The Research And Application Of The EMD And Fuzzy Neural Network In The Fault Diagnosis Of The Roller Bearing

The roller bearing is the damageable part in the mechanical equipment, according to the statistics, rotating machinery fault that 30% is caused by the roller bearing fault. If the roller bearing is fault the mechanical equipment's work precision will drop, or even lead to mechanical equipment does not work normally, and cause serious accident. It can be said that the quality of roller bearing has a great impact on the entire electromechanical equipment system. Therefore, the roller bearing, as one of the most commonly components in the mechanical equipment, fault monitoring and diagnosis is the subject to which the domestic and foreign technology domain paid many attention all along. According to a large number of research facts proved, at present, that the analysis of vibration signals is the most practical method in the condition monitoring and fault diagnosis of roller bearings.In this paper, from the roller vibration mechanism of the fault, summed up the corresponding changes in the value of spectrum when the roller bearing had the partial failure, and on the faults diagnosis laboratory bench in the laboratory, we simulate four kind of conditions about the roller bearing work. The four kind of conditions is normal,damage in outer track,damage on ball and damage in inner track, and we have collected these four vibration signals.In this paper, the self-adaptive method of EMD has applied in the eigenvalue extraction of the roller bearing fault, and put forward eigenvector extraction method based on IMF energy moment for the first time. This paper has proved that the eigenvector extraction method based on IMF energy moment will more well display the differences between non-stationary signals, compared to the wavelet packet frequency band energy method. And in this paper has also proved that, in the roller bearing fault identification, the combination of IMF energy moment and fuzzy neural network will remarkable increase the precision and the real-time characteristic, compared to the simple neural network or fuzzy diagnosis methods...