Multitechnique monitoring of fatigue damage in adhesively bonded composite lapjoints

Multitechnique monitoring of fatigue damage in adhesively bonded composite lapjoints Oleksii Karpenko, Ermias Koricho and, Anton Khomenko, Gerges Dib, Mahmoodul Haq, and Lalita Udpa Citation: AIP Conference Proceedings 65, 2 (25); doi:.63/.979 View online: http://dx.doi.org/.63/.979 View Table of Contents: http://aip.scitation.org/toc/apc/65/ Published by the American Institute of Physics

Multitechnique Monitoring of Fatigue Damage in Adhesively Bonded Composite Lap-Joints Oleksii Karpenko, Ermias Koricho, Anton Khomenko, Gerges Dib, Mahmoodul Haq,, and Lalita Udpa, Department of Electrical and Computer Engineering, Michigan State University, East Lansing, MI 882, USA Composite Vehicle Research Center, 2727 Alliance Dr, Lansing, MI 89, USA Department of Civil and Environmental Engineering, Michigan State University, East Lansing, MI 882, USA Corresponding author: karpenko@msu.edu Abstract. The requirement for reduced structural weight has driven the development of adhesively bonded joints. However, a major issue preventing their full acceptance is the initiation of premature failure in the form of a disbond between adherends, mainly due to fatigue, manufacturing flaws or impact damage. This work presents the integrated approach for in-situ monitoring of degradation of the adhesive bond in the GFRP composite lap-joint using ultrasonic guided waves and dynamic measurements from strategically embedded FBG sensors. Guided waves are actuated with surface mounted piezoelectric elements and mode tuning is used to provide high sensitivity to the degradation of the adhesive layer parameters. Composite lap-joints are subjected to fatigue loading, and data from piezoceramic transducers are collected at regular intervals to evaluate the progression of damage. Results demonstrate that quasi-static loading affects guided wave measurements considerably, but FBG sensors can be used to monitor the applied load levels and residual strains in the adhesive bond. The proposed technique shows promise for determining the post-damage stiffness of adhesively bonded joints. INTRODUCTION Adhesively bonded joints are being extensively used in aerospace, marine and automotive industries due to their light weight and excellent mechanical properties []. Bonded joints distribute forces over large areas and eliminate stress concentrations. Despite the many advantages, one of the major concerns for bonded joints is the fatigue degradation of the adhesive layer, which reduces the load carrying capacity and initiates failure in the form of a disbond between adherends. Hence, the development of cost efficient methods for structural health monitoring (SHM) is essential to ensure integrity of adhesively bonded joints and meet the reliability requirements for complex structures. Recently, guided wave (GW) techniques have demonstrated great potential for SHM of fiber-reinforced plastics [2]. Guided waves can be actuated and sensed by small, unobtrusive surface-mounted or embedded piezoelectric wafers. At the same time, GW facilitate rapid screening of structural components due to their ability to propagate far distances and detect superficial and internal flaws. It has been also demonstrated that guided waves are sensitive to degradation of shear modulus of the adhesive layer and disbonds between adherends in lap-joints [3, ]. However, bonded joints are usually designed to connect different components of the structure and transfer significant loads, which affects the propagation characteristics of ultrasonic waves [5, 6]. This work presents a hybrid concurrent measurement technique that uses a combination of guided waves and fiber Bragg-grating sensors (FBG) to monitor the fatigue degradation of the adhesive bond in composite lap-joints at different quasi-static loading levels. Pitch-catch guide wave measurements are performed with two PZT transducers placed at different sides of the adhesive bond. The FBGs are embedded into the adhesive layer to provide accurate and localized strain readings during fatigue tests, since their small size does not affect the intrinsic properties of the adhesive. Additional data from FBG sensors allow for redundancy in measurements and show potential in estimation of the applied tensile loads, which is highly essential in SHM of bonded lap-joints. The first section of the paper describes the procedure for manufacturing of composite samples, and the experimental set-up for fatigue testing and data collection. In the next section, the propagation of guided waves in the lap-joint is analyzed in order to identify the best modes and actuation frequencies for fatigue damage detection. The observed pitch-catch signals are demonstrated for different fatigue cycles and three quasi-static loading levels. Finally, the st Annual Review of Progress in Quantitative Nondestructive Evaluation AIP Conf. Proc. 65, 2- (25); doi:.63/.979 25 AIP Publishing LLC 978--735-292-7/$3. 2

last section provides the wavelength and strain measurements obtained with strategically embedded FBG sensor to complement the guided wave features and determine the amount of fatigue damage. Manufacturing of Adhesively Bonded Lap-Joints A relatively large scale lap-joint was manufactured from two quasi-isotropic GFRP composite plates which were then bonded with FM-9K adhesive film. The composite adherend was infused with SC-5 epoxy and comprised of 8 layers of plain-weave S2-glass []. Material properties of the adherends and adhesive were experimentally obtained and are shown in Table and Table 2. The FBG sensor (Micron Optics, OS-) with Bragg wavelength of 56 nm was placed inside the adhesive bond of the lap-joint along its longer side. The lap-joint was then cured at 27 C for 2 hours followed by gradual cooling. TABLE. Elastic constants of GFRP laminate obtained from tensile tests. E E 22 E 33 G 3 G 23 G 2 ν 3 ν 23 ν 2 ρ 23. 23. 6.9 2.5 2.5 3.2.28.28.36 97 Young s moduli E and shear moduli G are in GPa; material density ρ is in kg/m 3. TABLE 2. FM-9K. E ν ρ..33 22 The manufactured lap-joint (3 5 mm) is shown in Fig.. The thickness of composite plates and adhesive layer were.7 mm and. mm, respectively. The bondline length of the lap-joint was 5 mm. Two PZT transducers (7 7.2 mm) were bonded to the surfaces of composite plates using commercial epoxy at the distances of mm and 6 mm from the edges of the bond. Experimental Set-up The experimental setup for dynamic data collection and fatigue testing is illustrated in Fig.a and Fig.b. Preliminary tests were performed on similar composite lap-joints without sensors to determine the ultimate tensile strength. The result of a tensile test at 3 mm/min loading rate is shown in Fig. 2a. The relative displacement between the two adherends was measured using appropriately placed laser reflectors and laser extensometer. The lap-joint failed at approximately 9 kn and it was decided to run the fatigue testing at the 5% of this load level with cyclic frequency of Hz. The sinusoidal load was applied in tension to avoid buckling effect. (a) (b) (c) (d) FIGURE. a) Data collection system; b) lap-joint in the grips of the MTS machine; c) lap-joint with surface-bonded PZT & embedded FBG sensors; d) sideview of the adhesive bond. In the present work, guided waves were actuated with the help of the arbitrary waveform generator Agilent 3322 and collected at receiving PZT by the oscilloscope Agilent DSOA with a charge amplifier. The system was 3

synchronized in such way that after a certain number of fatigue loading cycles the tensile load on the grips of the MTS machine could be set to a user specified level (e.g. kn, 2 kn and kn) in order to enable guided wave pitchcatch measurements. FBG data were collected continuously at a rate of 5 Hz using the Micron Optics SM 25-7 interrogator. Load, kn 5 5.5.5 2 2.5 3 3.5 Displacement, mm (a) Stiffness, N/mm 7 Load = 2 N 6 Load = N 5 3 2 2 6 8 2 6 (b) FIGURE 2. a) Load-displacement curve obtained from the tensile test; b) Stiffness reduction curves obtained at different quasistatic loading levels. Figure 2b demonstrates the stiffness reduction curves for the quasi-static load levels of 2 kn and kn computed from the force and displacement measurements of the MTS unit. It turns out that the stiffness of the lap-joint reduces dramatically during the first 2 fatigue cycles and later decays very gradually. Hence, it is essential to detect and quantify fatigue damage in the lap-joint in order to predict its strength and residual load-carrying capacity. At the same time, the hybrid PZT/FBG monitoring system should keep track of damage progression to assess the durability of the joint design as a part of a larger composite structure. GUIDED WAVE MONITORING OF THE ADHESIVE BOND Maximal sensitivity of guided waves to stiffness degradation of the adhesive bond can be achieved by careful selection of the mode and excitation frequency. Phase velocity dispersion curves for the composite plate and multilayer bondline (Fig. 3) were obtained using the transfer matrix method [7]. The cross-section of the lap-joint at the adhesive bond has the total thickness of approximately 9.8 mm, and it supports more GW modes at the same actuation center frequency than a.7 mm thick composite plate. Hence, it is essential to excite a single guided wave mode to simplify signal processing and minimize the effect of mode conversion at the step interface. 6 6 5 5 Phase velocity, m/s 3 2 Phase velocity, m/s 3 2 2 3 5 6 Frequency, khz (a) 2 3 5 6 Frequency, khz (b) FIGURE 3. Phase velocity dispersion curves: a) composite substrate; b) multilayer bondline. Figure demonstrates the effect of stiffness reduction in the adhesive layer on the fundamental S and A guided wave modes in the low frequency range. Shear modulus of the adhesive bond was reduced form GPa to.6 GPa and the group velocity curves were calculated using the following equation c gr = c 2 ph c ph fd c ph ( fd) () where c ph is the phase velocity, (m/s), f the actuation frequency, (Hz) and d the half-thickness of the host structure, m.

Phase velocity, m/s 5 3 2 A healthy bondline S healthy bondline A reduced stiffness S reduced stiffness Group velocity, m/s 5 3 2 A healthy bondline S healthy bondline A reduced stiffness S reduced stiffness 5 5 Frequency, khz (a) 5 5 Frequency, khz (b) FIGURE. curves. Adhesive bond before and after fatigue damage: a) phase velocity dispersion curves; b) group velocity dispersion It follows that the group velocities of the fundamental modes reduce at excitation frequencies of 8 khz and 25 khz. Therefore, the time-of-flight (ToF) of received wavepackets in pitch-catch measurements will increase proportionally to the degradation of the adhesive bond if no external loads are applied to the lap-joint. On the other hand, the amplitude transmission coefficients of S and A modes will decrease with fatigue cycles, since the microcracks in the bondline will scatter the incident waves more intensively. Hence, the ToF and amplitude transmission coefficients of fundamental A and S modes, excited at 8 khz and 25 khz, can be used as guide wave features sensitive to fatigue damage. Normalized displacement.5 5 5 2 25 Frequency, khz S A Thickness, mm 3 2 In plane Out of plane 5 5 Displacement, m x 2 (a) (b) (c) Thickness, mm 3 2 In plane Out of plane 2 2 Displacement, m x 2 FIGURE 5. a) Frequency mode tuning using the surface-bonded 7 7.2 mm PZT; b) A mode shape at f = 8 khz in composite plate; c) S mode shape at f = 28 khz in composite plate. Thickness, mm 8 6 2 In plane Out of plane 2 2 Displacement, m x (a) (b) (c) Thickness,xmm 2 8 6 In plane Out of plane 2 2 Displacement,xm x2 22 FIGURE 6. a) Cross-section of the bondline; b) mode shape of A mode at f = 8 khz in the bondline; c) mode shape of S mode at f = 28 khz in the bondline. Selective actuation of the A mode with respect to S mode and vice verse can be accomplished by examining the PZT strain transfer to the surface of the composite plate, derived from the pin-force actuation model [2]. Figure 5a illustrates the displacement induced into the host structure by the PZT transducers used in experiments. A mode is dominant at khz and the S mode can be selectively excited at 25 khz. Mode shapes in the vicinity of these frequencies are shown in Fig. 5 and Fig. 6. The fundamental symmetric mode interrogates the adhesive layer in axial direction and antisymmetric mode has dominant displacement in transverse direction. The particle motion for A and S modes across the thickness of the structure is very similar for the composite plate and the multilayer bondline, which ensures minimal energy leakage to other modes after the propagation through the step interfaces [3]. 5

Effect of Quasi-Static Loading on Guide Wave Features It has been recently demonstrated that uniform external loading introduces considerable effect on propagation of guided waves in metal and composite structures [6, 8]. Since guided waves are actuated with surface-coupled PZT elements, the axial stress applied to a lap-joint changes the magnitude of the PZT-induced strain. This follows from the linear equations of forward and reverse piezoelectric effect: ε i = Sijσ E j + d mi E m D m = d mi σ i + ξik σ E (2) k where σ is the stress vector applied to the PZT (N/m), ε is resulting strain vector, E is the vector of applied electric field (V /m), ξ is permittivity (F/m), d is the matrix of piezoelectric constants (m/v ), S is the compliance matrix (m 2 /N), D is the vector of electric displacement (C/m 2 ). The indices i, j =,2...6 and m,k =,2,3 refer to different directions within the material coordinate system (Fig. 7). FIGURE 7. Piezoelectric effect. The stress along the length of the lap-joint is not uniformly distributed and reaches its peak values near the step interfaces of the bondline. Therefore, PZT transducers should be placed far enough from these regions to avoid larger influence of axial stress on the amplitude of generated guided waves. Impact of the axial stress on the time-of-flight of the fundamental modes is mainly determined by changes in geometry and phase velocities: ΔToF = ΔToF G + ΔToF V (3) Due to the Poisson s effect, the lap-joint stretches in axial direction and contracts in transverse direction. This reduces the frequency-thickness product and makes the path between two transducers longer. At the same time, the quasi-static loading causes creeping, which leads to residual strains in the bondline. Therefore, this factor additionally contributes minor changes to ΔToF G even after the load has been released. Finally, the time difference ΔToF V determined by velocity change is dependent upon the acoustoelastic effect. For plate-like structures, the acoustoelasticity relates the change in phase velocity c ph to stress σ through the constant K: Δc ph = Kσ () c ph The constant K depends on the direction of the applied stress, guided wave mode as well as 2 nd order Lamé constants and 3 d order Murnaghan elastic constants. Numerical values for these parameters are not readily available for the materials used in the experimental work due to a complicated setup required for taking measurements. Guided Wave Pitch-Catch Measurements Obtained from Fatigue Tests Figures 8 5 demonstrate the guided wave data obtained experimentally during fatigue testing of adhesively bonded composite lap-joint. A and S modes were actuated using the narrowband Morlet wavelet with center frequencies of 8 khz and 28 khz, respectively. Every pitch-catch signal was averaged 32 times for better SNR level and passed through a zero-phase bandpass digital filter with cut-off frequencies of 5 khz and 7 khz. The ToF was measured as the time difference between peaks of the envelopes of the actuation waveform and the received wavepacket. Signals corresponding to kn quasi-static loading are not shown for brevity. Figure shows that the amplitude of the fundamental S mode decays exponentially with fatigue cycles at zero quasi-static load. The amplitude reduction gets larger on the load levels of 2 kn and kn. This can be explained by forced opening of microcracks in the bondline, which increases scattering considerably. A different trend can be observed for the ToF change, which monotonically increases with the progression of fatigue damage. The time delay in S mode arrival nearly saturates at about 2 cycles that agrees very well with the experimental stiffness reduction 6

Amplitude, V.5.5.2..6.8.2 Time, S x 2 5 3 8 6 2 5 75 5 26 FIGURE 8. Pitch-catch guided wave signals (S mode, f = 28 khz) obtained with zero quasi-static load. Amplitude, V.5.5.2..6.8.2 Time, S x 2 5 3 8 6 2 5 75 5 26 FIGURE 9. Pitch-catch guided wave signals (S mode, f = 28 khz) at quasi-static loading level of 2 kn. Amplitude, V..9.8.7 Load = N Load = 2 N Load = N.6.2..6.8.2..6.8 2 x FIGURE. levels). Amplitude transmission of S mode across the bondline after fatigue (measured at different quasi-static loading x 6.5 ToF increase, S.5 Load = N Load = 2 N Load = N.2..6.8.2..6.8 2 x FIGURE. Time-of-Flight of S mode across the bondline after fatigue (measured at different quasi-static loading levels). 7

Si ApTlt ume d.2.2.2.2 6 6.2 8 8.2.2 5Time, x FalTgt mcycpms 6 2 8 3 V 6 2 72 2 6V FIGURE 2. Pitch-catch guided wave signals (A mode, f = 28 khz) obtained with zero quasi-static load. Si ApTlt ume d.2.2.2.2 6 6.2 8 8.2.2 5Time, x FalTgt mcycpms 6 2 8 3 V 6 2 72 2 6V FIGURE 3. Pitch-catch guided wave signals (A mode, f = 28 khz) at quasi-static loading level of 2 kn. Amplitude, V.9.8.7.6.5. Load = N Load = 2 N Load = N.2..6.8.2..6.8 2 x FIGURE. levels). Amplitude transmission of A mode across the bondline after fatigue (measured at different quasi-static loading ToF increase, S x 5 2 Load = N Load = 2 N Load = N.5.5.2..6.8.2..6.8 2 x FIGURE 5. Time-of-Flight of A mode across the bondline after fatigue (measured at different quasi-static loading levels). 8

curve. The applied tensile load shifts the ToF curve up, but does not affect its shape. The features of A mode exhibit similar behavior in the unstressed state. However, the ToF change is several times larger compared to the S mode (Fig. 5). External loading affects the amplitude reduction of the A mode on the higher fatigue cycles making the values deviate from the curve obtained for the unstressed state (Fig. ). MONITORING OF THE ADHESIVE BOND USING EMBEDDED FIBER BRAGG-GRATING SENSORS Fiber-optic Bragg-grating sensors (FBG) can be efficiently used for measuring static and dynamic strains in materials and structures []. An FBG is a sandwich-like distributed reflector containing a periodic oscillation of refractive index that is embedded into the optical fiber (Fig. 6). Such a structure acts as an optical filter that transmits the entire spectrum of the light source and reflects back the resonant Bragg wavelength (Fig. 7a). The Bragg wavelength, λ B is given by the following equation: λ B = 2nΛ, (5) where n is the effective refractive index of the fiber core and Λ is the period of Bragg-grating, (nm). The spectrum bandwidth of the back-reflected radiation, which dictates the sensor resolution, also depends on the effective refractive index of the fiber core and Bragg-grating period. As one can see, any perturbation of the Bragggrating period results in a shift of Bragg wavelength. The strain response arises from both the physical elongation of the sensor and the change in fiber effective refractive index due to photoelastic effects, whereas a thermal response arises from the inherent thermal expansion of the fiber material and the temperature dependence of the refractive index. Reflected light Core Transmitted light Incident light Clad Bragg gratings Strain & Temperature FIGURE 6. Schematic of a fiber Bragg-grating sensor. The wavelength-encoded nature of the FBG output provides a built-in self-referencing capability for the sensor. Since the wavelength is the fundamental parameter, the output does not depend directly on the total light levels, losses in the connecting fibers and couplers, or source power. Moreover, the fiber and sensor have relatively small dimensions; therefore embedding an FBG sensor into a host specimen does not affect its intrinsic properties to any appreciable degree. These advantages of FBG sensors along with their immunity to electromagnetic interference, light weight, and high sensitivity make them very appealing for monitoring of the adhesive bonds in lap-joints. (a) (b) FIGURE 7. a) Single-peak Bragg wavelength after the infusion of the resin; b) multipeak Bragg wavelength after resin cured. Non-uniform resin shrinkage during the manufacturing of the lap-joint can affect the spectrum of back-reflected radiation, when the FBG is embedded in the bond. The residual stresses in the material deform the sensor, which results in multipeak Bragg-wavelength (Fig. 7b). Effectively, a single FBG is partitioned in a few randomly distributed 9

smaller sensors, which can be advantageous [9]. In this work, the shifts of three separate wavelength peaks were tracked simultaneously using the Micron Optics interrogator. Microstrain 2 FBG FBG2 FBG3 Fatigue loading - cycles 2 N N N 5 Time, S 5 (a) Microstrain 2 N 2 N Time, S (b) N 6 cycles cycles FIGURE 8. a) Multipeak Bragg wavelength (every single peak is marked as a separate FBG sensor), quasi-static and dynamic loading; b) monitoring of strain on different quasi-static loading levels (the central peak is used). Figure 8a demonstrates strains determined from the multipeak Bragg-wavelength on the initial stage of the fatigue test. Since the lap-joint was operated within its linear elastic range, the strain measurements correlated well with the applied quasi-static tensile load levels. With the progression of fatigue damage, FBG sensor has shown sensitivity to stiffness reduction of the lap-joint. Figure 9 illustrates the shift of the resonant Bragg wavelength as a function of fatigue cycle number at a constant loading level of kn. Clearly, the wavelength change is more pronounced in the beginning of the test, and it accurately captures the overall trend in the experimental S N curve of the lap-joint, shown in Fig. 2b. In addition, the result agrees with the A and S mode features obtained using guided wave technique. Post-damage residual strain at zero quasitatic load was also observed (Fig. 8b). FIGURE 9. Shift of the central peak of the Bragg-wavelength corresponding to different fatigue cycles and quasi-static loading of kn. The data from a few fatigue cycles on the curve were missing due to synchronization issues in the experimental set-up. Their corresponding values were interpolated using splines. Hence, the embedded FBG sensors provide valuable information about the amount of fatigue damage in the adhesive bond and the axial loading applied to a lap-joint. The latter is of great importance, because the guided wave features show strong dependence on tensile stresses. On the other hand, if the load level is known, the ambiguity in estimation of fatigue damage using ToF and amplitude transmission of A or S is minimized. The quasi-static load level can be determined more accurately by the FBG, if it is embedded in composite plate away from the adhesive bond or other regions of the structure that undergo plastic deformations. In addition to the quasitatic measurements, the FBG response during the sinusoidal cyclic loading can be used as an extra feature to determine the health of the bondline. However, in the current experimental set-up the dynamic FBG data were aliased due to low sampling rate of the interrogator and this issue will be addressed in the future work of the authors. CONCLUSIONS This paper presented a hybrid technique based on guided wave and FBG measurements for monitoring of fatigue damage in adhesively bonded composite lap-joints. Experimental results revealed that the stiffness of the lap-joint reduces dramatically during the first few hundred fatigue cycles and then decays very gradually. This trend is well

followed by guided wave features obtained from pitch-catch signals, such as ToF and amplitude of the fundamental A and S modes. However, guided wave measurements are considerably affected by the axial loading, which makes it challenging to estimate the post-damage stiffness using a single SHM technique. Data from FBG sensors complement the guided wave features, because the FBG embedded in the bondline provides measurements of strain introduced by the fatigue damage. This follows from the monotonous shift of the resonant Bragg wavelength with increasing number of fatigue cycles. On the other hand, if FBG is embedded in some other section of the lap-joint that is not affected much by the fatigue damage, it can be used to determine the approximate level of the external stress and take it into account when analyzing guide wave data. The future work of the authors will focus on the estimation of post-damage stiffness of the lap-joint using statistical data analysis that will integrate multiple features from both techniques. REFERENCES. M. Haq, L. Drzal, and E. Patterson, Health Monitoring of Composite Pi-Joints using Fiber Optic Sensors, in Amercian Society of Composites, 25th Annual Technical Conference, Dayton, Ohio, USA, 2. 2. V. Giurgiutiu, Structural Health Monitoring with Piezoelectric Wafer Active Sensors, Academic Press, 28, pp. 32 33. 3. F. L. di Scalea, P. Rizzo, and A. Marzani, Propagation of Ultrasonic Guided Waves in Lap-Shear Adhesive Joints, in SEM, 2, vol. 539 of X International Congress & Exposition on Experimental & Applied Mechanics, pp. 78 88.. N. Quaegebeur, P. Micheau, P. Masson, and A. Maslouhi, Structural Health Monitoring of Bonded Composite Joints using Piezoceramics, in International Workshop, Smart Materials, Strutures and NDT in Aerospace Conference, Montreal, Canada, 2. 5. B. Mi, J. Michaels, and T. Michaels, The Journal of the Acoustical Society of America 9, 7 85 (26). 6. N. Gandhi, J. Michaels, and S. Lee, The Journal of the Acoustical Society of America 32, 28 93 (22). 7. A. Nayfeh, The Journal of the Acoustical Society of America 89, 52 53 (99). 8. F. Chen, and P. Wilcox, Ultrasonics 7, 22 (27). 9. M. Haq, A. Khomenko, L. Udpa, and S. Udpa, Fiber Bragg-Grating Sensor Array for Health Monitoring of Bonded Composite Lap-joints, in Experimental Mechanics of Composite, Hybrid, and Multifunctional Materials, Conference Proceedings of the Society for Experimental Mechanics Series, 23, pp. 89 96.