Study On Civil Turbofan Engine Health Intelligent Monitoring Technologies

Civil aeroengine health management is a key issue for civil administrationauthorities and airlines, and engine condition mornitoring, fault diagnosis and lifepredication are key parts of civil aviation engine heath managment system, which arethe important means to reduce operation and maintenance costs and avoid seriousaccidents, then achieving on-condition maintenance. Because of the high complexsystem structures and enormous components, the aeroengine faults vary a lot as wellas the fault consequences and causes, especially in the gas-path fault which therelevant parameters change very slightly, thus it is hard to make an accurate diagnosisonly by working experiences. Civil aviation engine fault diagnosis and life predicationbased on intelligent technologies are studied in this dissertation, The main researchesare listed as follows:1. QAR date are used for civil aeroengines fault detection amd fault diagnosis.PCA and information entropy method are employed to monitor aeroengine health.Thegaspath performance sort and health assessing are developed based on PCA,.andinformation entropy method is used to analyze relation between fault symptoms andcause.in order to develop the main cause of engine performance deterioration, then toassess engine health condition.2. In fault detection, the Hyper-plane SVM is employed to train and test thenormal samples and make a data description, model parameters are selected todetermine best classification boundary, then to decide samples classification bycalculating distance between unknown samples and support vectors. Parameters aresensitive to accuracy in detection modle,so cross validation is used for modelselection. The research shows that the diagnostic accuracy reaches93.2%.3. In fault diagnosis, the multi-classification algorithms are constructed andcompared by simulation samples based on the Least-square SVM because ofinsufficient real civil engine fault samples, The three types of kernel functions arealso studied, including the influence of function parameters and model parameters onthe classification results, and finally compared the results with Back PropagationNeural Network. The research shows that radial base kernel function gets betterresules than others.4. A sythetical assess method according to mean square root method based onFuzzy AHP method is proposed to predict removal time for Aero-engines, in whichdate from airline,shop, test cell and other information are analyze to determine the main factors influencing engine life-on-wing by AHP. then remaining time of engineis forecasted by a new assessing parameter. The research shows that sythetical assessmethod is according with reality..5) LS-SVM regression under Bayesian evidence framework is analysized and thecivil aviation engine life prediction model with error bars is built using the test celldata of overhauled engines. the results show the feasibility of prediction model.