Damage Detection of Trusses Utilizing Free Vibration Signals and Convolutional Neural Network Relied on Model Order Reduction

This chapter explores an innovative approach to damage detection in truss structures by combining free vibration signals and convolutional neural networks (CNN). The study introduces a method that relies on model order reduction to minimize the number of degrees of freedom in the analysis, making the process more efficient. The modal strain energy-index is employed to identify truss members with high damage probabilities, reducing the data dimension for the CNN model. Through numerical examples involving a 31-bar truss, the study demonstrates the effectiveness of the proposed method in accurately detecting damage with a small dataset. The results highlight the reliability and efficiency of the approach, which can be extended to other structural types such as plates, shells, and frames. The convergence of the loss function during the training process further validates the robustness of the CNN model. This chapter provides valuable insights into the application of machine learning techniques for structural health monitoring, offering a promising solution for early damage detection and prevention of structural failures.