Citation:
Abstract:
In most industrial environments, vibration analysis is widely used for fault diagnosis of rolling bearings. The vibration signal measured from a bearing represents a mixture of motor vibration, rolling vibration, noise, and other sources. Due to the high cost of devices and limited space, only one sensor can be installed to measure this signal. In this paper, a feature extraction method based on Single Channel Blind Source Separation (SCBSS) and Normal Distribution (ND) is proposed for vibration monitoring of rolling element bearings. To decompose the bearing signal, SCBSS is applied for separating the different sources. Because ND is sensitive to the type of fault, it is used as criterion to find an output that contains the maximum information about the fault by removing the other sources. In fact, the obtained signal contains other vibrations which affect the correct source of fault. A second SCBSS filter is, therefore, proposed to decompose the selected source and thus improves the performance of fault diagnosis. The application of the proposed method is carried out on a deep groove ball bearing with outer race fault, ball fault, and inner race fault in order to better validate the diagnosis results.