, l. Finally, the
information of each mode shape is combined into a single damage index β ij by
using the definition proposed by Cornwell et al. [21], which is a summation over
the number of modes Nmodes considered:
,
Nmodes h . i Nmodes h . i
(n) (n) (n)
β ij = ∑ γ̃ij γ̃ ∑ γij γ(n) . (4.9)
n =1 n =1
β ij − µ
Zij = . (4.10)
σ
This normalized indicator provides a statistical measure for the detection of outliers
in the damage index distributions, which are expected to be the damaged elements.
A strength of the MSE-DI algorithm is that it utilizes mode shape curvatures,
which showed to be a relatively sensitive damage feature compared to other modal
parameters [15, 35, 36]. The algorithm also combines damage related information
of a range of modes. This allows for a more robust damage index distribution,
according to Equation (4.9), because every mode is affected differently by the local
damage [18]. Weak points are the number of required data points to accurately
describe the mode shapes and the numerical errors introduced with the calculation
4.4. Experimental set-up and signal processing 69
of the second derivative [15, 36, 50, 51]. Although not extensively considered in
this chapter, many different publications address the problem of finding optimum
positions of the measurement points. Savananoz [52] particularly focused on
the estimation of the optimum spatial sampling interval for the calculation of
curvatures in beams. The present formulation of the MSE-DI algorithm also does
not allow to directly link the value of the damage index β ij to a damage severity
in terms of a stiffness loss ratio [22]. The ratio between the damage indices β ij at
the damaged and intact area showed to be a measure for the relative change in the
dynamic behavior of the structure due to the local damage [18]. The damage index
β ij is merely a qualitative and mathematical rather than a physical quantity [43].
11.
40
30
12.
Magnitude [dB]
20
10
0
4. -10
-20
0 200 400 600 800 1000 1200
3. Frequency [Hz]
9.
10. 5.
F(t) 8.
7.
v(t)
6.
y
2. z
1.
Figure 4.6 Schematic representation of the experimental set-up and data acquisition systems used for the
dynamic measurements of both stiffened structures.
This criterion provides a measure for the degree of consistency between two modal
vectors. A value close to one indicates a good correspondence between the modal
vectors, whereas a value close to zero indicates that the modal vectors are not
consistent. The MAC criterion is defined as [56]:
A T B ∗ 2
(ψm ) (ψn )
MAC (m, n) = A )T ( ψA ) ∗ ( ψB )T ( ψB ) ∗
, (4.11)
(ψm m n n
with ψmA the modal vector of mode m obtained by a first measurement (case A),
B
ψn the modal vector of mode n obtained by a second measurement (case B), ∗ the
complex conjugate and T the transpose. Average MAC values of 0.997 and 0.996
were obtained for successive measurements at the two and three stiffener structure
respectively. The MAC values exceeded the 0.968 in all cases.
The residue mode shapes were subsequently cubic spline interpolated (piecewise
function) and normalized to their maximum amplitude. The interpolation is
used to improve the calculation of the second derivatives of the mode shapes at
4.5. Experimental results 71
discrete points in x- and y-direction. The mode shape curvatures from the pristine
and damaged structures were combined in the MSE-DI algorithm to investigate
the effect of the damage scenarios on the damage index distributions β ij or Zij ,
according to the equations presented in Section 4.3.
1 1
1 1
MAC [−]
MAC [−]
0.5 0.5
0.5 0.5
0 0
5 5
10 20 10 15
15 15 10
20 10 0 15 5 0
5
Mode number Mode number Mode number Mode number
pristine structure damaged structure pristine structure damaged structure
(a) Two stiffener structure: damage scenario I. (b) Two stiffener structure: damage scenario II.
1
1
MAC [−]
0.5
0.5
0
5 25
10 20
15 15
20 10 0
5
Mode number 25 Mode number
pristine structure damaged structure
(c) Three stiffener structure: damage scenario III.
Figure 4.7 Comparison of the experimental mode shapes, in terms of MAC values, of the pristine and
impact damaged structures. The mode numbers correspond to the modes within the considered frequency
ranges, i.e. 500–2000 Hz and 250–1000 Hz for the two and three stiffener structure respectively.
of the 2D formulation, shown in Figure 4.9(a). The highest damage indices are
obtained at the damaged area, indicating the presence and location of the damage
correctly. However, the ratio between the damage index β ij of the damaged area
and the intact area is rather low. The relatively noisy damage index distribution also
tends to show a propensity for false damage predictions (indication of damage, but
no actual change in the structure). These false positives can be caused by aspects
such as inaccuracies in the mode shape data, misalignment of measurement points,
errors introduced by the numerical methods used in the algorithm, but also by
considering a small number of modes in the analysis.
The presence and location of the impact damage is predicted in a more pronounced
way by the 1D formulation in x-direction, as shown in Figure 4.9(b). The most
distinct damage indices are obtained by considering the measurement points on
top of the skin-stiffener connection. The damage index distribution also provides
a rough estimation of the dimensions of the damaged area. A delamination length
of roughly 105 mm is found. Earlier research by authors [18] showed that the
accuracy of this result highly depends on the number of measurement points and
4.5. Experimental results 73
2 2
1.8 1.8
5 5
1.6 1.6
1.4 0 1.4
0
0.2 0.3 1.2 0.2 0.3 1.2
0.2 0.2
0.1 0.1 0.1 0.1
0 0 x−coordinate [m] 1 0 0 x−coordinate [m] 1
y−coordinate [m] y−coordinate [m]
(a) 2D formulation. (a) 2D formulation.
5 5
m
5m
~10
Damage Index β [−]
3 3
0 0
0.3 2 0.3 2
0.2 0.2
0.2 0.2
0.1 0.1 0.1 0.1
0 0 x−coordinate [m] 1 0 0 x−coordinate [m] 1
y−coordinate [m] y−coordinate [m]
(b) 1D formulation in x-direction. (b) 1D formulation in x-direction.
5 5
Damage Index β [−]
4 4
5 5
3 3
0 0
0.3 2 0.3 2
0.2 0.2
0.2 0.2
0.1 0.1 0.1 0.1
0 0 x−coordinate [m] 1 0 0 x−coordinate [m] 1
y−coordinate [m] y−coordinate [m]
(c) 1D formulation in y-direction. (c) 1D formulation in y-direction.
Case I Case II
y y
x x
Figure 4.8 Damage index β ij distribution for Figure 4.9 Damage index β ij distribution for
damage scenario I in the two stiffener structure. All damage scenario II in the two stiffener structure.
modes within 500–2000 Hz and 40×30 MSE-DI All modes within 500–2000 Hz and 40×30 MSE-
elements are used in the analysis. DI elements are used in the analysis.
74 Chapter 4. Damage identification in skin-stiffener structures based on curvatures
the frequency range considered in the analysis. The real damage length (≈ 120 mm)
is underestimated by approximately 12% for the situation considered here. Despite
the fact that the MSE-DI algorithm does not allow for a physical interpretation of
the damage severity in terms of mechanical properties [22, 43], the elements with
the highest damage index values are predicted at the center of the damaged area.
These elements belong to the region within the damaged area with the most severe
change in the local dynamic behavior.
Figure 4.9(c) presents the results obtained by the 1D formulation in y-direction.
Slightly higher damage indices are predicted at the location of the skin-stiffener
damage. However, also several other potential damage locations are predicted,
hence showing a propensity for false damage predictions.
The results presented until now show that the sensitivity of the method depends on
the extent to which the dynamic behavior is affected by the local damage. This effect
is defined by the failure mechanism and the location of the damage with respect
to the structural design. Hence, the highest sensitivity is obtained for skin-stiffener
damage by considering the 1D formulation underneath and in the direction of the
stiffeners. This is caused by the fact that mainly the flexural stiffness in this direction
is affected by the skin-stiffener interface failure. The flexural stiffness perpendicular
to the stiffener is hardly affected by the interface damage. This directional behavior
is also clearly illustrated by comparing the normalized mode shape curvatures in
x- and y-direction of the pristine ( f n = 996.4 Hz) and damaged ( f n = 987.3 Hz)
structure in Figure 4.10. Locally higher amplitudes are found near the damaged
interface for the curvature in x-direction, while the curvature in y-direction seems
hardly affected. This difference in x- and y-direction also explains the reduced
sensitivity of the 2D formulation of the MSE-DI algorithm, since data obtained in
both directions is used in that case.
1 1
y−coordinate [m]
y−coordinate [m]
0.2 0.5 0.2 0.5
0 0
0.1 0.1
−0.5 −0.5
0 −1 0 −1
0 0.1 0.2 0.3 0 0.1 0.2 0.3
x−coordinate [m] x−coordinate [m]
(a) Normalized mode shape curvature [-] in (b) Normalized mode shape curvature [-] in
x-direction of the pristine structure. x-direction of the damaged structure.
1 1
y−coordinate [m]
0 0
0.1 0.1
−0.5 −0.5
0 −1 0 −1
0 0.1 0.2 0.3 0 0.1 0.2 0.3
x−coordinate [m] x−coordinate [m]
(c) Normalized mode shape curvature [-] in (d) Normalized mode shape curvature [-] in
y-direction of the pristine structure. y-direction of the damaged structure.
. .
(n) (n)
Figure 4.10 Normalized mode shape curvatures ∂2 uz ∂x2 and ∂2 uz ∂y2 of the (a, c) pristine
( f n = 996.4 Hz) and the (b, d) damaged ( f n = 987.3 Hz) two stiffener structure for damage scenario II.
shows for the 2D formulation that the damage index values Zij at the damaged
region deviate up to 9 times the standard deviation σ from the mean value µ,
indicating the significance of the obtained result. Comparison of Zij results between
different cases is often rather difficult due to its dependency on the number of
elements selected in the MSE-DI algorithm.
In contrast to the obtained result, it must be noticed that the modal parameters
themselves are hardly affected by the impact damage. The average change in the
natural frequency was ∆ f n = 0.06% (maximum: ∆ f n = 0.31%) within the considered
frequency range. The mode shapes within this range provide an average MAC
value of 0.992 (minimum: MAC = 0.930), as shown in Figure 4.7(c).
In the previous section it was shown that, in the case of skin-stiffener damage,
the best results were obtained by considering the 1D formulation of the MSE-DI
algorithm in the direction of the stiffeners. The opposite behavior is shown for
the impact damage at the stiffener run-out as considered here. Figures 4.11(b)
76 Chapter 4. Damage identification in skin-stiffener structures based on curvatures
2 10
1.8 8
Damage Index β [−]
3 1.6 10 6
2 5
1.4 4
1 1.2 1.2
1 0 1
0 0.8 1.2 0.8 2
0.6 0.6
0.3 0.4 0.3 0.4
0.2 x−coordinate [m] 1 0.2 x−coordinate [m] 0
0 0 0 0
y−coordinate [m] y−coordinate [m]
(b) 1D formulation in x-direction. (b) 1D formulation in x-direction.
2 10
1.8 8
Damage Index Z [−]
Damage Index β [−]
3 1.6 10 6
2 5
1.4 4
1 1.2 1.2
0.9 0 0.9
0 1.2 2
0.6 0.6
0.2 0.3 1 0.2 0.3 0
0 0 x−coordinate [m] 0 0 x−coordinate [m]
y−coordinate [m] y−coordinate [m]
(c) 1D formulation in y-direction. (c) 1D formulation in y-direction.
Damaged area Damaged area
Impact location Impact location
m m
.1 m .1 m
Case III t=6 Case III t=6
m m
.2 m .2 m
y x t=2 y x t=2
m m
1m 1m
t = 4. t = 4.
Figure 4.11 Damage index β ij distribution for Figure 4.12 Normalized damage index Zij for
damage scenario III in the three stiffener structure. damage scenario III in the three stiffener structure.
All modes within 250–1000 Hz and 60×20 MSE- All modes within 250–1000 Hz and 60×20 MSE-
DI elements are used in the analysis. DI elements are used in the analysis.
4.5. Experimental results 77
y−coordinate [m]
y−coordinate [m]
t = 4.14 mm t = 2.21 mm t = 6.08 mm t = 4.14 mm t = 2.21 mm t = 6.08 mm
0.4 1 0.4 1
0.2 0 0.2 0
0 −1 0 −1
0 0.5 1 0 0.5 1
x−coordinate [m] x−coordinate [m]
(a) Local torsion ( f n = 538.2 Hz). (b) Local bending ( f n = 617.6 Hz).
(n)
Figure 4.13 Normalized mode shapes uz ( x, y) of the pristine structure with (a) local torsion
( f n = 538.2 Hz) and (b) local bending ( f n = 617.6 Hz) deformation around the central stiffener. Both
mode shapes show a wavelength transition in x-direction at the thicker sections of the structure.
and 4.11(c) show the damage index β ij distributions obtained by applying the MSE-
DI algorithm in respectively x- and y-direction. Figures 4.12(b) and 4.12(c) represent
the same directions in the case of the normalized damage index Zij . Several
false positives appear for the x-direction, while the y-direction clearly predicts the
presence and the location of the impact damage correctly.
The most likely explanation for the reduced effect in x-direction is threefold. Firstly,
the damage is located at a region that already shows a geometrical discontinuity in
x-direction (end of stiffened section) for the pristine case. This causes a transition in
the local deformation patterns and reduces the possibility for the run-out damage to
be expressed in the measured curvatures in this direction. Secondly, the damage is
located at the thickest part of the skin. Generally, the thickest skin section showed a
relatively large wavelength of the mode shape in x-direction, analogous to lower
vibration modes, compared to the wavelength in x-direction at the mid-section
and the wavelength in y-direction. The reduced sensitivity for the x-direction is
understandable, since lower vibration modes are in general less affected by local
damage compared to higher vibration modes [18]. The differences in wavelength
are clearly depicted in Figures 4.13(a) and 4.13(b). Thirdly, a substantial amount
of mode shapes show a local torsional deformation around the central stiffener in
x-direction, similar to Figure 4.13(a). An earlier study on a composite T-beam [18]
revealed that the extent of the influence depends on the type of local deformation
around the damaged area. The mode shape in x-direction will be hardly affected if
the deformation around the damage represents local torsion in the yz-plane, since
the damage is located close to the local rotational axis. Figure 4.14 schematically
demonstrates the difference between local torsion and local bending behavior
around a skin-stiffener connection. Any potential nonlinear effect (e.g. opening
and closing behavior of the delamination) in the measured structural response is
not considered in the MSE-DI analysis, since it relies on a linear structural model.
The modal data obtained from the random excitation reflects a linear and stationary
representation of the system.
The more pronounced effect of the damage in the MSE-DI results for the y-direction
78 Chapter 4. Damage identification in skin-stiffener structures based on curvatures
y y y y
z z Interface z z Interface
failure failure
Figure 4.14 Schematic representation of (a, b) local torsion and (c, d) local bending deformation around a
skin-stiffener connection for a (a, c) pristine and (b, d) damaged structure.
can be explained by the widening of the stiffener towards its end, as shown in
Figure 4.3(c). The delamination underneath the titanium insert locally reduces the
bending stiffness in y-direction. Moreover, the wavelengths in y-direction are for
many mode shapes relatively short compared to the wavelengths in x-direction at
the damaged area. As explained before, this causes a larger sensitivity to changes
by local damage.
4.6 Discussion
The derivation of the damage identification algorithm, presented in Section 4.3,
relies on a number of assumptions and simplifications. One of the most important
simplifications concerns the 1D and 2D formulation of the algorithm. The 2D
formulation assumes an isotropic description of the plate flexural rigidity. On the
other hand, the 1D formulation assumes that the flexural rigidity in one direction is
dominant and that all twisting and Poisson related effects are neglectable. In reality,
the stiffeners cause a local anisotropic design of the structures. As a consequence,
the validity of the 2D formulation is essentially only justified at the unstiffened skin
sections, whereas the 1D formulation provides the most realistic representation
at the stiffened skin sections. The results obtained for the two stiffener structure
complies with this decomposition, since the 1D formulation outperforms the 2D
formulation at the stiffeners. The damage of the three stiffener structure showed a
different behavior due to the fact that the damage is located at the stiffener run-out.
The results presented in Section 4.5 demonstrated the capabilities of the modal
curvature based approach to identify damage in composite skin-stiffener structures.
Several researchers [16, 23], however, question the suitability of modal data for
damage identification due to a relatively low sensitivity. The present work showed
that a refinement of this statement is justified. First of all, there is the trade-
off between the sensitivity to damage of the algorithm and its noise rejection
capability. A high sensitivity is desired to identify damage at an early stage. On
the other hand, it also increases the sensitivity to uncertainties, hence potentially
causing false damage identifications. Secondly, the minimum required sensitivity
of an approach is generally defined by the allowable damage limit (ADL) of
4.7. Conclusions 79
4.7 Conclusions
Damage
identification
method
Damage
Structure
scenario
Figure 4.15 The multidisciplinary framework associated with the development of a structural health
monitoring approach.
Acknowledgments
The authors kindly acknowledge the support of Fokker Aerostructures B.V.,
Hoogeveen, The Netherlands, for manufacturing the composite panels used in this
research. This work is carried out in the framework of the European project Clean
Sky, Eco-Design ITD (grant agreement number CSJU-GAM-ED-2008-001).
References
[1] C. Boller, F.-K. Chang, and Y. Fujino. Encyclopedia of structural health monitoring. John
Wiley & Sons, Ltd, Chichester, UK, 2009.
[2] K. Diamanti and C. Soutis. Structural health monitoring techniques for aircraft
composite structures. Progress in Aerospace Sciences, 46(8):342–352, 2010.
[3] C.-P. Fritzen and P. Kraemer. Self-diagnosis of smart structures based on dynamical
properties. Mechanical Systems and Signal Processing, 23(6):1830–1845, 2009.
[4] C.R. Farrar and D.A. Jauregui. Comparative study of damage identification algorithms
applied to a bridge: I. Experiment. Smart Materials and Structures, 7:704–719, 1998.
[5] C.R. Farrar and D.A. Jauregui. Comparative study of damage identification algorithms
applied to a bridge: II. Numerical study. Smart Materials and Structures, 7:720–731, 1998.
[6] J.J. Lee and C.B. Yun. Damage diagnosis of steel girder bridges using ambient
vibration data. Engineering Structures, 28(6):912–925, 2006.
[7] C.-P. Fritzen. Vibration-based structural health monitoring – Concepts and
applications. Key Engineering Materials, 293-294:3–18, 2005.
[8] N. Stubbs, S. Park, C. Sikorsky, and S. Choi. A global non-destructive damage
assessment methodology for civil engineering structures. International Journal of
Systems Science, 31(11):1361–1373, 2000.
[9] Z. Su, L. Ye, and Y. Lu. Guided lamb waves for identification of damage in composite
structures: a review. Journal of Sound and Vibration, 295(3-5):753–780, 2006.
[10] A. Raghavan and C.E.S. Cesnik. Review of guided-wave structural health monitoring.
The Shock and Vibration Digest, 39(2):91–114, 2007.
[11] J.-B. Ihn and F.-K. Chang. Pitch-catch active sensing methods in structural health
monitoring for aircraft structures. Structural Health Monitoring, 7(1):5–19, 2008.
[12] W.J. Staszewski, S. Mahzan, and R. Traynor. Health monitoring of aerospace composite
structures – Active and passive approach. Composites Science and Technology,
69(11-12):1678–1685, 2009.
82 Chapter 4. Damage identification in skin-stiffener structures based on curvatures
[13] G. Park and D.J. Inman. Structural health monitoring using piezoelectric impedance
measurements. Philosophical Transactions of the Royal Society A: Mathematical, Physical and
Engineering Sciences, 365(1851):373–392, 2007.
[14] D. Montalvão, N.M.M. Maia, and A.M.R. Ribeiro. A review of vibration-based
structural health monitoring with special emphasis on composite materials. Shock and
Vibration Digest, 38(4):295–326, 2006.
[15] W. Fan and P. Qiao. Vibration-based damage identification methods: a review and
comparative study. Structural Health Monitoring, 10(1):83–111, 2011.
[16] E.P. Carden and P. Fanning. Vibration based condition monitoring: a review. Structural
Health Monitoring, 3(4):355–377, 2004.
[17] M. Dilena and A. Morassi. Dynamic testing of a damaged bridge. Mechanical Systems
and Signal Processing, pages 1–23, 2011.
[18] T.H. Ooijevaar, R. Loendersloot, L.L. Warnet, A. de Boer, and R. Akkerman. Vibration
based structural health monitoring of a composite T-beam. Composite Structures,
92(9):2007–2015, 2010.
[19] R. Loendersloot, T.H. Ooijevaar, L.L. Warnet, A. de Boer, and R. Akkerman. Vibration
based structural health monitoring and the modal strain energy damage index
algorithm applied to a composite T-beam. In C.M.A. Vasques and J.D. Rodrigues,
editors, Vibration and Structural Acoustics Analysis: Current Research and Related
Technologies, Chapter 6, pages 121–150. Springer, 2011.
[20] N. Stubbs and C.R. Farrar. Field verification of a nondestructive damage localization
and severity estimation algorithm. In Proceedings of the 13th International Modal Analysis
Conference (IMAC XIII), pages 210–218, 1995.
[21] P.J. Cornwell, S.W. Doebling, and C.R. Farrar. Application of the strain energy damage
detection method to plate-like structures. Journal of Sound and Vibration, 224(2):359–374,
1999.
[22] A. Alvandi and C. Cremona. Assessment of vibration-based damage identification
techniques. Journal of Sound and Vibration, 292(1-2):179–202, 2006.
[23] J. Humar, A. Bagchi, and H. Xu. Performance of vibration-based techniques for the
identification of structural damage. Structural Health Monitoring, 5(3):215–241, 2006.
[24] J.-M. Ndambi, J. Vantomme, and K. Harri. Damage assessment in reinforced concrete
beams using eigenfrequencies and mode shape derivatives. Engineering Structures,
24(4):501–515, 2002.
[25] W. Lestari, P. Qiao, and S. Hanagud. Curvature mode shape-based damage assessment
of carbon/epoxy composite beams. Journal of Intelligent Material Systems and Structures,
18(3):189–208, 2007.
[26] Y.Y. Li. Hypersensitivity of strain-based indicators for structural damage identification:
a review. Mechanical Systems and Signal Processing, 24(3):653–664, 2010.
[27] P. Cawley and R.D. Adams. The location of defects in structures from measurements of
natural frequencies. Journal of Strain Analysis, 14(2):49–57, 1979.
[28] O.S. Salawu. Detection of structural damage through changes in frequency: a review.
Engineering Structures, 19(9):718–723, 1997.
[29] W.J.B. Grouve, L.L. Warnet, A. de Boer, R. Akkerman, and J. Vlekken. Delamination
detection with fibre Bragg gratings based on dynamic behaviour. Composites Science
and Technology, 68(12):2418–2424, 2008.
[30] E. Parloo, P. Guillaume, and M. van Overmeire. Damage assessment using mode shape
sensitivities. Mechanical Systems and Signal Processing, 17(3):499–518, 2003.
References 83
[31] L.M. Khoo, P.R. Mantena, and P. Jadhav. Structural damage assessment using vibration
modal analysis. Structural Health Monitoring, 3(2):177–194, 2004.
[32] A.K. Pandey and M. Biswas. Damage detection in structures using changes in
flexibility. Journal of Sound and Vibration, 169(1):3–17, 1994.
[33] T. Toksoy and A.E. Aktan. Bridge-condition assessment by modal flexibility.
Experimental Mechanics, 34(3):271–278, 1994.
[34] F.N. Catbas, M. Gul, and J.L. Burkett. Damage assessment using flexibility and
flexibility-based curvature for structural health monitoring. Smart Materials and
Structures, 17(1):1–12, 2008.
[35] A.K. Pandey, M. Biswas, and M.M. Samman. Damage detection from changes in
curvature mode shapes. Journal of Sound and Vibration, 145(2):321–332, 1991.
[36] M.M. Abdel Wahab and G. De Roeck. Damage detection in bridges using modal
curvatures: application to a real damage scenario. Journal of Sound and Vibration,
226(2):217–235, 1999.
[37] H. Hu and C. Wu. Development of scanning damage index for the damage detection
of plate structures using modal strain energy method. Mechanical Systems and Signal
Processing, 23(2):274–287, 2009.
[38] P. Qiao, K. Lu, W. Lestari, and J. Wang. Curvature mode shape-based damage
detection in composite laminated plates. Composite Structures, 80(3):409–428, 2007.
[39] H. Hu and J. Wang. Damage detection of a woven fabric composite laminate using a
modal strain energy method. Engineering Structures, 31(5):1042–1055, 2009.
[40] A.A. Cury, C.C. Borges, and F.S. Barbosa. A two-step technique for damage assessment
using numerical and experimental vibration data. Structural Health Monitoring,
10(4):417–428, 2010.
[41] S. Banerjee, F. Ricci, E. Monaco, and A. Mal. A wave propagation and vibration-based
approach for damage identification in structural components. Journal of Sound and
Vibration, 322(1-2):167–183, 2009.
[42] R. Loendersloot, T.H. Ooijevaar, L.L. Warnet, A. de Boer, and R. Akkerman. Vibration
based structural health monitoring of a composite plate with stiffeners. In Proceedings
of International Conference on Noise and Vibration Engineering (ISMA2010), pages 909–924,
2010.
[43] R. Loendersloot, T.H. Ooijevaar, A. de Boer, and R. Akkerman. Development of a
damage quantification model for composite skin-stiffener structures. In C. Boller and
H. Janocha, editors, New Trends in Smart Technologies, pages 99–108. Fraunhofer Verlag,
2013.
[44] A. Offringa, J. List, J. Teunissen, and H. Wiersma. Fiber reinforced thermoplastic butt
joint development. In Proceedings of the International SAMPE Conference, pages 1–16,
2008.
[45] A. Offringa, J.W. van Ingen, and A. Buitenhuis. Butt-joined, thermoplastic
stiffened-skin concept development. SAMPE Journal, 48(2):7–15, 2012.
[46] J. Baaran. Visual inspection of composite structures. Technical report, Institute of
Composite Structures and Adaptive Systems, DLR Braunschweig, 2009.
[47] T.A. Duffey, S.W. Doebling, C.R. Farrar, W.E. Baker, and W.H. Rhee. Vibration-based
damage identification in structures exhibiting axial and torsional response. Journal of
Vibration and Acoustics, 123(1):84, 2001.
[48] J. Stewart. Calculus: early transcendentals. 5th edition. Brooks/Cole - Thomson
84 Chapter 4. Damage identification in skin-stiffener structures based on curvatures
Learning, 2003.
[49] R.E. Shiffler. Maximum Z scores and outliers. The American Statistician, 42(1):79–80,
1988.
[50] A.P. Adewuyi, Z. Wu, and N.H.M. Kammrujaman Serker. Assessment of
vibration-based damage identification methods using displacement and distributed
strain measurements. Structural Health Monitoring, 8(6):443–461, 2009.
[51] B. Samali, J. Li, F.C. Choi, and K. Crews. Application of the damage index method for
plate-like structures to timber bridges. Structural Control and Health Monitoring, pages
1–23, 2009.
[52] E.S. Sazonov and P. Klinkhachorn. Optimal spatial sampling interval for damage
detection by curvature or strain energy mode shapes. Journal of Sound and Vibration,
285(4-5):783–801, 2005.
[53] B.J. Schwarz and M.H. Richardson. Experimental modal analysis. In CSI Reliability
Week, Orlando, FL, pages 1–12, 1999.
[54] M.H. Richardson and B.J. Schwarz. Modal parameter estimation from operating data.
Sound and Vibration, 37(1):28–39, 2003.
[55] M.H. Richardson and D.L. Formenti. Global curve fitting of frequency response
measurements using the rational fraction polynomial method. In 3rd International
Modal Analysis Conference, pages 1–8, 1985.
[56] R.J. Allemang. The modal assurance criterion – Twenty years of use and abuse. Journal
of Sound and Vibration, 1:14–21, 2003.
[57] C.S. Hamey, W. Lestari, P. Qiao, and G. Song. Experimental damage identification of
carbon/epoxy composite beams using curvature mode shapes. Structural Health
Monitoring, 3(4):333–353, 2004.
Chapter 5
Abstract
One of the key issues in composite structures for aircraft applications
is the early identification of damage. Often, service induced damage
does not involve visible plastic deformation, but internal matrix
related damage. A wide range of technologies, comprising global
vibration and local wave propagation methods, can be employed
for health monitoring purposes. Traditional modal analysis based
methods are linear methods. The effectiveness of these methods
is sometimes limited since they rely on a stationary and linear
description of the system. The nonlinear interaction between a
low frequency wave field and a local impact induced damage in
a composite skin-stiffener structure is experimentally demonstrated
in this work. The different mechanisms linked to the distorted
waveforms are separated with the help of phase portraits. The
harmonic waveform distortions are concentrated at the damaged
region and increased for higher excitation amplitudes. It is shown
that linear damage identification methods are feasible for low
excitation amplitudes, but that the presence of nonlinear dynamic
effects cannot remain silent for higher amplitudes. Analyzing the
damage induced nonlinear effects can provide useful information
about the current state of the structure.
1 Reproduced from: T.H. Ooijevaar, M.D. Rogge, R. Loendersloot, L.L. Warnet, R. Akkerman,
T. Tinga, Nonlinear dynamic behavior of an impact damaged composite skin-stiffener structure.
In preparation for: Composite Science and Technology, 2014.
85
86 Chapter 5. Nonlinear dynamic behavior of an impact damaged skin-stiffener structure
5.1 Introduction
Composite skin-stiffener structures are a typical aerospace structural component
used to increase the bending stiffness of a component without a severe weight
penalty [1]. These skin-stiffener structures are considered to be safety-critical
components. Small defects at the skin-stiffener connection, caused by for example
impact, can significantly affect the performance of such a component. Often, the
damage does not involve visible plastic deformations, but is barely visible and
internal matrix related. This makes skin-stiffener connections a suitable candidate
for health monitoring.
A wide range of technologies, comprising global vibration and local wave
propagation methods, can be employed to monitor the integrity of structural
components [2–6]. The selection of the best technology is often a matter of
compromise. Traditional low frequency modal analysis based damage identification
methods are linear methods [3, 7]. Damage sensitive features are extracted by
fitting a mathematical model to the responses measured. Despite various successful
applications [8, 9], the effectiveness of these methods is sometimes limited since
they rely on a stationary and linear description of the system. A more realistic
damage description requires a physical understanding of the effect of the damage
on the dynamic behavior of the structure. The interaction between damage and an
acoustic wave field can yield dynamic phenomena that exhibit complicated material
(e.g. nonlinear elasticity) and geometrical (e.g. contact and friction) nonlinear
behavior [10, 11]. The symptoms related to these nonlinear phenomena can be
utilized for nonlinear system characterization and damage identification purposes.
The most commonly used nonlinear feature is probably the generation of higher
harmonics in the structural response. Selective imaging of these higher harmonics
was used by Krohn et al. [12] and Solodov et al. [13] to localize and characterize
small defects. Alternative nonlinear features are the generated sub-harmonics,
DC offsets, wave form distortions, frequency shifts, coherence functions, signal
modulations, etc. [3, 14]. Various researchers [15–17] state the potentially higher
sensitivity involved with these characteristics over the classical linear methods.
However, difficulties concern the experimental evidence and understanding of the
physical mechanisms related to the nonlinearities combined with the diversity of the
nonlinear mechanisms originating at different scales of the composite component
(micro-, meso- and macroscopic) [10]. Similar nonlinear effects can be a result
of different mechanisms, hence hampering the separation of the mechanisms.
Nonlinear elasticity and contact nonlinearities can, for example, both yield higher
harmonics in a dynamic response [14].
Researchers try to obtain a better understanding of the mechanisms involved with
the help of modeling. Della and Shu [18] reviewed models for the vibration of
delaminated structures, while Friswell and Penny [19] did the same for crack-like
damage. The complexity of the nonlinear effects introduced when non-bonded
5.2. Composite skin-stiffener structure 87
Bottom view
100 mm
25 25 25
20
Piezoelectric
diaphragm 2
1 3
2
32 Impact
40 mm
360 mm
y
x Impact
z Node line X1
2.2
location
Piezoelectric Shaker
Point (x,y) =
180 mm
diaphragm 1 connection
(25,120) mm
y
x
20
Node line Y1 Y2 Y3
Figure 5.1 Three dimensional and bottom view of the composite skin-stiffener structure with a butt-joint
stiffener. The dimensions, the measurement points (dots), the position of the shaker and the impact location
are indicated.
The damage scenario analyzed in this study comprises an impact damage located
at the connection between skin and stiffener. This connection is considered as a
safety-critical area, because damage at this connection can have major implications
for the integrity of the entire component. Damage was introduced by utilizing
a falling weight impact device and applying a repeated impact up to 15 J. The
ultrasonic C-scan in Figure 5.2 reveals a relatively complex, but typical fiber
reinforced material related damage scenario. The damage consists of delamination
at the interface between the skin and stiffener accompanied by a limited amount
of damage between the first and second ply of the skin. Local delaminations were
also introduced underneath one of the supports that was used during the impact
testing.
The experimental process used to investigate the interaction between the dynamic
wave field and the local damage consists of two steps. The global dynamic behavior
of the pristine and damaged skin-stiffener structure is characterized in the first step.
These dynamic characteristics are used as an input for the second step to study the
nonlinear dynamic behavior of the damaged structure in detail. The current section
5.3. Experimental work 89
Supports (4×)
y
during impact
x
Stiffener 10 mm
and filler
79 mm
Impact Cracked
75 mm
location Skin
filler
Point (x,y) = delamination
(25,120) mm
85 mm
Figure 5.2 Ultrasonic C-scan of the impact damaged skin-stiffener structure showing a complex
combination of failure mechanisms near the skin-stiffener interface.
addresses both steps, but starts with an introduction of the experimental set-up
used in this work.
The set-up and data acquisition systems used for all experiments are schematically
illustrated in Figure 5.3. Responses were measured before and after the impact
damage was introduced in the composite skin-stiffener structure. The structure
was freely suspended by an elastic wire, glued to the butt-joint, in order to isolate
the structure from environmental vibrations. An electromechanical shaker was
connected by a stringer and a force transducer to a corner of the structure. This
position is selected in order to excite a large number of vibrational modes. A
laser vibrometer, mounted on an x/y traverse system, measured the velocities at
different points at the skin of the structure. The measurement points are ordered
according to three lines in y-direction (‘Y1’, ‘Y2’, ‘Y3’) and one line in x-direction
(‘X1’), as shown in Figure 5.1. The location of line ‘X1’ matches with the center of
the damaged area. A data acquisition system, controlled by a Labview application,
was utilized to simultaneously send an excitation signal to the shaker and acquire
the force and velocity responses of the structure.
In the first set of experiments, the global dynamic behavior of the structure is
determined in terms of natural frequencies and the frequency dependent deflection
patterns, referred to as operational deflection shapes (ODS). An excitation signal
90 Chapter 5. Nonlinear dynamic behavior of an impact damaged skin-stiffener structure
11.
x 10
4 12.
Acceleration [mm/s ]
2
5
−5
3.
10. 9.
5. 8.
7.
F(t)
v(t)
6.
y
2. z
1.
composed of a linear sweep between 150 and 3050 Hz was sent to the shaker. The
force response was measured at the fixed excitation point. The velocities were
acquired at 51 (3×17) points, matching the node lines in y-direction according
to Figure 5.1. All signals were acquired for 2.62 seconds (217 samples) at a rate
of 50 kHz. A measurement at each point was repeated 10 times. The acquired
time responses were subsequently windowed, Fourier transformed and averaged to
obtain the power spectral densities. The auto-power spectral density S Fi Fi (ω ) and
cross-power spectral density S Fi v j (ω ) are used to calculate the mobility frequency
response functions HFi v j (ω ) between the fixed point of excitation i and the roving
measurement points j according to [22]:
S Fi v j (ω )
HFi v j (ω ) = (5.1)
S Fi Fi (ω )
Figure 5.4 shows the magnitude of the total set of frequency response functions
(FRFs) of the pristine structure. Sharp peaks correspond to the natural frequencies
of the structure.
The operational deflection shapes of the structure are extracted at the natural
5.3. Experimental work 91
2
10
|H | [(mm/s)/N] 0
Fv 10
−2
10
Figure 5.4 The magnitude of the frequency response functions for all 51 (3 × 17) measurement points of
the pristine structure.
frequencies with the help of peak picking. At these frequencies, the complex
operational deflection shapes become predominantly real valued and are a close
approximation of the associated mode shapes [22]. Bending and torsion dominant
operational deflection shapes are distinguished by considering the dominant real
part of the operational deflection shapes, as shown in Figure 5.5. Table 5.1 provides
an overview of the natural frequencies of the first six bending and torsion shapes
for the pristine and damaged structure.
The present work concentrates on bending dominant operational deflection shapes
in the yz-plane. Earlier research [8] showed that skin-stiffener damage hardly affects
20 4
0.3 0.3 0.3 0.3
30
2 15
0.25 20 0.25 0.25 0.25
10 2
y−coordinate [m]
y−coordinate [m]
y−coordinate [m]
y−coordinate [m]
10
0.2 0.2 0 0.2 5 0.2
0
0.15 0.15 0.15 0 0.15 0
−10
−2 −5
0.1 −20 0.1 0.1 0.1
−10 −2
−30
0.05 0.05 −4 0.05 0.05
−40 −15
0 0 0 0 −4
0 0.05 0 0.05 0 0.05 0 0.05
x−coordinate [m] x−coordinate [m] x−coordinate [m] x−coordinate [m]
(a) ℜ( HFv ) in (b) ℑ( HFv ) in (c) ℜ( HFv ) in (d) ℑ( HFv ) in
[(mm/s)/N] at [(mm/s)/N] at [(mm/s)/N] at [(mm/s)/N] at
(4) (4) (4) (4)
f T = 662 Hz. f T = 662 Hz. f B = 1456 Hz. f B = 1456 Hz.
Figure 5.5 Real (a,c) and imaginary (b,d) part of the operational deflection shapes of the pristine structure
at the natural frequencies of the 4th torsion mode (a,b) and the 4th bending mode (c,d).
92 Chapter 5. Nonlinear dynamic behavior of an impact damaged skin-stiffener structure
Table 5.1 Natural frequencies of the 1st till the 6th bending and torsion modes of the pristine and damaged
(indicated by a tilde) structure.
torsional deformation shapes around the y-axis in these kinds of structures due to
the small distance between the damage and the rotational axis.
20
Velocity [mm/s]
Velocity [mm/s]
5
10
0 0
−10
−5
−20
0.8 0.801 0.802 0.803 0.804 0.8 0.801 0.802 0.803 0.804
Time [s] Time [s]
(a) Velocity. (a) Velocity.
4 5
x 10 x 10
4
Acceleration [mm/s ]
Acceleration [mm/s ]
2
2
5
2
0 0
−2
−5
−4
0.8 0.801 0.802 0.803 0.804 0.8 0.801 0.802 0.803 0.804
Time [s] Time [s]
(b) Acceleration. (b) Acceleration.
Figure 5.6 Velocity and acceleration responses Figure 5.7 Velocity and acceleration responses
measured at location (x,y) = (0.025,0.120) m of the measured at location (x,y) = (0.025,0.120) m of the
pristine structure for a forced excitation at the 4th damaged structure for a forced excitation at the 4th
(4) (4)
bending mode ( f B = 1456 Hz) and five excitation bending mode ( f˜B = 1455 Hz) and five excitation
amplitude levels. amplitude levels.
The velocity and acceleration responses of the pristine structure show nearly purely
harmonic oscillations for all excitation amplitudes, while the waveforms are clearly
distorted in the damaged case. The amount of harmonic waveform distortion
is most pronounced in the acceleration response and increases for higher shaker
excitation amplitudes.
The velocity and acceleration responses are combined to obtain the phase portraits
in Figure 5.8. Each trajectory represents a different excitation amplitude. The
concentric circles around an equilibrium for the pristine structure indicate that the
motion is periodic, stable and purely harmonic. The damaged structure also shows
periodic and stable behavior, but the pure harmonic motion is distorted by the skin-
stiffener damage. The motion approaches nearly fundamental harmonic behavior
for the lowest excitation amplitude, while the harmonic distortion increases with
increasing excitation amplitude. Higher harmonics are also generated in the case
of the higher excitation amplitudes. The damage causes the motion of the skin at
94 Chapter 5. Nonlinear dynamic behavior of an impact damaged skin-stiffener structure
4 5
x 10 x 10
4
Acceleration [mm/s ]
Acceleration [mm/s ]
2
2
5
2
0 0
−2
−5
−4
−5 0 5 −20 0 20
Velocity [mm/s] Velocity [mm/s]
(a) Pristine structure. (b) Damaged structure.
Figure 5.8 Phase portraits measured at location (x,y) = (0.025,0.120) m of the pristine and damaged
(4) (4)
structure for a forced excitation at the 4th bending mode ( f B = 1456 Hz, f˜B = 1455 Hz) and five excitation
amplitude levels. Each trajectory represents a different excitation amplitude.
the damaged region to behave nonlinearly for the bending mode considered. The
nonlinearities mainly occur at one side of the phase portrait while the rest of the
motion keeps behaving rather linearly with respect to the excitation amplitude.
The skin at the damaged region behaves dynamically nonlinearly for the bending
mode considered. The physical mechanisms related to these nonlinear effects are
still unknown. The phase portrait revealed that the nonlinearities mainly occurred
at one side of the phase portrait, but did not link these nonlinear effects to physical
mechanisms. Therefore, the displacement, velocity and acceleration signals are
analyzed in more detail.
The skin-stiffener damage, although fairly complex, locally separates the skin from
the stiffener. This separation will potentially allow the damage to open under a
forced excitation at a bending dominant operational deflection shape. The relative
motion between skin and stiffener at the damaged area is required to verify the
opening of the delamination. The current set-up did not allow to directly measure
the deformation of the stiffener and to calculate the relative motion between the
skin and the stiffener. Alternatively, the motion of the stiffener at the damaged
region is approximated by analyzing the motion of the skin at multiple spatial
points underneath the skin-stiffener connection, i.e. 34 points at node line ‘Y2’
according to Figure 5.1. The motion of the stiffener at the undamaged interface
regions is equal to the motion of the skin, since both are connected. The high
flexural rigidity of the stiffener controls the skin motion at the undamaged regions.
At the damaged region, the skin and stiffener behave more independently. It is
assumed that the changes in the flexural motion of the stiffener due to the damage
5.3. Experimental work 95
y
x Damaged region
1 2 3 4 5 6 8 10 14 18 22 26 28 29 30 31 32 33 34
Figure 5.9 Interpolated nodes to approximate the motion of the stiffener at the damaged region.
are minor compared to the changes in the motion of the skin. This is justified by the
fact that the stiffener has a significantly higher flexural rigidity (approximately 150
times) than the skin. As a result, the motion of the stiffener at the damaged region
is approximated by spatially interpolating the motion of the skin at the undamaged
regions on both sides of the damaged region. Figure 5.9 schematically shows the
interpolated region within the node set. A piecewise cubic spline interpolation,
based on nodes 1–5 and 26–34, is utilized to estimate the velocity and acceleration
patterns as a function of the position and the time for nodes 6–25.
Figure 5.10 shows the spatial velocity and acceleration distributions of the skin and
(4)
stiffener for the forced excitation at the 4th bending mode ( f˜ = 1455 Hz) and B
the highest excitation amplitude. The three snapshots match with three different
time instances according to the phase portrait presented in Figure 5.11. Despite the
fact that the interpolation only provides an approximation of the stiffener motion,
the snapshots reveal a significant local difference in the velocity and acceleration
amplitudes between the skin and the stiffener. A difference in speed between the
skin and the stiffener causes the delamination to open or close.
These physical mechanisms are linked to sections in the phase portrait with the
help of the associated displacement, velocity and acceleration waveforms for the
skin at the damaged region, as shown in Figure 5.11. A relative estimation of the
skin displacement is obtained by time integrating the velocity response. The skin
at location (x,y) = (0.025,0.120) m showed a 5 µm peak-to-peak displacement for
the highest excitation amplitude considered. A detailed analysis of the waveforms
makes it possible to distinguish three damage related dynamic mechanisms within
each oscillation:
Opening phase | A
The first mechanism, assigned to section A, represents the movement of
the skin towards the laser vibrometer. In this phase, the velocity of
the skin is positive, while the acceleration is getting negative up to the
point where the skin obtains its maximum positive displacement. The
snapshot in Figure 5.10(a) shows that the skin at the damaged region reaches
significantly higher velocities than those of the stiffener. This positive
difference indicates the opening of the delamination at the skin-stiffener
96 Chapter 5. Nonlinear dynamic behavior of an impact damaged skin-stiffener structure
5
x 10
4
Skin Skin
Acceleration [mm/s ]
2
20
Stiffener Stiffener
Velocity [mm/s]
2
10
0 0
−10
−2
−20 t = 0.80008 s t = 0.80008 s
−4
0 0.1 0.2 0.3 0 0.1 0.2 0.3
y−coordinate [m] y−coordinate [m]
(a) Velocity at t = t0 . (b) Acceleration at t = t0 .
5
x 10
4
Skin Acceleration [mm/s ]
2 Skin
20
Stiffener Stiffener
Velocity [mm/s]
2
10
0 0
−10
−2
−20 t = 0.80021 s t = 0.80021 s
−4
0 0.1 0.2 0.3 0 0.1 0.2 0.3
y−coordinate [m] y−coordinate [m]
(c) Velocity at t = t1 . (d) Acceleration at t = t1 .
5
x 10
4
Skin Skin
Acceleration [mm/s ]
2
20
Stiffener Stiffener
Velocity [mm/s]
2
10
0 0
−10
−2
−20 t = 0.80031 s t = 0.80031 s
−4
0 0.1 0.2 0.3 0 0.1 0.2 0.3
y−coordinate [m] y−coordinate [m]
(e) Velocity at t = t2 . (f) Acceleration at t = t2 .
y y
x x
Figure 5.10 Snapshot of the skin and (approximated) stiffener velocity and acceleration distribution at
(4)
node line ‘Y2’ for the forced excitation at f˜B = 1455 Hz and the highest excitation amplitude. The three
time instances match the time stamps presented in the phase portrait in Figure 5.11.
5.3. Experimental work 97
−3
x 10
Displacement [mm] 4 t1 t1
2 t2 t0 t2 t0
Composite
0 B C A B C A
Node line structure
Y2
t3 t3
Damaged area
0.7995 0.8 0.8005 0.801
Shaker Laser vibrometer
t0 t0
Velocity [mm/s]
20 v(t)
y
10 x
0
t1 t1 Fz z
t3 t3
−10 Force
transducer
−20
t2 t2
0.7995 0.8 0.8005 0.801
5 5
x 10 x 10
Acceleration [mm/s ]
Acceleration [mm/s ]
4 4
2
C
2 2
t3 t3 t3
t2 t0 t2 t0 t0
0 0 t2
B A
−2 −2
t1 t1 t1
−4 −4
0.7995 0.8 0.8005 0.801 −20 0 20
Time [s] Velocity [mm/s]
Figure 5.11 The displacement, velocity and acceleration responses measured at location
(x,y) = (0.025,0.120) m of the damaged structure for a single tone harmonic excitation at the 4th
(4)
bending mode ( f˜B = 1455 Hz) and the highest excitation amplitude. The capitals refer to the opening (A),
closing (B) and contact (C) phase of the skin stiffener failure. The time stamps t0 – t3 correspond to the
zero-crossings of the velocity and acceleration responses.
The opening, closing and contact mechanisms are repeated periodically. Analyzing
the different phases of this periodic motion revealed that the nonlinear dynamic
behavior of the skin is mainly introduced when the skin and stiffener are interacting
with each other. The motion of the skin at the damaged region is barely distorted
during the opening and closing phase.
Phase portrait
Figure 5.12 shows five phase portraits, each containing five trajectories, at different
positions along the damaged skin-stiffener interface for a single tone excitation at
the 4th bending mode, together with a snapshot of the velocities at t = t2 . The most
severe asymmetric distortions of the phase portraits are obtained for the skin at the
damaged region as illustrated by Figure 5.12(c). The motion of the skin is barely
affected at the undamaged regions and only shows some small distortions for the
5.3. Experimental work 99
20
Velocity [mm/s]
10
0
−10
−20 t = 0.80031 s
0 0.1 0.2 0.3
y−coordinate [m]
Acceleration [mm/s ]
2
5 5 5 5 5
x 10 (a) x 10 (b) x 10 (c) x 10 (d) x 10 (e)
4 4 4 4 4
2 2 2 2 2
0 0 0 0 0
−2 −2 −2 −2 −2
−4 −4 −4 −4 −4
−20 0 20 −20 0 20 −20 0 20 −20 0 20 −20 0 20
Velocity [mm/s] Velocity [mm/s] Velocity [mm/s] Velocity [mm/s] Velocity [mm/s]
Figure 5.12 Phase portraits measured at different positions at node line ‘Y2’ of the damaged structure for a
(4)
forced excitation at f˜B = 1455 Hz and five excitation amplitude levels. Each trajectory represents a different
excitation amplitude.
Figures 5.13 and 5.14 show the velocity and acceleration time responses at node
line ‘Y2’ of the pristine and damage structure when excited in the 4th bending
mode. Note that the amplitudes are indicated by different color scales. The velocity
and acceleration amplitudes gradually vary around the nodal point at y = 0.23 m.
All points at the skin-stiffener connection move exactly in phase or out of phase
by 180 degrees for the pristine case. The responses change considerably in the
damaged case. The velocity distribution in Figure 5.14(a) shows a local increase in
the negative and positive amplitudes at the damaged region, while the acceleration
only shows a locally increased negative amplitude. The positive amplitude remains
in the same order of magnitude for all points at node line ‘Y2’. The interaction
between skin and stiffener restricts the acceleration of the skin to locally increase to
higher positive values.
The motion of the skin at the damaged region also shows a local phase shift with
100 Chapter 5. Nonlinear dynamic behavior of an impact damaged skin-stiffener structure
0.802 0.802
10 20
0.8015 5 0.8015
10
Time [s]
Time [s]
0.801 0 0.801 0
−5 −10
0.8005 0.8005
−10 −20
0.8 0.8
0 0.1 0.2 0.3 0 0.1 0.2 0.3
y−coordinate [m] y−coordinate [m]
(a) Velocity [mm/s]. (a) Velocity [mm/s].
5 ‘Closing’ ‘Distortion’ 5
x 10 x 10
0.802 0.802
1
0.8015 0.5 0.8015
0
Time [s]
Time [s]
0.801 0 0.801
−2
0.8005 −0.5 0.8005
−1
0.8 0.8 −4
0 0.1 0.2 0.3 0 0.1 0.2 0.3
y−coordinate [m] ‘Opening’ y−coordinate [m]
Figure 5.13 Velocity and acceleration responses Figure 5.14 Velocity and acceleration responses
measured at node line ‘Y2’ of the pristine structure measured at node line ‘Y2’ of the damaged structure
(4) (4)
for a forced excitation at f B = 1456 Hz and the for a forced excitation at f˜B = 1455 Hz and the
highest excitation amplitude. highest excitation amplitude.
respect to the undamaged regions of node line ‘Y2’. This phase shift is a result of
two effects. Firstly, the skin and the stiffener behave relatively independently at the
damaged region underneath the stiffener, while the bending motion of the skin at
the undamaged regions is controlled by the high flexural rigidity of the stiffener.
Secondly, the damage is located at a region where the motion of the skin at both
sides of the stiffener shows an antinode. These high amplitudes on both sides of the
stiffener are clearly illustrated by the operational deflection shape in Figure 5.5(c).
They drive the motion of the skin at the damaged region underneath the stiffener.
5.3. Experimental work 101
Skin−stiffener interface
0.802
200
0.8015 100
Time [s]
0.801 0
−100
0.8005
−200
0.8
C-scan of the structure First-ply failure
Measurement
points at
node line X1 0 0.025 0.05
x−coordinate [m]
Figure 5.15 Velocity response [mm/s] measured at node line ‘X1’ of the damaged structure for a forced
(4)
excitation at f˜B = 1455 Hz. The skin at the damaged skin-stiffener connection moves nearly in phase with
the motion of the skin on both sides of the stiffener.
As a consequence, the skin-stiffener damage opens and closes nearly in phase with
the motion of the skin on both sides of the stiffener, as demonstrated by the velocity
time response at node line ‘X1’ in Figure 5.15.
5 6
x 10 Wrinkles x 10
Acceleration [mm/s ]
2
Acceleration [mm/s ]
1
2
1
0.5
0 0
−0.5 Wrinkles
−1
−1
−5 0 5 −50 0 50
Velocity [mm/s] Velocity [mm/s]
(a) Pristine. (b) Damaged.
Figure 5.16 Phase portraits measured at location (x,y) = (0.025,0.120) m of the pristine and damaged
(6) (6)
structure for a forced excitation at the 6th bending mode ( f B = 2328 Hz, f˜B = 2340 Hz) and five amplitude
levels. Each trajectory represents a different excitation amplitude.
case of the 6th bending mode is better described by a smoother ‘rolling’ motion.
This ‘rolling’ motion is demonstrated by the acceleration distribution of the skin in
Figure 5.18(b). The diagonal lines in the zoomed inset indicate traveling acceleration
waves from left to right at the delaminated region during the closing phase. This
propagating effect matches the wrinkles around the zero crossing of the acceleration
in the phase portrait.
The smooth ‘rolling’ motion is attributed to the position of the damage with respect
to the nodes and antinodes in the motion of the skin and the stiffener. Figure 5.19
shows that the damaged region is centered with respect to the nodal points on both
sides (l1 ≈ l2 ) in case of the 4th bending mode. In contrast, the 6th bending mode
shows an off-centered position (l1 < l2 ) of the damaged region, causing the rolling
motion in the acceleration of the skin.
5.4 Discussion
The present section provides a discussion of two aspects. Firstly, the underlying
physical phenomena associated with the nonlinear dynamic behavior are discussed.
Secondly, the higher harmonic components are addressed in more detail.
0.801 0.801
100
10
50
5
Time [s]
Time [s]
0.8005 0 0.8005 0
−5 −50
−10
−100
0.8 0.8
0 0.1 0.2 0.3 0 0.1 0.2 0.3
y−coordinate [m] y−coordinate [m]
(a) Velocity [mm/s]. (a) Velocity [mm/s].
5 6
x 10 x 10
0.801 2 0.801
1
1
Time [s]
Time [s]
0
0.8005 0 0.8005
−1 −1
0.8 −2 0.8 −2
0 0.1 0.2 0.3 0 0.1 0.2 0.3
y−coordinate [m] y−coordinate [m]
Figure 5.17 Velocity and acceleration responses Figure 5.18 Velocity and acceleration responses
measured at node line ‘Y2’ of the pristine structure measured at node line ‘Y2’ of the damaged structure
(6) (6)
for a forced excitation at f B = 2328 Hz and the for a forced excitation at f˜B = 2340 Hz and the
highest excitation amplitude. highest excitation amplitude.
100
Node Node Node Node
20
Velocity [mm/s]
l1 l2 l1 l2
Velocity [mm/s]
50
10
0 0
−10
−50
−20 t = 0.80031 s t = 0.80045 s
−100
C-scan of the structure C-scan of the structure
Figure 5.19 Snapshots at time instance t = t2 of the velocity distributions at node line ‘Y2’ of the damaged
structure for the forced excitation at the 4th and 6th bending mode and five amplitude levels. The position of
the damaged region with respect to the associated nodal points is indicated by l 1 and l 2 .
100 100
f f
2f 2f
Velocity [mm/s]
Velocity [mm/s]
4f
3f 3f 5f
1 1
4f
5f
0.01 0.01
0.0001 0.0001
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Frequency [Hz] 4 Frequency [Hz] 4
x 10 x 10
(4) (6)
(a) f˜B = 1455 Hz. (b) f˜B = 2340 Hz.
Figure 5.20 Fourier transforms, indicating multiple higher harmonics, of the responses measured at
location (x,y) = (0.025,0.120) m of the damaged structure for a single tone harmonic excitation at the (a) 4th
and (b) 6th bending mode and the highest excitation amplitude.
order to the original time response. The associated phase portraits are shown in
Figure 5.21. The differences between the original and the filtered phase portraits
give an impression of the information that is discarded when the higher harmonics
are not included in the analysis. The largest differences are obtained for the
acceleration responses. This is caused by the fact that the majority of the distortions
is introduced around the zero-crossing of the velocity response.
An FRF extracted for the damaged structure is dominated by fundamental
harmonic components, but also contains higher harmonic information for a range of
excitation frequencies. A classical FRF based operational deflection shape analysis,
as was shown in Section 5.3.2, only considers the amplitude and phase at one
selected frequency. The higher harmonic information associated with this frequency
5 6
x 10 x 10
4
Acceleration [mm/s ]
Acceleration [mm/s ]
2
2 1
Original
0 0
Original f, 2f, 3f, 4f
f, 2f, 3f f, 2f, 3f
−2 −1
f, 2f f, 2f
−4 f f
−20 0 20 −50 0 50
Velocity [mm/s] Velocity [mm/s]
(4) (6)
(a) f˜B = 1455 Hz. (b) f˜B = 2340 Hz.
Figure 5.21 Phase portraits of the original time response and several low pass filtered time responses
comprising a selection of harmonics. The responses are measured at location (x,y) = (0.025,0.120) m for a
single tone harmonic excitation at the (a) 4th and (b) 6th bending mode and the highest excitation amplitude.
106 Chapter 5. Nonlinear dynamic behavior of an impact damaged skin-stiffener structure
is therefore not included in the operational deflection shape. Modal analysis based
damage detection approaches are subject to similar issues. They rely on a stationary
and linear representation of the system as was shown in Chapter 2. Nonlinear
effects in the structural response are not identified, while they can provide very
useful information about the health of the structure as shown in the present study.
amplitudes, but the presence of nonlinear effects cannot remain silent for higher
amplitudes.
Additional research is required on the separation of the underlying physical
nonlinear phenomena. A model representation of the skin-stiffener damage can
help to distinguish the most dominant physical phenomena and to provide a
quantification of the experimental observations. Future research will also focus
on methods to embed the nonlinear dynamics in a structural health monitoring
method.
Acknowledgments
This material is based on work supported by National Aeronautics and Space
Administration, Langley Research Center under Research Cooperative Agreement
No. NNL09AA00A awarded to the National Institute of Aerospace.
The authors kindly acknowledge the support of Fokker Aerostructures B.V.,
Hoogeveen, The Netherlands, for manufacturing the composite structure used in
this research. This work is funded by the European research project Clean Sky,
Eco-Design ITD (grant agreement number CSJU-GAM-ED-2008-001).
References
[1] A. Offringa, J.W. van Ingen, and A. Buitenhuis. Butt-joined, thermoplastic
stiffened-skin concept development. SAMPE Journal, 48(2):7–15, 2012.
[2] C. Boller, F.-K. Chang, and Y. Fujino. Encyclopedia of structural health monitoring. John
Wiley & Sons, Ltd, Chichester, UK, 2009.
[3] K. Worden, C.R. Farrar, J. Haywood, and M. Todd. A review of nonlinear dynamics
applications to structural health monitoring. Structural Control and Health Monitoring,
15:540–567, 2008.
[4] D. Montalvão, N.M.M. Maia, and A.M.R. Ribeiro. A review of vibration-based
structural health monitoring with special emphasis on composite materials. Shock and
Vibration Digest, 38(4):295–326, 2006.
[5] K. Diamanti and C. Soutis. Structural health monitoring techniques for aircraft
composite structures. Progress in Aerospace Sciences, 46(8):342–352, 2010.
[6] A. Raghavan and C.E.S. Cesnik. Review of guided-wave structural health monitoring.
The Shock and Vibration Digest, 39(2):91–114, 2007.
[7] M.H. Richardson and B.J. Schwarz. Modal parameter estimation from operating data.
Sound and Vibration, 37(1):28–39, 2003.
[8] T.H. Ooijevaar, R. Loendersloot, L.L. Warnet, A. de Boer, and R. Akkerman. Vibration
based structural health monitoring of a composite T-beam. Composite Structures,
92(9):2007–2015, 2010.
[9] M. Dilena and A. Morassi. Dynamic testing of a damaged bridge. Mechanical Systems
and Signal Processing, pages 1–23, 2011.
108 Chapter 5. Nonlinear dynamic behavior of an impact damaged skin-stiffener structure
[10] A. Klepka, W.J. Staszewski, R.B. Jenal, M. Szwedo, T. Uhl, and J. Iwaniec. Nonlinear
acoustics for fatigue crack detection - Experimental investigations of vibro-acoustic
wave modulations. Structural Health Monitoring, 11(2):197–211, 2011.
[11] I. Solodov, N. Krohn, and G. Busse. CAN: an example of nonclassical acoustic
nonlinearity in solids. Ultrasonics, 40:621–625, 2002.
[12] N. Krohn, R. Stoessel, and G. Busse. Acoustic non-linearity for defect selective
imaging. Ultrasonics, 40(1-8):633–7, 2002.
[13] I. Solodov, D. Döring, and G. Busse. New opportunities for NDT using non-linear
interaction of elastic waves with defects. Journal of Mechanical Engineering,
57(03):169–182, 2011.
[14] K.-Y. Jhang. Nonlinear ultrasonic techniques for nondestructive assessment of micro
damage in material: a review. Precision Engineering and Manufacturing, 10(1):123–135,
2009.
[15] P.B. Nagy. Fatigue damage assessment by nonlinear ultrasonic materials
characterization. Ultrasonics, 36(1-5):375–381, 1998.
[16] K.E.-A. Van Den Abeele, P.A. Johnson, and A. Sutin. Nonlinear elastic wave
spectroscopy (NEWS) techniques to discern material damage, part I: nonlinear wave
modulation spectroscopy (NWMS). Research in Nondestructive Evaluation, 12:17–30,
2000.
[17] K.E.-A. Van Den Abeele, J. Carmeliet, J.A. Ten Cate, and P.A. Johnson. Nonlinear
elastic wave spectroscopy (NEWS) techniques to discern material damage, part II:
single-mode nonlinear resonance acoustic spectroscopy. Research in Nondestructive
Evaluation, 12:31–42, 2000.
[18] C.N. Della and D. Shu. Vibration of delaminated composite laminates: a review.
Applied Mechanics Reviews, 60(1-6):1–20, 2007.
[19] M.I. Friswell and J.E.T. Penny. Crack modeling for structural health monitoring.
Structural Health Monitoring, 1(2):139–148, 2002.
[20] I. Müller. Clapping in delaminated sandwich-beams due to forced oscillations.
Computational Mechanics, 39(2):113–126, 2005.
[21] I. Müller, A. Konyukhov, P. Vielsack, and K. Schweizerhof. Parameter estimation for
finite element analyses of stationary oscillations of a vibro-impacting system.
Engineering Structures, 27(2):191–201, 2005.
[22] M.H. Richardson. Is it a mode shape, or an operating deflection shape? Sound and
Vibration, pages 1–11, 1997.
[23] C.J. Pye and R.D. Adams. Detection of damage in fibre reinforced plastics using
thermal fields generated during resonant vibration. NDT International, 14(3):111–118,
1981.
Chapter 6
Abstract
Vibro-acoustic modulation based damage identification relies on
the modulation of a high-frequency carrier signal by an intenser
low-frequency vibration signal due to damage induced structural
nonlinearities. A time domain analysis of the vibro-acoustic
modulation phenomena was presented at multiple spatial locations
in an impact damaged composite skin-stiffener structure. The
instantaneous amplitude and frequency of the carrier velocity
response were extracted to analyze the intermodulations between
the two excitation signals. Increased amplitude modulations at the
damaged region revealed the presence, location and length of the
skin-stiffener damage. The damage hardly modulated the frequency
of the carrier response. This difference in behavior was attributed
to the nonlinear skin-stiffener interaction introduced by the periodic
opening and closing of the damage, according to earlier research
by authors on the same structure. A parametric study showed
that the amplitude and phase of the amplitude modulation are
dependent on the selected carrier excitation frequency, and hence
the high frequency wave field that is introduced. The present work
demonstrates the potential, but also the complexity of the vibro-
acoustic modulation based damage identification approach.
1 Reproduced from: T.H. Ooijevaar, M.D. Rogge, R. Loendersloot, L.L. Warnet, R. Akkerman,
T. Tinga, Vibro-acoustic modulation based damage identification in a composite skin-stiffener
structure. In preparation for: Structural Health Monitoring, 2014.
109
110 Chapter 6. Vibro-acoustic modulation based damage identification
6.1 Introduction
(a)
Force [N]
Pump and carrier 0
excitation signals. f = 2 kHz
p
f = 50 kHz
c
−5
1 1.001 1.002
Time [s]
Linear Nonlinear
system Fp(t) + Fc(t) system
Linear response vl(t) vnl(t) Nonlinear response
Velocity v [mm/s]
Velocity v [mm/s]
4
2 2
(b) (c )
Superposition 0 0 Modulated
of responses. −2 response.
−2
−4
1 1.001 1.002 1 1.001 1.002
Time [s] Time [s]
Velocity |V| [mm/s]
3 f 3 f
p p (e)
2 2 Fundamental
(d) harmonics fp and fc,
Fundamental fc 2fp fc
1 1 higher harmonics
harmonics fc-fp fc+fp
3fp and sidebands.
fp and fc.
0 0
0 2 4 6 8 0 2 4 6 8
Frequency [Hz] 4 Frequency [Hz] 4
x 10 x 10
1
[mm/s]
1
(g)
(f) Bandpass Bandpass Modulated
bp
response. response.
−1
−1
1 1.001 1.002 1 1.001 1.002
Time [s] Time [s]
Figure 6.1 A schematic illustration of the vibro-acoustic modulation concept. Two harmonic excitation
signals (a) are simultaneously applied to the system. A linear system response (b, d, f) will be the superposition
of its response to each of the signals individually. The high-frequency system response (c) is modulated (e, g)
by the low-frequency pump excitation in case nonlinearities are present.
presented in Figure 6.1(d) will therefore only exhibit the two fundamental frequency
components. In the presence of nonlinearities, the system response does not
consist anymore of the linear superposition of the fundamental responses. The
two signals interact in such a way that the carrier signal is modulated by the
pump signal in amplitude, frequency, or both amplitude and frequency (see
Figure 6.1(c) ). The Fourier spectrum of the response signal reveals additional
components, higher harmonics (n f p and n f c , with n = 1, 2, 3, . . .) and sidebands
( f c + n f p and f c − n f p , with n = 1, 2, 3, . . .) around the high-frequency component,
as shown in Figure 6.1(e).
112 Chapter 6. Vibro-acoustic modulation based damage identification
Despite the fact that many studies showed that the amount of modulation is
correlated to the severity of the damage [16], the underlying physical phenomena
of the modulations are still not well understood. The main difficulty concerns the
diversity of the nonlinear phenomena originating at several material length scales
(micro-, meso- and macroscopic) combined with the fact that similar nonlinear
effects in an acoustic response can be a result of different physical phenomena.
This consequently hampers the separation of these underlying phenomena. Initially,
researchers claimed that the modulations are created by changes in the contact area
at the damage interface caused by the pump signal [17]. Studies by Zaitsev et al. [15]
and Klepka et al. [14], however, revealed that modulation can occur even when the
cracks are not being opened fully by the pump signal. They concluded that energy
dissipation based phenomena at the damaged area constitute the major mechanism
behind the modulations.
Regardless of the exact physical phenomena that cause the nonlinear modulations,
it has been demonstrated that a vibro-acoustic modulation based approach can be
employed to detect defects in a wide variety of structures ranging from aluminum
beams and plates [18, 19], bolt connections [20], to geometrically more complex
structures such as metallic automotive as well as aircraft subcomponents [10, 21].
The amount of work focused on composite materials is, however, rather limited [16,
22].
The majority of current vibro-acoustic modulation based investigations rely on
Fourier transformed responses obtained at one or a few measurement points [16,
22–24]. The sideband amplitudes of the carrier frequency are monitored to
detect and track the progression of damage. Limitations of this approach are,
firstly, the fact that the sidebands are represented by sharp peaks, while the
frequency resolution is often limited. Consequently, it is rather difficult to
obtain an accurate estimation of the amplitude of the sidebands. Secondly, the
sideband amplitudes can be the combined result of two phenomena, amplitude
and frequency modulations [25]. Analyzing only the sideband amplitudes is
insufficient to separate these effects and hence complicates the understanding of
the modulation phenomena. Alternative methods are required to overcome these
difficulties. Nearly all studies regarding vibro-acoustic modulation are limited to
one or a few measurement points. The spatial dependency of the modulations
has hardly been investigated. Consequently, these studies are mainly focused on
detection and quantification of defects, rather than on localizing them.
The current section provides a more thorough description of the wave modulation
phenomena. First, a basic theoretical description of the vibro-acoustic modulation
phenomena is presented. The development of the sideband components is
demonstrated based on a nonlinear single degree of freedom system. This is
followed by a description of the signal decomposition approach that is used to
extract the instantaneous characteristics of the modulated carrier response. The
last part essentially combines the two previous parts. A single degree of freedom
system is numerically solved for four different types of nonlinearities. The signal
decomposition approach is used to qualitatively analyze the system responses.
where q(t) is the displacement, ω0 the natural frequency, f (q, q̇) is a nonlinear
function. This nonlinear function is controlled by the parameter ǫ. For a weakly
nonlinear system, the parameter ǫ is considered relatively small such that the
solution of this system can be approximated by utilizing a perturbation technique
based on a power series [27]:
which can be solved recursively. The steady state solution of Equation (6.4) is:
in which:
Fp Fc
Gp = and Gc = .
ω02 − ωp2 ω02 − ωc2
Following from the nature of the excitation, the response q1 (t) consists of a linear
combination of harmonic components with frequencies equal to: 2ωp , 2ωc , ωc − ωp ,
ωc + ωp and a constant term. The fundamental harmonic frequency components
6.2. Theoretical description 115
ωp and ωc are added in the case of a first-order approximation of the solution q(t)
according to Equation (6.2). The carrier component ωc and its associated sidebands
(i.e. ωc − ωp , ωc + ωp ) are utilized for damage identification purposes. The system
response is therefore analyzed in a narrow frequency band around the carrier
frequency ωc . Consequently, the first-order steady state solution of q(t) reduces
to the narrow band response qbp (t) given by:
in which:
−ǫGp Gc −ǫGp Gc
Ac = Gc , Asb1 = and Asb2 = .
ω02 (ωc
− ωp ) 2 ω02 (ωc
+ ωp ) 2
The time domain analysis utilized in the present work is schematically illustrated
in Figure 6.2. This approach consists of two steps. In the first step, a zero-phase
bandpass FIR (i.e. finite impulse response) filter in a narrow frequency band
around the carrier frequency is applied to separate the carrier response and its
potential sidebands from the rest of the system response. Subsequently, the Hilbert
transform is applied to this filtered response vbp (t) to extract the instantaneous
amplitude Ainst (t) (i.e. the signal envelope) and frequency f inst (t). A description
of the Hilbert transform is presented in Appendix C. The instantaneous amplitude
and frequency reveal whether amplitude and/or frequency modulation effects are
present. These instantaneous characteristics are constant in the case of a purely
linear system response, but start to oscillate when nonlinearities are present, as
was illustrated in Figure 6.1. The peak-to-peak values of the oscillations in the
instantaneous amplitude and instantaneous frequency, represented by Ma and Mf
respectively, are used as a measure for the amount of modulation.
116 Chapter 6. Vibro-acoustic modulation based damage identification
Signal decomposition
Modulated Instantaneous Amplitude/
Actuation velocity Filtered velocity amplitude/ frequency
(2x single-tone) response response frequency modulation
Fp(t) v(t) vbp(t) Ainst(t) Ma
Bandpass Hilbert
Structure
Fc(t) filter, fc transform finst(t) Mf
Figure 6.2 Time domain signal analysis process of the response signal v(t) containing a bandpass filter
around the carrier frequency f c and the Hilbert transform to extract the instantaneous amplitude Ainst (t)
and frequency f inst (t). The peak-to-peak values of these instantaneous characteristics are used as a measure
for the amount of amplitude modulation Ma and frequency modulation Mf .
fp 2fp fp
fc 3fp fc
3fp 5fp
−5 −5
Velocity [m/s]
fc-nfp f +nf fc-2nfp f +2nf
Velocity [m/s]
10 c p 10 c p
−10 −10
10 10
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
Frequency [Hz] Frequency [Hz]
Figure 6.3 Fourier transformed response signals for a system containing a (a) quadratic and (b) cubic
nonlinearity with respect to displacement q(t) showing sidebands at f c ± n f p and f c ± 2n f p (with
n = 1, 2, 3, . . .) respectively.
This result is analogous to the fact that the response of an undamped system is
in phase with its excitation, whereas the response of a viscously damped system
can exhibit a phase deviation. A nonlinearity in terms of velocity can consequently
result in varying phase deviations. Frequency modulations are obtained because
the instantaneous frequency is calculated as the time derivative of the instantaneous
phase as shown in Appendix C. Note that the frequency of the modulation effect
is different for the quadratic and cubic nonlinear systems. The responses of
the systems containing a quadratic nonlinearity show a dominant superimposed
−4 −4 −4 −4
x 10 x 10 x 10 x 10
3.2 3.195
Ainst (t) [m/s]
3.195 3.192
3.19 3.19
3.191
3.19
3.18
15 15.1 15.2 15 15.1 15.2 15 15.1 15.2 15 15.1 15.2
Time [s] Time [s] Time [s] Time [s]
Figure 6.4 Instantaneous amplitude Ainst (t) (top) and frequency f inst (t) (bottom) variations obtained
for a system containing four different types of nonlinearity: (a) quadratic and (b) cubic with respect to
displacement q(t) and (c) quadratic and (d) cubic with respect to velocity q̇(t). Each case contains three
response signals corresponding to three different values for nonlinearity factor ǫ.
118 Chapter 6. Vibro-acoustic modulation based damage identification
frequency equal to f p , while this is 2 f p for the cubic nonlinear systems. This
inherently corresponds to the frequency differences between the sidebands as
shown in Figure 6.3.
Bottom view
100 mm
25 25 25
20
Piezoelectric
diaphragm 2
1 3
2
32 Impact
40 mm
360 mm
y
x Impact
z
2.2
location
Piezoelectric Shaker
Point (x,y) =
180 mm
diaphragm 1 connection
(25,120) mm
y
x
20
Node line Y1 Y2 Y3
Figure 6.5 Three dimensional and bottom view of the composite skin-stiffener structure with a butt-joint
stiffener. The dimensions, the measurement points (dots) and the impact location are indicated.
Supports (4×)
y
during impact
x
Stiffener 10 mm
and filler
79 mm
Impact Cracked
75 mm
location Skin
filler
Point (x,y) = delamination
(25,120) mm
85 mm
Figure 6.6 Ultrasonic C-scan of the impact damaged skin-stiffener structure showing a complex
combination of failure mechanisms near the skin-stiffener interface.
120 Chapter 6. Vibro-acoustic modulation based damage identification
13. 14.
5
Velocity [mm/s]
0
5. −5
0.8 0.802
Time [s]
0.804 0.806
12. 10. 4.
MODEL 240L POWER AMPLIFIER
11.
6. 9.
F(t) 8.
7. v(t)
1. y
3. z
2.
In the first set of experiments, the global dynamic behavior of the structure is
determined in terms of natural frequencies and the frequency dependent deflection
patterns, referred to as operational deflection shapes (ODS). An excitation signal
composed of a linear sweep between 150 and 3050 Hz was sent to the shaker. The
force response was measured at the fixed excitation point. The velocities were
acquired at 51 (3×17) points, matching the node lines in Figure 6.5. All signals
were sent and acquired for 2.62 seconds at a rate of 50 kHz (217 samples). A
measurement at each point was repeated 10 times. These time responses were
subsequently windowed, Fourier transformed and averaged to obtain the power
spectral densities. The cross-power spectral density S Fi v j (ω ) is divided by the auto-
power spectral density of the input force S Fi Fi (ω ) to obtain the mobility frequency
response functions HFi v j (ω ) between the fixed point of excitation i and the roving
measurement points j according to [30].
Figure 6.8 shows the magnitude of the total set of frequency response functions of
the damaged structure. Sharp peaks correspond to the natural frequencies of the
structure. The bending and torsion dominant operational deflection shapes of the
structure are extracted at the natural frequencies with the help of peak picking.
The natural frequencies of the first six bending and torsion dominant operational
deflection shapes of the damaged structure are presented in Table 6.1. Note that
the first torsion mode was not measured. The natural frequency of this mode is
lower than the lowest frequency used in the excitation. Earlier research [8] showed
that bending dominant deformations in the yz-plane are more affected by skin-
stiffener damage than torsional deformations around the y-axis. The present work
is, therefore, focused on the vibro-acoustic modulation effects introduced by the
bending dominant operational deflection shapes.
2
10
|HFv| [(mm/s)/N]
0
10
−2
10
Figure 6.8 The magnitude of the frequency response functions for all 51 (3 × 17) measurement points of
the damaged structure.
122 Chapter 6. Vibro-acoustic modulation based damage identification
Table 6.1 Natural frequencies of the first six bending and torsion modes of the damaged (indicated by a
tilde) structure.
The second set of experiments are the actual vibro-acoustic modulation measurements.
Two single-tone harmonic signals with different amplitude and frequency were
simultaneously sent to the shaker and the piezoelectric diaphragm. The pump
excitation signal introduced by the shaker had a frequency f p corresponding to
one of the bending frequencies listed in Table 6.1 and was varied in strength. The
piezoelectric diaphragm was subjected to a weaker 50 kHz ultrasonic excitation
signal. The amplitude of this carrier signal Ac was kept constant. The carrier
frequency f c is chosen in such a way that the carrier sideband components are well
separated from the higher harmonics of the pump frequency, but that the high-
frequency wave field can still be described by a limited number of measurement
points. The shaker excitation force and the velocity response were simultaneously
acquired at a sample rate of 1 MHz for 1.05 seconds (220 samples). No averaging was
applied. Only the steady state response is considered in the analysis. The structure
vibrates at steady state after the transient (start-up) response has disappeared.
A period of 0.8 seconds was revealed to be sufficiently long for these transient
effects to become negligible. The captured responses are processed by the time
domain signal decomposition approach as explained in Section 6.2.2. The results are
presented in the next section and are used to analyze the modulation phenomena
in the composite skin-stiffener structure in detail.
The experimental results are presented in the current section. The results obtained
for a single measurement point are discussed first, followed by the results obtained
for multiple points underneath the stiffener. Possible explanations for the measured
results are formulated and discussed. The effect of the underlying dynamic
behavior on the measured modulations is addressed in the last subsection.
6.4. Experimental results and discussion 123
The steady state velocity response v(t) of the damaged structure caused by two
single-tone harmonic excitation signals is shown in Figure 6.9(a). The excitation
consists of an intense pump excitation at the 4th bending frequency ( f p = 1455 Hz)
and a weaker carrier excitation ( f c = 50 kHz). The velocity response v(t) is measured
at location (x,y) = (25,120) mm according to the coordinate system presented in
Figure 6.5. This point is located at the skin near the skin-stiffener connection and
coincides with the damaged region. Low and high frequency components can be
distinguished in the velocity response. The amplitude of the Fourier spectrum of
this response, i.e. |V ( f )|, is shown in Figure 6.9(b) and reveals higher harmonic
components n f p and multiple carrier sidebands f c ± n f p (with n = 1, 2, 3, . . .). These
additional components are indicative of a nonlinear response.
The low frequency part of the response matches the nonlinear response that was
20
fp
1
20 fc
15
|V(f)| [mm/s]
v(t) [mm/s]
0.5
0 10
0
fc 4 5 6
5 x 10
4
−20
0
0.8 0.801 0.802 0.803 0.804 0 2 4 6 8 10
Time [s] Frequency [Hz] 4
x 10
(a) Original velocity response v(t). (b) Fourier spectrum of the original velocity
response v(t).
5
v (t) [mm/s]
0
bp
−5
0.8 0.801 0.802 0.803 0.804
Time [s]
(c) Bandpass filtered velocity response vbp (t).
Figure 6.9 The (a) original and (c) bandpass filtered (40-60 kHz) velocity response measured at location
(x,y) = (25,120) mm of the damaged structure for two single-tone excitation signals with f p = 1455 Hz and
f c = 50 kHz and the highest shaker excitation amplitude. The Fourier spectrum of the original response (b)
shows the higher harmonic components n f p and carrier sidebands f c ± n f p (with n = 1, 2, 3, . . .).
124 Chapter 6. Vibro-acoustic modulation based damage identification
measured and analyzed in earlier research by authors [26] on the same specimen
using a single-tone harmonic shaker excitation. The distorted harmonic response
of the skin in that study was linked to the opening, closing and contact phase of
the skin-stiffener damage. In the present work, however, the high frequency of the
response is utilized. The carrier response and its dominant sideband components
are separated from the rest of the response by applying a bandpass filter within a
f c ± 10 kHz frequency range. The resulting narrow band velocity response vbp (t)
is depicted in Figure 6.9(c). The oscillating signal envelope clearly indicates that
amplitude modulation effects are present.
The nonlinear modulation effects in the bandpass filtered velocity response are
extracted by utilizing the Hilbert transform, as was described in Section 6.2.2. The
periodic behavior of the instantaneous amplitude Ainst (t) and frequency f inst (t)
variations in Figures 6.10(a) and 6.10(b) shows that both amplitude and frequency
modulation effects are present. The amount of modulation, represented by the
4
x 10
6 5.1
Ma
5.5 Mf
(t) [mm/s]
5.05
(t) [Hz]
5
5
4.5
inst
inst
4.95
A
3.5 4.9
0.8 0.801 0.802 0.803 0.804 0.8 0.801 0.802 0.803 0.804
Time [s] Time [s]
(a) Instantaneous amplitude Ainst (t). (b) Instantaneous frequency f inst (t).
0.8 250
1455 Hz
(f)| [mm/s]
200 1455 Hz
0.6
(f)| [Hz]
150
0.4
100
inst
inst
|f
0.2
|A
50
0 0
0 2000 4000 6000 8000 10000 0 2000 4000 6000 8000 10000
Frequency [Hz] Frequency [Hz]
(c) Fourier spectrum of the instantaneous (d) Fourier spectrum of the instantaneous
amplitude Ainst (t). frequency f inst (t).
Figure 6.10 The (a) instantaneous amplitude Ainst (t) and (b) instantaneous frequency f inst (t) of the
bandpass filtered velocity response vbp (t) measured at location (x,y) = (25,120) mm of the damaged structure
for five shaker excitation amplitudes. The associated Fourier spectra (c, d), illustrated for the highest shaker
excitation amplitude, reveal a dominant frequency equal to f p = 1455 Hz.
6.4. Experimental results and discussion 125
peak-to-peak value, increases for higher amplitude levels of the pump excitation.
The dominant frequency of the amplitude and frequency modulation phenomena
matches the pump excitation frequency f p = 1455 Hz, as shown in Figures 6.10(c)
and 6.10(d). According to the numerical results presented in Section 6.2.3, this
indicates that the nonlinearity in the structure tends to be more a quadratic rather
than a cubic type of nonlinearity.
I II III
0.802 0.802
10
20
Time [s]
Time [s]
0.801 0 0.801 0
−20 −10
0.8 0.8
0 0.1 0.2 0.3 0 0.1 0.2 0.3
y−coordinate [m] y−coordinate [m]
(a) Original velocity response v(t) [mm/s]. (b) Bandpass filtered velocity response
vbp (t) [mm/s].
y y
x x
Figure 6.11 The (a) original and (b) bandpass filtered (40–60 kHz) velocity response measured at node
line ‘Y2’ of the damaged structure for two single-tone harmonic excitation signals with f p = 1455 Hz and
f c = 50 kHz and the highest shaker excitation amplitude.
126 Chapter 6. Vibro-acoustic modulation based damage identification
x-coordinate [m]
y
(a) x
t = 0.801005 s
(b) 0
−20
0 0.05 0.1 0.15 0.2 0.25 0.3
Velocity [mm/s]
0.8011
10
Time [s]
(c ) 0.8010 0
-10
0.8009
0 0.05 0.1 0.15 0.2 0.25 0.3
y-coordinate [m]
Figure 6.12 Comparison of the (a) damage location and geometry with the (b, c) bandpass filtered velocity
distribution vbp (t) of the skin at node line ‘Y2’.
the interface between the skin and the injection molded filler is not delaminated
over the entire width of the filler. A more detailed comparison between the damage
geometry and the bandpass filtered velocity response is presented in Figure 6.12.
The instantaneous amplitude and frequency of the bandpass filtered velocity
distribution are subsequently extracted according to the procedure described in
Section 6.2.2. The peak-to-peak values of the oscillations in these instantaneous
characteristics, represented by Ma and Mf , are utilized as a measure for the amount
of modulation. Figures 6.13(a) and 6.13(b) show the results obtained for a pump
excitation frequency equal to the 4th bending mode ( f p = 1455 Hz) and three pump
excitation amplitudes Fp . Figures 6.13(c) and 6.13(d) show the same results in the
case of the 6th bending mode ( f p = 2340 Hz). Although the structure exhibits
both amplitude and frequency modulation effects over the entire length of the
structure, increased amplitude modulation effects are measured at the damaged
area. The damage seems to hardly modulate the frequency of the carrier signal.
Figure 6.15 provides a closer look to the bandpass filtered time responses measured
at three successive points (i.e. y = 115 mm, y = 120 mm, and y = 125 mm) at the
damaged region. The dominant frequency of the amplitude modulation patterns
again matches the pump excitation frequency f p . These figures also reveal that
the phase of the amplitude modulation can significantly vary between the different
measurement points.
The frequency modulation in Figures 6.13(b) and 6.13(d) shows several higher
peaks. These peaks are attributed to a local low amplitude of the fundamental
6.4. Experimental results and discussion 127
3
x 10
6 4
Fp= 0.9 N Fp= 0.9 N
Fp= 5.4 N Fp= 5.4 N
Ma [mm/s]
Mf [Hz]
Fp= 9.4 N Fp= 9.4 N
2
2
0 0
0 0.1 0.2 0.3 0 0.1 0.2 0.3
y−coordinate [m] y−coordinate [m]
(a) Amplitude modulation for f p = 1455 Hz. (b) Frequency modulation for f p = 1455 Hz.
4
x 10
5 6
Fp= 0.9 N Fp= 0.9 N
4
Fp= 5.4 N Fp= 5.4 N
Ma [mm/s]
4
3 Fp= 9.4 N Mf [Hz] Fp= 9.4 N
2
2
1
0 0
0 0.1 0.2 0.3 0 0.1 0.2 0.3
y−coordinate y−coordinate [m]
(c) Amplitude modulation for f p = 2340 Hz. (d) Frequency modulation for f p = 2340 Hz.
y y
x x
Figure 6.13 The (a, c) amplitude and (b, d) frequency modulation distributions of the carrier response
measured at node line ‘Y2’ of the damaged structure caused by two single-tone harmonic excitation signals
for three pump excitation amplitudes Fp . The results for f p = 1455 Hz are shown in (a, b), while (c, d) show
the results for f p = 2340 Hz. The carrier frequency f c was 50 kHz for all cases.
carrier response (e.g. near a nodal point) combined with a relatively large amount
of amplitude modulation, almost leading to over-modulation effects [31]. For
example, the peak at y = 250 mm in Figure 6.13(d). Figure 6.14 shows that
the associated bandpass filtered velocity response is almost fully modulated in
amplitude. Consequently, poor estimations of the instantaneous frequency are
obtained at the time instances where the envelope of the response approaches zero.
1
5
( t) [mm/s]
(t) [mm/s]
0.5
0 0
(x,y) = (25,125) mm
−0.5
bp
bp
−5 (x,y) = (25,120) mm
v
v
0.8 0.8005 0.801 0.8015 0.802 0.8 0.8005 0.801 0.8015 0.802
Time [s] Time [s]
Figure 6.14 Severe amplitude modulations of Figure 6.15 Phase differences in the amplitude
the bandpass filtered (40-60 kHz) velocity response modulations of the bandpass filtered (40-60 kHz)
measured at location (x,y) = (25,250) mm of the velocity response measured at three successive
damaged structure for two single-tone harmonic points on the damaged area for two single-tone
excitation signals with f p = 2340 Hz and harmonic excitation signals with f p = 2340 Hz
f c = 50 kHz and the highest shaker excitation and f c = 50 kHz and the highest shaker excitation
amplitude. amplitude.
0.5
v(t) Damaged
0
(c) Closed state area
−0.5
−1 y
Carrier
II z v(t) wave
0 0.5 1 1.5 2
Time [s] −3 field
x 10
Figure 6.16 A simplified and schematic explanation of the (a) carrier modulation principle introduced by
the periodic opening and closing of the damage under an intense low-frequency pump excitation. In the (b)
open state the skin is free to vibrate, whereas the carrier amplitudes are compressed in the (c) closed situation.
amplitude modulations and the hardly modulated frequency of the carrier signal
at the damaged region. The applied low-frequency pump excitation signal will
change the damage interface conditions. In an ideal situation, the skin-stiffener
damage is completely opened and closed by the flexural vibrations of the skin, as
schematically illustrated in Figure 6.16. During the open state (b), the skin at the
damaged region is free to vibrate under the simultaneously applied high-frequency
carrier wave field. The amplitude of this wave field is, however, expected be to
smaller during the closed state (c). As a result, the high-frequency carrier wave
predominantly experiences periodic modulations of the amplitude, while the effects
on the frequency are expected to be minor.
6 10
(t) [mm/s]
(t) [mm/s]
4
5 f = 48 kHz
c
F
inst
inst
2 c f = 50 kHz
c
A
f = 52 kHz
c
0 0
0.8 0.802 0.804 0.806 0.8 0.802 0.804 0.806
Time [s] Time [s]
(a) Variable Fc = 0.01–0.07 V, f c = 50 kHz. (b) Variable f c = 48/50/52 kHz, Fc = 0.07 V.
Figure 6.17 The effect of the carrier excitation amplitude Fc and frequency f c on the instantaneous
amplitude Ainst (t). The responses are measured at location (x,y) = (25,120) mm of the damaged structure
for a pump excitation frequency equal to f p = 1455 Hz.
6.5. Conclusions & future prospects 131
|H | [(mm/s)/V]
1
(a)
0.1
hf
(x,y) = (25,120) mm
0.01
100
∠H [ ]
o
(b)
hf
0
−100
Velocity [mm/s]
(c) 1
0.1
0.01
0.001
4.4 4.6 4.8 5 5.2 5.4 5.6
Frequency [Hz] 4
x 10
Figure 6.18 The (a) magnitude and (b) phase of the frequency response functions within the 44–56 kHz
range measured at location (x,y) = (15,120) mm on node line ‘Y2’ of the damaged structure. The high
modal density causes the sidebands of the (c) Fourier spectrum of the modulated velocity response at the same
location to coincide with different modes, and hence introduces amplitude and phase deviations between the
left sideband, the carrier and right sideband components. The Fourier spectra associated with three carrier
excitation frequencies f c and a pump excitation frequency f p equal to 1455 Hz are presented.
In the experimental situation, the high modal density causes the frequencies of the
carrier and its sideband components to coincide with different modes, as illustrated
in Figure 6.18. This can lead to an unequal distribution of sideband amplitudes
around the carrier as was shown in Figure 6.9(b), but can also cause phase
deviations between the different harmonic components. Consequently, the spectral
effects introduced by the underlying dynamic behavior affect the amplitude and
phase of the amplitude modulation for each selected carrier frequency, as shown in
Figure 6.17(b), but also for each measurement point, as was shown in Figure 6.15.
The variations in the phase of the amplitude modulation distributions shown in the
previous section are inherently expected to be a result of the underlying spectral
response of the damaged structure, which are intensified by the high modal density
and the spatial variability of the complex wave field.
The objective of this research was to analyze whether the vibro-acoustic modulation
method can be utilized to detect, localize and characterize impact damage
in a composite skin-stiffener structure. The work further aimed to obtain a
better understanding of the modulation phenomena. Vibro-acoustic modulation
132 Chapter 6. Vibro-acoustic modulation based damage identification
Acknowledgments
This material is based on work supported by National Aeronautics and Space
Administration, Langley Research Center under Research Cooperative Agreement
References 133
References
[1] C. Boller, F.-K. Chang, and Y. Fujino. Encyclopedia of structural health monitoring. John
Wiley & Sons, Ltd, Chichester, UK, 2009.
[2] K. Diamanti and C. Soutis. Structural health monitoring techniques for aircraft
composite structures. Progress in Aerospace Sciences, 46(8):342–352, 2010.
[3] Y. Zou, L. Tong, and G.P. Steven. Vibration-based model-dependent damage
(delamination) identification and health monitoring for composite structures – A
review. Journal of Sound and Vibration, 230(2):357–378, 2000.
[4] C.-P. Fritzen and P. Kraemer. Self-diagnosis of smart structures based on dynamical
properties. Mechanical Systems and Signal Processing, 23(6):1830–1845, 2009.
[5] K. Worden, C.R. Farrar, J. Haywood, and M. Todd. A review of nonlinear dynamics
applications to structural health monitoring. Structural Control and Health Monitoring,
15:540–567, 2008.
[6] N. Krohn, R. Stoessel, and G. Busse. Acoustic non-linearity for defect selective
imaging. Ultrasonics, 40(1-8):633–7, 2002.
[7] K.-Y. Jhang. Nonlinear ultrasonic techniques for nondestructive assessment of micro
damage in material: a review. Precision Engineering and Manufacturing, 10(1):123–135,
2009.
[8] T.H. Ooijevaar, R. Loendersloot, L.L. Warnet, A. de Boer, and R. Akkerman. Vibration
based structural health monitoring of a composite T-beam. Composite Structures,
92(9):2007–2015, 2010.
[9] P.B. Nagy. Fatigue damage assessment by nonlinear ultrasonic materials
characterization. Ultrasonics, 36(1-5):375–381, 1998.
[10] K.E.-A. Van Den Abeele, P.A. Johnson, and A. Sutin. Nonlinear elastic wave
spectroscopy (NEWS) techniques to discern material damage, part I: nonlinear wave
modulation spectroscopy (NWMS). Research in Nondestructive Evaluation, 12:17–30,
2000.
[11] K.E.-A. Van Den Abeele, J. Carmeliet, J.A. Ten Cate, and P.A. Johnson. Nonlinear
elastic wave spectroscopy (NEWS) techniques to discern material damage, part II:
single-mode nonlinear resonance acoustic spectroscopy. Research in Nondestructive
Evaluation, 12:31–42, 2000.
[12] N.C. Yoder and D.E. Adams. Vibro-acoustic modulation utilizing a swept probing
signal for robust crack detection. Structural Health Monitoring, 9(3):257–267, 2010.
[13] S. Vanlanduit, E. Parloo, and P. Guillaume. Combined damage detection techniques.
Journal of Sound and Vibration, 266(4):815–831, 2003.
[14] A. Klepka, W.J. Staszewski, R.B. Jenal, M. Szwedo, T. Uhl, and J. Iwaniec. Nonlinear
acoustics for fatigue crack detection - Experimental investigations of vibro-acoustic
wave modulations. Structural Health Monitoring, 11(2):197–211, 2011.
134 Chapter 6. Vibro-acoustic modulation based damage identification
[15] V. Zaitsev and P. Sas. Nonlinear response of a weakly damaged metal sample: a
dissipative modulation mechanism of vibro-acoustic interaction. Journal of Vibration and
Control, 6(6):803–822, 2000.
[16] F. Aymerich and W.J. Staszewski. Impact damage detection in composite laminates
using nonlinear acoustics. Composites Part A: Applied Science and Manufacturing,
41(9):1084–1092, 2010.
[17] D. Donskoy, A. Sutin, and A. Ekimov. Nonlinear acoustic interaction on contact
interfaces and its use for nondestructive testing. NDT & E International, 34:231–238,
2001.
[18] D. Dutta, H. Sohn, K.A. Harries, and P. Rizzo. A nonlinear acoustic technique for crack
detection in metallic structures. Structural Health Monitoring, 8(3):251–262, 2009.
[19] Z. Parsons and W.J. Staszewski. Nonlinear acoustics with low-profile piezoceramic
excitation for crack detection in metallic structures. Smart Materials and Structures,
15(4):1110–1118, 2006.
[20] F. Amerini and M. Meo. Structural health monitoring of bolted joints using linear and
nonlinear acoustic/ultrasound methods. Structural Health Monitoring, 10(6):659–672,
2011.
[21] A. Zagrai, D. Donskoy, A. Chudnovsky, and E. Golovin. Micro- and macroscale
damage detection using the nonlinear acoustic vibro-modulation technique. Research in
Nondestructive Evaluation, 19(2):104–128, 2008.
[22] N.A. Chrysochoidis, A.K. Barouni, and D.A. Saravanos. Delamination detection in
composites using wave modulation spectroscopy with a novel active nonlinear
acousto-ultrasonic piezoelectric sensor. Journal of Intelligent Material Systems and
Structures, 22(18):2193–2206, 2011.
[23] M. Meo and G. Zumpano. Nonlinear elastic wave spectroscopy identification of impact
damage on a sandwich plate. Composite Structures, 71(3-4):469–474, 2005.
[24] U. Polimeno and M. Meo. Detecting barely visible impact damage detection on aircraft
composites structures. Composite Structures, 91(4):398–402, 2009.
[25] H.F. Hu, W.J. Staszewski, N.Q. Hu, R.B. Jenal, and G.J. Qin. Crack detection using
nonlinear acoustics and piezoceramic transducers – Instantaneous amplitude and
frequency analysis. Smart Materials and Structures, 19(6):065017, 2010.
[26] T.H. Ooijevaar, M.D. Rogge, R. Loendersloot, L.L. Warnet, R. Akkerman, and T. Tinga.
Nonlinear dynamic behavior of an impact damaged composite skin-stiffener structure.
In preparation for: Composite Science and Technology, 2014.
[27] L. Meirovitch. Fundamentals of vibrations. McGraw-Hill Higher Education, New York,
NY, 2001.
[28] I. Solodov, D. Döring, and G. Busse. New opportunities for NDT using non-linear
interaction of elastic waves with defects. Journal of Mechanical Engineering,
57(03):169–182, 2011.
[29] A. Offringa, J.W. van Ingen, and A. Buitenhuis. Butt-joined, thermoplastic
stiffened-skin concept development. SAMPE Journal, 48(2):7–15, 2012.
[30] M.H. Richardson. Is it a mode shape, or an operating deflection shape? Sound and
Vibration, pages 1–11, 1997.
[31] Q. Li and L. Atlas. Over-modulated AM-FM decomposition. In Franklin T. Luk, editor,
Proceedings of SPIE Volume 5559, Advanced Signal Processing Algorithms, Architectures, and
Implementations XIV, pages 172–183, 2004.
Chapter 7
Discussion
The work presented in previous chapters concerned the identification of damage in advanced
composite skin-stiffener structures by employing changes in the dynamic behavior. This
chapter broadens the discussions from the previous chapters with respect to the objective of
this research. Firstly, the design of a vibration based structural health monitoring approach
to identify damage in composite skin-stiffener structures is discussed. The results obtained in
previous chapters are combined to extract additional recommendations for the development
of a structural health monitoring approach. Secondly, a more general discussion on the
practical application of the vibration based methodologies is presented.
135
136 Chapter 7. Discussion
Damage
identification
method
Damage
Structure
scenario
Figure 7.1 The multidisciplinary framework for the design of a structural health monitoring strategy.
The first step is illustrated in Figure 7.2(b) and concerns the analysis of the potential
failure mechanisms and the potential damage scenarios of the structure that is
under investigation. Knowledge about the loading conditions, historic failure
data or structural models can provide useful information for this step. This step
inherently encourages the development of a structural health monitoring approach
to become an integral part of the design process of a component. This structural
design process can provide information about, for example, the safety-critical
damage locations or estimates of the damage tolerance of the component. Both
aspects set the requirements for the performance of the structural health monitoring
approach.
The need for repair of aircraft components is commonly assessed based on visual
inspection according to the BVID criterion [1]. Visual inspection allows for the
identification of macroscopic damage rather than microscopic defects. Three impact
induced damage scenarios were considered for the skin-stiffener structures that
7.1. Design of an SHM strategy 137
Sub-components
(a)
Structure
Step 2 τ
Shear Open/close (c)
Understanding mode Delamination
mode
effect of damage on τ
dynamic behavior. τ
Delamination Stiffener Skin
Structural
dynamics
Delaminated area
2
Acceleration [-]
1
d u/dz [−]
identification
-1
method. Pristine
−2
0 0.2 0.4 0.6 0.8 1 -1 0 1
Damaged Position z/L [−] Velocity [-]
Damage
identification
method
Level 1 Level 2 Level 3
Step 4 detection localization severity/extent
Performance (e)
Method A
classification for
each method. Method B
...
Structural health
monitoring strategy
for sub-component
Figure 7.2 The scenario based design procedure of a vibration based structural health monitoring strategy.
138 Chapter 7. Discussion
were used in the present work, i.e. damage at the skin, damage at the skin-
stiffener connection and damage at the stiffener run-out. These damage scenarios
are all classified as barely visible impact damage based on the dent depth. This
means that they may remain undetected according to the BVID criterion. The
skin related damage is indeed considered to be less safety-critical and is therefore
of lower importance. However, the damage at the skin-stiffener connection and
the skin-stiffener run-out can significantly affect the performance of the structure.
An evaluation of the potential consequences involved in the damage scenarios is
therefore required in the first step of the design procedure in order to assess the
importance of detection.
The second step is focused on the relation between the damage scenario and the
dynamic behavior of the component, as depicted in Figure 7.2(c). This step aims to
obtain a physical understanding of the effect of the potential damage scenarios
on the dynamic behavior of the component. Both analytical and experimental
analyses were used in the present research to investigate the effect of the damage
on the dynamic behavior. The damage at the skin-stiffener connection is usually
dominated by a delamination type of damage. The analysis of a delaminated beam
model in Chapter 2 revealed that the effect of a delamination is more pronounced in
the mode shape curvatures than in the mode shapes themselves. The mode shape
curvatures can show clear discontinuities at the boundaries of a delamination in the
case of flexural vibrations. The two driving phenomena behind these discontinuities
in the mode shape curvatures are, firstly, the change in shear stress distribution (see
Figure 7.3(a) ), and secondly, the potential opening and closing of the delaminated
segments (see Figure 7.3(b) ). The work presented in Chapter 5 experimentally
confirmed the opening and closing behavior and revealed severe nonlinear dynamic
effects associated with this motion. The study in Chapter 5 also showed that the
opening and closing motion of delaminated segments is magnified by the motion of
the skin at both sides of the stiffener, as schematically illustrated in Figure 7.3(c) ).
These geometrical effects can, on the one hand, result in larger discontinuities in
the mode shape curvature during the opening phase. On the other hand, they can
increase the nonlinear effects that are introduced during the contact phase.
The third step of the scenario based design procedure concentrates on the selection
of a suitable damage identification method, i.e. damage features and classifiers.
This step is schematically illustrated in Figure 7.2(d) and is an important step
in the process to obtain an effective damage identification strategy. The damage
identification method should match the observed changes in the dynamic behavior.
7.1. Design of an SHM strategy 139
x x x
z τ z y
τ
δ
τ δ
Figure 7.3 Dynamic phenomena related to delamination type of damage in skin-stiffener structures.
The last step is a classification of the performance of each method in terms of its
detection, localization and characterization capabilities, as depicted by Figure 7.2(e).
This classification can be used to define a structural health monitoring strategy
according to the desired performance level.
The work presented in Chapters 3 and 4 experimentally demonstrated the feasibility
of using mode shape curvatures combined with the modal strain energy damage
index algorithm to detect, localize and roughly estimate the size of barely visible
impact damage in advanced composite skin-stiffener structures. The results showed
that the effective application of the method is limited to health monitoring of
140 Chapter 7. Discussion
Diagnostics
Statistical
classifier I
Feature
Actuation Response extractor I
Statistical
classifier II
Sensor Current state
Structure
system of the structure
Feature Statistical
extractor II classifier
Damage
(a) Combining different damage sensitive features (e.g. natural frequency, mode shape curvatures,
nonlinear dynamic behavior) and classifiers.
Technique I
Actuation Sensor Feature Statistical
system extractor classifier
Current state
Structure Technique II of the structure
Actuation
Sensor Feature Statistical
system extractor classifier
Damage
(b) Combining different damage identification techniques (e.g. structural vibration based methods,
acoustic emission, acousto-ultrasonics).
Figure 7.4 A schematic illustration of the diagnostic part of a structural health monitoring process showing
two ways to combine multiple damage identification methods: (a) for a combination of different damage
sensitive features and classifiers.
The results presented in this thesis contribute towards obtaining a higher level of
maturity in the structural health monitoring technologies. However, several issues
need to be considered before field application of the structural health monitoring
technologies is possible. Five major technology gaps were described in Section 1.3.4:
Sensor
Skin-stiffener Skin
damage damage
y z x
Actuator
Figure 7.5 The skin-stiffener structure that was analyzed in Chapter 4. A combination of damage
identification techniques (e.g. structural vibration based methods, acousto-ultrasonics) is required to identify
the skin and the skin-stiffener interface damage.
The first three challenges were addressed in the present research. The extension
to more complex composite structures was the main focus of the work presented
in Chapters 3 and 4. Chapters 5 and 6 provide a first step towards a higher
performance level by characterizing the nonlinear dynamic behavior introduced by
the damage. The selection of appropriate damage features and classifiers was one
of the main topics addressed in this thesis. The scenario based design procedure,
as described in the previous section, showed itself to be an important outcome
and provides a useful guideline for the development of a suitable structural health
monitoring solution for safety-critical structures. The remaining two challenges, i.e.
the integrated sensing and the environmental effects, are discussed in the remainder
of this chapter.
to-noise ratio as was found by Grouve et al. [3]. Despite this limitation, successful
applications are found for low-frequency vibration sensing [4, 5] as well as high-
frequency sensing of propagating waves [6]. New developments of optical fiber
distributed sensing systems provide the ability for strain sensing along the optical
fiber with an extremely high spatial resolution [7]. A low sample rate, however,
limits the current application of this technology to quasi-static measurements.
Alternatively, piezoelectric elements are widely used for actuation and sensing
purposes in health monitoring [8–10]. These piezoelectric elements have the ability
to convert mechanical energy into electrical energy, and vice versa [11]. They utilize
the direct piezoelectric effect to sense structural deformations and the inverse effect
to actuate the structure. One of the main advantages of piezoelectric elements is
that they can be used over a broad frequency range. This makes them applicable for
nearly all dynamics based damage identification techniques, i.e. structural vibration
based methods [12], electromechanical-impedance based techniques [13], acoustic-
emission [14] as well as acousto-ultrasonics [8]. Other advantages include the cost
effectiveness, as well as the ability to harvest energy [15].
Distributed strain sensing based on fiber optic or piezoelectric sensors is potentially
advantageous for the mode shape curvature based approach. The direct relation
between bending strain and curvature can be beneficial to overcome the numerical
errors involved in the calculation of the second derivative of the displacement mode
shapes. This issue is addressed by several studies. Lestari et al. [16] and Qiao et
al. [17] used MSE-DI algorithm in combination with strain mode shapes measured
by PVDF film sensors in order to identify damage in a composite specimen.
Adewuyi et al. [18] performed similar experiments utilizing long-gage optical fiber
Bragg grating sensors.
An important challenge regarding the practical application of the vibration based
damage identification techniques is the number and the position of the sensors.
The vibration based methods utilize standing wave patterns. Consequently, high
spatial resolution monitoring requires a relatively dense sensor network to avoid
spatial aliasing (except for the natural frequencies and modal damping based
methods). This contradicts the practical application of a vibration based technology,
since the number of sensors is often limited. The optimal sensor placement has
received considerable attention in the literature [19–21]. Savananoz [22] particularly
addresses the optimum spatial sampling interval for damage detection based on
curvatures.
There are some other ways to deal with the sensor issue. The first and most
straightforward approach is to utilize new sensing technologies. Optical fiber
distributed sensing systems as well as other distributed sensor technologies are
under development [7]. The number of successful practical applications of these
emerging technologies is, however, limited. Secondly, and the more feasible, is to
monitor only the most safety-critical locations of the structure, which is referred
to as ‘hotspot’ monitoring. These critical locations can follow from historic failure
144 Chapter 7. Discussion
2
0
0.2 0.3
0.2 1
0.1 0.1 x−coordinate [m]
y−coordinate [m] 0 0
Skin-stiffener
damage
Measurement
y z x points
Figure 7.6 Damage index β ij distributions (1D formulation in x-direction) of the MSE-DI algorithm
obtained by utilizing only the measurement points underneath the stiffeners of the skin-stiffener structure
that was analyzed in Chapter 4.
data or from the structural design process. Chapter 4 showed that the effective
application of the mode shape curvature based approach is limited to health
monitoring of the skin-stiffener connections. Moreover, defects located at the skin
are considered to be less safety-critical. Consequently, the number of measurement
points can be reduced to only the measurement points underneath the skin-stiffener
connection, as shown in Figure 7.6. This solution can provide an effective way
of monitoring the most critical areas of a skin-stiffener structure. However, this
requires good insight into the failure behavior and mechanisms of the structure
and the most safety-critical damage locations. Finally, utilizing techniques based on
propagating waves can also allow for a reduced number of sensors to cover large
areas. This is also where the vibro-acoustic modulation based method, studied in
Chapter 6, can potentially be advantageous. Although the current work is focused
on the feasibility and the understanding of the approach in the case of the steady
state behavior, it opens the ability to use pulsed carrier excitation signals traveling
between an actuator and sensor, as was shown by Kazakov et al. [23]. In that case,
the approach essentially corresponds to the more conventional acousto-ultrasonics
techniques (pitch-catch or pulse-echo) [8, 24, 25] supplemented with an additional
lower-frequency pump source in order to produce the modulations.
References 145
References
8.1 Conclusions
The major conclusions of this work with respect to the objective of this thesis are
presented below.
General
• Vibration based structural health monitoring is a design related problem.
The development of a damage identification strategy is made-to-measure
work and requires a thorough physical understanding of the potential failure
mechanisms, the critical damage locations and their effect on the dynamic
behavior.
• A combination of several damage identification techniques, damage features
and classifiers is required to identify the wide variety of damage scenarios
that can occur in a composite skin-stiffener structure.
149
150 Chapter 8. Conclusions and recommendations
8.2 Recommendations
With respect to the work presented in this thesis further research is recommended
on the following aspects:
153
154 Appendix A. Dynamics based nondestructive testing techniques
Velocity
response v(t)
Frequency
Response
(a) Structural vibration
|Hij| [dB]
and acoustics 50
pristine
F(t) 0 damaged
Excitation
0 500 1000
force Frequency [Hz]
I(t)
Mechanical
Electrical k
(b) Electro-mechanical
PZT
V(t) m
impedance
c
Actuation Structural
& sensing impedance
’Passive’
approach Loading
F(t)
(c) Acoustic emission
Damage
or Plate
impact
event Sensor
’Active’
approach
(d) Acousto-ultrasonics
Damage Plate
Actuator Sensor
Coupling
medium Transducer
Damage Structure
Figure A.1 Schematic representation of the most commonly used nondestructive testing techniques that
utilize structural dynamic principles to identify damage.
155
element that is attached to the structure (see Figure A.1(b) ). An excitation signal is
applied to the piezoelectric element. The applied voltage V (t) and the current I (t)
that flows through the element are measured simultaneously to determine the
mechanical impedance Zst . This approach is low cost and easy to apply, but is
generally lacking in the physical interpretation of the measured deviations [3].
requires access to the structure. Common formats to present the responses are
known as A-scan, B-scan and C-scan presentations [4]. Advantages of this approach
include the high sensitivity to structural discontinuities and the high accuracy
to determine the position, size and shape of subsurface defects. The required
accessibility of the surface and the time-consuming scanning process makes this
technique less suitable for health monitoring applications.
References
[1] C.-P. Fritzen and P. Kraemer. Self-diagnosis of smart structures based on dynamical
properties. Mechanical Systems and Signal Processing, 23(6):1830–1845, 2009.
[2] C. Liang, F.P. Sun, and C.A. Rogers. An impedance method for dynamic analysis of
active material systems. Journal of Intelligent Material Systems and Structures,
116(4):323–334, 1994.
[3] W. Yan and W.Q. Chen. Structural health monitoring using high-frequency
electromechanical impedance signatures. Advances in Civil Engineering, 2010:1–11, 2010.
[4] C. Hellier. Handbook of nondestructive evaluation. McGraw-Hill, 2003.
[5] T. Kundu, S. Das, and K.V. Jata. Detection of the point of impact on a stiffened plate by
the acoustic emission technique. Smart Materials and Structures, 18(3):035006, 2009.
[6] A. Raghavan and C.E.S. Cesnik. Review of guided-wave structural health monitoring.
The Shock and Vibration Digest, 39(2):91–114, 2007.
[7] R.P. Dalton, P. Cawley, and M.J.S. Lowe. The potential of guided waves for monitoring
large areas of metallic aircraft fuselage structure. Journal of Nondestructive Evaluation,
20(1), 2001.
Appendix B
Time domain
Time response / waveform
Statistical time series analysis
COR Correlation functions (non-parametric) Fassois 2007 [1]
AR Autoregressive models (parametric) Fassois 2007 [1]
Time-frequency analysis
WA Wavelet transform based methods Rucka 2006 [2]
HT/HHT Hilbert(-Huang) transform based methods Hu 2010 [3]
Frequency domain
Fourier / power spectra
Frequency response function / operational deflection shapes
FRFCH Frequency response function change Kessler 2002 [4]
FRFDI Frequency response function damage index Banerjee 2009 [5]
FRFSHP Frequency response function shape method Liu 2009 [6]
DRQ Detection and relative damage quantification Sampaio 2009 [7]
indicator
FRAC Frequency response assurance criterion Nefske 1996 [8]
FDAC Frequency domain assurance criterion Pascual 1997 [9]
157
158 Appendix B. Damage features and classifiers
Modal domain
Natural frequencies
CAC Cawley-Adams criteria Cawley 1979 [20]
DLAC Damage location assurance criteria Messina 1998 [21]
MDLAC Multiple damage location assurance criteria Messina 1998 [21]
Mode shape
MS Mode shape amplitude change Ho 2003 [13]
MSS Squared mode shape slope change Ho 2003 [13]
RMS Relative mode shape change Khoo 2004 [22]
MAC Modal assurance criteria Parloo 2003 [23]
IMAC Inverse modal assurance criteria Allemang 2003 [24]
COMAC Co-ordinate modal assurance criteria Parloo 2003 [23]
FD Fractal dimension method Hadjileontiadis 2005 [25]
GFD Generalized fractal dimension Wang 2007 [26]
MSRC Mode shape rotation change Abdo 2002 [27]
Mode shape curvature
MSC Mode shape curvature method Pandey 1991 [28]
MSC Mode shape curvature squared method Maia 2003 [13]
CDF Curvature damage factor method Wahab 1999 [29]
References 159
References
[1] S.D. Fassois and J.S. Sakellariou. Time-series methods for fault detection and
identification in vibrating structures. Philosophical Transactions of the Royal Society A:
Mathematical, Physical and Engineering Sciences, 365(1851):411–48, 2007.
[2] M. Rucka and K. Wilde. Application of continuous wavelet transform in vibration
based damage detection method for beams and plates. Journal of Sound and Vibration,
297(3-5):536–550, 2006.
[3] H.F. Hu, W.J. Staszewski, N.Q. Hu, R.B. Jenal, and G.J. Qin. Crack detection using
nonlinear acoustics and piezoceramic transducers – Instantaneous amplitude and
frequency analysis. Smart Materials and Structures, 19(6):065017, 2010.
[4] S.S. Kessler, S.M. Spearing, M.J. Atalla, C.E.S. Cesnik, and C. Soutis. Damage detection
in composite materials using frequency response methods. Composites Part B:
Engineering, 33(1):87–95, 2002.
[5] S. Banerjee, F. Ricci, E. Monaco, and A. Mal. A wave propagation and vibration-based
approach for damage identification in structural components. Journal of Sound and
Vibration, 322(1-2):167–183, 2009.
[6] X. Liu, N.A.J. Lieven, and P.J. Escamilla-Ambrosio. Frequency response function
shape-based methods for structural damage localisation. Mechanical Systems and Signal
Processing, 23(4):1243–1259, 2009.
[7] R. Sampaio and N. Maia. A Simple Correlation Factor as an Effective Tool for
Detecting Damage. In Modeling, Simulation and Control of Nonlinear Engineering
Dynamical Systems, pages 233–242. Springer, 2009.
160 Appendix B. Damage features and classifiers
[8] DJ Nefske and SH Sung. Correlation of a coarse-mesh finite element model using
structural system identification and a frequency response assurance criterion. In
Proceedings of 13th International Modal Analysis Conference, pages 597–602, 1996.
[9] R. Pascual, J.C. Golinval, and M. Razeto. A frequency domain correlation technique for
model correlation and updating. In Proceedings of the 15th International Modal Analysis
Conference (IMAC XV), pages 587–592, 1997.
[10] R. Pascual, J.C. Golinval, and M. Razeto. On-line damage assessment using operating
deflection shapes. In Proceedings of the 17th International Modal Analysis Conference
(IMAC XVII), pages 238–243, 1999.
[11] C. Zang, M.I. Friswell, and M. Imregun. Structural health monitoring and damage
assessment using frequency response correlation criteria. Engineering Mechanics,
133(September):981–993, 2007.
[12] R.P.C. Sampaio, N.M.M. Maia, and J.M.M. Silva. Damage detection using the
frequency-response-function curvature method. Journal of Sound and Vibration,
226(5):1029–1042, 1999.
[13] N. Maia, J.M.M. Silva, E.A.M. Almas, and R.P.C. Sampaio. Damage detection in
structures: from mode shape to frequency response function methods. Mechanical
Systems and Signal Processing, 17(3):489–498, 2003.
[14] G. Park, H. Sohn, C.R. Farrar, and D.J. Inman. Overview of piezoelectric
impedance-based health monitoring and path forward. The Shock and Vibration Digest,
35(6):451–463, 2003.
[15] V. Giurgiutiu and A. Zagrai. Damage detection in thin plates and aerospace structures
with the electro-mechanical impedance method. Structural Health Monitoring,
4(2):99–118, 2005.
[16] T.J. Johnson and D.E. Adams. Transmissibility as a differential indicator of structural
damage. Journal of Vibration and Acoustics, 124(4):634, 2002.
[17] M. Dilena and A. Morassi. The use of antiresonances for crack detection in beams.
Journal of Sound and Vibration, 276(1-2):195–214, 2004.
[18] N. Krohn, R. Stoessel, and G. Busse. Acoustic non-linearity for defect selective
imaging. Ultrasonics, 40(1-8):633–7, 2002.
[19] D.M. Donskoy and A.M. Sutin. Vibro-acoustic modulation nondestructive evaluation
technique. Journal of Intelligent Material Systems and Structures, 9(9):765–771, 1998.
[20] P. Cawley and R.D. Adams. The location of defects in structures from measurements of
natural frequencies. Journal of Strain Analysis, 14(2):49–57, 1979.
[21] A. Messina, E.J. Williams, and T. Contursi. Structural damage detection by a sensitivity
and statistical-based method. Journal of Sound and Vibration, 216(5):791–808, 1998.
[22] L.M. Khoo, P.R. Mantena, and P. Jadhav. Structural damage assessment using vibration
modal analysis. Structural Health Monitoring, 3(2):177–194, 2004.
[23] E. Parloo, P. Guillaume, and M. van Overmeire. Damage assessment using mode shape
sensitivities. Mechanical Systems and Signal Processing, 17(3):499–518, 2003.
[24] R.J. Allemang. The modal assurance criterion – Twenty years of use and abuse. Journal
of Sound and Vibration, 1:14–21, 2003.
[25] L.J. Hadjileontiadis, E. Douka, and A. Trochidis. Fractal dimension analysis for crack
identification in beam structures. Mechanical Systems and Signal Processing,
19(3):659–674, 2005.
[26] J. Wang and P. Qiao. Improved damage detection for beam-type structures using a
uniform load surface. Structural Health Monitoring, 6(2):99–110, 2007.
References 161
[27] M.A.-B. Abdo and M. Hori. A numerical study of structural damage detection using
changes in the rotation of mode shapes. Journal of Sound and vibration, 251(2):227–239,
2002.
[28] A.K. Pandey, M. Biswas, and M.M. Samman. Damage detection from changes in
curvature mode shapes. Journal of Sound and Vibration, 145(2):321–332, 1991.
[29] M.M. Abdel Wahab and G. De Roeck. Damage detection in bridges using modal
curvatures: application to a real damage scenario. Journal of Sound and Vibration,
226(2):217–235, 1999.
[30] C. Ratcliffe. Damage detection using a modified laplacian operator on mode shape
data. Journal of Sound and Vibration, 204(3):505–517, 1997.
[31] N. Stubbs and C.R. Farrar. Field verification of a nondestructive damage localization
and severity estimation algorithm. In Proceedings of the 13th International Modal Analysis
Conference (IMAC XIII), pages 210–218, 1995.
[32] Z. Shi and S.S. Law. Structural damage localization from modal strain energy change.
Journal of Sound and Vibration, 218(5):825–844, 1998.
[33] F.J. Guild and R.D. Adams. The detection of cracks in damaged composite materials.
Journal of Physics D: Applied Physics, 12(1):5–42, 1981.
[34] A.K. Pandey and M. Biswas. Damage detection in structures using changes in
flexibility. Journal of Sound and Vibration, 169(1):3–17, 1994.
[35] Q. Lu, G. Ren, and Y. Zhao. Multiple damage location with flexibility curvature and
relative frequency change for beam structures. Journal of Sound and Vibration,
253(5):1101–1114, 2002.
[36] Z. Zhang and A.E. Aktan. Application of modal flexibility and its derivatives in
structural identification. Research in Nondestructive Evaluation, 10(1):43–61, 1998.
[37] D. Wu and S.S. Law. Damage localization in plate structures from uniform load surface
curvature. Journal of Sound and Vibration, 276(1-2):227–244, 2004.
[38] S. Choi, S. Park, S. Yoon, and N. Stubbs. Nondestructive damage identification in plate
structures using changes in modal compliance. NDT & E International, 38(7):529–540,
2005.
[39] A. Deraemaeker and A. Preumont. Vibration based damage detection using large array
sensors and spatial filters. Mechanical Systems and Signal Processing, 20(7):1615–1630,
2006.
[40] N.A. Chrysochoidis and D.A. Saravanos. Assessing the effects of delamination on the
damped dynamic response of composite beams with piezoelectric actuators and
sensors. Smart Materials and Structures, 13:733–742, 2004.
Appendix C
Hilbert transform
The Hilbert transform H[ x (t)] of a real-valued function x (t) extending from −∞ <
t < ∞ is a real-valued function x̂ (t) defined by [1]:
Z +∞
1 x (τ )
H[ x (t)] = x̂ (t) = dτ . (C.1)
π −∞ t−τ
where Ainst (t) is the envelope signal and φ(t) the instantaneous phase signal. The
envelope Ainst (t) is extracted from the original time signal x (t) and the Hilbert
transformed signal x̂ (t) by:
q
Ainst (t) = x2 (t) + x̂2 (t) . (C.3)
x̂ (t)
φ(t) = arctan = ℑ ln z(t) , (C.4)
x (t)
and can be used to calculate the instantaneous frequency ωinst (t) [2]:
dφ(t) d 1 dz(t)
ωinst (t) = = ℑ ln z(t) = ℑ . (C.5)
dt dt z dt
References
[1] M. Feldman. Hilbert transform in vibration analysis. Mechanical Systems and Signal
Processing, 25(3):735–802, 2011.
[2] J.F. Claerbout. Earth soundings analysis: processing versus inversion. Blackwell Scientific
Publications, 2004.
163
Dankwoord
165
In the third year of my PhD research, I received the opportunity to conduct research
at the Nondestructive Evaluation and Sciences Branch (NESB) of NASA Langley
Research Center in Hampton, Virginia, USA. My special thanks goes to Matthew
Rogge, who provided me this opportunity. It was an honor to collaborate with you
for a period of approximately 7 months. I have learned a lot from your views and
approaches regarding the challenges I have faced during that time. Also the support
of Cara Leckey, James Ratcliffe and Wade Jackson was really helpful to make this
experience a success.
Het onderzoek dat gepresenteerd is in dit proefschrift was onmogelijk zonder de
onvoorwaardelijke steun van Fokker Aerostructures. Alle constructies die in dit
proefschrift zijn gebruikt, zijn door jullie geproduceerd en beschikbaar gesteld. In
het bijzonder wil ik Jaap Willem van Ingen bedanken. Zelfs als ik proefstukken
wilde gebruiken voor experimenten in de Verenigde Staten kon dat op korte termijn
gerealiseerd worden. Geweldig!
Frank Grooteman van het Nationaal Lucht- en Ruimtevaartlaboratorium (NLR) wil
ik bedanken voor de samenwerking op het gebied van met name optische sensor
technologie. De uitdagingen waar we voor gestaan heb ik als uitermate leerzaam
ervaren. Helaas is maar een klein deel van het werk in dit boekje terecht gekomen.
Hierbij wil ik tevens de onlangs gestarte promovendus Jason Hwang (tevens NLR)
veel succes wensen met het vervolgonderzoek.
Ik kan dit proefschrift niet afsluiten zonder alle (oud-)collega’s binnen de
vakgroepen Toegepaste Mechanica (TM) en Productie Technologie (PT) te bedanken.
De talloze discussies hebben bijgedragen aan de realisatie van dit werk. In het
bijzonder wil ik mijn kantoorgenoten Arnoud van der Stelt, Emiel Drenth en Shaojie
Liu bedanken. Ik heb veel baat gehad bij de inhoudelijke discussies. Ook onze
gesprekken over van alles en nog wat waren uitermate plezierig. Arnoud, wanneer
gaan we een broodjeszaak openen...?
De experimenten die ik heb verricht zouden waarschijnlijk een mislukking zijn
geworden zonder de technische ondersteuning van Bert Wolbert, Bert Vos, Gert-
Jan Nevenzel en Laura Vargas Llona. Wouter Grouve wil ik graag bedanken voor
het beschikbaar stellen van zijn thesis template. Tevens bedank ik afstudeerder
Matthijs Oomen (inmiddels promovendi) voor de prettige samenwerking. Speciale
dank gaat uit naar de secretaresses Debbie Zimmerman van Woesik en Belinda
Bruinink. Jullie zijn van onschatbare waarde om het reilen en zeilen binnen de
vakgroepen op rolletjes te laten verlopen.
Gedurende de eerste jaren van mijn promotieonderzoek heb ik het geluk gehad dat
ik mijn sportieve ambities nog redelijk heb kunnen nastreven. De balans tussen het
sporten enerzijds en het promoveren anderzijds heeft me in die tijd in veel opzichten
geholpen. De diverse hoogtepunten waren niet mogelijk geweest zonder de steun
van mijn teamgenoten. Aanvullend daarop wil ik de (schaats)trainingsgroep
bedanken voor de opbeurende woorden als ik weer eens een uurtje kwam ’harken’
op mijn inline-skates of schaatsen.
Tot slot bedank ik mijn familie en vrienden. Jullie waren nooit te beroerd om mijn
’geneuzel’ aan te horen op de momenten dat het nodig was.
Ted Ooijevaar
Publications
Journal articles
Book chapters
169
Conference proceedings
1. T.H. Ooijevaar, R. Loendersloot, M.D. Rogge, R. Akkerman and T. Tinga.
Vibro-acoustic modulation based damage identification in a composite skin-
stiffener structure. Submitted to: 7th European Workshop on Structural Health
Monitoring (EWSHM 2014), Nantes, France, 2014.
2. T.H. Ooijevaar, M.D. Rogge, R. Loendersloot, L.L. Warnet, R. Akkerman
and A. de Boer. Nonlinear dynamic behavior of impact damage in a
composite skin-stiffener structure. 9th International Workshop on Structural
Health Monitoring (IWSHM 2013), Stanford, CA, USA, 2013.
3. T.H. Ooijevaar, L.L. Warnet, R. Loendersloot, R. Akkerman, A. de Boer.
Vibration based damage identification in a composite T-beam utilising low
cost integrated actuators and sensors. 6th European Workshop on Structural
Health Monitoring (EWSHM 2012), Dresden, Germany, 2012.
4. R. Loendersloot, T.H. Ooijevaar, A. de Boer, R. Akkerman. Development of
a damage quantification model for composite skin-stiffener structures. 5th
ECCOMAS Thematic Conference on Smart Structures and Materials (SMART’11),
Saarbrucken, Germany, 2011.
5. T.H. Ooijevaar, R. Loendersloot, L.L. Warnet, R. Akkerman, A. de Boer.
Structural health monitoring of an advanced composite aircraft structure
using a modal approach. 8th International Workshop on Structural Health
Monitoring (IWSHM 2011), Stanford, CA, USA, 2011.
6. T.H. Ooijevaar, F.P. Grooteman, L.L. Warnet, R. Loendersloot, R. Akkerman,
A. de Boer. Dynamic characterisation of a damaged composite structure
with stiffeners employing fibre Bragg gratings. 5th International Conference
on Composites Testing and Model Simulation (CompTest 2011), Lausanne,
Switzerland, 2011.
7. R. Loendersloot, T.H. Ooijevaar, L.L. Warnet, A. de Boer, R. Akkerman.
Vibration based structural health monitoring of a composite plate with
stiffeners. International Conference on Noise and Vibration Engineering, ISMA2010,
Leuven, Belgium, 2010.
8. T.H. Ooijevaar, L.L. Warnet, R. Loendersloot, R. Akkerman, A. de Boer.
Vibration based structural health monitoring of a composite plate structure
with multiple stiffeners. 5th European Workshop on Structural Health Monitoring
(EWSHM 2010), Sorrento, Italy, 2010.
9. R. Loendersloot, T.H. Ooijevaar, L.L. Warnet, R. Akkerman, A. de Boer.
Vibration based structural health monitoring in fibre reinforced composites
employing the modal strain energy method. 3rd International Conference on
Integrity, Reliability and Failure (IRF2009), Porto, Portugal, 2009.
10. T.H. Ooijevaar, R. Loendersloot, L.L. Warnet, A. de Boer, R. Akkerman.
Vibration based structural health monitoring in fibre reinforced composites
employing the modal strain energy method. 15th International Conference on
Composite Structures (ICCS15), Porto, Portugal, 2009.
ISBN: 978-90-365-3624-0