A Post-Seismic Damage Detection Strategy in Time Domain for a Suspension Bridge with Neural Networks

XU Bin, WU Zhishen, YOKOYAMA Koichi

doi:10.2208/journalam.6.1149

A neural-network-based damage detection approach with the direct use of actual incomplete time series of earthquake response is developed for a suspension bridge. Two neural networks are constructed and trained using the segment of the timeseries of seismic responses on several locations of the bridge, when earthquakeis in small level, to identify the transversal and vertical velocities responsesat the deck in the middle of the main span of the suspension bridge. The two neural networks are assumed as nonp rametric models for the bridge in health condition before the earthquake occurred. The performance of the trained emulator neural network models for the suspension bridge is evaluated by numerical simulation. The RRMS error between the forecast responses and the measurements in different stages are decided. Results show that the RRMS errors corresponding to the transversal and vertical velocities in the middle of the m in span have a variance in different segments. This results mean that occurrence of damages in some structural members is possible. This analysis result is testified by inspection result that broken stay cables have been found after the earthquake. The proposed approach is a non-parametric damage detection strategy, in which a prior information about the exact model of the suspension bridge is not needed. The proposed strategy has a significant advantage when dealing with large-scale structures inreal-word.

A Post-Seismic Damage Detection Strategy in Time Domain for a Suspension Bridge with Neural Networks

Bibliographic Information

Abstract

Journal

Keywords

Details 詳細情報について

Export

Report a problem

A Post-Seismic Damage Detection Strategy in Time Domain for a Suspension Bridge with Neural Networks

Bibliographic Information

Abstract

Journal

Keywords

Details 詳細情報について

Export

Report a problem

Project list