Early-Warning System in Bridge Monitoring Based on Acceleration and Displacement Data Domain

Abstract

Bridges should be monitored periodically in order to assess the bridge health at any given time. The sensors send the acceleration and displacement data of a bridge response under earthquakes loading to the system server. This study aims to conduct the early-warning intelligent system based upon the performance of the acceleration and displacement data. The damage detection in the system applied the Neural Networks for prediction of a bridge condition at the real time. The architecture of Neural Networks’ model used one input layer, which consists of acceleration and displacement data domain, two hidden layers and an output layer with four neurons consist of safety level, Immediate Occupancy (IO), Life Safety (LS) and Collapse Prevention (CP). The IO, LS and CP are the bridge condition which indicates the extent of bridge health condition ranging from the light damage until high-risk level during and after subject to six earthquakes data. The training activation used the Gradient Descent Back-propagation and activation transfer function used Log Sigmoid function. The early-warning system is applied on 3 spans of box girder bridge model which is monitored in the local and remote server. The result showed that the evaluation of bridge condition using alert-warning in the bridge monitoring system can help the bridge authorities to repair and maintain the bridge in the future.