PSI - Issue 80

Haomiao Fang et al. / Procedia Structural Integrity 80 (2026) 53–64 H. Fang et al./ Structural Integrity Procedia 00 (2025) 000 – 000

54 2

performance, leading to potential safety risks and substantial maintenance costs. Therefore, Non-Destructive Testing (NDT) becomes essential to monitor defects and assess repairs, ensuring the airframe’s reliability and facilitating regular maintenance [5, 6]. Ultrasonic guided wave testing (UGWT) is one of Non-Destructive Testing (NDT) techniques that employs ultrasonic waves to inspect and evaluate the condition of structures [7]. Due to its long propagation range and low energy loss, this technique holds great potential for detecting a range of defects, including delamination, cracks, and corrosion. [8]. In recent years, guided wave-based structure health monitoring (GWSHM) has gained significant attention and experienced rapid advancement across various engineering fields. However, guide waves are significantly influenced by the changes in environmental and operational conditions (EOCs), especially temperature and humidity variations [9]. To address this, temperature compensation techniques have been proposed to mitigate temperature effects. Ren et al. have presented a novel framework for compensating temperature effects on guided waves in composite structures with different thicknesses, based on both theoretical and experimental investigations [10]. This approach requires less baseline temperature data, enhances the damage localization performance and is well-suited for industrial settings where panels of varying thicknesses are commonly utilized. Unlike baseline approaches relying on pristine data, baseline-free methods enable the identification and assessment of BVIDs without the need for prior knowledge of the undamaged state, becoming focus of research. The current leading baseline-free approaches include time-reversal (TR) [11, 12], reciprocity [13, 14], instantaneous baseline (IB) [15]. However, these baseline-free techniques typically approximate wave propagation using limited physics-based hypotheses [16-18], making accurate modelling and interpretation of wave behaviour without a baseline significantly more challenging. Additionally, it has been found that changes in EOCs can still affect the performance of these baseline-free techniques in damage identification [17]. To address these challenges, we previously proposed a novel scalable data-driven K-SVD based framework, learning universal patterns based on a large training dataset and transferring these statistical patterns to a more complex structure based on building block (BB) philosophy [18, 19]. However, this framework can be quite time-consuming, especially when dealing with a substantial number of training data. In this study, a simple alternative is proposed for damage identification in composite coupon panels. Unlike the approach in [18], which necessitates an extensive number of training data, the proposed alternative relies solely on an analytical 5-cycle Hanning-windowed toneburst signal. This not only simplifies the process and reduces the problem's dimensionality but also minimizes data collection efforts and significantly lowers computational time. The core hypothesis supporting the methodology is that the analytical signal can be segmented and reorganized into a series of time sequences, referred to as ‘shapelets’ [20]. These shapelets refer to specific waveform patterns or sub-sequences that can represent the universal characteristics of the Lamb wave pristine signals. Detailly, a substantial number of time sequences are extracted from the entire time series, referred to as ‘shapelet candidates.’ However, these extracted candidates do not necessarily carry any distinctive patterns or representative information. To learn the most representative shapelets among these candidates, the K-SVD algorithm [18, 21, 22] has been implemented in this study, looking for a dictionary — a collection of basic functions capable of sparsely representing the data. Once completed, these shapelets are reorganized and subsequently used to reconstruct the anomalous signal for effective damage identification. This study demonstrates the effectiveness and robustness of the proposed baseline-free K-SVD method through experimental comparisons with TR, IB, and reciprocity approaches. The method successfully detects BVIDs across a temperature range of [20: 5: 50] ℃, exhibiting high accuracy and consistent localization performance in various damage scenarios. The manuscript is organized as follows: In section 2, the backgrounds and fundamentals of Lamb waves are introduced. In section 3, the methodology used in this work has been detailed, including a typical post processing pipeline to be followed and fundamentals of the proposed K-SVD baseline-free approach. In section 4, the robustness and effectiveness of the proposed methodology are validated based on damage identification performance of composite coupon panels across temperature variations.