Remote inspection system for impact damage in largecomposite structure

By C. H. Zhong*, A. J. Croxford and P. D. Wilcox

Department of Mechanical Engineering,University of Bristol,Queens Building,

University Walk, Bristol, BS8 1TR, UK

ABSTRACT

A significant opportunity for reducing the weight of composite aircraft is through the development of an

economically-efficient method to detect barely-visible or invisible impact damage sustained in service. In this

paper, a structurally-integrated, inert, wireless system for rapid, large-area impact damage detection in compos-

ite is demonstrated. Large-area inspection from single sensors using ultrasonic guided waves is achieved with

a baseline-subtraction technique. The wireless interface uses electromagnetic coupling between coils in the em-

bedded sensor and inspection wand. Compact encapsulated sensor units are built and successfully embedded

into composite panels at manufacture. Chirp-based excitation is used to enable single-shot measurements with

high signal-to-random-noise ratio to be obtained. Signal processing to compensate for variability in inspection

wand alignment is developed and shown to be necessary to obtain adequate baseline subtraction performance

for damage detection. Results from sensors embedded in both glass fibre and carbon fibre reinforced composite

panels are presented. Successful detection of a 10 J impact damage in the former is demonstrated at a range

of 125 mm. Quantitative extrapolation of this result suggests that the same level of impact damage would be

detectable at a range of up to 1000 mm with an inspection wand alignment tolerance of 4 mm.

Keywords: Wireless, Embedded sensor, Composites, NDE

1. INTRODUCTION

Composite materials are susceptible to barely visible impact damage (BVID) caused by low-velocity impacts

(energy level of 10-30 J) such as tool drops, bird strikes and hailstones.1,2 By definition, BVID is hard to

detect through routine visual inspection. Although ultrasonic bulk wave and thermal imaging non-destructive

testing (NDT) techniques are able to detect BVID, neither are ideal for performing rapid in-service inspections.

For example, ultrasonic inspection requires water coupling and the ultrasonic transducer must be scanned over

* Author for correspondence C.H.Zhong (E-mail: chenghuan.zhong@bristol.ac.uk)

www.ndt.net/?id=17809Vol.20 No.5 (May 2015) - The e-Journal of Nondestructive Testing - ISSN 1435-4934

the complete surface to be inspected. Thermography is an inherently large-area technique, but due to the low

thermal diffusivity of composite, thermographic methods are primarily limited to near-surface damage.3

The current approach adopted by most of the aerospace industry is to over-design composite structures so

that they can tolerate the existence of BVID4 (more serious impact damage is assumed to be detectable during

routine visual inspections between flights). Consequently, there is a real opportunity for reducing the weight of

composite aircraft by developing an efficient method of detecting BVID between flights. The potential weight

reductions would also help to meet the new environmental targets set for the aerospace industry.5 Conceptually,

a fully-integrated structural health monitoring (SHM) system using permanently-installed sensors is the ideal

solution to provide rapid BVID detection. Such a system could either detect impacts directly (e.g. via acoustic

emission) or the damage resulting from an impact.6–9 Such SHM systems have been widely demonstrated in the

laboratory and in some cases trialled on actual aircraft. However, there are two major challenges preventing the

widespread deployment of such SHM systems in the aerospace industry.

The first challenge is the fundamental problem of making the necessary physical measurements with suitable

levels of reliability, sensitivity and selectivity. For large-area monitoring with a limited number of sensors,

ultrasonic guided waves provide arguably the only generally-applicable physical measurement mechanism. In a

structure with any degree of complexity, some form of baseline signal comparison approach becomes necessary

when guided waves are used in order to discriminate between benign structural features and damage. Significant

progress continues to be made in this area,10–12 with increasingly robust detection sensitivity being demonstrated

even in the presence of environmental variability.

The second challenge is the more practical problem of data and power connectivity to a distributed sensor

network in a large aerospace structure. The easiest connectivity solution is to use wires and a number of strategies

for ultrasonic composite SHM based on wired transducers have been investigated.13,14 The main limitation of a

wired system is that a large number of conductive routes and transducer control units such as multiplexers need

to be integrated into the structure as well. For a large structure, the amount of wiring and electronics not only

increases weight, manufacturing and installation costs,15 but potentially decreases the reliability of the system,

due to the possibility of failures in the wiring. A widely-researched alternative is to use wireless RF protocols (e.g.

wifi, ZigBee) for communication,16,17 but the major challenge for wireless systems is power consumption. If the

power source is a battery or other energy storage device then when the power source is exhausted, replacement

presents a significant problem if the sensors are embedded or integrated into a structure such as a composite

aircraft.15 An alternative solution is to use power-harvesting technology such as photovoltaic, thermoelectric,

piezoelectric devices and rectenna arrays for the wireless unit. However, in order to provide enough power for

sensing electronics and data communication, current power-harvesting requires complex, bulky devices.18

In this paper, a rapid, impact-damage detection solution is proposed that falls between conventional NDT

and fully-integrated SHM. Rather than attempting to embed sensors, power sources and data communication

into the structure, only the sensing elements are embedded and measurements are made using an external

device. The advantage of this approach over a fully-integrated SHM system is that the sensors are inert when

not in use and no energy storage or long-range data transfer need be considered. The potential advantages over

conventional NDT (where the sensors as well as instrumentation are separate to the structure) is (a) the ease and

speed with which measurements can be made and (b) the fact that sensors remain at precisely fixed positions

in the structure, enabling baseline signal comparison techniques to be employed if required. The concept of the

proposed system for impact detection in a composite structure is shown in Fig. 1(a). A distributed array of

inert sensors is permanently embedded in the composite structure and they are probed using an external device

termed the inspection wand. The inspection wand could be deployed either manually, as shown in the figure, or

using a platform such as a unmanned aerial vehicle. When the inspection wand is in close physical proximity to

a sensor, the sensor is activated and a measurement is made. A key attraction of the system is that it enables

quantitative measurements to be rapidly obtained from sensors at precisely-defined locations using an inspection

wand that does not have to be precisely positioned or even physically coupled to the structure.

An inductively-coupled transducer system (ICTS) was initially proposed by Greve et al.,19,20 and inductive

coupling is employed in the current paper to realize the concept shown in Fig. 1(a). The same technique has

also been used by Wu et al. for through-thickness inspection with flexible transducers.21,22 The proposed ICTS

is shown in Fig. 1(b) and consists of three coils. One coil (termed the transducer coil) is physically connected

to a piezoelectric transducer and forms a sensor unit, which is embedded in the structure. The other two coils

(termed the transmitting and receiving coils) are in the separate inspection wand, where they are connected

to the output and input channels of ultrasonic instrumentation. Importantly, a completely-embedded sensor

benefits from the physical protection provided by the host structure and also, if manufactured correctly, exhibits

superior ultrasonic coupling to the structure compared to a surface-mounted sensor.

Previously, a model of an ICTS parameterised by physical constants, such as the number of turns and

diameter of each coil, was developed and validated.23 In the current paper, this electromagnetic coupling model

was used to develop the optimised design for guided wave detection of impact damage described in Section 2.2.

A process to practically embed an optimised sensor into composite is demonstrated in Section 2.3. In order to

enable single-shot measurements with high signal-to-random-noise ratio and to compensate for the effect of wand

misalignment, the linear shift between the axes of the coils, several signal processing techniques are described in

Compositestructure

Embeddedsensors

Embeddedsensor

Inspectionwand

InspectionwandTransmitter

coil

Receivercoil Piezoelectric

transducer

Transducercoil

(a)Fromsignal

generator

Todigital

oscilloscope

Inductive

coupling

(b)

Figure 1. (a) Concept of the inductively-coupled transducer system (ICTS) and (b) detail of the inductive coupling

between inspection wand and embedded sensor.

Section 2.4. The consistency of embedded sensors, the performance of misalignment compensation methods and

finally the use of the system to detect impact damage are reported in Section 3.

2. SYSTEM DESIGN

2.1 Methodology and Objectives

The inspection target for the system described in the current paper is to detect the damage from a 10 J low-

velocity impact in a composite panel at distance of up to 1000 mm from a sensor location.1,2 Because the

intended use of the proposed system is for rapid routine manual inspection, it was decided that the inspection

target should be achievable in a single-shot measurement with a realistic degree of inspection wand misalignment

relative to the sensor.

Guided elastic waves (Lamb waves) have been widely acknowledged as one of the most promising tools for

long range detection of various types of defect in different materials and structures,24 and deployable guided

wave systems have been successfully used for industrial NDT inspections of simple structures such as pipes25–27

and rails.28,29 Because a guided wave sensor can potentially detect damage over a significant surrounding

area, a permanent sparse array of guided wave sensors in pitch-catch mode is an ideal candidate for large-area

SHM. In the current paper, a sparse array of guided wave sensors, each operating independently in pulse-echo

mode, is used as the basis of the composite impact damage detection system. The damage in the structure

is detected and localised by identifying the appearance of new reflected signals in the pulse-echo mode instead

of new diffracted signals in pitch-catch mode. The pulse-echo configuration is more practical for use in this

particular application, requiring only a single inspection wand to be aligned with a single sensor to perform a

measurement. To achieve a reasonable inspection range and sensitivity to damage, the fundamental extensional

(S0) mode in the 0.35-0.5 MHz mm frequency-thickness product range is used. The motivation for this is that at

higher frequency-thickness products, higher-order modes exist making signals harder to interpret, while at lower

frequency-thickness products, the sensitivity to localised defects is diminished. The attenuation of the S0 mode

due to material damping is also significantly less than the fundamental flexural (A0) mode30 and the S0 mode

is also less susceptible to attenuation due to energy leakage if the structure has liquid-loading (as is the case in,

for example, wet-wing designs). Due to the complexity of aerospace composite structure, it is assumed that the

system will operate in a baseline subtraction mode to suppress reflections from benign structural features, such

edges and stiffeners.31,32

The embedded sensor unit should be of minimal size, especially thickness, to reduce its impact on structural

integrity. For the reasons outlined above, it is necessary to use the S0 Lamb wave mode, which requires a

transducer that is sensitive to in-plane motion. As complete structural coverage in terms of defect detection

is desired, the transducers should have omni-directional sensitivity to this mode. Therefore thin, disk-shaped

piezoelectric transducers operating primarily in an axi-symmetric radial mode are used (specifically 16 mm

diameter, 0.3 mm thick devices manufactured from NCE51 piezoelectric material supplied by Noliac Group,

Kvistgaard, Denmark).

It is worth noting that the electromagnetic coupling and ultrasonic measurement are considered as indepen-

dent challenges in this paper. This allows well-understood electromagnetic modelling33 and ultrasonic guided

wave SHM experience34 to be adopted separately in the design process.

2.2 Electromagnetic design

Fig. 2(a) shows a typical ICTS pulse-echo time-domain response.35 Cross-talk due to direct coupling between

the transmitting and receiving coils gives rise to a dead-zone in the coverage provided by a sensor unit, as the

cross-talk obscures signals from nearby reflectors, as indicated in Fig. 2(b). Inductively-coupled systems are

susceptible to incoherent electromagnetic interference (EMI) from other circuits causing incoherent or random

noise. The signal prior to the cross-talk in Fig. 2(a) gives an indication of the level of such noise. Obtaining an

adequate signal-to-random-noise ratio, SR, is an important consideration for the design of the ICTS. However,

simply designing to maximise the amplitude of the received signal will result in a resonant circuit, which is how

most inductive power transfer systems are designed.36,37 In fact, signal fidelity is of equal if not greater concern

in the design of ICTS than SR. Resonance is manifested as long signal ring-down which has several undesirable

effects, one of which is poor temporal resolution. More importantly, resonance increases the duration of cross-

talk and consequently the size of the dead-zone around each sensor. To quantify the signal fidelity of an ICTS

response from a sensor on a particular structure, the signal-to-cross-talk ratio, SX , of a received signal, v(t), is

defined as:

SX =|H {v(t)}|t=t1

1δ

∫ 2δδ|H {v(t)}|dt

(1)

where H {.} indicates a Hilbert transform, t1 is the arrival time of a reference reflection signal, and δ is the

duration of the input signal used to drive the ICTS whose peak is assumed to start at t = 0. The numerator in

this expression is signal amplitude and the denominator is the mean cross-talk signal amplitude in the period,

yellow region in Fig. 2, after the input signal has finished. The latter would be theoretically zero if the ICTS

had no resonance and is consequently a measure of signal fidelity.

In practice, it has been found that an ICTS designed to maximise SX also achieves an acceptable SR (i.e.

one that is not a limit on system performance). Therefore, SR is not considered in the electromagnetic design

process, but will be discussed further in the context of signal processing.

As noted, the electromagnetic design and ultrasonic measurements are considered as independent challenges

in this paper. The objective of the electromagnetic design here is therefore to optimise the electromagnetic

Figure 2. (a) Typical pulse-echo response from ICTS and (b) inspection coverage of single ICTS sensor. Yellow shading

is (a) refer to a δ to 2δ region

coupling aspect of an ICTS for a specified piezoelectric transducer and ultrasonic instrument. In other words,

the electromagnetic optimisation is performed on the basis that the specified piezoelectric transducer would

yield a suitable ultrasonic signal if it were physically wired to an appropriate ultrasonic pulser-receiver. Here,

a combined digital-oscilloscope signal-generator unit (HS3, TiePie Engineering, Sneek, Netherlands) is used as

the ultrasonic instrumentation. An optimisation procedure is applied to electromagnetic design to maximize

SX for the ICTS for a specified input signal (a 5 cycle Gaussian windowed toneburst at a centre frequency of

160 kHz). The predefined constants of the system such as output impedance of signal generator, impedance of

digital oscilloscope and manufacturing limitations are listed in Table. 1(a) along with system information such

as thickness of PCB substrate and input signal.

In order to reduce the physical size of the embedded sensor, the secondary coil in the sensor is a double-sided

planar coil fabricated on a printed circuit board (PCB) that surrounds the piezoelectric disc. Therefore, the

inner diameter of this coil is constrained to the diameter of the piezoelectric disc plus 4 mm radial clearance for

connection soldering, thus giving a minimum inner coil diameter of 24 mm. Also, in order to decrease the size

of all coils and the number of design parameters, the PCB manufacturing resolution, k, is used to define both

the PCB track width as well as the distance between the two neighbouring tracks. Because the transmitting and

receiving coils are built into a opposite side of a single PCB, these coils are single sided.

The specifications of the optimised three coil design based on the constraints given before are listed in

Table. 1(b). Fig. 3 shows SR, SX of the optimised system at different reading distances. It can be seen that

although SX does decrease with reading distance, the decrease is fairly modest comparing to SR. SR drops

exponentially and approach to zero as reading distance increased. For this optimized system, the maximum

Table 1. (a) List of predefined constants for electromagnetic design and (b) specifications of optimized coils after electro-

magnetic design.

reading distance is 40 mm, where signal-to-random-noise ratio is equal to 2 (6 dB). It worth noting that the

experimentally measured noise level was used to calculate SR.

2.3 Embedded sensor design

Fig. 4(a) shows the process developed to manufacture an encapsulated sensor unit for embedding in a composite

structure. First, the piezoelectric disc is placed in a circular cut-out in the centre of a double-sided PCB containing

the secondary coil with dimensions determined from the optimisation procedure (24 mm inner diameter, 38.4 mm

outer diameter, double-sided with 19 turns on each side). Then, the assembled unit is cured with 0.075 mm

thick insulation layers (Pyralux LF, Dupont, Wilmington, USA) on both sides. The insulation layers consist of

polyimide film coated on both sides with acrylic adhesive. The result is an encapsulated sensor unit in the form

of a 0.45 mm thick, 55 mm diameter disc, which is ready for embedding into a composite material.

Sensor units were embedded into a glass fibre composite panel (Hexcel 950, Hexcel, Stamford, CT, USA)

cured at 125 ◦C and a carbon fibre composite panel (Hexcel M21, Hexcel, Stamford, CT, USA) cured at 180 ◦C

with 7 bar pressure applied. The glass fibre and carbon fibre composite panel lay-up sequences are shown in

Figs. 4(b) and (c) respectively and the finished panels are shown in Fig. 5. The glass fibre panel is used for impact

damage detection tests and the carbon fibre panel for sensor performance studies. The embedded ICTS can be

visually identified in the glass fibre panel, but their presence in the carbon fibre material is only detectable via

a slight thickening on the non-tool side. In the carbon fibre plate there are three ICTSs and one wired unit for

reference. The effect of embedding on the strength of composite structure is open research topics, but has been

0 10 20 30 40 500

50

100

150

200

250

Reading distance (mm)

SR

0 10 20 30 40 500

1

2

3

4

5

6

7

Reading distance (mm)

SX

(b)(a)

Figure 3. (a) Signal-to-random-noise ratio, SR, and (b) Signal-to-cross-talk ratio, SX , of the optimized system given in

Table 1(b) as a function of the reading distance in linear scale. Experimentally measured random noise level was used

to calculated SR. The dashed lines in (a) and (b) refer to optimum reading distance where maximum SX is achieved;

the dash-dotted lines in (a) and (b) refer to a reading distance where the signal is 2 times (6dB) higher than the random

noise level.

studied by Yang et al.38 It has been found that embedding a sensor with polyimide bond ply does not decrease

the strength of the composite structure.

2.4 Signal processing

In this section the signal processing associated with the embedded ICTS is described. The signal processing

is required for two purposes: to allow single-shot measurements with high signal-to-random-noise ratio and to

allow successful baseline signal subtraction despite signal variability due to inspection wand misalignment.

2.4.1 Excitation and pre-processing

Let u(t) be the desired input signal to the ICTS, where t is time. This is typically a short, limited-bandwidth

toneburst of duration T . In order to enable single-shot measurements with high signal-to-random-noise ratio,

the system response, v(t), to u(t) is recovered from the system response, vc(t), to a narrowband chirped input

signal, uc(t), of duration NcT . In the following, F{.} and F−1{.} are the forward and inverse Fourier transform

operators and ω is angular frequency. The necessary chirp excitation is given by:

uc(t) = F−1{U(ω)C(ω)} (2)

Figure 4. (a) Manufacture of encapsulated unit for embedding application. (b) The lay-up sequence of Hexcel 950 glass

fibre and (c) Hexcel M21 carbon fibre composite laminated plate with embedded ICTS.

Figure 5. (a) Glass fibre composite panel with embedded ICTS, (b) carbon fibre composite panel with three embedded

ICTSs. The circle and inset picture in (a) show the position where 10 J impact damage was introduced and the delami-

nation on the reverse side of the panel. In (b), the dotted circles represent the positions of ICTSs, and the stars represent

laser measurement points on the edge of the panel.

where U(ω) = F{u(t)} is the frequency spectrum of the desired input signal and C(ω) = F{c(t)} is the

frequency spectrum of a linear chirp signal, c(t), defined over the interval 0 ≤ t ≤ NcT given by:

c(t) = sin(ω1t+(ω2 − ω1)t2

NcT) (3)

where ω1 and ω2 are the start and end frequencies of the chirp. The bandwidth of c(t) (i.e. from ω1 to ω2)

must span the bandwidth of u(t). The stretch factor, Nc, specifies the duration of the chirp signal relative to

the duration of the desired input signal. The increase in SR associated with a chirp is proportional to√Nc.

39

The frequency-domain response of the system, V (ω), to the desired input signal is extracted from the system

response to the chirp excitation, vc(t), through the following frequency-domain deconvolution augmented with a

bandpass filter, G(ω), to remove out-of-band noise:

V (ω) =F{vc(t)}G(ω)

C(ω). (4)

Thus the final time-domain response of the system to the desired input signal is:

v(t) = F−1{V (ω)}. (5)

2.4.2 Effect of coil misalignment

In an infinitely-large, featureless structure, impact damage could be directly detected by the appearance of a

reflection in v(t). However, in a real structure, large signals are present due to reflections from benign structural

features and these must be suppressed in order to be able to detect much smaller signals due to impact damage.

The detection of new impact damage therefore requires comparison of the current signal with an earlier baseline

signal. Here, baseline signal subtraction is employed as this has the advantage of being a linear operation with

predictable performance.34 It has been found that the amplitude of a received signal from an ICTS varies

dramatically with changes in the relative position of coils,19,23 which here is due to imprecise positioning of

the inspection wand between repeated measurements at the same sensor. This presents a major challenge for

baseline signal subtraction that must be overcome in order to achieve reliable damage detection.

Fig. 6(a) shows waveforms prior to baseline subtraction from the ICTS embedded in glass fibre plate for

coil separations of 12 and 17 mm. The coil separation is defined as the distance between the plane of the

sensor and the plane of inspection wand under the assumption of perfect coaxial and angular alignment. The

effective input signal was a Gaussian-windowed, 5-cycle (defined by the −40 dB points of the Gaussian window)

toneburst with centre frequency 160 kHz and the results were obtained using chirp excitation with Nc = 1300.

The change in amplitude of the first echo signal for a range of coil separations is shown in Fig. 6(b). As the coil

separation increases, the signal amplitude drops and decreases by about 65% of its original amplitude when the

coil separation is increased by 8 mm.

However, amplitude variation is not the only effect resulting from coil misalignment. The corresponding

spectra for the first echo signals shown in Fig. 6(a) are shown in Fig. 6(c). There is a small change in the

frequency of the spectral peak for the two coil separations and this is plotted for a range of separation distances

in Fig. 6(d). Similar results are found for lateral misalignment.

0 100 200 300 400 500−1.5

−1

−0.5

0

0.5

1

1.5

Distance8away8from8sensor8zmmd

Arb

.8uni

ts

Dps=12mmDps=17mm

12 13 14 15 16 17 18 19 200.2

0.4

0.6

0.8

1

Dps zmmd

|Hzv

ztdd

| t=t 1

12 13 14 15 16 17 18 19 201.79

1.8

1.81

1.82

1.83

1.84

1.85 x8105

Dps zmmd

f p

1 1.5 2 2.5x8105

0

2000

4000

6000

8000

10000

Frequency8zHzd

Arb

.8uni

ts

Dps=12mmDps=17mm

zad zbd

zddzcd

Figure 6. (a) Time-domain signals obtained from ICTS in glass fibre composite for coil separations of 12 and 17 mm and

(b) amplitude of first echo as a function of coil separation. (c) Frequency spectra of first echo signals for coil separations

of 12 and 17 mm and (d) frequency of spectral peak as a function of coil separation.

2.4.3 Received signal compensation

Pure amplitude variations can be effectively compensated by normalisation. In this approach to compensation

prior to baseline subtraction, both the received signal, v(t), and baseline, v0(t), are normalised to the maximum

amplitude of the envelope of the first echo. Therefore, the residual signal using amplitude normalisation is:

∆norm(t) =v(t)

|H {v(t)}|t=t1− v0(t)|H {v0(t)}|t=t1

. (6)

Since spectral shape as well as signal amplitude are changed by the coil misalignment, amplitude normalisation

is generally inadequate. Here, an inverse filter40 is employed to achieve better compensation for coil misalignment

effects. This is calculated from the measurement and baseline spectra of the first echo in each signal:

I(w) =F{v0(t1 − δ ≤ t ≤ t1 + δ)}F{v(t1 − δ ≤ t ≤ t1 + δ)}

(7)

Consequently, the residual signal obtained using the inverse filter method is:

∆inv(t) = F−1{V (ω)I(ω)− V0(ω)} (8)

where V (ω) = F{v(t)} and V0(ω) = F{v0(t)}.

2.4.4 Beam spread compensation

In order to quantify the damage detection capability of the system, it is desirable for the response to a given

level of damage to be independent of its distance from a sensor. In a complex structure, this would most likely

require some form of experimental calibration procedure. However, in a simple plate, the reduction in signal

amplitude with distance is predictable and due to a combination of dissipative and geometric attenuation. As the

attenuation of the S0 mode in composites, particularly along the fibre direction, is low,30 geometric attenuation

due to the divergence of both transmitted and scattered waves is the dominant effect. Both result in an amplitude

reduction proportional to the square root of propagation distance. Hence the amplitude of the response from

localised damage for a single sensor operating in pulse-echo mode is expected to be proportional to the reciprocal

of the distance of damage from the sensor. Therefore, it is instructive to scale the amplitude of residual signals

with distance from sensor, d, and to only consider the envelope of the signal. Consequently, the quantity plotted

in subsequent residual signal graphs is

rnorm,inv = d|H {∆inv(2d/vg)}|, (9)

where vg is the group velocity of the S0 mode at the center frequency of the input signal, 160 kHz.

3. SYSTEM TESTING

In this section, the embedded ICTS and signal processing techniques described previously are investigated for

consistency, performance in the presence of inspection wand misalignment and ultimately for their ability to

detect impact damage.

3.1 Sensor consistency

The performance consistency of ICTS is initially tested through the investigation of the frequency response of the

devices embedded in carbon fibre plate described in Section 2.3. The three embedded ICTS and a surface-bonded

ICTS are wirelessly probed using the inspection wand at a coil separation distance of 12 mm. The effective input

signals are 5-cycle, Gaussian-windowed tonebursts with various centre frequencies. They are applied using chirp

excitation signals with Nc = 1300 and 24 V peak-peak voltage. Meanwhile, a laser vibrometer (OFV-505,

Polytec, Harpenden, Hertfordshire, UK) is used to measure the in-plane guided wave displacement generated by

each ICTS on the edge of the plate nearest to each sensor. In-plane displacement measurements are necessary

as the S0 mode has very low out-of-plane displacement in this frequency range. The sensor positions and laser

measuring points are shown in Fig. 5. A typical signal measured by the laser interferometer is shown in Fig. 7(a).

The set-up for the laser vibrometer is fixed for all the measurements and the edge of the carbon fibre plate is

polished at the measurement points.

−5 0 5 10 15xA10−5

−0.1

−0.05

0

0.05

0.1

0.15

TimeA(s)

In−p

lane

Adis

plac

emen

tA(ar

b.Au

nits

)

1 1.5 2 2.5 3xA105

0

0.05

0.1

0.15

FrequencyA(Hz)

Am

plitu

deA(a

rb.u

nits

)

DirectAattachedAICTSEmbeddedAICTSA1EmbeddedAICTSA2EmbeddedAICTSA3

FirstAarrival

(a) (b)

Figure 7. (a) Typical in-plane displacement due to ICTS measured using laser interferometer for an effective input signal

with centre frequency of 160 kHz. (b) Spectra of first arrival signals for three embedded ICTSs and one surface-bonded

ICTS.

Fig. 7(b) shows the frequency response of both embedded and surface-bonded ICTS. It can be seen from this

figure that consistent performance in terms of both signal amplitude and peak response frequency is obtained with

the three embedded ICTS. Furthermore, the signal amplitude obtained by embedding the ICTS into composite

is almost twice that of a surface-bonded ICTS. This is thought to be because a better bond layer between the

unit and structure is achieved through the curing process and because both sides of sensor are bonded to the

structure. The frequency shift of the peak response between the embedded ICTS and directly-bonded ICTS is

due to the fact that the capacitance between the neighbouring tracks in the transducer coil is increased due to

the increased permittivity contributed by the carbon fibre. It is worth noting that the consistency of embedded

ICTS in absolute terms is potentially very attractive for an SHM system as it makes system calibration to an

absolute level of impact damage much more straightforward.

3.2 Baseline signal subtraction

Enveloped subtraction results obtained using both simple normalisation, Eq. 6, and the inverse filter technique,

Eq. 8, are shown in Fig. 8 for various degrees of coil misalignment. These results were obtained from the ICTS in

the damage-free, glass-fibre composite plate. Again, the input signal was a 5-cycle, Gaussian-windowed toneburst

with centre frequency 160 kHz and the results were obtained using chirp excitation with Nc = 1300. For the

baseline signal, the inspection wand was placed 12 mm above the embedded transducer and aligned laterally

to maximise the signal amplitude. It worth noting that the residual signals are normalised to the maximum

amplitude of the impact damage echo for impact damage study in the next section.

From Fig. 8, it can be seen that the residual signal level achieved by using the inverse filter technique is much

better than the the results obtained using the amplitude normalisation technique, with more than 3 times (10 dB)

better suppression of reflections from benign features achieved for all cases. However, the question remains as to

whether the residual signal is low enough to allow impact damage to be detected. Also, it is possible that the

inverse filter could suppress the signal introduced by the impact damage. These issues are discussed in the next

section.

100 200 300 400 500 600 700 800 900 10000

0.5

1

1.5

Distancelawaylfromlsensorlbmm)

Nor

m.lr

esid

ualls

igna

l

100 200 300 400 500 600 700 800 900 10000

0.5

1

1.5

Distancelawaylfromlsensorlbmm)

Nor

m.lr

esid

ualls

igna

l

100 200 300 400 500 600 700 800 900 10000

0.5

1

1.5

Distancelawaylfromlsensorlbmm)

Nor

m.lr

esid

ualls

igna

l

100 200 300 400 500 600 700 800 900 10000

0.5

1

1.5

Distancelawaylfromlsensorlbmm)

Nor

m.lr

esid

ualls

igna

l

100 200 300 400 500 600 700 800 900 10000

0.5

1

1.5

Distancelawaylfromlsensolrbmm)

Nor

m.lr

esid

ualls

igna

l

100 200 300 400 500 600 700 800 900 10000

0.5

1

1.5

Distancelawaylfromlsensorlbmm)

Nor

m.lr

esid

ualls

igna

l

ResiduallsignallwithlnormalizationResiduallsignallwithlinverselfilter10Jlimpactlthreshold

ba) bd)

be)

bf)bc)

bb)

Figure 8. Enveloped residual signals after baseline subtraction with (a) 3 mm, (b) 6 mm, (c) 9 mm coil horizontal

misalignment, and (e) 3 mm, (f) 6 mm, (g) 9 mm coil vertical separation variation by using normalization (grey line),

and inverse filter (black line) signal processing techniques. The dashed line represents the 10 J impact damage threshold

discussed in Section 3.3

.

3.3 Impact damage detection

The glass fibre panel was subjected to a 10 J impact at a distance of 125 mm from the ICTS, resulting in a

delamination around 20 mm diameter, as shown in the inset in Fig. 5 (a). Fig. 9 shows the residual signals after

baseline subtraction using the inverse filter method, when the coils are perfectly aligned with 12 mm separation.

The residual signals are normalised to the maximum amplitude of the impact damage echo.

From Fig. 9(b), it can be seen that the reflected signal from a 10 J impact can be clearly identified at the

0 100 200 300 400 500 600 700 800 9000

0.5

1

1.5

Distanceuawayufromusensoru(mm)

Nor

mal

ised

ures

idua

lusig

nal

0 100 200 300 400 500 600 700 800 9000

0.5

1

1.5

Distanceuawayufromusensoru(mm)

Nor

mal

ised

ures

idua

lusig

nal

Residualusignal10Juimpactudamageuthreshold

(b)

(a)

Figure 9. Residual signals from (a) damage free structure, and (b)structure with 10 J impact damage. Grey lines in (a)

and (b) refer to the location of damage.

expected distance away from the transducer of 125 mm. This qualitatively shows that the embedded ICTS is

capable of detecting impact damage in a composite plate and that the inverse filter did not suppress the signal.

Note that the increased residual signals after 125 mm are due to forward scattering from the impact interfering

with subsequent edge reflections. Although not directly interpretable, the appearance of these signals provides

another opportunity for detecting damage.41

Fig. 9(a) shows that the normalised residual signal is still significantly smaller (6 dB) than the 10 J threshold

(the dashed horizontal lines) even at a distance of 1000 mm away from the sensor. The amplitude of the threshold

is defined as the maximum value of the 10 J impact damage echo. This suggests that the inspection range of the

embedded ICTS is around 1000 mm, when the inspection wand is perfectly aligned. The 10 J impact damage

threshold is also plotted in Fig. 8 (the dashed horizontal lines). From this it can be seen that as the misalignment

increases, the performance of the inverse filter decreases and the coherent noise level increases due to worsening

suppression of reflections from structural features. However, when the coil misalignment is smaller than 5 mm

the residual signal is still about 4.5 dB lower than the threshold at a distance of 1000 mm away from the sensor.

To maintain a signal-to-coherent-noise ratio of 6 dB for 10 J impact damage at a distance of 1000 mm, more

detailed experiments reveal that the maximum horizontal and vertical coil misalignments are 10 mm and 4 mm.

4. CONCLUSION

This paper has demonstrated a structurally-integrated sensing system for large-area impact-damage detection

in composite panels. The system uses a inert, guided wave sensor embedded in the composite panel, which is

externally activated from an inductively-coupled inspection wand. The design of the inductive coupling link has

been optimised for the application using a previously-developed model. To enable highly-consistent, single-shot

measurements to be made rapidly and without precise alignment of the inspection wand several signal processing

steps have been employed: chirp excitation, deconvolution and band-pass inverse filtering. The last step of inverse

filtering enables baseline signal subtraction techniques to be employed to detect the occurrence of relatively weak

scattered signals from impact damage in the presence of large reflections from benign structural features. The

system has been experimentally demonstrated on two composite panels and 10 J impact damage successfully

detected. It is estimated that the system could detect this level of impact damage at a range of up to 1000 mm.

It is worth noting that, although the inverse filter technique is specifically developed here for compensation

of ICTS coil misalignment, the principles are general and could be applied to compensation of signal changes

due to environmental variation, such as temperature. These are major issues when using the baseline signal

subtraction for general guided wave SHM applications.

Here, the construction of the inverse filter is based on the assumption that the echo signals used to generate it

only change due to coil misalignment effects. This raises a potential concern that damage signals may be masked

if they interact with this echo. This has not caused a problem in the impact detection example presented here

even though forward scattered signals from the damage must have interacted with the edge echo. Although the

presence of these forward scattered signals has been suppressed (as evidenced by the low residual signal level at

the position of the first edge echo in Fig. 9(b) at 225 mm), the presence of the new direct echo from the impact

at 125 mm is separate and is not suppressed. However, there would be a problem if the direct echo from the

impact damage itself had overlapped with the first edge echo. In a real system, it would therefore be desirable

for all points in a structure to be within the inspection range of more than one sensor.

REFERENCES

1. K. S. Ahmed, S. Vijayarangan, and A. Kumar, “Low velocity impact damage characterization of woven

jute-glass fabric reinforced isothalic polyester hybrid composites,” Journal of Reinforced Plastics and Com-

posites 26, pp. 959–976, 2007.

2. M. V. Hosur, F. Chowdhury, and S. Jeelani, “Low-velocity impact response and ultrasonic nde of woven

carbon/epoxy-nanoclay nanocomposites,” Journal of Composite Materials 41, pp. 2195–2212, 2007.

3. N. P. Avdelidis, D. P. Almond, A. Dobbinson, B. C. Hawtin, C. Ibarra-Castanedo, and X. Maldague,

“Aircraft composites assessment by means of transient thermal ndt,” Progress in Aerospace Scienes 40,

pp. 143–162, 2004.

4. D. Zenkert, A. Shipsha, P. H. Bull, and B. Hayman, “Damage tolerance assessment of composite sandwich

panels with localised damage,” Composites Science and Technology 65, pp. 2597–2611, 2005.

5. B. Beral and H. Speckmann, “Structural health monitoring (shm) for aircraft structures: A challenge for sys-

tem developers and aircraft manufacturers,” Proc. 4th Int. Workshop on Structural Health Monitoring 1,2,

pp. 12–29, 2001.

6. M. A. Hamstad, “A review - acoustic emission, a tool for composite material studies,” Experimental Me-

chanics 26, pp. 7–13, 1986.

7. J. M. Carlyle, H. L. Bodine, S. S. Henley, R. L. Dawes, R. Demeski, and E. V. Hill, “Practical ae methodology

for use on aircraft,” American Society for Testing and Materials Special Technical Publication 1353, pp. 191–

205, 1999.

8. C. R. Farrar and K. Worden, “An introduction to structural health monitoring,” Proceedings of The Royal

Society of London Series A-Mathematical Physical and Engineering Sciences 365, pp. 303–315, 2007.

9. C. Boller, “Next generation structural health monitoring and its integration into aircraft design,” Interna-

tional Journal of Systems Science 31, pp. 1333–1349, 2000.

10. K. Worden, C. R. Farrar, G. Manson, and G. Park, “The fundamental axioms of structural health mon-

itoring,” Proceedings of The Royal Society of London Series A-Mathematical Physical and Engineering

Sciences 463, pp. 1639–1664, 2007.

11. T. Clarke, F. Simonetti, and P. Cawley, “Guided wave health monitoring of complex structures by sparse

array systems: Influence of temperature changes on performance,” Journal of Sound and Vibration 329,

pp. 2306–2322, 2010.

12. Z. Lu, S. J. Lee, J. E. Michaels, and T. E. Michaels, “On the optimization of temperature compensation for

guided wave structural health monitioring,” Review of Progress in Quantitative Nondestructive evaluation

1211, pp. 1860–1867, 2010.

13. X. L. Qing, A. Kumar, C. Zhang, I. F. Gonzales, G. P. Guo, and F. K. Chang, “A hybrid piezoelectric/fiber

optic diagnostic system for structural health monitoring,” Smart Material and Structure 14, pp. 98–103,

2005.

14. M. Lin, X. Qing, A. Kumar, and S. J. Beard, “Smart layer and smart suitcase for structural health monitoring

applications,” in SPIE — Smart Structures and Materials, Proc. SPIE 4332, 2001.

15. J. P. Lynch and K. J. Loh, “A summary review of wireless sensors and sensor networks for structure health

monitoring,” The Shock and Vibration Digest 38(2), pp. 91–128, 2006.

16. S. Jang, H. Jo, S. Cho, K. Mechitov, J. A. Rice, S. H. Sim, H. J. Jung, C. B. Yun, J. Spencer, F. Billie, and

G. Agha, “Structural health monitoring of a cable-stayed bridge using smart sensor technology: deployment

and evaluation,” Smart Structures and Systems 6, pp. 439–459, 2010.

17. P. P. Liu, S. F. Yuan, and L. Qiu, “Development of a pzt-based wireless digital monitor for composite impact

monitoring,” Smart Materials and Structures 21, p. 035018.1035018.10, 2012.

18. X. L. Zhao, T. Qian, G. Mei, C. Kwan, R. Zane, C. Walsh, T. Paing, and Z. Popovic, “Active health moni-

toring of an aircraft wing with an embedded piezoelectric sensor/actuator network: ii. wireless approaches,”

Smart Materials and Structures 16, pp. 1218–1225, 2007.

19. D. W. Greve, H. Sohn, C. P. Yue, and I. J. Oppenheim, “An inductively coupled lamb wave transducer,”

IEEE Sensors 7(1-2), pp. 295–301, 2007.

20. P. Zheng, D. Greve, and I. Oppenheim., “Crack detection with wireless inductively-coupled transducers,” in

SPIE — Sensors and Smart Structures Technologies for Civil Mechanical, and Aerospace Systems, 69321,

2008.

21. M. Kobayashi, K. T. Wu, L. Song, C. K. Jen, and N. Mrad, “Structural health monitoring of composites

using integrated and flexible piezoelectric ultrasonic transducers,” Journal of Intelligent Material Systems

and Structures 20, pp. 969–977, 2009.

22. K. Wu, C. Jen, and N. Mrad., “Non-contact local and global damage detection with integrated ultrasonic

transducers,” in SPIE — Sensors and Smart Structures Technologies for Civil Mechanical, and Aerospace

Systems, 69321, 2008.

23. C. H. Zhong, A. J. Croxford, and P. D. Wilcox, “Investigation of inductively coupled ultrasonic transducer

system for nde,” IEEE Transactions on Ultrasonics Ferroelectrics and Frequency Control 60, pp. 1115–1125,

2013.

24. Z. Q. Su, L. Ye, and Y. Lu, “Guided lamb waves for identification of damage in composite structures: A

review,” Journal of Sound and Vibration 295, pp. 753–780, 2006.

25. M. J. S. Lowe, D. N. Alleyne, and P. Cawley, “Defect detection in pipes using guided waves,” Ultrosonic 36,

pp. 147–154, 1998.

26. D. N. Alleyne, B. Pavlakovic, M. J. S. Lowe, and P. Cawley, “Rapid, long range inspection of chemical

plant pipework using guided waves,” Review of Progress in Quantitative Nondestructive Evaluation 557,

pp. 180–187, 2001.

27. R. Long, P. Cawley, and M. J. S. Lowe, “Acoustic wave propagation in buried iron water pipes,” Proceedings

of The Royal Society of London Series A-Mathematical Physical and Engineering Sciences 459, pp. 2749–

2770, 2003.

28. J. L. Rose, M. J. Avilioli, and W. J. Song, “Application and potential of guided wave rail inspection,”

Insight Non-Destructive Testing and Condition Monitoring 44, pp. 353–358, 2002.

29. P. D. Wilcox, M. Evans, B. Pavlakovic, D. Alleyne, K. Vine, P. Cawley, and M. J. S. Lowe, “Guided wave

testing of rail,” Insight–NDT and Condition Monitoring 45, pp. 413–420, 2003.

30. R. A. Badcock and E. A. Birt, “The use of 0-3 piezocomposite embedded lamb wave sensors for detection

of damage in advanced fibre composites,” Smart Material and Structures 9, pp. 291–297, 2000.

31. Y. Kagawa, T. Tsuchiya, B. Fujii, and K. Fujioka, “Discrete huygens’ model approach to sound wave

propagation,” Journal of Sound and Vibration 218, pp. 419–444, 1998.

32. Y. H. Lu and J. E. Michaels, “A methodology for structural health monitoring with diffuse ultrasonic waves

in the presence of temperature variations,” Ultrasonics 43, pp. 717–731, 2005.

33. Y. Lee, “Rfid coil design,” application note, Microchip Technology Inc., 1998.

34. A. J. Croxford, P. D. Wilcox, B. W. Drinkwater, and G. Konstantinidis, “Strategies for guided-wave struc-

tural health monitoring,” Proceedings of The Royal Society of London Series A-Mathematical Physical and

Engineering Sciences 463, pp. 2961–2981, 2007.

35. C. Zhong, P. Wilcox, and A. Croxford, “Inductively coupled transducer system for damage detection in

composites,” in SPIE — Health Monitoring of Structural and Biological Systems, 8348, 2012.

36. B. Lenaerts and R. Puers, “Inductive powering of a freely moving system,” Sensors and Actuators A:

Physical 123-24, pp. 522–530, 2005.

37. B. Lenaerts and R. Puers, “An inductive power link for a wireless endoscope,” Biosensors and Bioelectron-

ics 22, pp. 1390–1395, 2007.

38. S. M. Yang, C. C. Hung, and K. H. Chen, “Design and fabrication of a smart layer module in composite

laminated structures,” Smart Materials and Structures 14, pp. 315–320, 2005.

39. J. E. Michaels, S. J. Lee, A. J. Croxford, and P. D. Wilcox, “Chirp excitation of ultrasonic guided waves,”

Ultrasonics 53, pp. 265–270, 2013.

40. R. C. Gonzales, R. E. Woods, and S. L. Eddins, Digital Image Processing Using Matlab, Prentice Hall, 2003.

41. E. B. Flynn, M. D. Todd, P. D. Wilcox, B. W. Drinkwater, and A. J. Croxford, “Maximum-likelihood

estimation of damage location in guided-wave structural health monitoring,” Proceedings of The Royal

Society of London Series A-Matematical Physical and Engineering Sciences 467, pp. 2575–2596, 2011.