A fault diagnosis method for underwater thruster based on RFR-SVM

A fault diagnosis method for underwater thruster based on Random Forest Regression (RFR) and Support Vector Machine (SVM) is proposed in this paper. Aiming at the problem of insufficient fault diagnosis accuracy caused by the extremely unbalanced scale of normal samples and fault samples, a data argumentation method of fault samples based on RFR is proposed. Considering the over-fitting phenomenon of machine learning in the case of small samples, tsfresh package, and kernel principal component analysis (KPCA) are used to extract features from thruster time series data, and then the SVM is used to train the thruster fault diagnosis model. Finally, the effectiveness of the proposed method is verified by experiment in a pool environment.