Early Shock Pulse Detection Method for Rolling Bearings Condition Monitoring

The detection of incipient bearing faults in rolling element bearings (REBs) as early as possible has significant economic value when operating large, low-speed machines (LLSM). One observation made in this study is that common commercial systems based on envelope analysis methods have an unreliable fault detection performance when applied on machines with strong vibrations from multiple internal and external sources. Because of this insufficient performance, multiple costly preventive maintenance are commonly applied today. To solve the problem in hand, we have developed a new algorithm for early and reliable bearing fault detection. It is based on adaptive division of the vibration signal into a number of frequency bands, time-domain segmentation algorithm and high-resolution maximum-likelihood frequency estimation to discover small vibration pulses excited by the defect in the bearing. Furthermore, sequential probability ratio tests are used to make the final decision about the incipient bearing fault. Testing with data from real operation conditions shows that the developed method works reliably.