Fault diagnosis of ship fuel system based on DPC-GMM algorithm

  Objectives  The traditional Gaussian Mixture Model(GMM)has the inherent shortcoming of slow convergence which can easily lead to over-fitting and cause the parameter calculation to fall into a local optimum. As such, it is not suitable for the fault diagnosis of marine fuel systems.

  Methods  A fault diagnosis method for a ship fuel system based on the DPC-GMM algorithm is proposed. First, the GMM and parameter estimation algorithms are analyzed. Combined with the Density Peaks Clustering(DPC) algorithm, GMM parameters corresponding to the state of the fuel system of ship are calibrated to achieve the unsupervised diagnosis of the failure of a ship's fuel system. Based on the obtained fuel system failure data, the proposed method is verified.

  Results  The experimental results show that this method has higher recognition accuracy and faster recognition speed than the traditional Back Propagation(BP)neural network and Support Vector Machine(SVM) diagnosis algorithm.

  Conclusions  The analysis results have important guiding significance for the fault diagnosis of marine fuel systems.