PSI - Issue 80

Alessandro De Luca et al. / Procedia Structural Integrity 80 (2026) 403–410 D Alessandro De Luca / Structural Integrity Procedia 00 (2019) 000 – 000

409

7

4. Conclusions This study explored the potential of feature-based unsupervised clustering for detecting low-velocity impact damage in Carbon Fibre Reinforced Polymer (CFRP) panels using Ultrasonic Guided Waves (UGW). A comprehensive experimental campaign was carried out on ten panels subjected to controlled impacts at their geometric centers and monitored via a network of piezoelectric transducers (PZTs). A broad set of scalar features was extracted from the acquired UGW signals, and all unique pairwise and triplet feature combinations were systematically analyzed using Gaussian Mixture Model (GMM) clustering. Clustering performance was rigorously evaluated through multiple metrics, including silhouette score, purity, and cluster balance. Results showed that specific features, most notably feature 10, consistently enhanced clustering performance when when applied to UGW data collected along the vertical propagation path crossing the panel centre (I path) at an excitation frequency of 300 kHz. Feature combinations involving feature 10 achieved high purity values, reflecting a strong alignment between the unsupervised cluster assignments and the actual structural health states (pristine or damaged). Additionally, these combinations yielded well-balanced cluster sizes and acceptable silhouette scores, indicating effective and interpretable damage classification. These findings underscore the importance of carefully selecting features and propagation paths to enable effective unsupervised damage detection in composite structures. Future work will aim to leverage these findings by discarding less informative features, paths, and frequencies, thereby reducing computational demands and enable the analysis of a larger number of panels. This approach will support the extension of clustering to higher-dimensional feature spaces and facilitate studies on more complex damage scenarios. References Boratto, T., Bernardino, H. S., Vieira, A. B., Gontijo, T. S., Bodini, M., Martyushev, D. A., Saporetti, C. M., Cury, A., Barbosa, F., & Goliatt, L. (2025). An Agglomerative Clustering Combined with an Unsupervised Feature Selection Approach for Structural Health Monitoring. Infrastructures 2025, Vol. 10, Page 32 , 10 (2), 32. De Luca, A. D., Perfetto, D., Petrone, G., de Fenza, A., & Caputo, F. (2018). Guided-Waves in a Low Velocity Impacted Composite Winglet. Key Engineering Materials , 774 , 343 – 348. De Luca, A., Perfetto, D., Polverino, A., Minardo, A., & Caputo, F. (2023). Development and validation of a probabilistic multistage algorithm for damage localization in piezo-monitored structures. Smart Materials and Structures . Du, F., Zeng, L., Sha, B., Huang, L., & Xiao, J. (2022). Damage Imaging in Composite Laminates Using Broadband Multipath Lamb Waves. IEEE Transactions on Instrumentation and Measurement , 71 . El Mountassir, M., Yaacoubi, S., Mourot, G., & Maquin, D. (2021). An adaptive PCA-based method for more reliable ultrasonic guided waves SHM: Data-driven modeling and experimental validation in high attenuating medium. Structural Control and Health Monitoring , 28 (1), e2634. Ferri Aliabadi, M. H., & Khodaei, Z. S. (2017). Structural health monitoring for advanced composite structures. Structural Health Monitoring for Advanced Composite Structures , 1 – 274. García-Macías, E., & Ubertini, F. (2022). Integrated SHM Systems: Damage Detection Through Unsupervised Learning and Data Fusion. Structural Integrity , 21 , 247 – 268. Güemes, A., Fernandez-Lopez, A., Pozo, A. R., & Sierra-Pérez, J. (2020). Structural health monitoring for advanced composite structures: A review. Journal of Composites Science , 4 . Hassani, S., Mousavi, M., & Gandomi, A. H. (2021). Structural Health Monitoring in Composite Structures: A Comprehensive Review. Sensors 2022, Vol. 22, Page 153 , 22 (1), 153. Ikotun, A. M., Ezugwu, A. E., Abualigah, L., Abuhaija, B., & Heming, J. (2023). K-means clustering algorithms: A comprehensive review, variants analysis, and advances in the era of big data. Information Sciences , 622 , 178 – 210. Liu, J., Li, Q., Li, L., & An, S. (2024). Structural damage detection and localization via an unsupervised anomaly detection method. Reliability Engineering & System Safety , 252 , 110465. Lomazzi, L., Junges, R., Giglio, M., & Cadini, F. (2024). Unsupervised data-driven method for damage localization using guided waves. Mechanical Systems and Signal Processing , 208 , 111038. Luo, K., Zhu, J., Li, Z., Zhu, H., Li, Y., Hu, R., Fan, T., Chang, X., Zhuang, L., & Yang, Z. (2024). Ultrasonic Lamb Wave Damage Detection of CFRP Composites Using the Bayesian Neural Network. Journal of Nondestructive Evaluation , 43 (2), 1 – 15. Nerlikar, V., Mesnil, O., Miorelli, R., & D’Almeida, O. (2024). Damage detection with ultrasonic guided waves using machine l earning and aggregated baselines. Structural Health Monitoring , 23 (1), 443 – 462. Perfetto, D., Pezzella, C., Fierro, V., Rezazadeh, N., Polverino, A., & Lamanna, G. (2024). FE modelling techniques for the simulation of guided waves in plates with variable thickness. Procedia Structural Integrity , 52 , 418 – 423. Perfetto, D., Rezazadeh, N., Aversano, A., De Luca, A., & Lamanna, G. (2023). Composite Panel Damage Classification Based on Guided Waves and Machine Learning: An Experimental Approach. Applied Sciences , 13 (18), 10017. Qiu, L., Fang, F., & Yuan, S. (2019). Improved density peak clustering-based adaptive Gaussian mixture model for damage monitoring in aircraft structures under time-varying conditions. Mechanical Systems and Signal Processing , 126 , 281 – 304. ASTM D7136/D7136M-15. (2015). Test Method for Measuring the Damage Resistance of a Fiber-Reinforced Polymer Matrix Composite to a Drop-Weight Impact Event .