Research on the fault monitoring method of marine diesel engines based on the manifold learning and isolation forest

Abstract In this paper, an innovative hybrid fault monitoring scheme integrating the manifold learning and the isolation forest was established to monitor the state of marine diesel engine. The manifold learning was used to extract the useful feature and realize data dimension reduction, and these extracted features could ameliorate the fault monitoring process. Then, the isolation forest only utilized the normal operating data to realize the fault monitoring, and with manifold learning, the hybrid model can reduce computation complexity and improve diagnostic accuracy. However, the conventional isolation forest may ignore some fault information and cannot provide satisfactory fault detection performance. Therefore, a threshold based on partial monitoring fault data and the clustering algorithm was set to provide more transparent and accurate diagnostic results. For validating the proposed scheme, a two-stroke marine diesel engine was developed in MATLAB/Simulink environment based on zero-dimensional approach to represent a real engine behavior, and an in-service marine diesel engine provided reliable normal and fault condition datasets. Finally, comparisons of fault detection rate and false alarm rate of other state-of-the-art methods on simulated and measured datasets demonstrated the excellent performance of the proposed hybrid fault monitoring scheme.