Original Manuscript

Structural Health Monitoring

1–10

� The Author(s) 2017

Reprints and permissions:

sagepub.co.uk/journalsPermissions.nav

DOI: 10.1177/1475921716686772

journals.sagepub.com/home/shm

A photophone-based remotenondestructive testing approach tointerfacial defect detection infiber-reinforced polymer-bondedsystems

Tin Kei Cheng1 and Denvid Lau1,2

AbstractExternally bonded fiber-reinforced polymer is an increasingly popular material to be used in strengthening and retrofittingaging structures. In such structures, debonding defects may occur at or near the interface between fiber-reinforced poly-mer and concrete. As such debonding in fiber-reinforced polymer-bonded systems is generally brittle in nature, there is aneed of a reliable inspection technique that can provide early warning of interfacial defects such that premature failure offiber-reinforced polymer-strengthened structures can be avoided. A remote nondestructive testing approach based on theworking principle of a photophone is presented here as an economical alternative to laser Doppler vibrometry for detect-ing interfacial defects. Concrete specimens retrofitted with fiber-reinforced polymer are excited acoustically by whitenoise, while the surface of the structure is illuminated by a light source. If an interfacial defect exists beneath the surface,the surface will exhibit a frequency response different from an intact surface. The surface of the fiber-reinforced polymerportrays the role of flexible mirror in a photophone, which encodes information about surface vibration into amplitude-modulated light signal. A light detector then captures the irradiance of the reflected beam, and the amplitude modulationis converted into frequency domain in post-processing. With this technique, defect dimensions and thus damage extentcan be inferred from the frequency spectrum obtained. The obtained results correspond well with the theoretical calcula-tion, demonstrating the robustness and the applicability of the proposed technique in civil infrastructure.

KeywordsAcoustic-laser, amplitude-modulated modal analysis, fiber-reinforced polymer, interfacial defect, nondestructive testing,photophone

Introduction

The maintenance of aging structures is a challenging issuefaced by many nations.1 In most cases, strengthening andretrofitting of these deteriorated structures may be moreeconomical than reconstruction, especially in situationsinvolving the preservation of historic sites. The use offiber-reinforced polymer (FRP) composites as an exter-nally bonded element to retrofit structural elements suchas beams, columns, slabs, and bridge decks in order torestore the design capacity and structural redundancy ofdeteriorated structures has emerged as a popularstrengthening and retrofitting technique. FRP, as a thinsheet of composite material, has a lot of desirable proper-ties when compared with the conventional bulk construc-tion materials, such as high strength-to-weight ratio,

good environmental resistance, and excellent resistanceagainst fatigue. Because externally bonded FRP becomesmore popular in retrofitted structures, the failure modesof FRP-bonded systems have thus been studied exten-sively to better understand the mechanical behavior ofsuch composite structures.2–6 Failures in FRP-bondedreinforced concrete (RC) beams can be classified into

1Department of Architecture and Civil Engineering, City University of

Hong Kong, Hong Kong2Department of Civil and Environmental Engineering, Massachusetts

Institute of Technology, Cambridge, MA, USA

Corresponding author:

Denvid Lau, Department of Architecture and Civil Engineering, City

University of Hong Kong, Hong Kong.

Email: denvid@mit.edu

two major mechanisms: (1) flexural failures of critical sec-tions, such as crushing of concrete and FRP rupture and(2) debonding of FRP from the RC beams.2–6 The near-surface debonding defects resulting from the latter casewill be referred to as interfacial defects in this article, andthey are the focus of this study. Interfacial defects oftenoccur in regions with high stress concentration and/orshear stress discontinuity, which are positively correlatedwith (1) regions of material discontinuity, (2) poor instal-lation including inadequate surface preparation of thesubstrate, and (3) the presence of cracks (e.g. propagationof flexural or flexural-shear cracks from the concrete sub-strate to the FRP-concrete interface, resulting in delami-nation). The failures modes due to interfacial defects arebrittle in nature and will result in premature failures ofFRP-strengthened elements if not promptly resolved.2–6

Current nondestructive testing (NDT) approachesdeveloped for civil engineering application include theuse of ultrasound, X-ray and neutron radiography,near-field and far-field radar, and infrared thermogra-phy.7–19 However, these techniques may not always beoptimal for the remote detection of interfacial defects.1

Passive thermography typically has insufficient spatialresolution to resolve small interfacial defects.18

Radiography involves the use of hazardous ionizingradiation.10,11 The radar approach may have issues ifcarbon fiber–reinforced polymer (CFRP) sheets beingpenetrated by radar are not arranged in a parallelorientation as CFRP is a highly conductive material atmicrowave frequencies.14 Finally, while near-field radarand ultrasound are promising approaches, they requirea close proximity to the subject.14,17,19 Recently, high-speed video in conjunction with motion magnificationhas been applied to remotely evaluate modal behavior.Despite being promising in detecting larger scale dam-ages, as of the moment of writing, the maximum fre-quency measurement demonstrated by the high-speedvideo technique is 2500 Hz, which is below the typicalfundamental frequencies of interfacial defects asexplored in this article.20 To complement existing NDTapproaches, an acoustic-laser technique has beendevised specifically to detect interfacial defects from adistance based on their vibrational behavior.21–24 It hasbeen shown that interfacial defects in FRP-bonded sys-tems can be modeled as a combination of a clampedplate (representing the FRP surface) and an acousticcavity (representing the air gap at the interface), resem-bling the physics of a drum. In such defects, the funda-mental mode of vibration lies typically at the order ofkHz.21 However, the dedicated setup used in previousacoustic-laser studies may cost a million US dollarsand might prove an obstacle to the commercializationof the acoustic-laser technique. This study presents aneconomical augmentation to the previous works basedon the design of a photophone. Electronic parts

required to assemble the NDT device put forward inthis article can readily be purchased from consumerelectronics retailers, and the total cost of a functioningsetup is around a hundredth of the original work.

Originally conceived by Alexander Graham Bell whoinvented the telephone25 in 1880, a typical photophoneconcentrates sunlight onto a flexible mirror whichreflects the sunlight to a remote receiver (Figure 1).26,27

Excited by the speaker’s voice, the flexible mirror mod-ulates the reflected sunlight’s radiant intensity by flex-ing convex and concave according to the changingsound pressure, dispersing, and converging the beam ofsunlight. This amplitude-modulated radiant intensity isconverted back to an audio signal when detected by thereceiver. The photophone eventually achieved the firstwireless telephone communication, predating the firstspoken radio transmission by almost two decades andlaying the groundwork for fiber-optic communication acentury later. Although considered by Bell to be one ofhis greatest achievements, even more superior than thetelephone, the photophone never went into mainstreamusage and it is found mostly in military use.25,28,29

Nowadays, the photophone is commonly known as the‘‘spy microphone’’ in hobbyist projects, with few real-life applications. In this article, the interest in photo-phone is renewed by considering an NDT applicationin the detection of interfacial defects in FRP-bondedsystems. The working principle of the modified photo-phone will first be introduced with the relation to previ-ous acoustic-laser studies, followed by experiments withartificially induced delamination-like interfacial defectson FRP-bonded concrete specimens. Finally, benefitsand limitations of this photophone-based technique willbe addressed based on experimental results, and futureresearch direction including an extension of this tech-nique to general modal analysis will be discussed.

Figure 1. A representative arrangement of the photophone.26

A beam of sunlight is concentrated upon a flexible mirror on thetransmitter side. After reflection, the beam is again madeparallel by means of another lens. The beam is then received ata distant location upon a parabolic reflector, the focus of whichis placed on a selenium photoresistor, connected in a localcircuit with a battery and telephone.

2 Structural Health Monitoring

Working principle

The acoustic-laser technique

Although previous acoustic-laser studies focus ondetecting debonding defects located at or near the inter-face between a sheet of FRP and its substrate, it is infact a general NDT method for remote evaluation ofstructures that can be excited acoustically and wouldinduce a vibration on an exposed measurable sur-face.21,30 Defects are identified based on the frequencyresponse of the surface vibration. Previous worksinvolve exciting the FRP-bonded structure by a fre-quency sweep from a high-power standoff parametricacoustic array (PAA), and the resulting vibration ismonitored remotely by a laser Doppler vibrometer(LDV) with the peaks in the frequency response spec-trum corresponding to vibration modes of the defect.30

The benefits of this design include the ability of long-range measurements up to 30 m, the robustness againstenvironmental noise as the technique is based on inter-ference, and the ability to connect measurement result(surface vibration velocity) with defect characteristicsfor intuitive interpretation of experiment results.

However, the technique measures only surface vibra-tions. Therefore, to obtain a holistic view of the struc-ture’s health, the acoustic-laser technique must becombined with other NDT techniques to reveal subsur-face defects.

The photophone-based approach and experimentsetup

In this study, the measurement setup is refined todemonstrate the flexibility in deployment and commer-cial viability of the acoustic-laser technique(Figure 2(a)). An ordinary loudspeaker is used in placeof the PAA as the acoustic excitation source given thatthe fundamental modes of interfacial defects have beenshown to be in the order of kHz,21 thus can be excitedby consumer-grade electronics. White noise is intro-duced for acoustic excitation instead of a frequencysweep to cover a wide range of frequencies in a shortperiod of time. The LDV is replaced by the low-costphotophone-based measurement system proposed inthis article, exchanging frequency modulation (FM) inan LDV system with amplitude modulation (AM), with

Figure 2. An illustration of the photophone-based experiment setup. (a) An FRP-bonded concrete slab with artificially inducedinterfacial defects is excited acoustically by white noise, while a point on the surface of the structure is illuminated by a laser. Asspecular reflection is strong in our specimens, a lens system is not installed in the light detector, though a collimator is used toisolate low-frequency noise due to ambient lighting. (b) The surface of the FRP portrays the role of flexible mirror in a photophone,alternately focusing and diverging the reflected light, encoding information about surface mechanical vibration into amplitude-modulated light signal, which is captured by a light detector. (c) The amplitude modulation of irradiance is converted into frequencydomain in post-processing. If an interfacial defect exists beneath the surface, the surface will exhibit a frequency response differentfrom an intact surface. Defect dimensions and damage extent can be inferred from the obtained frequency spectrum.

Cheng and Lau 3

a tradeoff in accuracy and precision. Although thesetup in previous works involving PAA and LDV pro-duces highly precise results, the cost of the setup is overa million dollars. At around a hundredth of the originalcost, the current setup provides comparable perfor-mance, albeit at the expense of the resistance to envi-ronmental noise, as expected in an AM system.

Although the photophone was intended for voicetransmission, it is realized that the transmission carriesnot only the voice but also information regarding thevibrational characteristics of the flexible mirror. Throughwhite noise excitation or a frequency sweep, the modesof resonance of the mirror can be probed. Furthermore,it is noted that all reflective surfaces can act as the flex-ible mirror, allowing modal analysis on a general objectbased on the principle of a photophone. If a surface isdull, a retroreflective sticker or reflective paint can beapplied to the surface to enhance reflection. If a surfaceis rough, a lens system can be included to focus the dif-fuse light reflected from the vibrating surface.

The working principle of our photophone-basedmethod is illustrated in Figure 2. The specimen isexcited acoustically by white noise and is simultane-ously illuminated by a beam of laser for vibration mea-surement at a single point (Figure 2(a)). If a structure isacoustically ideal in an open area and it is excited bywhite noise, the resulting vibration would have equalenergy at every frequency bin, reproducing the incidentwhite noise. However, in reality, all objects have a setof resonance modes, and therefore, the frequencyresponse is not flat. To be precise, spikes in the fre-quency response correspond to resonance frequenciesof the object, save for a minor shift due to the effect ofdamping. The vibration of the surface results in theperiodic variation of the surface normal, which in turncauses the reflected beam of light to converge anddiverge at the same frequency as the vibration, resultingin an amplitude-modulated beam of reflection carryinginformation about the surface vibration (Figure 2(b)and (c)). The reflected light is picked up by a photo-diode circuit, amplified, and collected by a data loggeras voltage signal. The frequency-domain view of themodulated signal is obtained via fast Fourier transform(FFT), and vibration modes are identified by locatingresponse peaks in a plot of the frequency spectrum.

Defect dimensions and damage extent are inferredfrom the frequency spectrum as interfacial defects canbe modeled as a clamped plate with an acoustic cavityunderneath.21,31 In the following equation, the variableE is designated as Young’s modulus, n as Poisson’sratio, r as volumetric density, h as thickness of theFRP-epoxy composite layer, and D = Eh3=12ð1� v2Þ asthe bending stiffness. Furthermore, A is defined as thedefect area; Lz as the depth of the air gap; vair as the

speed of sound in air; l, m, and n as positive integers ofvibration mode numbers along the x, y, and z axes of aCartesian coordinate system with the x and y axesoriented along the orthogonal edges of a rectangulardefect. For a square isotropic defect, its area can beestimated by

A =ll;mh

vl;m

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiE

12rð1� v2Þ

s’

1:65h

f1;1

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiE

rð1� v2Þ

sð1Þ

where ll;m is a numerically calculated constant withl1;1’35:99. A table of ll;m for higher modes can befound in the work by Leissa.31 However, since FRP isintrinsically a transversely isotropic material, the quali-tative behavior rather than the exact outcome of theabove relation should be noted. The depth of the defectcan be identified by the following relation

Lz =nvair

2fn’

170

f1ð2Þ

It should be noted that the latter relation is linear inits harmonics, and thus can be decoupled from the fre-quency response of equation (1). However, as noted byCheng and Lau,21 the vibration magnitude of theacoustic cavity is generally one to two orders of magni-tude weaker than the vibration of plate, as the sur-roundings of the cavity (FRP, epoxy, and concrete) aremuch stiffer and denser than air.

Although a laser is used for single-point measure-ment, the efficiency would be low when an area is to beprobed and the location of the defect is not known inadvance. In this scenario, the FRP surface could beilluminated by sunlight, a light emitting diode (LED)array, or equivalent even lighting setup (Figure 3(a)).Focusing the reflected rays with a lens system, the sin-gle photodiode detector can be replaced by a light sen-sor array, such as a high-speed camera, to obtain animage of vibration activity over the entire area underinspection. Significant response peaks observed in thespectrum would indicate existence of defects (Figure3(b)), prompting further investigation using the single-point laser technique to localize the defect. It should benoted that in principle the illumination source is notlimited to visible light but applicable to the entire elec-tromagnetic (EM) wave spectrum.

Experiment

Specimens

Three square concrete slabs with edge lengths ofroughly 300 mm are casted with various interfacialdefects manifested as near-surface voids induced byinserting Styrofoams onto the mold of the concrete

4 Structural Health Monitoring

(Figure 4). The use of air voids is a common strategy toemulate interfacial defects.7,8,14–16,30 The specimens arecured for approximately 28 days in a water tank andthen dried at 50�C in a oven for 3 days. The surfaces ofthe specimens are sanded to reduce surface roughnessand are bonded to three layers of CFRP (TORAYUT70-30) using epoxy (Sikadur�-300). The bondedspecimens are further allowed to cure for a minimumof 7 days before measurements are made.

Procedure

A white noise recording with a sample rate of 48 kHz isplayed over a period of 11 s at 102.1 6 0.2 dB (soundpressure level (SPL)), while the environmental noise isfluctuating around 65–71 dB (SPL). Two controlrecordings without acoustic excitation of the sameduration are recorded before and after the experimentto identify the influence of environmental noise. Overthe entire period of recording, the FRP surface underinspection is illuminated by either a beam of 532 nmlaser at a power of 80 mW or direct sunlight. When alaser is used for illumination, the source of acousticexcitation, a studio monitor (Yamaha MSP7 Studio), isplaced 0.96 6 0.01 m away from the specimen surfaceat an angle + 15� 6 5� from the surface normal, whilethe laser and light detector are placed 1.80 6 0.01 and1.7 6 0.1 m, respectively, away from the specimen andlocated at + 25� 6 1� and 220� 6 10� on either side ofthe specimen. The position of the light detector has alarger variation as the reflection path is dependent onthe specimen surface normal. A laser line filter and col-limator are required as the light detector can easily besaturated by the 100 Hz or 120-Hz frequency of ambi-ent lighting (50 or 60 Hz mains electrical supply recti-fied), which would have limited indoor use. Undersunlight, the studio monitor and the light detector areplaced 1.0 6 0.1 m in front of the three specimens. Aphotodiode connected to an amplification circuit isplaced in the path of the reflected light ray from the spe-cimens. The data are recorded on a computer via a datalogger system operating at a sample rate of 50 kHz, giv-ing a Nyquist frequency of 25 kHz. In post-processing,

Figure 3. Simultaneous detection of multiple defects over alarge area. (a) A broad-beamed light source replaces the laser inFigure 2 as the illumination source. Focusing the reflected rayswith a lens system, the single photodiode detector can bereplaced by a light sensor array to obtain an image of vibrationactivity over the entire area under inspection. (b) Vibrationsignatures of multiple defects can be captured.

Figure 4. FRP-bonded concrete slab specimens with artificialinterfacial defects are shown here. The FRP is bonded to thesurface facing the reader but is omitted in this illustration toreveal the underlying defects. Interfacial defects of three types ofarea: 25 3 25, 37.5 3 37.5, and 50 3 50 mm2 are investigated,together with five types of depths: 5, 10, 20, 30, and 40 mm.

Cheng and Lau 5

the data are converted into frequency domain usingFFT of size 1024, multiplied by Hanning window, andaveraged using Welch’s method with an overlappingregion of 50%. The FFT size is chosen to optimizesmoothness of the resulting spectrum rather than maxi-mizing signal-to-noise ratio (SNR), as the latter wouldresult in a highly jagged curve, a well-known artifact ofFFT. As the specimen is small in scale, the concrete slabhas a high fundamental mode of vibration around2 kHz, resulting in low-frequency noise during the mea-surement. Furthermore, aliasing toward the high-frequency end of the frequency spectrum is strong.Therefore, on each end of the frequency spectrum, a2.5 kHz portion was cropped from the result.

Results

The measurement results from single-point laser mea-surement with white noise excitation are shown inFigure 5. The measurements of 50 3 50, 37.5 3 37.5,and 25 3 25 mm2 defects are plotted in Figure 5(a) to(c), respectively. Each solid curve represents a measure-ment of a different defect specimen, while the dottedline represents the average of all frequency responses ofeach defect area. The response peaks appear to cluster

together even though the scale of signal magnitude var-ies across specimens. As this photophone-based tech-nique depends on measured irradiance, and the voltageoutput of the photodiode depends heavily on the reflec-tivity of the measured surface and the positioning ofthe light detector. Thus, reproducing the result in termsof absolute irradiance is difficult to achieve as the mea-sured specimen is not a smooth surface (i.e. an FRPfabric consisting of interweaving fibers). Althoughrepeatability in terms of measured voltage is degradedand the noise floor is not constant across measure-ments, the clarity and consistency in the frequencydomain are high. In Figure 5(d), the frequency of eachpeak in the cluster is plotted as a ratio to the peak withthe lowest frequency, further visualizing the effect ofthe clustering of peaks and implying that the distribu-tion of peaks is not random. Taking the peak with thelowest frequency to be the fundamental mode of vibra-tion, it is plotted against the inverse of defect area(Figure 5(e)). A linear relationship with R2 = 0.971 isobserved, indicating strong correspondence to platetheory. However, as the FRP, epoxy, and concrete areall much stiffer and denser than the trapped pocket ofair at the interface, the SNR is too low for defect depthto be decoupled from the current measurement.

Figure 5. Measurement results using the photophone-based technique. (a–c) The frequency response of 50 3 50, 37.5 3 37.5,and 25 3 25 mm2 defects, respectively, are shown. The behavior of each specimen is shown in the background as translucentcurves, and the averaged response for each set of defect areas is displayed as a dotted line. (d) The frequency of each peak in thecluster is plotted as a ratio to the fundamental mode, further visualizing the effect of this clustering and implying that the distributionof peaks is not random. (e) The peak with the lowest frequency is taken to be the fundamental mode of the defect and is plottedagainst the inverse of defect area for obtaining a linear relationship with R2 = 0.971.

6 Structural Health Monitoring

In Figure 6, the magnitude of acoustic excitation isvaried. Although generally a stronger excitation willresult in a larger response, the increase in signalstrength is not linear. Occasionally, a higher mode mayhave a stronger response than the fundamental mode(e.g. the third mode in Figure 6(a)). This observation isdue to the fact that even though the photophone-basedtechnique has a high correlation with the specimen’sresonance modes, it is an indirect measurement of thespecimen’s vibration. A mode that results in the highestcontrast in its reflection of the illumination source tothe light detector is favored. When a flat surface is dis-placed (e.g. due to vibration), the direction of reflectionremains unchanged. However, as the FRP surface isrough, the displacement of the surface amplifies thevariation of the reflected light intensity, which mayexplain the nonlinear increase in SNR (Figure 6(b)).It is also possible that the laser fall in between two car-bon fibers, and the direction of reflection is erratic(Figure 6(b)).

In Figure 7, the excitation duration is varied. It isobserved that the longer the excitation, the smootherthe response. In the particular signal processing methodused in this study (i.e. Welch’s method), however, eventhough the duration is increased exponentially, the sig-nal strength only improved marginally. Judging fromthe result, an exposure of 2.6 s is optimal for manualtrigger, while an excitation duration below 0.2 s is pos-sible for computer trigger. It should be noted that the

latter is only a loose upper bound as the excitation andrecording are manually synchronized in this study;results below 0.2 s can be significantly influenced byhuman error.

In Figure 8, the measurement of the specimens illu-minated by sunlight is shown. The excitation durationis reduced to 5 s, and the response from a frequencysweep from 100 to 20,000 Hz is also shown for compar-ison. Resonance modes of various defects are found tohave superimposed on each other. However, a minorfrequency offset of around 500 Hz is observed, which ispossibly due to the specimens being clamped differentlyfrom the indoors single-point laser measurement,thereby altering the damping coefficient of the system.Moreover, it is found that white noise excitation resultsin a higher SNR than using frequency sweep, justifyingthe use of white noise for excitation in place of fre-quency sweep.

Discussion

From the consistency in inferring defect dimensions,the capacity of the measurement technique introduced

Figure 6. (a) Variation of frequency response under differentmagnitudes of acoustic excitation. Generally, a strongerexcitation will result in a larger response but the increase inSNR is not linear. (b) The change in the direction of reflection(gray line) of an incident laser (green line) on a rough surfacewith small displacement.

Figure 7. (a) Change of frequency as a result of varyingexcitation duration. (b) A magnified section to better visualizethe variation. It is observed that the longer the excitation, thesmoother the observed response. However, even though theduration is increased exponentially, the signal strength is onlyincreased marginally.

Cheng and Lau 7

in this article as a low-cost alternative to the originalacoustic-laser technique is validated. Although onlysquare interfacial defects are investigated in this study,it is anticipated that a similar relationship between reso-nance frequencies and defect dimensions would existfor other defect shapes. Indeed, supported by the evi-dence that known analytical solutions to other defectshapes (e.g. circular, triangular) behave similarly,31 ithas been suggested empirically that an inverse relation-ship exists for all regular polygon-shaped interfacialdefects between resonance frequencies and defect area,which differs only by a shape-dependent factor.21

From the review of existing remote NDT approachesin the introduction of this article, it can be seen thatmany of such techniques penetrate the specimen. Inradiography and infrared thermography, the resultingimage contains information regarding the entire depthof the structure; in radar imaging, although the penetra-tion depth can be reduced by adjusting the frequency ofthe radar, the measurement still covers a certain depthof the structure. As concrete is a material containingcoarse aggregates, the resulting scattering would mostcertainly result in a degradation in such techniques’ res-olution. It may thus be difficult to distinguish a subsur-face defect from an interfacial defect, which hasdifferent failure modes, if access is restricted to the ante-rior surface (e.g. when measuring a wall). Fortunately,the acoustic-laser technique, the current modificationinclusive, is based entirely on surface reflection. Thisensures that the measured response is predominantlydue to interfacial defects. Therefore, a full picture ofboth interfacial and subsurface defects can be obtainedwhen the acoustic-laser technique is paired with otherpenetrating NDT techniques. Previous attempts have

been made to combine different NDT techniques tooffer a more comprehensive understanding of structuralhealth,8,9 however, the techniques utilized remain pene-trating in nature.

As the design of the photophone-based technique isin an early stage, multiple improvements are possiblefor future studies. Due to the simplicity of the design,the setup can be further miniaturized as an all-in-onehandheld device. If sunlight is used for illumination,the device can be plugged into a smartphone via itsmicrophone input as the signal is compatible by design,with the defect being excited via the smartphone’sspeaker. The frequency spectrum can be obtained viaexisting audio-processing software with modal identifi-cation in real time. A prototype has already beendemonstrated while still using laser for illuminationand is capable of providing a comparable level of mea-surement (Figure 9). If a stabilized drone is available,the device can be mounted on the drone for automatedairborne areal measurement. Environmental noise suchas highway traffic can also be considered as a source ofacoustic excitation. Moreover, noting that FRP is athin material, the loudspeaker or PAA can in fact beomitted in the setup and replaced by a modulating illu-mination source, utilizing photoacoustic effect to gener-ate mechanical vibration from a long distance withminimal dispersion.26 In this scenario, from the experi-ence of war-time development with the aid of high-powered light sources, the device is potentially capableof NDT over a range of kilometers.28,29 As a furtherimprovement, instead of a single light detector, theFRP surface of interest can be focused by a lens systemonto a detector array to obtain a two-dimensional (2D)vibration map of the entire surface, visualizing themode shape without the need of a scanning laser systemas in a scanning LDV. Similarly, the recording of ahigh-speed camera can be processed by treating eachpixel as an independent light detector for the same pur-pose, or the entire video is processed collectively usingthe technique of motion magnification.32–34

It is envisioned that this photophone-based tech-nique can be applied to a wider range of NDT applica-tions, including general modal analysis. The currentapplication of the technique in FRP, which is a fabricmaterial, is already one of the worst possible scenariosfor technologies based on surface reflection. Althoughthe absolute comparison across results measured in dif-ferent conditions is challenging for FRP surfaces, therobustness and consistency in frequency domain arehigh. Hence, this technique is most suited for measure-ments focusing on the frequency domain. In retrospect,if interferometry is considered as a frequency-modulated measurement technique, this proposedphotophone-based technique has the potential to be

Figure 8. Measurement of multiple defects under plain sunlightillumination. Resonance modes of various defects are found tohave superimposed on each other. Moreover, it is found thatwhite noise excitation results in a higher SNR than usingfrequency sweep, justifying the use of white noise for excitationin place of frequency sweep.

8 Structural Health Monitoring

developed into the amplitude-modulated counterpartof interferometry.

Conclusion

A low-cost augmentation to the acoustic-laser tech-nique is developed and demonstrated in this study byapplying the technique for the detection of interfacialdefects in FRP-bonded systems. As part of the

modification, a photophone-based measurement deviceis conceived for analyzing surface mechanical vibrationat a distance. Various parameters and their influenceon SNR are investigated. The FRP-bonded specimen isexcited acoustically and its frequency response is mea-sured remotely using the aforementioned photophone-based device. The result corresponds well with theoreti-cal prediction, allowing defect dimensions to be inferredbased on the measured frequency response. If interfero-metry is considered as a frequency-modulated tech-nique, this proposed technique has the potential to bedeveloped into the amplitude-modulated counterpart ofinterferometry.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest withrespect to the research, authorship, and/or publication of thisarticle.

Funding

The author(s) disclosed receipt of the following financial sup-port for the research, authorship, and/or publication of thisarticle: This work was financially supported by the CroucherFoundation through the Start-up Allowance for CroucherScholars with the Grant No. 9500012 and the ResearchGrants Council (RGC) in Hong Kong through the EarlyCareer Scheme (ECS) with the Grant No. 139113.

References

1. Wu HF and Wan Y. New vision of the post-NIST civil

infrastructure program: the challenges of next-generation

construction materials and processes. In: Proceedings of

the SPIE smart structures/NDE, San Diego, CA, 8–12

March 2014.

2. Buyukozturk O, Gunes O and Karaca E. Progress on

understanding debonding problems in reinforced con-

crete and steel members strengthened using FRP compo-

sites. Constr Build Mater 2004; 18(1): 9–19.3. Buyukozturk O and Yu T-Y. Understanding and assess-

ment of debonding failures in FRP–concrete systems. In:

Proceedings of the seventh international congress on advances

in civil engineering, Istanbul, 11–13 October 2006.4. Kotynia R. Debonding failures of RC beams strength-

ened with externally bonded strips. In: Proceedings of the

international symposium on bond behavior of FRP in struc-

tures, Hong Kong, 7–9 December 2005.5. Smith ST and Teng JG. FRP-strengthened RC beams. I:

review of debonding strength models. Eng Struct 2002;

24(4): 385–395.6. Teng JG and Chen J. Debonding failures of RC beams

strengthened with externally bonded FRP reinforcement:

behaviour and modelling. In: Proceedings of the first

Asia-Pacific conference on FRP in structures (APFIS

2007), Hong Kong, 12–14 December 2007.

Figure 9. (a) Top view of the setup of real-time measurementof a 50 3 50 3 30 mm3 defect using a smartphone, with thelight detector plugged in via a microphone port (white cable).(b) The cyan curve is the real-time measurement and the yellowcurve is an averaged result. The observed peaks agreed with theprevious results (Figure 5(a)). Inset: A copy of Figure 5(a) forcomparison.

Cheng and Lau 9

7. Feng MQ, De Flaviis F and Kim YJ. Use of microwavesfor damage detection of fiber reinforced polymer-wrappedconcrete structures. J Eng Mech 2002; 128(2): 172–183.

8. Maierhofer C, Brink A, Rollig M, et al. Detection of shal-low voids in concrete structures with impulse thermogra-phy and radar. NDT&E Int 2003; 36(4): 257–263.

9. Maierhofer C, Krause M and Wiggenhauser H. Non-destructive investigation of sluices using radar and ultra-sonic impulse echo. NDT&E Int 1998; 31(6): 421–427.

10. Owen RD. Portable linear accelerators for X-ray andelectron-beam applications in civil engineering. NDT&E

Int 1998; 31(6): 401–409.11. Pugliesi R and Andrade MLG. Study of cracking in con-

crete by neutron radiography. Appl Radiat Isotopes 1997;48(3): 339–344.

12. Shaw M, Millard S, Molyneaux T, et al. Location of steelreinforcement in concrete using ground penetrating radar

and neural networks. NDT&E Int 2005; 38(3): 203–212.13. McCann DM and Forde MC. Review of NDT methods

in the assessment of concrete and masonry structures.NDT&E Int 2001; 34(2): 71–84.

14. Akuthota B, Hughes D, Zoughi R, et al. Near-fieldmicrowave detection of disband in carbon fiber reinforcedpolymer composites used for strengthening cement-basedstructures and disbond repair verification. J Mater Civil

Eng 2004; 16(6): 540–546.15. Li J and Liu C. Noncontact detection of air voids under

glass epoxy jackets using a microwave system. SubsurfSens Technol Appl 2001; 2(4): 411–423.

16. Rhim HC and Buyukozturk O. Wideband microwaveimaging of concrete for nondestructive testing. J Struct

Eng 2000; 126(12): 1451–1457.17. Popovics JS and Rose JL. A survey of developments in

ultrasonic NDE of concrete. IEEE T Ultrason Ferr 1994;41(1): 140–143.

18. Clark MR, McCann DM and Forde MC. Application ofinfrared thermography to the non-destructive testing ofconcrete and masonry bridges. NDT&E Int 2003; 36(4):265–275.

19. Ham S and Popovics JS. Application of contactless ultra-sound toward automated inspection of concrete struc-tures. Automat Constr 2015; 58: 155–164.

20. Cha YJ, Chen JG and Buyukozturk O. Motion magnifi-cation based damage detection using high speed video.

In: Proceedings of the 10th International Workshop on

Structural Health Monitoring (IWSHM), Palo Alto, CA,USA, 1–3 September 2015.

21. Cheng TK and Lau D. Vibrational characteristics ofFRP-bonded concrete interfacial defects in a low fre-quency regime. In: Proceedings of the SPIE smart struc-

tures/NDE, San Diego, CA, 8–12 March 2014.22. Lau D and Qiu Q. A review of nondestructive testing

approach using mechanical electromagnetic waves. In:Proceedings of the SPIE smart structures/NDE, LasVegas, NV, 20–24 March 2016.

23. Qiu Q and Lau D. The sensitivity of acoustic-laser tech-nique for detecting the defects in CFRP-bonded concretesystems. J Nondestruct Eval 2016; 35: 33.

24. Yu T-Y, Cheng TK, Zhou A, et al. Remote defect detec-tion of FRP-bonded concrete system using acoustic-laserand imaging radar techniques. Constr Build Mater 2016;

109: 146–155.25. Bruce RV. Bell: Alexander Graham Bell and the conquest

of solitude. 1st ed. Boston, FL: Little, 1973.26. Bell AG. On the production and reproduction of sound

by light. Am J Sci 1880; 118: 305–324.27. Bell AG. Selenium and the photophone. Nature 1880;

22(569): 500–503.28. Case TW. Infra red telegraphy and telephony. J Opt Soc

Am 1922; 6(4): 398–406.29. Huxford WS and Platt JR. Survey of near infra-red com-

munication systems. J Opt Soc Am 1948; 38(3): 253.30. Yu T-Y and Haupt R. Damage inspection of fiber rein-

forced polymer-concrete systems using a distant acoustic-laser NDE technique. In: Proceedings of the SPIE smart

structures and materials nondestructive evaluation and

health monitoring, San Diego, CA, 7 March 2010.31. Leissa AW. Vibration of plates. Washington, DC: Scien-

tific and Technical Information Division, National Aero-nautics and Space Administration, 1969.

32. Chen JG, Wadhwa N, Cha YJ, et al. Modal identificationof simple structures with high-speed video using motionmagnification. J Sound Vib 2015; 345: 58–71.

33. Wu HY, Rubinstein M, Shih E, et al. Eulerian video mag-nification for revealing subtle changes in the world. ACMT Graphic 2012; 65: 1–8.

34. Wadhwa N, Rubinstein M, Durand F, et al. Phase-basedvideo motion processing. ACM T Graphic 2013; 80: 1–9.

10 Structural Health Monitoring