Research On Fault Detection And Diagnosis Method For Marine Diesel Engine

Under the impetus of modern science and technology,ships are constantly developing in the direction of networking,digitization,and intelligence.This requires higher demands on the reliability and management efficiency of ships.Diesel engine as a ship's core equipment,its operating status directly affects the safety of the ship's navigation,and its serious accident may cause catastrophic consequences.However,the traditional methods of monitoring alarms and manual inspections obviously can't meet the requirements for ship intelligence.Therefore,the intelligent detection and diagnosis method for marine diesel engine faults have gradually become a hot topic in ship intelligent research.This thesis takes the 7K98MC diesel engine as the research object,builds the simulation model of diesel engine working process based on Matlab/Simulink platform,and uses the principal component analysis(PCA)and BP neural network to study the fault detection and diagnosis method of the diesel engine.The research work completed in this thesis includes the following aspects:(1)The working process of the diesel engine and the mathematical model of each subsystem are introduced.The simulation model of each system module of the7K98MC diesel engine is built using Matlab/Simulink.The simulation results are compared with the bench test data to verify the accuracy and rationality of the model.At the same time,the thesis uses the diesel engine simulation model to simulate the diesel engine faults,such as compressor fault,intercooler fault,fuel injection timing fault,fuel injection fault,exhaust pipe fault and other six common diesel engine faults.The characteristics of various parameters with different faults and different fault levels are obtained.The results show that the simulation results of the model are consistent with the actual theoretical results and have high simulation accuracy.(2)The principle of fault detection in PCA is described.Based on Q statistics and T~2 statistics,a comprehensive statistical fault detection method is proposed.The influence of variable factors under various faults is analyzed with a contribution graph,and through the fault simulation data of diesel engine to verify the effectiveness of PCA.The results show that PCA has better detection capability for diesel engine fault conditions,and the use of contribution graphs can better analyze the cause of the fault.(3)The thesis studies the model structure and algorithm principle of BP neural network,analyzes the searching ability and prediction accuracy of standard BP neural network and improved BP neural network,and proposes a method of PCA to optimize BP neural network fault diagnosis.In addition,the failure mode of the diesel engine is identified through normal and six kinds of fault simulation data.The results show that using PCA simplifies the model structure of BP neural network,improves the efficiency of fault diagnosis,and has high recognition accuracy.