Bridge Damage Identification Based on LSTM Network and Contact Point Response

Abstract

In this study, a bridge damage identification method based on the Long Short-Term Memory (LSTM) network and the contact point response is proposed, which utilizes the time behavior of the contact point response. The contact point response serving as the input to the LSTM network is obtained by calculating the vehicle-bridge interaction motion equation. A new bridge damage indicator titled the Euclidean Distance Damage Index (EDDI) is developed based on the difference between the actual and predicted values of the contact point response. Under ideal road surface conditions, the EDDI calculated in the healthy state of the bridge remains below 0.16, while exceeding 1.0 in the damaged state. When road roughness is considered, the EDDI is calculated to be less than 0.12 in the healthy state of the bridge and higher than 0.65 in the damaged state. The results show that the EDDI is more effective in distinguishing between the damaged and the healthy states of the bridge. Meanwhile, road roughness has a negative effect on the damage sensitivity of EDDI.