Detection of Damage-induced Fatigue Response Based on Structural Health Monitoring Data of In-service Steel Bridges Using Artificial Neural Network

Abstract

Successful signal-processing of the collected SHM data can provide information for early detection of the structural performance variation, which can indicate damage or deterioration. Fatigue assessment is a critical component of condition assessment for inservice bridges, especially steel bridges with fracture-critical elements. In addition, at an early stage damage, the measured damage-induced fatigue response can be misinterpreted as the other measured deviant fatigue responses. This study aims to develop an Artificial Neural Network (ANN) model to predict the normalized fatigue response over a discrete period of data collection at the healthy condition. The proposed ANN model is used for a case-study vertical lift truss bridge, the Memorial Bridge, located in Portsmouth, NH. A 12-month data collection period, from the long term SHM program of the bridge, is used to investigate the variability of the traffic pattern and the impact on the measured fatigue response. Through a validated global finite element model of the bridge, a physical damage case is simulated to compute the time-history stress response. The computed numerical stress response is applied to evaluate the capability of the ANN model in recognizing the measured fatigue response in the damaged conditions.