Previous work has reported how a phased array can be used in two operating modes to detect nonlinear features. Specifically firing each element individually, as in a classic full matrix capture, and firing all elements at once with an appropriate delay to form a physical beamform, as in classic beamforming. If the difference in the energy of these two approaches is compared then nonlinearity in the material can be measured.
This approach has proven effective, but greater sensitivity is desired. This paper studies how the phase of the received signal can be used to characterise nonlinearity. The underlying approach is first investigated and it is shown that phase information is preserved in the diffuse ultrasonic field. Then analytical and numerical models are presented to show that a tightly closed crack in a part should lead to an offset in the measured phase. Experimental results are then presented to validate these conclusions and the benefits of such an approach discussed.
The development of low-cost bonded assembly of composite aerospace structures ideally requires an NDE method to detect the presence of poor quality, weak bonds or kissing bonds. Such interfaces can introduce nonlinearity as a result of contact nonlinearity where an ultrasonic wave is distorted when it interacts with the interface. In general, the nonlinear elastic behaviour of these interfaces will generate harmonics but they can be lost among the harmonics generated by other nonlinearities present in the experimental system. The technique developed in this research is a non-collinear method; this involves the interaction of two ultrasonic beams, and it allows the removal of virtually all system nonlinearity except for that produced in the region where the two beams overlap. The frequencies of the two beams and the angle between are varied during the experiment. By measuring the nonlinear mixing response as these two parameters are swept through a ‘fingerprint’ of the nonlinear properties in the interaction region can be obtained. This fingerprint has been shown to contain information about the bulk material and the interface status. Work is ongoing to understand which features in the fingerprints reliably correlate with particular material or interface properties. To build this understanding a greatly simplified kissing bond, a compression loaded aluminium-aluminium interface, has been tested. Modelling of the nonlinear behaviour of the aluminium interface has also been conducted.
This study demonstrates the capability of inductively coupled piezoelectric sensors to monitor the state of health throughout the lifetime of composite structures. A single sensor which generated guided elastic waves was embedded into the stacking sequence of a large glass fiber reinforced plastic plate. The progress of cure was monitored by measuring variations in the amplitude and velocity of the waveforms reflected from the plate’s edges. Baseline subtraction techniques were then implemented to detect barely visible impact damage (BVID) created by a 10 Joule impact, at a distance of 350 mm from the sensor embedded in the cured plate. To investigate the influence of mechanical loading on sensor performance, a single sensor was embedded within a glass fiber panel and subjected to tensile load. The panel was loaded up to a maximum strain of 1%, in increments of 0.1% strain. Guided wave measurements were recorded by the embedded sensor before testing, when the panel was under load, and after testing. The ultrasonic measurements showed a strong dependence on the applied load. Upon removal of the mechanical load the guided wave measurements returned to their original values recorded before testing. The results in this work show that embedded piezoelectric sensors can be used to monitor the state of health throughout the life-cycle of composite parts, even when subjected to relatively large strains. However the influence of load on guided wave measurements has implications for online monitoring using embedded piezoelectric transducers.
A technique is presented for imaging acoustic nonlinearity within a specimen using ultrasonic phased arrays. Acoustic nonlinearity is measured by evaluating the difference in energy of the transmission bandwidth within the diffuse field produced through different focusing modes. The two different modes being classical beam forming, where delays are applied to different element of a phased array to physically focus the energy at a single location (parallel firing) and focusing in post processing, whereby one element at a time is fired and a focused image produced in post processing (sequential firing). Although these two approaches are linearly equivalent the difference in physical displacement within the specimen leads to differences in nonlinear effects. These differences are localized to the areas where the amplitude is different, essentially confining the differences to the focal point. Direct measurement at the focal point are however difficult to make. In order to measure this the diffuse field is used. It is a statistical property of the diffuse field that it represents the total energy in the system. If the energy in the diffuse field for both the sequential and parallel firing case is measured then the difference between these, within the input signal bandwidth, is largely due to differences at the focal spot. This difference therefore gives a localized measurement of where energy is moving out of the transmission bandwidth due to nonlinear effects. This technique is used to image fatigue cracks and other damage types undetectable with conventional linear ultrasonic measurements.
The layer wise construction of laminated composites offers the potential to embed sensors within composite structures.
One possible solution is the embedding of sensors that are inductively coupled to an external probe; which allows for the
efficient contactless transfer of electrical signals to the sensor. Embedding sensors within structures is an attractive
option, due to the physical protection offered to the sensor by the host structure. However, for embedding sensors to be
viable, sensor integration must result in minimal degradation of the laminates mechanical performance. This work
focuses on designing embedded inductively coupled sensors for structural performance. A suitable sensor coating for the
sensor unit was identified using interlaminar shear strength testing. Sensors were then embedded into quasi-isotropic
four-point bend flexural strength specimens, and different embedding strategies demonstrated. In addition to providing
the sensor with physical protection, embedding sensors within a composite host offers the additional benefit of
monitoring the curing process of the surrounding composite. A single inductively coupled sensor was embedded into a
large glass fiber epoxy plate, and the measured guided wave pulse echo response used to monitor the curing process.
This novel cure monitoring technique was then benchmarked against direct scanning calorimetry.
The use of Lamb waves in Non-Destructive Evaluation (NDE) and Structural Health Monitoring (SHM) is gaining popularity due to their ability to travel long distances without significant attenuation, therefore offering large area inspections with a small number of sensors. The design of a Lamb-wave-based NDE/SHM system for composite materials is more complicated than for metallic materials due to the directional dependence of Lamb wave propagation characteristics such as dispersion and group velocity. Propagation parameters can be theoretically predicted from known material properties, specifically the stiffness matrix and density. However, in practice it is difficult to obtain the stiffness matrix of a particular material or structure with high accuracy, hence introducing errors in theoretical predictions and inaccuracies in the resulting propagation parameters. Measured Lamb wave phase velocities can be used to infer the stiffness matrix, but the measurements are limited to the principal directions due to the steering effect (different propagation directions of phase and corresponding group velocities). This paper proposes determination of the stiffness matrix from the measured group velocities, which can be unambiguously measured in any direction. A highly anisotropic carbon-fibre-reinforced polymer plate is chosen for the study. The influence of different stiffness matrix elements on the directional group velocity profile is investigated. Thermodynamic Simulated Annealing (TSA) is used as a tool for inverse, multi variable inference of the stiffness matrix. A good estimation is achieved for particular matrix elements.
The laminated construction of composite offers the possibility of permanently embedding sensors into structure,
for example, ultrasonic transducers which can be used for NDE applications. An attractive and simple solution
for probing embedded sensors wirelessly is via inductive coupling. However, before this can be achieved it is
necessary to have a full understanding and proper design strategy for the inductively coupled system. This
paper presents the developments of both system design procedure and a computer program for one dimensional
inductively coupled transducer system mounted on a solid substrate. The design strategy in this paper mainly
focuses on issues of localization of transducers, and optimizing the signal to noise level. Starting from a three
coil equivalent circuit, this paper also explains how the measured impedance of a bonded piezoelectric disc
is implemented into the system model representing a transducer bonded to an arbitrary solid substrate. The
computer programme using this model provides immediate predictions of electrical input impedance, acoustic
response and pulse-echo response. A series of experiments and calculations have been performed in order to
validate the model. This has enabled the degree of accuracy required for various parameters within the model,
such as mutual inductance between the coils and self-inductance of coils, to be assessed. Once validated, the
model can be used as a tool to predict the effect of physical parameters, such as distance, lateral misalignment
between the coils, and the coil geometry on the performance of an inductively coupled system.
Presented is an approach to damage localization for guided wave structural health monitoring (GWSHM) in plate-like
structures. In this mode of SHM, transducers excite and sense guided waves in order to detect and characterize the
presence of damage. The premise of the presented localization approach is simple: use as the estimated damage location
the point on the structure with the maximum a posteriori probability (MAP) of being the location of damage (i.e., the
most probable location given a set of sensor measurements). This is accomplished by constructing a minimally-informed
statistical model of the GWSHM process. Parameters of the model which are unknown, such as scattered wave
amplitude, are assigned non-informative Bayesian prior distributions and averaged out of the a posteriori probability
calculation. Using an ensemble of measurements from an instrumented plate with stiffening stringers, the performance of
the MAP estimate is compared to that of what were found to be the two most effective previously reported algorithms.
The MAP estimate proved superior in nearly all test cases and was particularly effective in localizing damage using very
sparse arrays of as few as three transducers.
If sparse arrays are attached to structures for the purposes of structural health monitoring it is likely that there will be
variation in the placement of the sensors, resulting in deviation from the assumed locations. In addition, poor knowledge
of the material through which the signals are propagating can result in the use of incorrect velocities, or failing to take
account of delays inherent in the equipment. These deviations will result in reduced performance in terms of defect
detectability and characterisation. This paper outlines an autofocus approach whereby the transducer locations and
material properties can be estimated from the experimental data to ensure the highest levels of defect detectability. The approach is validated using both models and a more complex real world structure. The performance of the approach is considered across a range of potential operating conditions to demonstrate its robustness. Finally limitations and potential solutions to these are addressed.
Conventional ultrasonic NDT techniques are limited in their ability to detect small defects by the diffraction limit, that is
there is much reduced sensitivity to defects smaller than the wavelength of the interrogating ultrasonic wave. While not a
major issue for most inspection, this problem becomes particularly significant for the detection of fatigue damage prior
to crack formation. In this regime conventional NDT has proven to be inadequate. For this reason significant effort has
been expended on the development of non-linear techniques. These techniques rely on deviations of the material from
linear stress strain behaviour which create harmonics in the resulting frequency response. Evidence suggests that changes
to a materials condition, such as fatigue damage, change this non-linear response. This paper presents a non-linear
inspection method using a non-collinear interaction. This technique has several advantages over other harmonic approaches in that there is spatial separation, modal separation and frequency separation of the non-linear signal. This allows the origin of the non-linear signal and underlying noise levels to be well defined. The capability of the technique is demonstrated using plastically strained material and samples subjected to low cycle fatigue.
It has been shown that guided waves can be used with sparse arrays of permanently attached sensors to detect the
presence of damage in structures. When applied with temperature compensation strategies complex structures can be
inspected over time and in the presence of varying conditions. Current analysis suggests a series of relationships for
individual sensor pairs but is difficult to expand to predict the signal to noise performance of a real world large network
of sensors. The result of this is that it is unclear as to what is the best sensor layout to detect damage. This paper
quantitatively and qualitatively investigates the performance of different sensor geometries to determine the signal to
noise ratio of different configurations. It is shown that using more than two sensors not only offers the ability to localize
damage but also produces enhanced signal to noise ratio over a single pair of transducers. It is shown that there is no
single optimum sensor layout, with the optimum layout dependant on the type of damage that is to be detected. However
a network of squares or hexagons offers excellent performance.
The use of permanently attached arrays of sensors has made it clear that guided waves can be used for the
SHM of structures. The approaches developed have relied on the use of reference signal subtraction to indicate
changes to the state of the structure, such as the appearance of damage. The limit of performance of any system
is defined by the post subtraction noise.
In order to confirm the basic principles at work the majority of this work has been carried out on simple
metallic plates. While important to confirm the levels of understanding, this is not sufficient for practical use.
This paper looks at the application of SHM techniques in more complex structures, more typical of those any
system would be used on in practise.
A rib from a BaE 146 aircraft is used to demonstrate the practical difficulties of applying guided wave SHM
methods to densely featured structures.
A model system comprising a plate with a single stringer is used to demonstrate a method for normalizing
signals to give responses directly related to the scattering properties of the change in the system, mitigating the
effect of the position of the change, and a method is proposed to generalize the approach to complex systems.
Preliminary tests in the region of the stringer are used to identify the experimental challenges to realizing the
calibration on complex systems.
The level of post-subtraction noise due to benign structural features limits the sensitivity that guided wave structural
health monitoring systems can achieve. Subtraction of reference signals without compensation leads to unacceptably
high post-subtraction noise in the presence of modest environmental changes, and in particular temperature. Hence some
form of compensation is necessary. In this paper, various compensation strategies are investigated and their performance
quantified. Factors such as the length of time-window considered, sensor variations and inhomogeneous temperature
variations are also addressed. It is concluded that the best performance that can currently be achieved is by (a) obtaining
the best matched signal from an ensemble of multiple reference signals recorded at different temperatures and (b) fine
tuning this signal by numerically stretching or compressing it.
It has been shown by many researchers that guided wave structural health monitoring is capable of detecting the presence
of damage in a structure. The requirements for grid spacing and sensitivity to temperature change have been established
and can be used to specify an array with a given signal to noise ratio. What is not clear at this point is how, given that
damage is detectable, its location should be found. This paper discusses two different imaging algorithms and
investigates the relative merits of each. This is initially done on the smallest possible array of three transducers. This is
then carried forward to larger sparse arrays to show how a larger structure with a distributed sensor network can be
imaged with several "units" of transducers working together. It is shown that in general using more transducers is
beneficial to the quality of imaging achieved. However it is still necessary to perform imaging using smaller arrays to
ensure that in the event of multiple damage sites occurring simultaneously each can be detected.
Changes in environmental conditions, and in particular temperature, limit the sensitivity of guided wave structural health
monitoring (SHM) systems that use reference signal subtraction. The limitation on sensitivity is the size of the residual
signal left after reference signal subtraction that arises from imperfect subtraction of the signals from benign structural
features. The sensitivity can be improved by decreasing the spacing between sensors but the effect of temperature is so
strong that it is doubtful whether the resulting SHM system is economically viable. This provides the motivation for
searching for alternative strategies to improve sensitivity. One possibility is to record an ensemble of reference signals
over a range of temperatures and then use the signal in the ensemble that best matches a subsequent signal for
subtraction. Experimental results show that this provides an improvement in sensitivity of around 35 dB. It does however
require a large database of signals and there is the potential concern that the subtraction of the best match signal may
somehow also remove a genuine signal from damage. Another possibility is signal processing to improve sensitivity. A
uniform temperature change to a structure results in a change in wave velocity and a dilation of the structure itself. The
net effect is a dilation of the arrival times of each wave-packet in a guided wave signal. An obvious strategy to
compensate for this effect is to apply the inverse dilation to the time-axis. However, this does not compensate for the
effect exactly since the temperature change does not dilate individual wave-packets. An alternative and exact
compensation scheme is presented and its practical application is discussed.
Proc. SPIE. 6532, Health Monitoring of Structural and Biological Systems 2007
KEYWORDS: Signal to noise ratio, Waveguides, Sensors, Nondestructive evaluation, Interference (communication), Wave propagation, Structural health monitoring, Aluminum, Signal detection, Temperature metrology
Deployable guided wave systems are commercially used for the inspection of long lengths of pipelines in non-destructive
testing (NDT) applications. On this basis it might seem that guided waves could be used in a structural health monitoring
system (SHM) that is able to detect damage anywhere in a structure with a relatively sparse array of permanently
attached sensors. Furthermore, while guided wave NDT is limited to simple structures because of the problem of signal
interpretation, reference signal subtraction can be applied to guided wave SHM hence apparently solving the problem of
structural complexity. Despite this, and considerable international research effort, there have been no serious commercial
applications of guided wave SHM. In this paper, the concept of guided wave propagation and reference signal
subtraction are examined at a fundamental level to analytically estimate the sensitivity of the reference signal subtraction
approach. It is argued that the limitation on sensitivity is the size of the residual signal left after baseline signal
subtraction. The subtraction is never perfect due to environmental changes and results in imperfect cancellation of the
signals from benign structural features, such as welds, edges, flanges etc. It is shown that the sensitivity decreases with
propagation distance and therefore sensor spacing. Examples of the required sensor pitch to detect a 6mm hole in a 3mm
thick aluminium plate subjected to a 1°C temperature change are given, and show the significant detrimental effect that
even small temperature changes can have. It is shown that a significant improvement (typically 20 dB) is possible if
signal envelopes rather than RF signals are subtracted but that this leads to the problem of sensitivity functions that vary
non-monotonically and which may even include blind spots.