Research On Fault Diagnosis For The Rotor System Of Civil Aeroengine Based On Vibration Analysis

The rapid development of the civil aviation industries has put forward more and more demands on aviation safety and economy. Civil aeroengine serves as the core component of civil aircraft, the condition of which has a significant effect on the operational safety and economic benefits of the whole industry. According to statistics, in the variety of mechanical failure that caused flight accidents, engine failure accounted for about 1/3, and 60%~70% of which were structure and strength faults of the rotor system. Therefore, to carry out fault diagnosis for engine rotor system is an important means to achieve engine health management and condition-based maintenance, thus to improve flight safety and reduce maintenance cost. It is of great scientific significance and application value to research and develop advanced diagnostic approaches and techniques. This work focus on several key issues of civil aeroengine rotor system fault diagnosis, in which, studies on the establishment of fault database, fault signal processing and feature extraction, fault recognition and decision-making were conducted thoroughly with simulation and experiment.According to the general database design method and the diagnostic requirements of rotor system, the overall structure of the fault database was designed. Aiming at the problem that fault data are not easy to acquire, besides collecting vibration data from engine ground test, rotor system dynamic simulation as well as experimental study was carried out. A dynamic model with typical faults was constructed to generate various fault signals. An experimental testrig was setup to simulate the fault conditions of the aeroengine rotor system, and huge data were collected. The establishment of fault database laid a solid foundation for further researches on diagnostic methods.In field of fault signal processing and feature extraction, studies on empirical mode decomposition(EMD), a widely applied signal processing method, were conducted. Aiming at the shortcoming of EMD including end effect, mode mixing and failing to extract the high frequency low energy features, modified approaches were proposed. Applying which to analyzing rolling element bearing fault signal and rotor fault signals, the results demonstrated the superiority of the proposed approaches in describing the time-frequency distribution of the signal as well as extracting fault features.In the study of fault pattern recognition, a method based on fuzzy support vector machine(FSVM) was presented to figure out the problem that traditional SVM is sensitive to noises and outliers. The adaptive kernel fuzzy clustering method was employed to assign fuzzy membership to samples, and genetic algorithm was adopted to optimize the model parameters. Then, the method was applied to diagnostics of rotor system with fault features extracted from multiple domains, and the comparisons with some other approaches validated the performance of the presented method in identifying fault type, fault degree as well as its antinoise ability.In terms of decision-making, considering the imprecision and uncertainty of the diagnostic result with a single source, diagnostic models based on multi-classifier fusion and multi-sensor fusion were put forward. Decision level fusion with evidence theory was studied, and DSm T was adopted to fuse the conflicted evidences. Diagnostic results of rotor faults indicated that both the models could improve the reliability and accuracy of the decision. Among which, multi-sensor fusion based model shown excellent robustness and antinoise capability for the reason that more basic information about fault could be provided from different aspects.Finally, based on Lab VIEW platform, combining database technology and MATLAB programming, fault diagnosis system for aeroengine rotor was designed and implemented, which facilitated the engineering application of the diagnostic technologies.