Location

London

Event Website

http://www.csce2016.ca/

Description

Despite offering a great promise for continuous and automated monitoring of civil infrastructure systems, vibration-based damage detection methods may yield false positives and negatives due to environmental and/or operational effects. This paper presents a method based on ARMAX residual error in conjunction with Artificial Neural Networks (ANNs) to eliminate the environmental effects from damage detection process. A finite element model of a bridge type structure was simulated with different damage scenarios under various temperatures. Damage features obtained from statistical process on ARMAX residual errors were then compared between with and without environmental effects. Artificial neural networks were trained to learn and predict damage features due to temperature change only, by subtracting which the final damage feature was obtained. It is shown that both damage location and damage severity can be accurately identified.


Share

COinS
 
Jun 1st, 12:00 AM Jun 4th, 12:00 AM

STR-960: ELIMINATING ENVIRONMENTAL INFLUENCES IN VIBRATION-BASED DAMAGE DETECTION USING ARMAX RESIDUAL ERROR AND ARTIFICIAL NEURAL NETWORKS

London

Despite offering a great promise for continuous and automated monitoring of civil infrastructure systems, vibration-based damage detection methods may yield false positives and negatives due to environmental and/or operational effects. This paper presents a method based on ARMAX residual error in conjunction with Artificial Neural Networks (ANNs) to eliminate the environmental effects from damage detection process. A finite element model of a bridge type structure was simulated with different damage scenarios under various temperatures. Damage features obtained from statistical process on ARMAX residual errors were then compared between with and without environmental effects. Artificial neural networks were trained to learn and predict damage features due to temperature change only, by subtracting which the final damage feature was obtained. It is shown that both damage location and damage severity can be accurately identified.

http://ir.lib.uwo.ca/csce2016/London/Structural/97