Chu Fulei

Incipient fault feature extraction of main bearing by iterative squared envelope analysis

Extracting weak features of incipient bearing fault from the collected vibration of rotating system is the basis of the fault diagnostics of main bearing in the aero engine. To monitor the running condition of the main bearing, a novel weak feature extraction method for bearing fault, named as iterative squared envelope analysis (ISEA) is proposed by extracting the fault characteristic orders of bearings. Both simulations and experiments, involving the outer and inner race faults, are performed to validate the efficacy of ISEA. It is shown that the ISEA can efficiently eliminate the vibrations produced by rotor and extract the bearing fault feature. Compared with the result obtained by the cepstrum pre-whiten method, both amplitude and cyclic feature can be reserved closer to the true values than that obtained by the cepstrum pre-whiten (CPW) method. Therefore, the ISEA is more powerful in the weak feature extraction of bearings than the CPW method.

Time Series-Neural Networks Diagnostics for the Fatigue Crack of the Large-scale Overloaded Supporting shaft

The time series-neural network is attempted to be applied in research on diagnosing the fatigue cracks degree based on the analysis of characteristics on the supporting shaft. By analyzing the characteristic parameter which is easy to be detected from the supporting shafts exterior, the time series model parameter which is hypersensitive to the situation of fatigue crack is the target input of neural network, and the fatigue cracks degree value of supporting shaft is the output. The BP network model can be built and trained after the structural parameters of network are selected. Furthermore, choosing the other two different group data can test the network. The test result will verify the validity of the BP network model. The result of experiment shows that the method of time series-neural network is effective to diagnose the occurrence and the development of the fatigue cracks degree on the supporting shaft.

Blind identification of threshold auto-regressive model for machine fault diagnosis

A blind identification method was developed for the threshold auto-regressive (TAR) model. The method had good identification accuracy and rapid convergence, especially for higher order systems. The proposed method was then combined with the hidden Markov model (HMM) to determine the auto-regressive (AR) coefficients for each interval used for feature extraction, with the HMM as a classifier. The fault diagnoses during the speed-up and speed-down processes for rotating machinery have been successfully completed. The result of the experiment shows that the proposed method is practical and effective.