A Fault Diagnosis Method for Typical Failures of Marine Diesel Engines Based on Multisource Information Fusion

Due to the fact that different components of a diesel engine may exhibit the same fault phenomenon, leading to a low fault identification rate caused by one effect with multiple causes, this paper acquires multiple thermodynamic parameters under various fault conditions through models. The k ‐nearest neighbors mutual information (KNN‐MI) method is used to calculate the correlation between different thermodynamic parameters, eliminate strong correlation coefficients, and select thermodynamic parameters with low correlation for fault diagnosis. Comprehensively comparing the downscaling effects of the four different downscaling methods, the data after downscaling by the t‐SNE are characterized by a smaller distance within the class, a larger distance between the classes, and a better classification effect, which is used for the downscaling of the fusion of the screened thermal parameters and vibration feature parameters, and finally, the feature fusion method is used. An accuracy rate of 98.7% was achieved and compared with the fault diagnosis method based on a single signal of vibration. The fusion of multisource information can effectively distinguish between different fault categories and improve the accuracy of diesel engine fault identification.