The data learning and anomaly detection based on the Rudder system testing facility

Abstract For the data analysis of existing rudder system testing facility (RSTF) being a manual process now, a machine learning (ML) method for fault diagnosis based on RSTF is proposed to realize intelligent data analysis. For this purpose, we have focused on developing a new decision-points-distribution and weight-assignment-oversampling method integrated with optimized Support Vector Machine (SVM) to conduct anomaly detection based on RSTF in this paper, which takes advantage of decision making derived from SVM and is combined with the cluster-based synthetic samples generation mechanism. It is proposed to solve the problem caused by imbalanced data collected from RSTF. Additionally, the SVM classifier is optimized by Perturbed Particle Swarm Optimization (PPSO) while avoiding the risk of falling into local optimization. Experiments are conducted on the imbalanced dataset collected from RSTF and the proposed strategy exhibits its superiority over some existing algorithms.