Study on cable damage identification of cable‐stayed bridges based on combined Convolutional Neural Network and Long Short‐Term Memory

Abstract Cable‐stayed bridges are one of the most popular bridge structures for mega bridge projects. In the safe operation of cable‐stayed bridge, it is necessary to accurately evaluate the state of the bridge structure and improve the accuracy of cable damage identification. In this paper, the combined damage index of frequency and total energy change rate is taken as input vector. Convolutional Neural Network (CNN) has the advantage of high‐dimensional feature extraction. Long Short Memory Network (LSTM) advantage in time series modeling ability. A combined CNN&LSTM method was proposed to recognize the cable‐stay damage of cable‐stayed bridges. Then, damage identification was performed on the cable‐stayed bridge model under single and two cable damage conditions. And varied damage conditions are selected to verify the accuracy of the combined CNN&LSTM model. The results show that the damage location and damage degree of cable‐stayed Bridges can be identified with high precision by using the combined CNN&LSTM model.