Microcrack Localization and Evaluation Using Nonlinear Lamb Wave-Mixing Technique and Lamb Transformer Network

• Published in: IEEE sensors journal Vol. 23; no. 16; pp. 18423 - 18433
• Main Authors: Xu, Xu, Ai, Xing, Liu, Tinghao, Lan, Jun, Li, Yifeng
• Format: Journal Article
• Language: English
• Published: New York The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 15-08-2023
• Subjects: Coders; Deep learning; Finite element method; Lamb waves; Localization; Microcracks; Sensor arrays; Training; Transformers
• ISSN: 1530-437X; 1558-1748
• DOI: 10.1109/JSEN.2023.3289954

The problem of locating and evaluating microcracks in plates simultaneously was investigated using the nonlinear Lamb wave-mixing technique and Lamb transformer (LT) network in this work. The analyses of the nonlinear effects induced by the interaction of Lamb waves and microcracks with different locations, directions, and lengths were performed using the finite-element simulation. During the analysis part, the stacked spectrum map was introduced to analyze and reveal the influence of these three factors on the spectrum of the signal received by the sensor array. In the learning stage, to further identify the relationships between the microcrack features and the stacked spectrum map, the LT network using the stacked spectrum map and transformer encoder unit was provided for the training and conducted for further prediction of the microcrack with a well-trained model. The training and prediction results show that the average correct rate obtained in the validation set is 98.72% and the generated model also performs well on the test set, achieving an accuracy of 99%. This outcome demonstrates that the proposed LT network can be employed to identify the complex connections between the spectrum of the sensor array and microcracks and to realize the simultaneous localization and evaluation of microcracks. In summary, the LT network together with the stacked spectrum map presented in this work provides a novel baseline-free method for the localization and evaluation of microcracks and will further advance the application of Lamb wave technique and deep learning methods in microcrack detection.