Identification of Composite Damages in Cement-Based Structures Using Convolutional Variational Autoencoder Combined with Surface Wave Dispersion Energy

The chapter investigates the application of convolutional variational autoencoders (CVAE) in identifying complex composite damages within multilayered cement-based structures. It focuses on the use of surface wave dispersion energy (SWDE) derived from ultrasonic signals to enhance the accuracy of structural health monitoring. The chapter begins by discussing the limitations of traditional non-destructive evaluation (NDE) methods, particularly the multi-channel analysis of surface waves (MASW) method, in addressing material heterogeneity and randomness in concrete structures. It then introduces the CVAE method, which combines convolutional neural networks (CNN) for feature extraction with the variational autoencoder (VAE) framework for data compression and reconstruction. The chapter provides a detailed explanation of the MASW method, including the generation of surface waves and the extraction of SWDE. It also describes the numerical solution of ultrasonic signals using the finite difference method and the modeling of multilayered cement-based structures. The results demonstrate the superior performance of the CVAE model in classifying different damage states, achieving high accuracy and stability. The chapter further explores the robustness of the proposed method under varying noise levels, highlighting its potential for practical applications in complex structural environments. The findings underscore the efficacy of integrating deep learning with traditional NDE methods for more accurate and reliable structural health monitoring.