Vibration analysis for defective bearings by blind source separation

An obstacle to diagnosing multi-component machines using multiple sensors to acquire vibration data first lies in the data acquisition itself. This is because the vibration signals collected by each sensor are a mixture of vibrations produced by different components and noise. In industrial environments, the rolling element bearing is the most important components of rotating machines which requires permanent or periodic monitoring. The signal acquired from a running bearing often presents a mixture of vibrations. The analysis of these vibrations is important to achieve precise diagnosis. For this purposea method based on blind source separation (BSS) is proposed in this work. The performance of theproposed method isverified using synthetic and real vibration signals acquiredfrom an accelerometer placed on a test bench. Theobtained results confirm the effectiveness of the proposed method.