An algorithm for determination of bearing health through automated vibration monitoring

Bearing fault diagnostic techniques have long been employed to deduce rolling element bearing faults in the paper, power, and chemical industries. Relatively low speed, constant load rotating machinery in these indus- tries lends itself to diagnosis of developing bearing faults. Application of this type of analysis to high speed, multi-state variable turbomachines is a relatively new concept, however.

Vibration analyses conducted following catastrophic failures of turbine engines suggested that the concepts of bearing fault diagnosis were indeed applicable to these turbomachines. If the thought processes of the vibration analyst could be automated, then developing faults indicative of pending failure might be recognized in time to alert engine operators prior to manifestation of failure.

The bearing health characterization algorithm devel- oped herein provides a means of statistically banding vibration responses for turbine engine families over a multitude of flight conditions and vibration sensor locations. Applying this algorithm to offline vibration data, the analyst may deduce meaningful limit criteria for incorporation into a real time health monitoring system.

The real time bearing health monitoring algorithm developed herein may be expanded to achieve the ultimate goal of catastrophic failure avoidance. The algorithm, once automated, will screen incoming data for potential indications of developing bearing faults. Alert and alarm annunciators would then be used to warn operators of a pending threat to the engine.