Energy Ratio Variation-Based Structural Damage Detection Using Convolutional Neural Network

In the field of structural health monitoring (SHM), with mature development artificial intelligence, deep learning-based damage identification techniques have attracted wide attention. this paper, convolutional neural network (CNN) is used to extract feature simple supported steel beams. Firstly, transient dynamic analysis beam carried out by finite element software, and acceleration response signals under different scenarios are obtained. Then, signal decomposed wavelet packet decomposition (WPD) band energy ratio variation (ERV) index as training sample CNN. Subsequently, vibration experiment a was out, results were compared numerical simulation results. The characteristic indexes obtained making corresponding changes signal, then, experimental data input into CNN predict effect detection. show that method can successfully detect intact structure, single damage, multiple damages an accuracy 95.14% impact load, performance better than support vector machine (SVM), good robustness.