Event-Based Intelligent Sensitivity Analysis and Performance Prediction With Application to Hydraulic System Monitoring

This study proposes an intelligent sensitivity analysis and performance prediction approach for complex systems based on novel integration of event clustering and random forest regression techniques. The event clustering method is used for real-time sensitivity analysis by quantifying the impact of critical events on system outputs as well as reducing the number of key features on system performance. Subsequently, the random forest regression model is deployed for efficient feature selection as well as performance prediction through iteratively optimizing selection rules. As such, the proposed approach is able to provide an effective data-driven solution to real-time analysis, monitoring and prediction for complex systems. The application to a hydraulic system validates the effectiveness of the proposed intelligent approach.