Research On Fault Diagnosis Of Aero-engine Oil System Based On Improved Support Vector Machine

Among the faults of aero-engines,the faults of the oil system takes up a large part.In view of the characteristics which the fault diagnosis data is deficient,the qualitative characteristics is ambiguous and the quantitative is lacking of the current oil system faults diagnosis.The fault diagnosis of the oil system is studied to facilitate the acquisition of faults data,timely judgment at the moment of failure,or prediction before the failure is formed,is of great significance for ensuring the safe operation of aero engines.Taking a certain type of aero engine as the research object,the fault diagnosis of oil system was realized by using improved support vector machine.The specific research content of the paper is as follows:(1)The structure of a certain type of aero-engine lubricating system is studied,then four typical faults of the oil system are selected based on the survey data,and the causes of the faults are analyzed one by one.(2)Based on the analysis results of faults,a bayesian network model of typical faults is established.Then,taking the fault of large oil consumption as an example,Hugin's algorithm,the bayesian network's precise inference algorithm,is used for inference.Finally,combining the three methods for evaluating the importance of components,the important elementary events in the cause of the fault were screened out.(3)Because the lubricating oil supply system of the aero engine is similar to the oil return system,the oil supply system is used as an example,and the model of the oil supply system is established using AMESim software.Then,the important elementary events selected above are simulated in the AMESim model,and the mapping relationship between the fault and the fault characteristic parameters,that is,the fault data,is obtained.(4)First,using the support vector machine to establish the fault diagnosis model of the oil system.Then,a method for calculating the mean influence value of the support vector machine on the classification problem-"average distance calculation method" is proposed.The method is used to reduce the dimension of the fault data,and the correctness of the method is proved by comparing with the principal component analysis method.In order to improve the accuracy of fault diagnosis,the grasshopper optimization algorithm is introduced to optimize the penalty factor C and gaussian kernel parameter g of the support vector machine,so that the accuracy of the fault diagnosis model is increased to 97.5%.