Fault Diagnostic Methods And Trend Prediction Research For Marine Power System

As the power source of the whole ship, marine power system plays an importantrole in the ship normal operation, which includes main propulsion system, auxiliarypower system and equipment ensuring the ship security, the crew’s life andenvironmental protection. Since the marine power system operates in a hostilecondition with strong time-variation, the faults in this system tend to cause severeconsequence. The safe and reliable operation of this system has a strong influence onthe normal operation of ships. With the development of the automation and artificialintelligence, it is an important way to identify the faults in the marine power systemwith the intelligentized methods. It is therefore essential to research on the faultdiagnostic methods for the marine power system.In this thesis, the development of the domestic and overseas fault diagnosticsystem were analysized and the intelligent fault diagnostic methods were exploredaccording to the problems in this area such as insufficient diagnostic capacity ofonline fault diagnosis and etc. Firstly, by using the rule engine technology, the systemscheme with Drools was designed, the reasoning mode of the fault diagnostic expertsystem for the marine power system was selected and the diagnostic rule base wasconstructed. Secondly, in order to solve the limitations in the application of the expertsystem like difficulty in knowledge acquirement and poor processing ability for theincomplete information, the research has been done on the data-driven diagnosticmethod. The diagnostic model with SOM neural network was constructed tocompensate for the shortage of the expert system. Thirdly, to realize the transitionfrom post-diagnosis to prognosis, the trend of the main state parameters of the powersystem was predicted with the ARMA model and wavelet neural network. Aftercomparing the characteristics and application scopes of the two methods, differentmethods were selected to predict the trends of different parameters respectively,which can forecast the abnormal variation of the parameters to guide the routinemaintanence of the power system.“Dong Hai Jiu117”, a salvaging ship was selected as the research object and thefault diagnostic function was realized with the data collected by the condition monitoring system in the ship. According to the research of the fault diagnosticmethods and trend predition methods, the function and application process of thediagnostic system were designed. Based on the Windows platform and compilationtool for Java-Eclipse, the function of the expert system was realized.