The Application Research Of Aero-engine Vibration Rating Using Multidimensional Data Mining Methods

Aero-engine vibration faults are the main problems of aero-engine.Moreover,because the complexity of aero-engine’s structure is so great,the engine may have several different failure phenomenons,such as imbalance,rubbing,misalignment,vibration jump and so on.This brings out enormous challenges for engine faults prediction and diagnosis.In general,the method of building faults analysis models with expert experience is adopted to predict and diagnose the vibration.Usually,the dimension of aero-engine faults data features is high,it is difficult to specify the details of the fault problem effectively.Recently,big data and data mining technology brings new ways to solve industry manufacture problems,and the result is effective in many factories and companies.As the crown of industry manufacture,aero-engine is famous for its complicated structure and extremely accurate standard.Therefore,how to combs the structure of aero-engine with the present methods in big data and data mining research field and how to predict and diagnose engine faults from the vibration faults data are urgent problems to be solved.Aiming to solve the above problems,this paper mainly works on:(1)Designed and implemented an aero-engine vibration analysis system.Starting from the system’s requirement analysis,this paper illustrates six modules include aero-engine structure entity,faults entity,assembly benchmark,assembly decomposition,test parameter,vibration analysis and functional or nonfunctional requirement includes performance,safety,extensibility.After,the paper focus on system architecture design,detailed design and implementation.The whole system implements on J2EE architect based B/S scheme,and EXTJS front end framework,Spring MVC framework and unstructured MongoDB database are used in the system during development phrase.This data management and analysis system is built on the basis of aero-engine faults data,and it provides a powerful tool and analysis footstone for aero-engine faults prediction and diagnosis.(2)This paper gives a vibration faults analysis method based on association analysis and clustering.Experiments are designed on engine imbalance fault analysis.First,the fault data is discretized by using a percentage technique,then an algorithm called MAFIA generates the most significant feature set from the discretized data set.Second,a two layer clustering method is provided.This step eliminates outliers and detects irregular clusters by DB SCAN algorithm,then the result is analyzed with kmeans for the second clustering.The effectiveness of the clustering indicates characteristcs that is relevant to the vibration.At last,compared the result data and the standard data set,the range of vibration data is partitioned,and the vibration level is determined.