Research On Fault Diagnosis Of Marine Refrigeration System Based On Information Fusion

As a large energy consumption,the research on energy conservation and environmental protection of ships is directly related to the development of green ships.The instability and failure of the ship’s refrigeration system will not only worsen the air quality in the cabin,damage the fresh-keeping of the food in the refrigerator,reduce the comfort of the crew’s life,but also waste a lot of energy,shorten the service life of the equipment,and reduce the economy of the shipping.This is obviously not conducive to the development of green ships,so a practical and effective fault diagnosis method of refrigeration system is urgently needed.With the development of big data and artificial intelligence,many scholars use data-driven method to study fault diagnosis,and have made a lot of achievements.However,most of the research methods have a single source of information,only based on a single diagnosis model for fault identification,which is one-sided.Based on the research experience of relevant scholars,this paper focuses on the research of fault diagnosis of refrigeration system by integrating multiple diagnosis models,and proposes a research method of fault diagnosis based on information fusion.The main work of this paper is as follows:(1)The operation principle of the refrigeration system is studied.The thermodynamic analysis of the ideal cycle and the actual cycle of the refrigeration system is carried out.The fault analysis of the refrigeration system is carried out,which lays the theoretical foundation for the fault diagnosis in the following paper.(2)A large number of researches on information fusion technology have been carried out,and a system framework based on feature level fusion and decision level fusion in information fusion level theory has been determined.In feature level fusion,multiple local diagnosis models are established to analyze the same information,and the analysis results are regarded as evidence input for decision-making level fusion;In the decision level fusion,the Output evidence of each local model is fused based on DS synthesis rules to make the final decision.(3)Based on the experimental samples,three kinds of classification models,BP neural network,support vector machine and probabilistic neural network,are established respectively.The modeling principle and recognition mechanism of each diagnosis model are analyzed,and the preliminary comparative analysis of each model is carried out.Then,genetic algorithm is used to optimize the parameters of the three models,and the optimization principle of each model is analyzed,and the optimized model is compared and divided Analysis.(4)Based on the optimized models,the output of each model is taken as the evidence body,and the probability distribution function is given to the evidence body by using the error function or recognition error rate of the model,and the information fusion of each group of evidence is carried out based on the traditional DS evidence synthesis rules.(5)The traditional DS evidence synthesis rules are helpless for the problem of high conflict of evidence.This paper proposes a decision-making level weighted fusion method based on evidence distance.By giving different influence weights to each group of evidence bodies,each group of evidence bodies has different trust,so as to make more effective fusion decision.All the modeling processes in this paper are realized by MATLAB.Through the simulation test,the unique advantages of this method are verified;the whole diagnosis system has strong fault-tolerant ability and expansion performance,can more accurately and effectively determine the type of fault,is conducive to the energy conservation and environmental protection of ships and shipping economy,and has certain engineering practical significance.