Multi-layer perceptron for damage detection in thin aluminium plate

Structural health monitoring (SHM) is essential for maintaining the reliability and safety of various engineering structures. This study investigates the application of a multi-layer perceptron (MLP) model for detecting damage in a thin aluminium plate by analysing Lamb wave responses. A 1.6 mm thin square aluminium plate with side length of 1200 mm serves as the test structure under both healthy and damaged conditions. The dataset has been constructed using a combination of finite element (FE) simulations and experimental measurements. The problem of damage detection has been framed as a binary classification task, which has been tackled using a MLP, with the aim of distinguishing between the two states based on the Lamb wave responses. The MLP model, featuring multiple fully connected layers, has been trained to learn the complex relationships within the dataset and identify damage with high accuracy. The trained MLP model has been tested on a completely unseen test dataset and, has achieved an accuracy of 99% in classifying the plate’s condition. The results underscore the MLP’s effectiveness in accurately detecting damage in aluminium structures, offering a robust, data-driven approach for real-time structural monitoring and fault detection.