Unbalance fault diagnosis of Marine gas turbine based on RDE-SVM algorithm

As ship automation gradually advances, the automatic maintenance of ship propulsion systems becomes increasingly important, and currently, common maintenance often takes place after a breakdown. This paper, through research on data-driven models for marine gas turbines, proposes a state maintenance and diagnostic method based on a combination of the Ranking-Based Differential Evolution (RDE) algorithm and support vector machine (SVM) algorithm. Initially, an anomaly detection model describing normal sample distribution is trained by using the SVM optimized by RDE algorithm. Then, a diagnostic model is separately trained for each fault to describe the decision boundary to classify gas-path faults. Comparison of the proposed method with traditional methods show that the proposed method can effectively address the data imbalance and overfitting issues, and it can detect abnormal conditions in marine gas turbine with only a few samples.