Non-stationary Vibration Signal Analysis Of Rotating Machinery Based On Blind Source Separation

Rotating machinery is one of the most commonly used equipments in industries; it is the key equipment in industry production, such as generators, compressors, steam turbines, blowers, etc. With the amazing development of modern technology and machinery's large-scale, high-speed, integrated, multi-functional tendency, the structure of machinery become more complex, and the automatic level of machinery increasing dramatically. A good management and maintenance for rotating machinery is required by modern enterprises. The normal running operation of these equipments has a significant interest for the production and the national economy.This thesis aim at investigating the non-stationed signal generated by rotating equipments using blind signal separation (BSS),envelop and order tracking processing technology. New methods of fault feature extraction and separation have been studied in order to extract fault features caused by bearing faults and separate the fault components which contain two different defects.For experimental study, the gear tooth broken failure mixed with the bearing outer race defect and the outer race detect combined with the rolling element defect are introduced in the thesis. As a result, a multi-faults separating method for gear boxes is presented. Firstly, envelop signals are extracted from the raw signal that contains more than one defects. Secondly, the ICA scheme is applied to separate each component from the mixed signal. Finally, the fault source of each component is obtained, and the extraction and separation of fault sources under a multi-default condition are achieved.Based on a though grasp to the ICA technology and the order envelope spectrum analysis technology, an alternative multiple vibration sources separation approach based on the ICA and the envelope order analysis is proposed in the paper. This method solved the problem that the ICA does not prior determine the number of independent vibration sources, and the applications of ICA may be limited in practice. In the scheme, the envelope extraction is utilized to reduce the dimension of vibration source in original data sets at first. Then, the even-angle sample of order track is applied for the envelope waveform. Subsequently, the ICA is employed to separate the even-angle sample signals according to the independent of vibration sources, which allows obtaining the fault features. The results indicate the presented method can realize the feature extraction and separation for outer faults of rolling element bearings and broken teeth faults of gears. This method has a good contribution to the development of features extraction technologies under multi-faults conditions of rotating machinery in academic fields.