Unsupervised-learning based Structural-health Evaluating Method of Station Structures

초록

Authors have been developing the statistical pattern recognition technique based structural-health evaluating method of station structures. This method makes us to know the health-state of station structures in real-time from measured signal data through smart sensor network and IT technologies. In this study, authors suggested the structural-health evaluating algorithm based on unsupervised-learning method such as outlier analysis and verified the algorithm through the numerical studies on Chungmuro station. The numerical results demonstrate that the suggested structural-health evaluating algorithm can effectively detect the damaged structural members and subsequently enable to do appropriate maintenance work promptly.