工程結構於服役期間易因環境與操作因素產生疲勞或損傷，結構健康監測（SHM）技術對確保結構安全至關重要。蘭姆波為板狀結構常用之主動式導波監測手段，然其固有之頻散與多模態特性，使傳統物理模型與淺層機器學習方法在特徵提取上面臨挑戰，且既有研究對頻散效應如何影響物理先驗特徵之設計仍缺乏探討。
本研究以COMSOL Multiphysics建立鋁-6061薄板之壓電耦合有限元素模型，模擬蘭姆波殘差訊號並施以資料增強，建立無須真實損傷數據之訓練資料庫。據此提出三種結合物理先驗與1D-CNN之缺陷定位架構：模型A採用未修正頻散之TDoA先驗；模型B引入動態頻散補償機制修正到達時間；模型C則融合兩者。三模型皆以交叉注意力融合物理先驗與資料驅動特徵，並經多次五折交叉驗證與統計檢定比較。
結果顯示，模型A訓練不穩定、誤差最高；模型B收斂穩定且精度顯著提升；模型C經集成預測後與模型B精度相當。本研究證實，物理先驗之「品質」較「數量」對模型性能更為關鍵，可為物理知情深度學習於結構健康監測之應用提供設計參考。

Structural health monitoring (SHM) is essential for detecting fatigue damage in engineering structures. Lamb waves are widely used for active guided-wave monitoring of plates, but their dispersion and multimodal behavior challenge feature extraction, and how dispersion affects physics-informed prior design remains underexplored. This study builds a piezoelectrically coupled finite element model of an aluminum-6061 plate in COMSOL Multiphysics to simulate and augment Lamb wave residual signals without real damage data. Three 1D-CNN architectures fusing physics-based priors via cross-attention are proposed: Model A uses an uncorrected TDoA prior; Model B applies dynamic dispersion compensation; Model C fuses both. Compared via five-fold cross-validation, Model A is unstable with the largest error, Model B converges stably with significantly improved accuracy, and Model C matches Model B after ensemble prediction. This confirms prior quality outweighs quantity, guiding physics-informed deep learning design for SHM.

[1] G. T. Webb, P. J. Vardanega, and C. R. Middleton, "Categories of SHM deployments: Technologies and capabilities," Journal of Bridge Engineering, vol. 20, no. 11, p. 04014118, 2015.
[2] C. R. Farrar and K. Worden, "An introduction to structural health monitoring," Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, vol. 365, no. 1851, pp. 303-315, 2007.
[3] H. Sohn et al., "A review of structural health monitoring literature: 1996–2001," Los Alamos National Laboratory, USA, vol. 1, no. 16, pp. 10-12989, 2003.
[4] S. Gholizadeh, Z. Leman, and B. Baharudin, "A review of the application of acoustic emission technique in engineering," Struct. Eng. Mech, vol. 54, no. 6, pp. 1075-1095, 2015.
[5] F. Magalhães, Á. Cunha, and E. Caetano, "Vibration based structural health monitoring of an arch bridge: From automated OMA to damage detection," Mechanical Systems and signal processing, vol. 28, pp. 212-228, 2012.
[6] D. Montalvao, N. M. M. Maia, and A. M. R. Ribeiro, "A review of vibration-based structural health monitoring with special emphasis on composite materials," Shock and vibration digest, vol. 38, no. 4, pp. 295-324, 2006.
[7] M. Mitra and S. Gopalakrishnan, "Guided wave based structural health monitoring: A review," Smart Materials and Structures, vol. 25, no. 5, p. 053001, 2016.
[8] Z. Su, L. Ye, and Y. Lu, "Guided Lamb waves for identification of damage in composite structures: A review," Journal of sound and vibration, vol. 295, no. 3-5, pp. 753-780, 2006.
[9] V. Giurgiutiu, "Tuned Lamb wave excitation and detection with piezoelectric wafer active sensors for structural health monitoring," Journal of intelligent material systems and structures, vol. 16, no. 4, pp. 291-305, 2005.
[10] B. Xu, L. Yu, and V. Giurgiutiu, "Advanced methods for time-of-flight estimation with application to Lamb wave structural health monitoring," in Proc. International Workshop on SHM, 2009, pp. 1202-1209.
[11] C. Hu, B. Yang, B. Xiao, F.-Z. Xuan, and Y. Xiang, "Damage localization in pressure vessel using guided wave-based techniques: Optimizing the sensor array configuration to mitigate nozzle effects," Applied Acoustics, vol. 185, p. 108393, 2022.
[12] H.-W. Kim and F.-G. Yuan, "Enhanced damage imaging of a metallic plate using matching pursuit algorithm with multiple wavepaths," Ultrasonics, vol. 89, pp. 84-101, 2018.
[13] Z. Yang et al., "A review on guided-ultrasonic-wave-based structural health monitoring: From fundamental theory to machine learning techniques," Ultrasonics, vol. 133, p. 107014, 2023.
[14] F.-G. Yuan, S. A. Zargar, Q. Chen, and S. Wang, "Machine learning for structural health monitoring: challenges and opportunities," Sensors and smart structures technologies for civil, mechanical, and aerospace systems 2020, vol. 11379, p. 1137903, 2020.
[15] Y. Lu, L. Ye, Z. Su, L. Zhou, and L. Cheng, "Artificial neural network (ANN)-based crack identification in aluminum plates with Lamb wave signals," Journal of Intelligent Material Systems and Structures, vol. 20, no. 1, pp. 39-49, 2009.
[16] Z. Su and L. Ye, "Lamb wave propagation-based damage identification for quasi-isotropic CF/EP composite laminates using artificial neural algorithm: Part I-methodology and database development," Journal of intelligent material systems and structures, vol. 16, no. 2, pp. 97-111, 2005.
[17] M. M. Azad, O. Munyaneza, J. Jung, J. W. Sohn, J.-W. Han, and H. S. Kim, "Damage localization and severity assessment in composite structures using deep learning based on lamb waves," Sensors, vol. 24, no. 24, p. 8057, 2024.
[18] T. Tong, X. Zhao, J. Hua, D. Chen, J. Tan, and J. Lin, "Lamb wave-based damage localization in composite laminates using LSTM networks trained with improved loss functions," Smart Materials and Structures, vol. 34, no. 6, p. 065010, 2025.
[19] H. Liu and Y. Zhang, "Deep learning based crack damage detection technique for thin plate structures using guided lamb wave signals," Smart Materials and Structures, vol. 29, no. 1, p. 015032, 2020.
[20] C. Su et al., "Improved damage localization and quantification of CFRP using Lamb waves and convolution neural network," IEEE Sensors Journal, vol. 19, no. 14, pp. 5784-5791, 2019.
[21] O. Abdeljaber, O. Avci, S. Kiranyaz, M. Gabbouj, and D. J. Inman, "Real-time vibration-based structural damage detection using one-dimensional convolutional neural networks," Journal of sound and vibration, vol. 388, pp. 154-170, 2017.
[22] A. Rai and M. Mitra, "Lamb wave based damage detection in metallic plates using multi-headed 1-dimensional convolutional neural network," Smart Materials and Structures, vol. 30, no. 3, p. 035010, 2021.
[23] G. Chen, W. Song, W. Shao, H. Sun, and X. Qing, "Damage presence, localization and quantification of aircraft structure based on end-to-end deep learning framework," IEEE Transactions on Instrumentation and Measurement, 2025.
[24] H. Xu, L. Liu, J. Xu, Y. Xiang, and F.-Z. Xuan, "Deep learning enables nonlinear Lamb waves for precise location of fatigue crack," Structural Health Monitoring, vol. 23, no. 1, pp. 77-93, 2024.
[25] J. Achenbach, Wave propagation in elastic solids. Elsevier, 2012.
[26] J. L. Rose, Ultrasonic guided waves in solid media. Cambridge university press, 2014.
[27] L. Zeng, L. Huang, and J. Lin, "Damage imaging of composite structures using multipath scattering Lamb waves," Composite Structures, vol. 216, pp. 331-339, 2019.
[28] IEEE Standard on Piezoelectricity, ANSI/IEEE Std 176-1987, IEEE, New York, NY, USA, 1988. [Online]. Available: https://ieee.org
[29] L. Cohen, "Time-frequency distributions-a review," Proceedings of the IEEE, vol. 77, no. 7, pp. 941-981, 1989.
[30] M. Feldman, "Hilbert transform in vibration analysis," Mechanical systems and signal processing, vol. 25, no. 3, pp. 735-802, 2011.
[31] M. H. Sazlı, "A brief review of feed-forward neural networks," Communications Faculty of Sciences University of Ankara Series A2-A3 Physical Sciences and Engineering, vol. 50, no. 01, 2006.
[32] X. Glorot and Y. Bengio, "Understanding the difficulty of training deep feedforward neural networks," in Proceedings of the thirteenth international conference on artificial intelligence and statistics, 2010: JMLR Workshop and Conference Proceedings, pp. 249-256.
[33] C. Nwankpa, W. Ijomah, A. Gachagan, and S. Marshall, "Activation functions: Comparison of trends in practice and research for deep learning," arXiv preprint arXiv:1811.03378, 2018.
[34] R.-Y. Sun, "Optimization for deep learning: An overview," Journal of the Operations Research Society of China, vol. 8, no. 2, pp. 249-294, 2020.
[35] D. E. Rumelhart, G. E. Hinton, and R. J. Williams, "Learning representations by back-propagating errors," nature, vol. 323, no. 6088, pp. 533-536, 1986.
[36] M. Raissi, P. Perdikaris, and G. E. Karniadakis, "Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations," Journal of Computational physics, vol. 378, pp. 686-707, 2019.
[37] F. Moser, L. J. Jacobs, and J. Qu, "Modeling elastic wave propagation in waveguides with the finite element method," Ndt & E International, vol. 32, no. 4, pp. 225-234, 1999.
[38] COMSOL. "Selecting a Mesh Size for Wave Propagation Problems." COMSOL Technical Support Knowledge Base. https://cn.comsol.com/support/knowledgebase/1272 (accessed February 18, 2026.
[39] R. Malaeb, E. Mahfoud, and M. Harb, "Decomposition of Fundamental Lamb wave Modes in Complex Metal Structures Using COMSOL®," in COMSOL conference proceedings, Lausanne, Switzerland, 2018, vol. 23.
[40] C. A. De Moura and C. S. Kubrusly, "The courant–friedrichs–lewy (cfl) condition," AMC, vol. 10, no. 12, pp. 45-90, 2013.
[41] T.-M. Dutschmann, L. Kinzel, A. Ter Laak, and K. Baumann, "Large-scale evaluation of k-fold cross-validation ensembles for uncertainty estimation: T.-M. Dutschmann et al," Journal of cheminformatics, vol. 15, no. 1, p. 49, 2023.
[42] C. Nadeau and Y. Bengio, "Inference for the generalization error," Advances in neural information processing systems, vol. 12, 1999.
[43] J. Demšar, "Statistical comparisons of classifiers over multiple data sets," Journal of Machine learning research, vol. 7, no. Jan, pp. 1-30, 2006.
[44] T. G. Dietterich, "Approximate statistical tests for comparing supervised classification learning algorithms," Neural computation, vol. 10, no. 7, pp. 1895-1923, 1998.
[45] A. Krogh and J. Vedelsby, "Neural network ensembles, cross validation, and active learning," Advances in neural information processing systems, vol. 7, 1994.