Research On Fault Diagnosis Of Marine Diesel Engine Based On Support Vector Machine

As the main power source of ship,the performance of diesel engine directly affects the safety and efficiency of ship operation.Because the diesel engine system structure is complex,the working environment is harsh and changeable,the running time is long,once appears the navigation breakdown to obtain the rescue very difficult.On the basis of bench test and fault simulation,the algorithm of fault diagnosis and its optimization for marine machinery are studied in this paper.The study contents and conclusions are as follows:1.Develop the cylinder pressure acquisition system of diesel engine.The cylinder pressure acquisition system of diesel engine is developed by using data acquisition card,cylinder pressure sensor,crankshaft angle sensor and Labview software,and the indicator diagram information of 4135 diesel engine is collected.2.Build and verify the diesel engine model.The 4135 diesel engine simulation model is built in the simulation software GT-Power,and the system boundary,intake and exhaust module,injector module,cylinder module and crankcase module are mainly set up,the accuracy of the model is verified by the actual indicator diagram.3.Fault type and thermal parameter selection and data acquisition.5 typical faults of diesel engine were selected and 6 thermal parameters were determined.1200 groups of data were obtained by fault simulation and preprocessed,including 900 groups of training data and 300 groups of test data.4.Research on fault diagnosis of marine machinery and determine the basic algorithm.BP neural network,Elman neural network and Support vector machine are used to classify the faults respectively.The accuracy of fault classification of Support vector machine is the highest,reaching 91.33%.5.Carry out the optimization research of the basic algorithm of fault diagnosis.By comparing the optimization results of gray wolf algorithm,whale algorithm and the improved whale algorithm,the improved whale algorithm achieves 96.67% accuracy in Support vector machine classification,and the optimization effect is the best,can be used in practical engineering.