Study on Semi-blind Signal Processing Method for Rolling Bearing Faulty Vibration Signal

Because of the cyclostationary nature of the vibration signal of rotating machine, the blind equalization technique based on output second order statistic (SOS) is introduced to recover “real” vibration signal from multi-sensor signals, which is essentially data level information fusion for homogeneous sensors signal. Based on the prior knowledge of the order spectral structure of the rolling element bearing faulty vibration signal, two performance indexes are proposed to optimize the blind equalization filter parameters. Finally, a semi-blind signal processing method for rolling bearing fault diagnosis is given. The performance of the method is verified with a single reduction gearbox test rig experiment data.