Inverse Characterization of Multilayered Composite Plates Using Ultrasonic Guided Waves and Machine Learning Algorithms

Ultrasonic guided waves are an effective method for detecting structural damage and characterizing material properties across various structures. However, their sensitivity to environmental and operational conditions can impact measurement reliability. This study integrates ultrasonic guided waves with a pre-trained ResNet architecture and a K-Nearest Neighbors (KNN) algorithm to address inverse problems and determine the material properties of multilayered composite plates. By leveraging advanced machine learning techniques and data-driven models, we accurately predict material properties such as mass density, shear modulus, and Lame’s constant. Our approach involves generating dispersion curves, utilizing the pre-trained ResNet model for feature extraction, and applying KNN for classification, validated by data from analytical models. The results demonstrate that our proposed machine learning model significantly enhances the efficiency and accuracy of material property determination, providing a promising tool for the rapid and non-destructive evaluation of multilayer structures.