Research On Intelligent Fault Diagnosis Of Marine Diesel Engine Based On Data Driven

As the most critical mechanical equipment of marine engine room,marine diesel engine has complex systems and structures.If the diesel engine fails,it will seriously affect the navigation safety of the ship.In order to reduce the loss caused by the failure of the marine diesel engine,it is necessary to perform timely and reliable diagnosis and maintenance based on fault diagnosis technology.The Intelligent Ship Code issued by China Classification Society(CCS)indicates the direction of future intelligent fault diagnosis and condition monitoring technology for intelligent engine rooms.The traditional fault diagnosis technology of marine diesel engines relies heavily on expert experience and low reliability.The data-driven fault diagnosis is based on the monitoring data of the equipment,and uses machine learning and artificial intelligence algorithms to train the diagnosis model and identify faults,avoiding the dependence on expert experience.However,data-driven methods that rely on small sample learning cannot learn complex mapping relationships from sample data,and the accuracy and reliability of fault diagnosis needs to be improved.Deep learning methods can obtain deep information from a large amount of sample data,which has become a research hotspot in the field of data-driven fault diagnosis.As a commonly used algorithm in the framework of deep learning,Deep belief network has good fusion ability with other algorithms,and requires less data.Therefore,this paper focuses on the application of fault diagnosis methods based on deep belief networks to marine diesel engines.Due to the high cost,small number,and single data type of the marine diesel engine failure samples actually obtained,the sample data cannot be effectively processed.This paper uses the simulation model of MAN 8L/5160DF marine diesel engine established in AVL BOOST software to simulate and analyze six typical thermal faults of marine diesel engines.Through marine diesel engine failure simulation experiments,typical thermal fault data of marine diesel engines are obtained.Then,the deep belief network algorithm is used to study the sample data,and the fault diagnosis model of marine diesel engine is obtained.In order to verify the effectiveness of the fault diagnosis method,the fault diagnosis algorithm is compared with the traditional fault diagnosis method based on BP neural network and support vector machine.The comparison results show that the method of marine diesel engine fault diagnosis based on deep belief network has higher recognition accuracy and generalization performance.During the operation of a marine diesel engine,multiple faults often occur together.Therefore,it is necessary to carry out research on the diagnosis of compound faults of marine diesel engines.The compound fault samples obtained through fault simulation experiments are studied and tested using the fault diagnosis method based on deep belief network.In the process of model learning,in order to reduce the complexity of the network model,correlation analysis is used to select the features of the sample.On the premise of ensuring the accuracy of model diagnosis,sample features with high correlation with the fault types are selected for constructing the fault diagnosis model.Finally,in order to reduce the losses caused by major faults,this paper studies the diagnosis methods of early small faults of marine diesel engines.Firstly,the deep belief network is used to establish a state diesel engine state parameter prediction model.After that,the kernel density estimation was used to analyze the residuals of the parameter prediction results and set the fault threshold to realize the early fault diagnosis of the marine diesel engine.The research results show that the early fault diagnosis method of marine diesel engine based on deep belief network has better fault recognition ability.