ChebNet-based Online Health Monitoring for Diesel Propulsion System

As an important means of marine traffic, the safety of ships has always been the focus of attention. The structure of the ship is complex and it is composed of multiple subsystems. The problems of each subsystem will affect the ship's normal operation, and even lead to serious consequences. Therefore, it is an important task to carry out health monitoring on the main subsystems or main components of the ship, which helps to alarm in time when problems happen. This paper takes the intake and exhaust system of the ship propulsion system as an example, sorts out the data collected by multiple sets of sensors arranged on the system, such as temperature and pressure difference, and excavates the correlation between the data to build the graph, and then uses the graph convolutional neural network to process it efficiently to complete the task of predicting the change trend of the data. When the data's change trend is abnormal, the system will be checked or maintained in time to prevent accidents and achieve the goal of health monitoring. In this paper, three different graph convolutional neural networks are used to predict the graph. After comparison, a method that is most suitable for this data set is obtained. It shows high efficiency and good accuracy in the prediction task. It plays an important role in monitoring the health status of the intake and exhaust system of the ship propulsion system.