Study On Remote Fault Diagnosis Technique For Marine Diesel Engine Based On Neural Network

Remote fault diagnosis of marine diesel engine is a strongly comprehensive subject which comprises sensor technology,data collection,signal analysis,data processing,data transmission,network and communication,marine diesel engine, fault diagnosis technology and so on.This paper,with the research of The Online Monitoring System for Ocean-going Vessel and Cargo Transportation from the COSCO Group,firstly introduces the development process and situation of the remote monitoring and fault diagnosis technology in both domestic and overseas.After carefully analyzed the structure of the shipping remote monitoring and fault diagnosis system,special emphasis has been placed on its basic framework and the key technology problems which are related to overall design of remote fault diagnosis system for the marine diesel engine.Secondly,the Radial Basis Function(RBF) neural network is used to analyze the remote fault diagnosis of marine diesel engine.Then the paper makes a deeply research to the principle and features of RBF neural network.Considering the characteristics of marine diesel engine fault diagnosis,the paper discusses the structure of RBF network as well as the training mode and relative parameters of the network in detail.Finally,the model of remote fault diagnosis for marine diesel engine is established by using the MATLAB Neural Network Toolbox.A large number of simulation data of the diesel engine turbocharger system are achieved with the help of the performance on low-speed marine diesel engine simulation program.Using the sample sets of symptoms and troubles the neural network is trained and the analysis of the remote fault diagnosis is finished.It has been demonstrated that remote fault diagnosis on ships conforms to the development trends of the ocean-going ship technology.Moreover,the application of the remote monitoring and fault diagnosis system will provide more reliability and safety for shipping companies significantly.