MULTIPLE CRACK DETECTION METHOD FOR THIN-PLATE STRUCTURES BASED ON DATA-DRIVEN ALGORITHM

In this paper, the scaled boundary finite element methods (SBFEM) are used to simulate the propagation process of Lamb wave in thin-plate structures, and the correlation between the defect parameters and the location of observation points when the Lamb wave propagates in the structure is studied by combining the SBFEM and the maximum information coefficient, which provides a basis for the selection of sensor location during defect inversion. On this basis, a multiple crack inversion method for thin-plate structure based on SBFEM data sets and deep learning is established. The multiple crack inversion is classified as a classification and regression prediction problem, which can inverse the number, location and size of cracks without any prior knowledge about the number of cracks. Finally, the performance of the model is verified by numerical examples. The proposed method can better classify the number of crack-like defects and identify their parameters.