In order to improve the fault diagnosis level of marine power systems, this paper studies the real-time fault diagnosis of a marine supercharged boiler based on a convolutional neural network (CNN).

First, the simulation program of the marine supercharged boiler is developed based on the GSE platform, and the simulation fault data is obtained. The fault diagnosis model of the boiler is then established using the CNN method. Next, through the change trends of temperature, flow and other parameters, combined with a priori knowledge and the machine learning method, fault identification is carried out. Lastly, the performance of the method is evaluated against criteria such as confusion matrix and accuracy.

According to the comparison results between the feature extracted dataset and the original dataset, the stability of the model output results and the generalization ability of the model are optimized and improved, with an overall fault classification accuracy reaching 99.53%.

The results of this study can provide valuable references for the intelligent monitoring of marine power systems.