reference-free ndt technique for debonding detection in cfrp-strengthened...

Reference-Free NDT Technique for Debonding Detectionin CFRP-Strengthened RC Structures

Seung Dae Kim1; Chi Won In2; Kelly E. Cronin3; Hoon Sohn4; and Kent Harries5

Abstract: This study attempts to develop a real-time debonding monitoring system for carbon fiber-reinforced polymer �CFRP� strength-ened structures by continuously inspecting the bonding condition between the CFRP layer and the host structure. The uniqueness of thisstudy is in developing a new concept and theoretical framework of nondestructive testing �NDT� in which debonding is detected withoutrelying on previously obtained baseline data. The proposed reference-free damage diagnosis is achieved based on the concept of timereversal acoustics �TRA�. In TRA, an input signal at an excitation point can be reconstructed if the response signal measured at anotherpoint is reemitted to the original excitation point after being reversed in the time domain. Examining the deviation of the reconstructedsignal from the known initial input signal allows instantaneous identification of damage without requiring a baseline signal representingthe undamaged state for comparison. The concept of TRA has been extended to guided wave propagations within the CFRP-strengthenedreinforced concrete �RC� beams to improve the detectability of local debonding. Monotonic and fatigue load tests of large-scale CFRP-strengthened RC beams are conducted to demonstrate the potential of the proposed reference-free debonding monitoring system.

DOI: 10.1061/�ASCE�0733-9445�2007�133:8�1080�

CE Database subject headings: Nondestructive tests; Bonding; Fiber reinforced polymers; Concrete structures; Bridge maintenance;Wave propagation.

Introduction

Carbon fiber-reinforced polymer �CFRP� composite materialshave become an attractive alternate material for retrofit and reha-bilitation of civil infrastructure systems due to their outstandingstrength, light weight, and versatility �Karbhari et al. 2000�. How-ever, the improvement of strength and stiffness in a host structurecan only be guaranteed when a reliable bonding condition be-tween the host structure and the added CFRP materials is main-tained. Therefore, a reliable nondestructive testing �NDT� systemis required to monitor the initial installation quality and the long-term efficiency of bonding.

There is a large volume of research on damage detection tech-

1Ph.D. Candidate, Dept. of Civil and Environmental EngineeringCarnegie Mellon Univ., Pittsburgh, PA 15213. E-mail: [email protected]

2M.S. Student, Dept. of Civil and Environmental Engineering,Carnegie Mellon Univ., Pittsburgh, PA 15213. E-mail: [email protected]

3Undergraduate Student, Dept. of Civil and EnvironmentalEngineering, Carnegie Mellon Univ., Pittsburgh, PA 15213. E-mail:[email protected]

4Associate Professor, Dept. of Civil and Environmental Engineering,Korea Advanced Institute of Science and Technology, Daejeon, Korea.E-mail: [email protected]

5Assistant Professor, Dept. of Civil and Environmental Engineering,Univ. of Pittsburgh, Pittsburgh, PA 15261. E-mail: [email protected]

Note. Associate Editor: Ahmet Emin Aktan. Discussion open untilJanuary 1, 2008. Separate discussions must be submitted for individualpapers. To extend the closing date by one month, a written request mustbe filed with the ASCE Managing Editor. The manuscript for this paperwas submitted for review and possible publication on April 24, 2006;approved on January 8, 2007. This paper is part of the Journal of Struc-tural Engineering, Vol. 133, No. 8, August 1, 2007. ©ASCE, ISSN

0733-9445/2007/8-1080–1091/$25.00.

1080 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / AUGUST 200

Downloaded 07 Jun 2010 to 143.248.122.147. Redistrib

niques for FRP strengthened concrete structures. To name afew, acoustic emission �Mirmiran et al. 1999�, ultrasonic pulsevelocities �Mirmiran and Wei 2001�, infrared thermography�Levar and Hamilton 2003�, fiber optic sensing �Ansari 2005�,electromechanical impedance spectrum �Giurgiutiu et al. 2003�,electrochemical impedance spectroscopy methods �Hong andHarichandran 2005�, and microwave sensing �Akuthota et al.2004; Ekenel et al. 2004; Feng et al. 2000� have been applied.These techniques are shown to successfully identify FRP debond-ing. However, data interpretation often needs to be manually per-formed by experienced engineers, and automation of data analysisremains largely unsolved. For continuous monitoring, it will becritical to reduce unnecessary interference by users and to auto-mate the data analysis process as much as possible. In addition,although many damage detection techniques are successfully ap-plied to scaled models or specimens tested in controlled labora-tory environments, the performance of these techniques in realoperational environments is still questionable and needs to bevalidated. Varying environmental and operational conditions pro-duce changes in the system’s dynamic response that can be easilymistaken for damage �Sohn 2007�. It is challenging to develop aNDT technique with minimal false positive and negative indica-tions of damage when the system is exposed to varying environ-mental and operational conditions. Few NDT systems have beendeveloped with the intent of deploying it for continuous monitor-ing for in-service structures.

The ultimate goal of this study is to develop an NDT techniquethat goes beyond the laboratory demonstration and can be de-ployed in the field on real-world structures. To achieve thisgoal, a new NDT technique is developed by applying the con-cept of time reversal acoustics �TRA� �Fink and Prada 2001�to guided wave propagations �Rose 1999; Viktorov 1967� withinCFRP-strengthened RC beams. Based on TRA, an input signal

at an excitation point can be reconstructed if the response sig-

7

ution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org

nal measured at another point is reemitted to the original excita-tion point after being reversed in the time domain. This timereversibility is based on linear reciprocity of elastic waves, andbreaks down when there is a source of nonlinearity along thewave propagation path. Because certain types of defects introducenonlinear responses, examining the deviation of the reconstructedsignal from the known initial input signal allows instantaneousidentification of damage without requiring a direct comparisonwith previously obtained baseline signal data. This novel conceptis extended to develop a NDT system that can be rapidlydeployed on laboratory specimens or in-field structures and au-tonomously perform local damage diagnoses at the presence ofoperational and environmental variation that in-service structuresencounter. Smart materials such as lead zirconate titanate �PZT�are used for both generating and measuring guided waves�Giurgiutiu and Lyshevski 2004�.

Wave Propagation in Target Structuresand Time Reversal Acoustics

Elastic waves propagating in solid media can be classified intobody and guided waves. All elastic waves including body andguided waves are governed by the same sets of Navier’s partialdifferential equations �Rose 1999�. The primary difference is that,while body waves are not constrained by any boundaries, guidedwaves need to satisfy the boundary conditions imposed by thephysical systems as well as the governing equations. Guidedwaves can be further divided into Lamb, Stoneley, and Rayleighsurface waves depending on specifics of the imposed boundaryconditions.

Lamb waves are one type of guided waves that are constrainedby two closely-spaced free surfaces �Viktorov 1967�. In spite oftheir unique dispersion and multimode characteristics shown inFig. 1, Lamb waves are widely used for defect detection in NDTapplications due to their relatively long sensing range �Ing andFink 1996; Kessler et al. 2003; Kim et al. 2005; Mal et al. 2005;Sohn et al. 2005�. However, as the thickness of the plate �or theproduct of the exciting frequency f and the thickness of the plated in Fig. 1� increases, it becomes very hard to distinguish wavecomponents because the fundamental symmetric �S0� and anti-symmetric �A0� Lamb modes converge to a Rayleigh surfacewave �Cr�, and additional higher symmetric and antisymmetricmodes appear.

Wave propagation characteristics are further complicatedwhen a thin layer is attached to a thick medium. A RC beam

Fig. 1. Typical dispersion curve of Lamb wave in a thin plate

with a CFRP layer is a good example of such a layered struc-

JOURNA


ture with two distinctively different thicknesses. Luangvilai et al.�2002� experimentally obtained the dispersion curve of a concretebeam with a CFRP layer and demonstrated the complexity ofits wave propagation characteristics. Therefore, a conventionalLamb wave approach may not be applicable for the monitoring ofCFRP reinforced concrete �RC� beam coupled structures, and anew approach, which can be used regardless of the complexity ofwaves, is necessary. To address this issue, the concept of TRA isproposed.

The origin of the proposed time reversal process traces back toTRA �Fink and Prada 2001�. An example of TRA can be found ineveryday life: If one screams “Hello” at the summit of a mountaintoward another mountain, the sound will hit the second mountainand bounce back. One hears “Hello” as an echo. This time revers-ibility of acoustic �or body� waves has found applications inlithotripsy, ultrasonic brain surgery, nondestructive evaluation,and acoustic communications �Fink 1999�. However, the time re-versibility does not work well for guided waves due to their mul-timode and dispersion characteristics. A combination of a specificnarrowband input waveform and multiresolution signal process-ing is employed so that the time reversibility of guided waves ispreserved within an acceptable tolerance for more complex con-figurations such as the layered structure presented in this study�Park et al. 2007�.

In the extended time reversal process, a narrowband input sig-nal can be reconstructed at an excitation point �PZT A� if anoutput signal recorded at another point �PZT B� is reemitted tothe original source point �PZT A� after being reversed and scaledin the time domain as shown in Fig. 2. This process is referredto as the time reversibility of waves and is based on the spatial

Fig. 2. Schematic concept of TRA-based damage identification thatdoes not require any pas baseline signals: �a� known input signal isapplied to PZT A; �b� corresponding response is measured at PZT B;�c� response at PZT B is reversed in the time domain and appliedback to PZT B; �d� final response is measured at PZT A

Fig. 3. Creation of multiple sidebands in a reconstructed signaldue to multimodes, multiple wave propagation paths, and reflectivewaves

L OF STRUCTURAL ENGINEERING © ASCE / AUGUST 2007 / 1081


reciprocity and time-reversal invariance of linear wave equations�Draeger et al. 1997�. Sohn et al. �2005� have shown that theexistence of multimodes, multiple reflective waves, and reflectedwaves creates sidebands in the reconstructed signal as shown inFig. 3, but the presence of the sidebands is irrelevant to the geo-metrical asymmetry of the structure. The existence of these side-bands is one of the unique characteristics of the time reversal inguided wave propagation that has not been observed in bodywaves. In addition, it is demonstrated that the reconstructed signalis symmetric with respect to the main peak in the middle regard-less of the configuration of the structure or the PZT transducerlayout as long as the linear reciprocity holds and the original inputsignal is symmetric. However, it is noted that the shape of the“main wave packet,” where most of energy converges, remainsidentical to the original input signal, although the amplitude ofthe reconstructed signal is smaller than that of the original inputsignal due to attenuation. In Figs. 3 and 4, the reconstructed sig-nal is scaled so that the shape of the main wave packet in thereconstructed signal can be better compared with that of the inputsignal.

Damage detection using the time reversal process is based onthe premise that if there are certain types of defect along the wavepropagation path, time reversibility breaks down. More precisely,the shape of the reconstructed signal’s main wave packet willdepart from that of the original input signal and the symmetry ofthe reconstructed signal is violated. By examining the deviationof the restored signal’s main wave packet from the known inputsignal or the violation of the reconstructed signal’s symmetry asshown in Fig. 4, certain types of damage can be identified withoutrequiring any previously obtained baseline signals. Based on thispremise, two indices are proposed for damage identification: timereversibility �TR� and symmetry �SYM� indices. The TR index,defined below, compares the waveform of the original input withthat of the reconstructed signal

Fig. 4. Definition of tl, to, and tr used in Eqs. �1� and �2�

Fig. 5. Comparison of two forwarding signals �A to B versus B to A�wave propagation: �a� forwarding signals before debonding; �b� forw



TR = 1 −��tltr I�t�V�t�dt�2��tltr I�t�2dt�tltr V�t�2dt��1�

where I�t� and V�t�=known input signal and the main wavepacket in the reconstructed signal, respectively. For the experi-mental study presented, a seven-peak toneburst signal is usedfor excitation; tl and tr represent the starting and ending timepoints of the toneburst signal as defined in Fig. 4. The value ofthe TR index becomes zero when the shape of the main wavepacket in the reconstructed signal is identical to that of the ori-ginal input signal. Note that the amplitude scaling differencebetween I�t� and V�t� does not affect the TR value. If V�t� devi-ates from I�t�, the TR index value increases and approaches 1.0,indicating the existence of damage along the wave propagationpath.

The SYM index measures the degree of symmetry of the re-constructed signal with respect to the main wave packet in themiddle.

SYM

= 1 −��totr L�− t�R�t�dt�2��tlto L�t�2dt�totr R�t�2dt��2�

where L�t� and R�t�=left-hand and right-hand sides of the recon-structed signal with respect to the main wave packet; to=centertime point of the main wave packet; and tl and tr=starting andending time points as defined for the TR index. All terms areshown in Fig. 4. Because time reversibility is based on the linearreciprocity of elastic waves, one propagating from PZT A to PZTB and the other from PZT B to PZT A should be identical for thesame input signal when the system stays in a linear regime. Thislinear reciprocity is shown to break down when there is a CFRPdebonding along the wave path as shown in Fig. 5. In Fig. 5�a�,the two forwarding signals, A-to-B and B-to-A, are comparedbefore introducing debonding. The two signals are almost identi-cal in a practical sense. Once debonding is introduced, noticeabledifferences between the two forwarding signals are observed �Fig.5�b��. When a CFRP layer was debonded from the concrete sub-strate, a top surface of the concrete substrate often came off withthe CFRP layer, causing an anisotropic behavior of the system. Itis speculated that this anisotropic behavior caused the discrepancybetween the two forwarding signals in Fig. 5�b�. More profound

mine the effect of the CFRP debonding on linear reciprocity of elasticsignals after debonding

to exaarding

7


material level research is warranted to investigate the source ofthe anisotropic behavior, but this was beyond the scope of thisstudy. In this study, the TR and SYM indices are used for thefollowing experimental study to identify the initiation and propa-gation of CFRP debonding.

Fig. 6. Test setup and configuration of strain gauge and active sensin�a� elevation of a CFRP strengthened RC beam with strain gauge locatRC beam showing the PZT sensors and the sensing zones; �c� testing

Fig. 7. Changes of the TR and SYM indices measured at selective lselective loading steps; �b� SYM indices at 6 selective loading steps

JOURNA


Descriptions of Experiments

The overall configuration of the test specimen is shown in Fig. 6.The test specimen consists of an RC beam with a surface-mounted preformed CFRP strip. The RC beam is 254-mm deep,

ces embedded/attached to the CFRP strengthened RC beam �in mm�:hown upside down�; �b� inverted plan view of the CFRP strengthenedguration; and �d� layout of the active sensing data acquisition

steps during the Case I monotonic loading test: �a� TR indices at 6

g deviions �s

confi

oading



Table 1. TR Index Values during the Monotonic Loading Test �Case I�

Steps�load, kN�

Sensing zones

1 2 3 4 5 6 7 8 9 10 11 12 13 14

1 �4.18� 0.017 0.040 0.159 0.026 0.073 0.057 0.121 0.038 0.070 0.041 0.059 0.084 0.027 0.036

2 �13.09� 0.020 0.050 0.044 0.021 0.101 0.170 0.470 0.073 0.076 0.086 0.068 0.008 0.010 0.0293 �22.02� 0.048 0.006 0.021 0.036 0.088 0.062 0.637 0.046 0.018 0.013 0.046 0.072 0.013 0.0094 �30.87� 0.062 0.034 0.080 0.069 0.026 0.093 0.501 0.249 0.024 0.049 0.101 0.105 0.017 0.0665 �39.68� 0.037 0.048 0.055 0.051 0.150 0.076 0.560 0.224 0.110 0.243 0.111 0.005 0.019 0.0596 �41.59� 0.025 0.045 0.046 0.036 0.187 0.040 0.817 0.274 0.169 0.136 0.099 0.005 0.024 0.0437 �45.37� 0.022 0.036 0.058 0.061 0.038 0.212 0.601 0.369 0.208 0.328 0.087 0.009 0.004 0.0198 �47.15� 0.021 0.021 0.056 0.195 0.059 0.129 0.607 0.374 0.259 0.440 0.330 0.029 0.025 0.0419 �48.99� 0.020 0.024 0.041 0.057 0.079 0.067 0.786 0.370 0.971 0.237 0.124 0.029 0.089 0.07210 �49.74� 0.016 0.027 0.067 0.050 0.080 0.146 0.730 0.835 0.796 0.869 0.273 0.027 0.114 0.08411 �50.32� 0.013 0.027 0.100 0.136 0.172 0.165 0.972 0.991 0.564 0.821 0.093 0.037 0.130 0.11512 �51.18� 0.011 0.027 0.089 0.126 0.209 0.351 0.925 0.974 0.913 0.459 0.124 0.039 0.117 0.12913 �51.82� 0.016 0.036 0.075 0.147 0.174 0.217 0.415 0.977 0.825 0.659 0.167 0.037 0.100 0.11314 �52.02� 0.021 0.044 0.092 0.167 0.388 0.190 0.981 0.747 0.908 0.497 0.268 0.042 0.107 0.12015 �52.22� 0.025 0.041 0.090 0.106 0.422 0.216 0.911 0.815 0.479 0.993 0.067 0.042 0.085 0.10216 �52.11� 0.029 0.046 0.076 0.129 0.367 0.673 0.691 0.617 0.963 0.681 0.765 0.027 0.082 0.10717 �51.92� 0.022 0.048 0.074 0.396 0.888 0.585 0.778 0.888 0.732 0.821 0.507 0.031 0.105 0.11318 �51.95� 0.032 0.045 0.069 0.166 0.699 0.550 0.807 0.727 0.588 0.745 0.371 0.033 0.102 0.13019 �50.51� 0.019 0.045 0.084 0.184 0.774 0.555 0.474 0.696 0.969 0.917 0.268 0.029 0.112 0.09520 �48.94� 0.016 0.048 0.080 0.140 0.981 0.961 0.855 0.570 0.324 0.688 0.304 0.053 0.108 0.08921 �48.27� 0.013 0.043 0.082 0.133 0.844 0.594 0.777 0.551 0.772 0.778 0.290 0.026 0.111 0.10922 �47.77� 0.010 0.047 0.063 0.212 0.811 0.940 0.629 0.981 0.858 0.815 0.478 0.044 0.107 0.06923 �46.86� 0.008 0.045 0.101 0.212 0.661 0.724 0.660 0.660 0.866 0.813 0.336 0.051 0.119 0.07924 �44.54� 0.007 0.044 0.056 0.114 0.860 0.617 0.278 0.682 0.681 0.900 0.307 0.091 0.105 0.064

Note: Numbers in bold represent the TR indices exceeding a threshold value of 0.20.

Table 2. SYM Index Values during the Monotonic Loading Test �Case I�

Steps�load, kN�

Sensing zones

1 2 3 4 5 6 7 8 9 10 11 12 13 14

1 �4.18� 0.000 0.000 0.000 0.001 0.006 0.000 0.065 0.002 0.000 0.015 0.015 0.000 0.001 0.000

2 �13.09� 0.000 0.000 0.000 0.005 0.028 0.001 0.465 0.012 0.004 0.010 0.016 0.002 0.002 0.0003 �22.02� 0.000 0.002 0.001 0.005 0.038 0.011 0.673 0.010 0.008 0.030 0.017 0.002 0.002 0.0024 �30.87� 0.001 0.001 0.001 0.058 0.014 0.113 0.494 0.312 0.016 0.074 0.196 0.001 0.001 0.0025 �39.68� 0.001 0.003 0.002 0.028 0.092 0.089 0.756 0.317 0.197 0.487 0.062 0.001 0.001 0.0066 �41.59� 0.002 0.013 0.018 0.067 0.251 0.095 0.438 0.115 0.195 0.121 0.141 0.002 0.004 0.0117 �45.37� 0.003 0.006 0.008 0.111 0.048 0.704 0.717 0.560 0.376 0.627 0.228 0.007 0.006 0.0088 �47.15� 0.002 0.004 0.007 0.196 0.093 0.289 0.170 0.657 0.662 0.854 0.183 0.009 0.006 0.0099 �48.99� 0.001 0.003 0.006 0.093 0.085 0.184 0.894 0.718 0.992 0.375 0.154 0.015 0.006 0.00710 �49.74� 0.003 0.007 0.011 0.059 0.117 0.296 0.556 0.882 0.818 0.317 0.198 0.006 0.005 0.01411 �50.32� 0.001 0.004 0.022 0.136 0.262 0.481 0.529 0.245 0.789 0.242 0.134 0.012 0.004 0.00712 �51.18� 0.002 0.003 0.010 0.239 0.253 0.710 0.982 0.625 0.304 0.608 0.341 0.003 0.002 0.00813 �51.82� 0.005 0.002 0.012 0.216 0.218 0.380 0.195 0.362 0.246 0.545 0.079 0.005 0.005 0.00914 �52.02� 0.002 0.003 0.011 0.318 0.927 0.257 0.421 0.284 0.756 0.314 0.215 0.006 0.007 0.01115 �52.22� 0.001 0.001 0.011 0.190 0.497 0.366 0.838 0.710 0.451 0.844 0.029 0.003 0.008 0.01216 �52.11� 0.007 0.002 0.012 0.238 0.339 0.395 0.720 0.901 1.000 0.489 1.000 0.026 0.005 0.01417 �51.92� 0.003 0.001 0.012 0.984 0.928 0.920 0.529 0.720 0.541 0.785 0.723 0.009 0.005 0.00518 �51.95� 0.010 0.002 0.011 0.150 0.735 0.148 0.686 0.842 0.770 0.737 0.577 0.016 0.005 0.00919 �50.51� 0.002 0.002 0.013 0.293 0.853 0.817 0.844 0.917 0.715 0.539 0.416 0.011 0.009 0.00820 �48.94� 0.003 0.002 0.016 0.326 0.612 0.904 0.614 0.532 0.263 0.340 0.115 0.028 0.004 0.00921 �48.27� 0.001 0.002 0.008 0.219 0.205 0.974 0.838 0.592 0.924 0.964 0.531 0.017 0.004 0.01622 �47.77� 0.002 0.002 0.006 0.295 0.944 0.958 0.473 0.619 0.901 0.333 0.817 0.013 0.004 0.01023 �46.86� 0.003 0.002 0.011 0.222 0.990 0.883 0.944 0.673 0.897 0.462 0.490 0.017 0.003 0.01924 �44.54� 0.003 0.002 0.030 0.118 0.762 0.883 0.102 0.585 0.912 0.338 0.147 0.027 0.004 0.008

Note: Numbers in bold represent the TR indices exceeding a threshold value of 0.20.


Downloaded 07 Jun 2010 to 143.248.122.147. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org

152-mm wide, and simply supported over 4,750 mm as shown inFig. 6. It is reinforced with three #4 �13-mm diameter� primaryand two #3 �9-mm diameter� compression reinforcing bars. Thesoffit-applied preformed CFRP strip is 102-mm wide and 1.3-mmthick, and the strip has a rupture strength and tensile modulus of155 and 2.8 MPa, respectively. This CFRP layer is applied overthe middle 3,750 mm of the beam span using an epoxy-basedstructural adhesive with a modulus of 2.2 MPa and a rupturestrain of 0.006. Two such beam specimens, shown in Fig. 7�c�,were used in this study �described subsequently�. The load andstrain data acquisition system �DAQ2 in Fig. 6�d�� includes �1�four electrical resistance strain gauges �labeled as strain gauges1–4 in Fig. 6�a�� attached to the internal reinforcing steel bars; �2�eight additional strain gauges mounted on the surface of theCFRP layer coincident with the previous strain gauges �labeled asStrain Gauges 5–12 in Fig. 6�a��; and �3� a load cell and displace-ment transducer for measuring the applied load and displacementat the midspan of the beam. Note that Fig. 6�a� shows the beam inan inverted position. These instruments were used to measurestrains and to identify the presence of debonding at the discretegauge locations. Details of the experimental setup and data analy-sis results based on strain measurements can be found in Reeve�2005� and Zorn �2006�.

A total of 15 square PZT wafers �2�2�0.0508 cm� wereattached on the free surface of the CFRP layer to form a distrib-uted active sensing system �Fig. 6�b��. The location of the PZTs isstrategically decided so that the TR index is not deteriorated bythe wave reflections from the structural boundary. Because thePZTs produce an electrical charge when deformed, the PZT wa-fers were used as dynamic strain gauges in DAQ1, shown in Fig.6�d�. The same PZT wafers are also used as actuators becauseelastic waves are produced when an electrical field is applied tothe wafers �Sun et al. 1995�.

The active sensing system �DAQ1 in Fig. 6�d�� was composedof an arbitrary function generator, a signal digitizer, and two mul-

Fig. 8. Debonded regions of the test specimen after the Case Imonotonic loading test �beam inverted�

Fig. 9. Changes of TR and SYM indices as a function of incrementaTR indices at Sensing Zones 10 and 14; �b� SYM indices at Sensing

JOURNA


tiplexers. In this experiment, excitation signals were applied toeven number PZT wafers, and responses were measured at oddnumber PZTs. For instance, a seven-peak toneburst signal wasapplied to PZT #2, and the corresponding forward signal wasmeasured at PZT #1. Then, the measured forward signal was timereversed and applied at PZT #2 again, and the reconstructed sig-nal was measured at PZT #1. This time reversal process wasrepeated for a total of 14 different path combinations �PZTs #2–1,PZTs #2–3,…, and PZTs #14–15�. These 14 sensing segmentswere referred to as Sensing Zones #1–14 in Fig. 6�b�. The drivingfrequency value and the sampling rate of the digitizer were set to45 and 5 MHz, respectively. In order to find the optimal excita-tion frequency, a frequency sweep test was performed from10 to 150 kHz with a 10-kHz incremental value. As the drivingfrequency increased over 50 kHz, the signal attenuation increasedsignificantly and the time reversibility deteriorated. On the otherhand, the frequency excitation below 40 kHz did not produce adesirable wavelength �resolution� that was small for disbond de-tection. During the time reversal process, 10,000 time data points,equal to 2 ms, were collected and time reversed. The length of thetime-reversed signal was selected to be long enough so that themajority of the mechanical responses were captured within thetime segment.

Three loading cases were investigated in this study �CasesI–III�. In Case I �Specimen L4 reported in Reeve 2005�, one ofthe two large-scale specimens was subjected to incrementalmonotonic loading, and the data from the active sensing systemwere collected at each loading step. The monotonic load wasgradually increased until the specimen failed. The loading wasinitially force controlled up to Loading Step 5 �40.03 kN� andthen switched to a displacement control as the beam “yielded”starting from Loading Step 6 �41.59 KN� to Loading Step 24�44.54 kN�.

The second specimen �Specimen L4F reported in Zorn 2006�was first subjected to fatigue loading in Case II and subsequentlyto monotonic loading to failure in Case III. In Case II, cyclicloads with a driving frequency of 1.3 Hz and an applied loadrange of 4.45 kN to 22.24 kN was applied. The specimen under-went a total of 2,000,000 fatigue load cycles over 16 days.Data from the active sensing system were gathered at se-veral loading cycles: N=0, 1, 100, 200, 500, 1,000, 2,000,5,000, 10,000, 162,330, 308,800, 439,880, 609,490, 721,990,891,160, 1,030,700, 1,175,690, 1,321,900, 1,428,920, 1,564,430,1,651,580, 1,786,920, 1,896,400, and 2,000,000 cycles. During

otonic loading measured at Sensing Zones 10 and 14 for Case I: �a�10 and 14

l monZones



the data collection from the active sensing system, the cyclic loadwas paused at the minimum load of 4.45 kN.

A monotonic load test similar to Case I was performed on thesecond specimen following the fatigue conditioning describedabove. This was referred to as Case III. Data were measured at 14loading steps �13.34–48.11 kN�. The force-controlled loadingwas initially applied up to Loading Step 4 �36.90 kN�, and theloading was changed to the displacement control up to LoadingStep 14 �48.11 kN�. In all three cases, the loading was applied atthe midpoint of the simply supported beam, and data from theactive sensing devices were collected while the load was heldconstant.

Experimental Results

The monotonic and fatigue load tests were performed on twoCFRP strengthened RC beams to introduce debonding betweenthe CFRP layers and the RC beams, and data were periodicallycollected from the active sensing system during the loading tests.Damage diagnosis based on the measured data is presented in thissection for all three loading cases �Cases I–III�.

Fig. 10. Forwarding and reconstructed signals measure

Fig. 11. Examination of PZT sensor functionality: new set of PZTsof the PZT sensors after the Case I monotonic loading test: �a� New PTR and SYM indices obtained from the old and new PZTs



Case I: Monotonic Loading

The damage diagnosis obtained during the monotonic loading testof the first CFRP-RC beam specimen �Case I defined in the pre-vious section� is presented in Fig. 7 and Tables 1 and 2. In Fig. 7,the TR and SYM indices defined in Eq. �1� and �2� are shownalong the length of the beam and computed at selective loadingsteps �Loading Steps 1, 2, 6, 7, 15, and 24�. All TR and SYMindices for Case I are shown in Tables 1 and 2, respectively, andthe numbers in bold indicate TR and SYM index values exceed-ing a threshold value of 0.2. The sensing zones in Fig. 7 andTables 1 and 2 are defined in Fig. 6�b�. It is observed from Fig. 7and Tables 1 and 2 that: �1� at Loading Step 2, there was a sig-nificant increase of the TR and SYM indices at the midspan point�Sensing Zone 7�; and �2� as loading increased, larger TR andSYM index values were observed near Sensing Zones 5 and 10. Itis speculated that the initial increases of the TR and SYM indicesat the midspan �Sensing Zone 7� did not result from debondingbut rather from initial cracking of the concrete beam. An addi-tional anomaly in Sensing Zone 7 was a 4.8-mm diameter steelrod embedded in the concrete used to connect the midspan dis-placement transducer whose presence may have affected the TR

at Sensing Zone 10; �b� at Sensing Zone 14 for Case I

ached side-by-side to PZTs #5–11 to inspect the proper performancettached next to PZTs #5–11 for signal verification; �b� comparison of

d �a�

are attZTs a

7


and SYM values. Further study is underway to better understandthe sources of these abnormal TR and SYM index values. In thisstudy, these TR and SYM outlier values not related to debondingwere disregarded. Overall, the findings from the proposed diag-nosis system agreed well with visual inspection performed afterthe completion of the monotonic load test and the data obtainedfrom the strain gauge system �Reeve 2005�. Fig. 8 shows thedebonding regions identified by either a visual inspection or acoin-tapping test after the completion of the test. The largest deb-onding was observed near Sensing Zones 5 and 10, which coin-cide with the results shown in Fig. 7 and Tables 1 and 2.

Fig. 12. Changes of the TR and SYM indices measured at selectivenumbers of cycles during the Case II fatigue loading test: �a� TRindices at 5 selective numbers of fatigue cyles; �b� SYM indices at 5selective numbers of fatigue cycles

Fig. 13. Changes of the TR and SYM indices as a function of increa�a� TR indices at Sensing Zones 10 and 14; �b� SYM indices at Sens

JOURNA


Next, the TR and SYM indices values at Sensing Zones 10 and14 are plotted as a function of the 24 loading steps in Fig. 9:Sensing Zone 10 is where the greatest debonding is observedalong the beam length, and Sensing Zone 14 is located at the endof the beam where no debonding is expected. Fig. 9 shows thatthe TR and SYM indices significantly increased at Loading Step 5near Sensing Zone 10, but they did not vary much near Sensing

tigue load cycles measured at Sensing Zones 10 and 14 for Case II:nes 10 and 14

Fig. 14. Forwarding and reconstructed signals measured �a� atSensing Zone 10; �b� at Sensing Zone 14 for Case II

sing faing Zo



Zone 14 throughout the entire loading steps as also reported inTables 1 and 2. This finding is again in good agreement with thevisual inspection performed after the completion of the mono-tonic loading test. In contrast, the conventional strain gaugeinstrumentation was only able to capture debonding as the deb-onding front passed the discrete location of the gauges. Reeve�2005� presents a method for determining debonding strain in theCFRP that requires posttest data reduction and some level ofqualitative interpretation of the data. In the tests reported here,this latter method was able to assess debonding strain with greataccuracy but only at discrete locations. As a result, debondingwas not observed using these conventional methods until later inthe monotonic test.

The forwarding and reconstructed time response signals mea-sured at Sensing Zones 10 and 14 are presented in Fig. 10 forselective loading steps. In these figures, the “forwarding” signal isthe response signal measured at the sensing PZT when the knownexcitation signal is applied to the exciting PZT. For example, theforwarding signal in Fig. 10�a� shows the response signal mea-sured at PZT #11 when the excitation waveform is applied to PZT#10. Then, the measured signal at PZT #11 is time reversed andapplied back to PZT #10. The corresponding response signal ismeasured at PZT #11; this is the “reconstructed” signal shown inFig. 10�a�. The following observations are made from Fig. 10: �1�as the load level increased, the static deflection of the beam in-creased near Sensing Zone 10; consequently, the strain of the PZTmaterial exceeded the maximum strain specified by the manufac-

Fig. 15. Changes of the TR and SYM indices measured at selective6 loading steps; �b� SYM indices at 6 loading steps

Fig. 16. Change of the TR and SYM indices as a function of increas�a� TR indices at PZTs #10–11 and 14–15; �b� SYM indices at PZTs



turer �so-called “strain saturation”�, and its response to the elasticwaves became less sensitive as shown in Fig. 10�a�; and �2� de-viation of the reconstructed signal from the original input signalwas observed at the increased load levels, indicating the initiationof debonding near Sensing Zone 10. To verify that the increasesof the TR and SYM indices were not caused by the large straindeformation of the PZTs, a new set of PZT wafers were attachedside-by-side to the existing PZTs #5–11 following the completionof the monotonic load test as shown in Fig. 11�a�. The resultsfrom this new set of PZT wafers are summarized in Fig. 11�b�.The data from the new PZT series was consistent with the posttestobservations presented in Fig. 11 and with the TR and SYM in-dices obtained from the original PZTs. Therefore, the changes ofthe TR and SYM indices are concluded to be related to debondingrather than PZT defects or strain saturation.

Qualitatively different results were obtained from SensingZone 14 located near one end of the beam. As mentioned previ-ously, no sign of debonding was expected or found near SensingZone 14. This is consistent with the finding that the reconstructedsignal in Fig. 10�b� did not change much throughout the entireloading history. However, it should be noted that the forwardingsignal continuously changed as loading progressed. This observa-tion clearly demonstrates the advantage of adopting the recon-structed signal for damage diagnosis rather than the conventionalmethod of using only the forwarding signal: while the forwardingsignal is sensitive to normal operational variations of the systemthat are not necessarily related to defects �increased beam shear

g steps during the Case III monotonic loading test: �a� TR indices at

onotonic loading measured at Sensing Zones 10 and 14 for Case III:1 and 14–15

loadin

ing m#10–1

7


alue of

force, in this case�, the reconstructed signal seems to be morerobust against these normal variations. The robustness of the pro-posed monitoring system against variations in temperature andtraffic loading is also investigated through field testing �Kim andSohn 2006�. The time-reversal process has been traditionally usedfor refocusing the energy to improve the signal-to-noise ratio.Furthermore, the signal-to-noise ratio is improved by repeatedlyapplying the same excitation and taking the average of thecorresponding response signals in the time domain. This timeaveraging of the response signals significantly improves thesignal-to-noise ratio. Therefore, it is expected that the proposeddebonding monitoring system based on the time reversal conceptis more suitable for field deployment and efficient for minimizingfalse-positive indications of defects. The robustness of the pro-posed monitoring system again temperature variation and ambienttraffic vibration is also investigated through field testing of Buf-falo Creek Bridge in Pennsylvania �Kim and Sohn 2006�.

Table 3. TR Index Values during the Monotonic Test after Fatigue Cycl

Load steps�load, kN� 1 2 3 4 5 6

1 �13.34� 0.063 0.021 0.079 0.005 0.077 0.138

2 �22.24� 0.065 0.017 0.049 0.012 0.153 0.159

3 �31.13� 0.081 0.020 0.043 0.055 0.064 0.2024 �36.89� 0.087 0.013 0.034 0.010 0.052 0.2455 �42.91� 0.083 0.016 0.044 0.005 0.050 0.194

6 �45.58� 0.071 0.037 0.094 0.033 0.062 0.140

7 �47.50� 0.061 0.064 0.157 0.054 0.089 0.160

8 �49.04� 0.058 0.091 0.096 0.097 0.125 0.2169 �50.01� 0.050 0.097 0.107 0.066 0.182 0.147

10 �50.93� 0.040 0.113 0.114 0.056 0.241 0.062

11 �51.12� 0.034 0.110 0.092 0.044 0.183 0.057

12 �50.85� 0.033 0.110 0.076 0.050 0.155 0.103

13 �47.99� 0.033 0.112 0.073 0.054 0.140 0.144

14 �49.16� 0.041 0.127 0.065 0.044 0.116 0.121

Note: Numbers in bold represent the TR indices exceeding a threshold v

Table 4. SYM Index Values during the Monotonic Test after Fatigue Cy

Load steps�load, kN� 1 2 3 4 5 6

1 �13.34� 0.005 0.012 0.009 0.001 0.011 0.020

2 �22.24� 0.010 0.019 0.019 0.014 0.017 0.023

3 �31.13� 0.006 0.010 0.011 0.012 0.014 0.023

4 �36.89� 0.001 0.003 0.009 0.010 0.008 0.039

5 �42.91� 0.002 0.001 0.001 0.001 0.001 0.015

6 �45.58� 0.001 0.001 0.004 0.001 0.012 0.041

7 �47.50� 0.001 0.001 0.000 0.002 0.014 0.016

8 �49.04� 0.001 0.000 0.001 0.001 0.013 0.051

9 �50.01� 0.002 0.000 0.001 0.003 0.010 0.065

10 �50.93� 0.004 0.003 0.002 0.003 0.013 0.013

11 �51.12� 0.004 0.004 0.003 0.005 0.011 0.024

12 �50.85� 0.005 0.006 0.010 0.008 0.013 0.009

13 �47.99� 0.006 0.006 0.007 0.007 0.008 0.072

14 �49.16� 0.007 0.006 0.003 0.005 0.004 0.018

Note: Numbers in bold represent the TR indices exceeding a threshold value of

JOURNA


Case II: Fatigue Cyclic Loading Test

The second specimen was subjected to 2,000,000 cycles of fa-tigue loading. As shown in Fig. 12, there was no sign of CFRPstrip debonding from the substrate concrete beam during the fa-tigue loading test nor was any evidence of debonding found fromvisual inspection or a coin tapping test performed following thefatigue loading test. Additionally, no debonding was identifiedusing the conventional strain gauge instrumentation �Zorn 2006�.The TR and SYM indices in Figs. 12 and 13 mostly remainedbelow 0.2 except in Sensing Zone 7. As previously described, theoutliers at Sensing Zone 7 are believed not to be related to deb-onding. In Fig. 14, it was observed that the reconstructed signaldid not change much although the forwarding signal varied sig-nificantly as the number of cyclic loading increased.

ding Test �Case III�

sing zones

8 9 10 11 12 13 14

8 0.014 0.009 0.056 0.025 0.177 0.020 0.0253 0.024 0.050 0.146 0.236 0.015 0.087 0.0099 0.038 0.037 0.075 0.124 0.031 0.062 0.0143 0.108 0.015 0.028 0.149 0.012 0.102 0.1024 0.143 0.032 0.027 0.073 0.011 0.097 0.0989 0.227 0.040 0.047 0.022 0.011 0.091 0.1003 0.366 0.169 0.035 0.055 0.016 0.095 0.1222 0.556 0.115 0.047 0.058 0.047 0.102 0.1153 0.177 0.128 0.095 0.051 0.052 0.125 0.1001 0.269 0.134 0.072 0.065 0.039 0.137 0.0958 0.220 0.102 0.266 0.046 0.043 0.133 0.0965 0.289 0.576 0.321 0.248 0.023 0.121 0.0945 0.246 0.197 0.222 0.044 0.033 0.123 0.0967 0.244 0.266 0.345 0.038 0.030 0.102 0.087

0.20.

ading Test �Case III�

sing zones

8 9 10 11 12 13 14

3 0.020 0.007 0.006 0.002 0.001 0.003 0.0039 0.026 0.045 0.016 0.004 0.005 0.006 0.0071 0.033 0.012 0.012 0.006 0.006 0.005 0.0073 0.013 0.008 0.008 0.006 0.007 0.005 0.0060 0.026 0.015 0.018 0.002 0.001 0.001 0.0008 0.041 0.015 0.020 0.001 0.001 0.001 0.0010 0.024 0.018 0.012 0.002 0.002 0.000 0.0018 0.068 0.005 0.009 0.009 0.001 0.004 0.0022 0.095 0.044 0.058 0.004 0.002 0.006 0.0079 0.016 0.193 0.012 0.007 0.003 0.003 0.0047 0.032 0.164 0.338 0.016 0.001 0.006 0.0075 0.064 0.350 0.278 0.012 0.004 0.005 0.0066 0.121 0.127 0.390 0.004 0.005 0.005 0.0007 0.010 0.150 0.261 0.006 0.004 0.005 0.006

ic Loa

Sen

7

0.660.870.870.690.920.870.610.590.560.840.960.540.830.73

clic Lo

Sen

7

0.980.460.910.490.970.770.720.580.530.480.610.880.380.32

0.20.



Case III: Monotonic Loading Test after FatigueLoading Test

After the Case II fatigue loading test, the same specimen wassubjected to a monotonic loading test similar to that describedpreviously. As shown in Fig. 15, the overall findings from CaseIII are similar to those described for Case I. The TR and SYMindices shown in Fig. 16 and reported in Tables 3 and 4 furthersuggest that the debonding was initiated after Loading Step 11near Sensing Zone 10. This finding is in agreement with thatreported by �Zorn 2006� based on conventional strain gauge data.However, the debonding became visible only after Loading Step13 �see Fig. 17�a��. This demonstrates that the proposed methodmay be able to provide an earlier warning of debonding thandetailed visual inspection. Soon after the initial debonding wasidentified, the CFRP strip suddenly separated from the RC beamin a brittle manner. Fig. 17�b� shows the fully detached CFRPlayer after loading Step 14. A closer examination of Fig. 15 re-veals that debonding initiated between PZT #9 and #10 andpropagated toward PZT #15.

Conclusion

In this study, a continuous monitoring system for detecting CFRPdebonding from a host RC structure is developed. The uniquenessof this study is in developing a new concept and theoreticalframework of NDT, in which debonding is detected without usingpreviously obtained baseline data. Conventionally, damage sensi-tive features are often defined by comparing a test signal with aprevious baseline signal. For instance, damage diagnosis based onnatural frequencies requires comparison of the current resonancefrequency with the reference frequency obtained when the systemwas in a known pristine condition.

On the other hand, the proposed technique does not requirethis type of the reference data to define a damage sensitive fea-ture. The potential of the proposed method is demonstrated usingexperimental data obtained from three loading tests of two CFRP-strengthened RC beams: a monotonic load test �Case I�, a fatigueload test �Case II�, and a monotonic test following the fatigueload test �Case III�. The proposed approach successfully esti-mated the initiation and region of debonding. The experimentalresults demonstrate the potential advantage of adopting the time-reversal process for damage diagnosis over conventional ap-proaches: it is expected that the proposed reference-free NDT

Fig. 17. Visual inspection of debonding during the Case III monotdebonding of the CFRP layer following load Step 14

technique is suitable for field deployment and is particularly ef-



fective in minimizing false-positive indications of defects. In ad-dition, the proposed approach is able to sense over a larger areafor debonding monitoring than conventional strain gauges or im-pedance measurements, which provide only discrete points ornear-field measurements.

Additional experiments are underway to address further re-search questions: although this study demonstrates that CFRPdebonding from the substrate concrete media introduces nonlinearresponses, further research is warranted to better understand theexact sources of nonlinearity. At this point, it is not clear howconcrete cracking and the interaction of cracking with debondingaffect nonlinear wave propagation responses. In addition, to com-plete the development of the reference-free NDT technique, dam-age classification should be also performed without relying onpredetermined decision boundaries. Currently, a statistical cluster-ing technique is being explored to develop a reference-freedecision-making procedure so that the entire process of debond-ing monitoring can be automated without relying on prior deci-sion boundaries or baseline signals for comparison.

Acknowledgments

This research was supported by an NSF Grant No. CMS-0529208, Pennsylvania Infrastructure Technology Alliance�PITA� program and Smart Infra-Structure Technology Center�SISTeC�. The writers would like to thank the CMS ProgramManager, Dr. Shi Lui, and the PITA codirector, Professor CristinaAmon for their supports. The writers also like to thank SeungBum Kim and Dena E. De Iuliis for assisting our experiments.The first author would like to acknowledge the Electric PowerNational Scholarship Program at the Ministry of Commerce, In-dustry and Energy �MOCIE� in Gwachon, South Korea. Thebeam testing was conducted by Andrew Zorn and BenjaminReeve in the Watkins-Haggart Structural Engineering Laboratoryat the University of Pittsburgh.

References

Akuthota, B., Hughes, D., Zoughi, R., Myers, J., and Nanni, A. �2004�.“Near-field microwave detection of disbond in carbon fiber reinforcedpolymer composites used for strengthening cement-based structuresand disbond repair verification.” J. Mater. Civ. Eng. 16�6�, 540–546.

st: �a� debonding identified between load Steps 13 and 14; �b� full
onic te
Ansari, F. �2005�. “Fiber optic health monitoring of civil structures using

7


long gage and acoustic sensors.” Smart Mater. Struct. 14�3�, S1–S7.Draeger, C., Cassereau, D., and Fink, M. �1997�. “Theory of the time-

reversal process in solids.” J. Acoust. Soc. Am., 102�3�, 1289–1295.Ekenel, M., Stephen, V., Myers, J. J., and Zoughi, R. �2004�. “Microwave

NDE of RC beams strengthened with CFRP laminates containing sur-face defects and tested under cyclic loading.” Proc., 16th World Conf.on Nondestructive Testing, Montreal, Canada.

Feng, M. Q., Flaviis, F., and Kim, Y. J. �2000�. “Use of microwaves fordamage detection of fiber reinforced polymer-wrapped concrete struc-ture.” J. Eng. Mech. 128�2�, 172–183.

Fink, M. �1999�. “Time-reversed acoustics.” Sci. Am., 281�5�, 91–97.Fink, M., and Prada, C. �2001�. “Acoustic time-reversal mirrors.” Inverse

Probl., 17, R1–R38.Giurgiutiu, V., Harries, K. A., Petrou, M. F., Bost, J., and Quattlebaum, J.

�2003�. “Disbond detection with piezoelectric wafer active sensors inRC structures strengthened with FRP composite overlays.” Earth-quake Eng. Eng. Vibration, 2�2�, 213–224.

Giurgiutiu, V., and Lyshevski, S. E. �2004�. Micromechatronics: Model-ing, analysis, and design with MATLAB, CRC, Boca Raton, Fla.

Hong, S., and Harichandran, R. �2005�. “Sensors to monitor CFRP/concrete bond in beams using electrochemical impedance spectros-copy.” J. Compos. Constr., 9�6�, 515–523.

Ing, R. K., and Fink, M. �1996�. “Time recompression of dispersive lambwaves using a time reversal mirror—Application to flaw detection inthin plates.” Proc., 1966 IEEE Ultrasonics Symp., 1, 659–663.

Karbhari, V. M., Chin, J. W., and Reynaud, D. �2000�. “Critical gaps indurability data for FRP composites in civil infrastructure.” Proc., 45thInt. Soc. for the Advancement of Materials and Process Engineering(SAMPLE) Symp. and Exhibition, 45, 549–563.

Kessler, S. S., Johnson, C. E., and Dunn, C. T. �2003�. “Experimentalapplication of optimized Lamb wave actuating/sensing patches forhealth monitoring of composite structures.” Proc., 4th Int. Workshopon Structural Health Monitoring, Stanford Univ., 429–436.

Kim, S., and Sohn, H. �2006�. “Application of time reversal guided wavesto field bridge testing for baseline free damage diagnosis.” Proc.,SPIE Int. Symp., Smart Structure and Materials and NondestructiveEvaluation for Health Monitoring and Diagnostics, SPIE.

Kim, S., Sohn, H., Greve, D., and Oppenheim, I. �2005�. “Application ofa time reversal process for baseline-free monitoring of a bridge steel

JOURNA


girder.” Proc., Int. Workshop on Structural Health Monitoring, SagePublications.

Levar, J., and Hamilton, H. �2003�. “Nondestructive evaluation of carbonfiber-reinforced polymer-concrete bond using infrared thermography.”ACI Mater. J., 100�1�, 63–72.

Luangvilai, K., Punurai, W., and Jacobs, L. �2002�. “Guided Lamb wavepropagation in composite plate/concrete component.” J. Eng. Mech.,128�12�, 1337–1341.

Mal, A. K., Shih, F. J., Ricci, F., and Sauvik, B. �2005�. “Impact damagedetection in composite structures using Lamb waves.” Proc. SPIE5768, 295–303.

Mirmiran, A., Shahawy, M., and Echary, H. �1999�. “Acoustic emissionmonitoring of hybrid FRP-concrete columns.” J. Eng. Mech., 125�8�,899–905.

Mirmiran, A., and Wei, Y. �2001�. “Damage assessment of FRP-encasedconcrete using ultrasonic pulse velocity.” J. Eng. Mech., 127�2�,126–135.

Park, H. W., Sohn, H., Law, K. H., and Farrar, C. R. �2007�. “Timereversal active sensing for health monitoring of a composite plate.”J. Sound Vib., 302�1–2�, 55–66.

Reeve, B. �2005�. “Effect of adhesive stiffness and CFRP geometry onthe behavior of externally bonded CFRP retrofit measures subject tomonotonic loads.” MS thesis, Dept. of Civil and Environmental En-gineering, Univ. of Pittsburgh, Pittsburgh, Pa.

Rose, L. J. �1999�. Ultrasonic waves in solid media, Cambridge Univ.Press, New York.

Sohn, H. �2007�. “Effects of environmental and operational variabilityon structural health monitoring.” Special Issue of PhilosophicalTransactions of the Royal Society on Structural Health Monitoring,365, 539–560.

Sohn, H., Kim, S., and Park H. �2005�. “Searching for a reference-freedamage detection technique.” Proc., 2nd Int. Workshop on AdvancedSmart Materials and Smart Structures Technology, Gyeong-ju, Korea.

Sun, F. P., Chaudhry Z., Liang, C., and Rogers, C. A. �1995�. “Trussstructure integrity identification using PZT sensor-actuator.” J. Intell.Mater. Syst. Struct., 6, 134–139.

Viktorov, I. A. �1967�. Rayleigh and Lamb waves, Plenum, New York.Zorn, A. �2006�. “Effect of adhesive stiffness and CFRP geometry on the

behavior of externally bonded CFRP retrofit measures subject to fa-tigue loads.” MS thesis, Dept. of Civil and Environmental Engineer-ing, Univ. of Pittsburgh, Pittsburgh, Pa.



reference-free ndt technique for debonding detection in cfrp-strengthened...

Documents