A Machine Learning Framework for Automated Functionality Monitoring of Movable Bridges

Functionality of movable bridge highly depends on the performance of the mechanical components including gearbox and motor. Therefore, on-going maintenance of these components are extremely important for uninterrupted operation of movable bridges. Unfortunately, there have been only a few studies on monitoring of mechanical components of movable bridges. As a result, in this study, a statistical framework is proposed for continuous maintenance monitoring of the mechanical components. The efficiency of this framework is verified using long-term data that has been collected from both gearbox and motor of a movable bridge. In the first step, critical features are extracted from massive amount of Structural Health Monitoring (SHM) data. Next, these critical features are analyzed using Moving Principal Component Analysis (MPCA) and a condition-sensitive index is calculated. In order to study the efficiency of this framework, critical maintenance issues have been extracted from the maintenance reports prepared by the maintenance personnel and compared against the calculated condition index. It has been shown that there is a strong correlation between the critical maintenance actions, reported individually by maintenance personnel, and the condition index calculated by proposed framework and SHM data. The framework is tested for the gearbox.