Structural Damage Detection Based on One-Dimensional Convolutional Neural Network

This paper proposes a structural damage detection method based on one-dimensional convolutional neural network (CNN). The can automatically extract features from data to detect damage. First, three-layer framework model was designed. Second, the displacement of each node collected under environmental excitation. Then, transformed into interlayer form dataset. Third, in order verify feasibility proposed method, datasets were divided three categories: single dataset, multiple and degree types dataset be classified by network. results showed that recognition accuracy is above 0.9274. Thereafter, visualization tool called “t-SNE” employed visualize raw output feature extraction ability CNN excellent. However, there are many hidden layers CNN. outputs these invisible. In last section, visualized understand how networks work.