Fault Diagnosis Methods Research For Aero-Engine Spindle Bearing

Spindle bearing is a key component of aero-engine and it affects the working condition of aero-engine to large extent directly. Because of working under the harsh conditions of high rotating speed and high temperature, spindle bearing is also one of the weakest parts of aero-engine. As the heart of aircraft, aero-engine provides power to flight. In case some fault occurs with aero-engine due to spindle bearing mar, the flight safety of aircraft will be suffered serious threat. Therefore, methods research about the fault of aero-engine spindle bearing is very important to avoid aero-engine fault and ensure safety flight of aircraft.In this paper, three theory stages have been studied to diagnose the fault types of aero-engine sindle bearing based on vibration signal, including vibration signal preprocessing, construction of characteristic parameters and pattern recognition diagnosis.An improved ITD algorithm based on mixed fitting of cublic spline interpolation and linear transform has been proposed against the defect of original method firstly. Then, through decomposing signal by using improved ITD algorithm and selecting the first PR as reconstructed signal, the process of vibration signal preprocessing is completed. The signal processing method can effectively reduce the noise and other frequency components, which is interfering with diagnostic process. In addition, it can also increase the fault feature frequency components of original signal. Besides, the method combining full vector envelope spectrum with improved ITD algorithm is researched to fuse fault feature of multi-channel data, which is applied to actual fault diagnosis of bearing later.Three basic types of characteristic parameters are constructed to characterize the running states of spindle bearing in this paper:time-frequency domain parameters, ITD fuzzy entropy and MFCCS. In order to describe the fault states more synthetically and accurately, optimal characteristic parameters with high sensitivity and regularity are extracted based on LTSA algorithm from the matrix of mixed domain characteristic parameters, which is constructed by the three basic characteristic parameters.VPMCD algorithm with the advantages of simple structure and fast operational efficiency is applied to fault diagnosis of aero-engine spindle bearing in the paper. This diagnosis method has more superiorities than traditional methods (BP Neural Network, SVM algorithm and so on) in the parts of recognition accuracy and efficiency via the experimental results. Meanwhile, the problem that VPMs is not stable is improved by using PCA to ameliorate the solving process of VPMs. In conclusion, the improved VPMCD algorithm can availably enhance accuracy of the VPMs and diagnosis results.Finally, a simple aero-engine spindle bearing fault diagnosis platform has been achieved to implement engineering practice of the theoretical results in this paper. The database and software interfaces has been designed and developed respectively. By the debugging results of each function module, the practical value of the platform has been verified.