Research On Blind Source Separation And Its Application In Rotaing Machinery Fault Diagnsis

Rotating machinery play an important role in the production, which is widely used in the petroleum, chemical, metallurgy, electric power, aerospace, and other engineering fields. Therefore, once the rotating machinery failure, will cause production downtime, bring huge economic losses, even making the machinery occur serious accidents, causing casualties, therefore, the condition monitoring and fault diagnosis is essential. With industrialization, information technology, and the development of economic globalization, the fault diagnosis technology of rotating machinery is constantly progress.In the practical application, the mechanical vibration signals are non-stationary, and are often faced with various interruptions or condition of fault coupling. How to effectively extract the expected fault feature from the observed signals is the bottleneck of mechanical fault diagnosis. Blind source separation can get source signals from the observed signals. Blind source separation is a new method of signal recognition, which is developed rapidly in recent years. With systematically studied the lack of traditional blind source separation algorithms, this paper presented three improved algorithms applied in different conditions. And put the algorithms into the practical rotating machinery fault diagnosis.Firstly, a new method of the blind source separation based on Gabor transform is proposed, it is used for separate correlated source signals, and it combines the advantages of Gabor transform and blind source separation. Simulation results show that even if the source signals are correlative, or there are more than one Gaussian signal in the sources, the new method can get better separation performance.Secondly, aiming at the problem that in the observation conditions of underdetermined, the traditional blind source separation methods produce poor result, thus, an under-determined blind source separation method of machine faults based on wavelet decomposition and time-frequency analysis is proposed. The simulation results confirm the feasibility and validity of the proposed method. Finally the proposed method is applied to the separation of mixed faults of rolling bearing, its fault features are fully detected and the effectiveness of the proposed method is verified.Thirdly, blind source separation methods include instantaneous and convolution blind source separation, instantaneous blind source separation is the basis for convolution blind source separation, however, in the actual mechanical failure, due to existing scattering, diffraction and other reasons in the signal transmission process, so that the way of signals mixing is closer to the convolution mixing. Therefore, the study for the convolution blind source separation is essential. The paper presents a blind source separation method based on frequency domain. The mixed signals in the time domain are transformed into instantaneous mixing in the frequency domain. The simulation and experiment testify the validity of the proposed method.