A novel nonlinear Lamb wave based approach for detection of multiple disbonds in adhesive joints

Abstract

This paper proposes a Lamb wave based technique for detecting multiple disbonds using novel damage features based on mode conversion and nonlinear breathing phenomena. An aluminum scarf joint with a top cap has been considered as the example structure. Generation and sensing of Lamb waves have been done through piezoelectric patches. Three possible cases of disbonds, i.e., horizontal disbond between the top cap and the scarf joint, inclined disbond at the scarf joint, and the simultaneous presence of both the horizontal and inclined disbonds have been considered. Numerical simulations and experiments have been run considering disbonds of different lengths at different locations. For extraction of damage features, the wave propagation response has been decomposed into symmetric and antisymmetric components. Next, damage features have been defined using the amplitudes of the superharmonics present in each of these components. Detection of disbond has been done in two steps – localization and quantification. The problem of disbond localization has been formulated as a classification problem and solved employing a Probabilistic Neural Network (PNN) using damage features extracted from numerical simulation data. Quantificationof disbond length has been done through Artificial Neural Networks (ANN) using the same damage features. Finally, this technique has been tested using experimental data.

Introduction

Lap joint, butt joint, scarf joint, etc., are some of the common adhesive joint types used in aerospace structures. However, these joints are vulnerable to disbonds that grow under repeated loading.This necessitates an in situ system for detection of these disbonds [1,2]. Various ultrasonic techniques based on A, B and C scans are widely used for such damage detection problems because of their high accuracy and sensitivity, high penetration depth, ability to detect volumetric and crack-like defects on any orientation, etc. However, detection techniques using Lamb waves, which are one of the types of guided waves are also in the focus of the researchers for application to thin structurese.g., thin plates because of several associated advantages. Lamb waves can propagate longer distance through a plate-like structure. The Lamb wave-based approach is fast and suitable for detecting both internal and surface defects in a large area.Lamb wave based techniques can be conveniently used to monitor inaccessible areas of a structure, facilitate the development of anintegrated onboard damage detection system that can reduce inspection time and cost [3,4].

Apart from that, the Lamb wave based techniquesdon’t require transducer motion and involves low energy consumption.Two types of Lamb wave modes, i.e., A0 (antisymmetric) and S0 (symmetric) are generally used for damage detection.

In the last few decades, several researchers have worked on different techniques for damage detection based on linear and nonlinear Lamb waves [[5], [6], [7]]. A detailedreview of these techniques can be found in Refs. [8,9].Thelinear Lamb wavetechniques are based on the rational that structural damage results inphenomena likemode conversion and wave scattering.Hence, a quantitative evaluation of damage location and severity can be achieved by analyzing the wave signals scattered by damage. Lu, Ye and Su [10] demonstrated a Lamb wave based crack identification technique onan aluminum plate using a piezoelectric sensor network.Due to the dispersive nature of Lamb waves, the Lamb wave signals were purified using wavelet transform. For crack position identification, a triangulation approach based on TOF was applied using a nonlinear least-squares optimization algorithm. Park, An, and Sohn [11] proposed laser ultrasonic wavefield imaging for detection and visualization of hidden debonding and delamination in composite structures. A Nd:YAG pulse laser beam was used to generate ultrasonic waves and a laser Doppler vibrometer wasused for measuring the corresponding ultrasonic responses. Thehidden damages were identified and visualized through adopting a standing wave filter, which can isolate damage-induced standing waves from the obtained wavefield. Ren and Lissenden [12] employed modal content-based damage features for disbond detection in the adhesive joint from guided waves. Phased array transducers were used for mode excitation along the preferred direction and multielement array sensors were used for modal content extraction. The modal amplitude was used to detect the size of the disbond. Puthillath and Rose [13] demonstrated a guided wave based technique for inspection of the adhesive bond between a titanium repair patch and analuminum aircraft skin. Gauthier et al. [14] presented a guided wave based technique to detect the level of adhesion at an aluminum/epoxy interface.Ren and Lissenden [4] proposed selectionof the defect sensitive combination of mode and frequency using a concept of dispersion curve matching.Ong et al. [15] proposed a baseline free technique based on Lamb wavesto detect disbond in lap joint. A set of strategically positioned sensors were used to detect the Lamb wave modes generated by disbond. Recently Gauthier et al. [16] proposed the use of Lamb wavesand vertical modes for the characterization of adhesive bonds. Wang et al. [17] proposed a modally selective ultrasonic guided wave for identifying and sizing disbond in a multilayer bonded structure. Three key issues including the ultrasonic guided wavesensibility/excitability, the mechanism involved in the interaction of the guided wave with disbondsof different sizes, and separation of the preferred guided wave mode from other modes, were addressed.

Linear Lamb wave based techniques are effective only when the scale of damage is comparable to the wavelength of a probing Lamb wave. The use of Lamb waves of higher frequency can be a way to increase the sensitivity. But, this introduces challenges in interpreting captured Lamb wave signals because of the co-existence of multiples modes at high frequencies. The nonlinear features of Lamb waves such as sub and superharmonics due to contact nonlinearity are relatively more sensitive to cracks, disbonds of a smaller size than the linear features and thus generally have a higher dynamic range. Hence, for a given size ofdamage, the nonlinear technique can be applied employing Lamb waves of relatively lower frequencies. Apart from that, the linear methods struggle to detect damage in the plate structures with stiffeners and other geometric features because of multiplereflections and mode conversions, making the temporal data much more chaotic [9,18].

Meo, Polimeno andZumpano [19] proposed two different nonlinear elastic wave spectroscopy (NEWS) methods: single-mode nonlinear resonance ultrasound (NRUS) and nonlinear wave modulation technique (NWMS) and showed application on various composite materials. The NRUS method was based on observation of resonance shift caused by an increment of driving amplitude. The NWMS method monitors the generation of harmonics and sidebands on the spectrum of acquired signals excited by a bi-tone signal (NWMS). A review of various nonlinear ultrasonic techniques can be found in Ref. [20]. Yelve, Mitra and Mujumdar [18] worked on the detection of disbond beneath a stiffened plate using Lamb wave based nonlinear technique. Harmonics presentbecause ofcontact nonlinearity were used to form damage index for the detection of the disbond length.Yan, Drinkwater and Neild [21] studied wave propagation through kissing bonds present in adhesive joints. When propagation of ultrasonicwave takes place through the kissing bond area, nonlinearity is induced. They observed that under light compressive loads, significant nonlinearity is exhibited and with an increase in compressive load there is a rapid reduction in the level of nonlinearity.Yelve, Tse andMasurkar [22] proposed a nonlinear Lamb wave based techniquefor delamination detection in a composite laminate. A spectral damage index was formedusing higher harmonics to detect the delamination size. The delamination was located using a hybrid method using both spectral and temporal information. Scarselli et al. [23] analyzed the nonlinear content of structural response of an adhesive lap joint.Depending on the excitation frequency, super and subharmonics with different levels of amplitude were observed to be present in the measured spectra due to contact acoustic nonlinearities or nonlinear oscillations of bubbles present in theadhesive lap joint. Mandal and Banerjee [24] identified breathing disbonds in metallic plates with stiffener using a nonlinear Lamb wave based technique. Clock-like array sensors of piezoelectric wafer transducers were used for sensing. The second harmonic component due to contact acoustic nonlinearity was extracted from received signals through bandpass filtering followed by wavelet transformation. Disbond detection wasdone in two steps, i.e., localization and quantification.Nicassio, Carrino, and Scarselli [25] proposed a nonlinear Lamb wave and local defect resonance (LDR) based method to locate and evaluate disbonds present in single lap joints. The wave-damage interaction was observed to reach maximaat particular excitation frequencies, which enhance the nonlinear response through LDR. The subharmonics were exhibited by Lamb wavesat LDR frequency due to defects. An algorithm based on propagation properties and time of flight of subharmonics packet was used to localize the defect.Inspite of significant research progress, Lamb wave based techniques, particularly the nonlinear techniques, still don’t find wide application in industrial practice due to several reasons. To enhance the practical applicability of these techniques research effort is needed to address the complexities of the inverse problem associated with the presence of multiple damages of one or more types and geometric complexities, reduce the requirement of sensors and actuators [8,9]. Other problems often faced with the nonlinear technique are due to the distortion of signals, especially of the superharmonic components making it difficult to extract the damage features properly.

In the present work, a Lamb wave based nonlinear technique has been used for detection and localization of disbond in scarf joint with acap at the top. Such a joint involves multiple adhesive bonds. The possibility of disbonds in these multiple joints at different orientation makes the detection problem relatively more challenging. Literature dealing with such a problem involving multiple adhesive bonds has been observed to be rather limited by the authors. To solve this complex inverse problem using a limited number of sensors, novel features have been formulated from the mode conversion and superharmonic generation characteristics of the disbonds.Usingthese features, the disbonds detection problem has been solved in two steps – localization and quantification. The localization problem has been solved as a classification problem through a PNNand the quantification problem has been solved as a regression problem througha set of ANNs using damage features extracted from numerical simulation data. Such a data-based approach generally requires a huge database. Creating such a database through experiments or expensive 3D simulations may not always be feasible. Several researchers [18,26] have used 2D plane strain simulation due to its reduced computational complexity. However, the robustness of a damage detection system trained using 2D simulation in detecting damages from experimental data has not been adequately discussed in the literature to the best of the authors’ knowledge. In this paper, it has been shown that the PNN and ANNs trained using the proposed novel damage features extracted from 2D simulation data are robust enough in localizing and quantifying damages from experimental data.