Investigation into backup bearing life using delevitation severity indicators

Abstract

Active magnetic bearings (AMBs) are inherently flawed in terms of possible failure from either mechanical, electronic or software components. Component failure could induce a rotor delevitation event (RDE) during operation and possibly damage the backup bearing (BB) system. To improve BB reliability and safety, the applicability of using quantified delevitation severity indicators DVAL, VVAL and AVVAL for quantifying degradation and predicting BB life is investigated. A small-scale AMB system is used to generate BB degradation data by subjecting steel-caged rolling-element bearings to multiple RDEs. The RDEs are induced at specific initial conditions to analyse bearing failure distribution. Delevitation severity indicators are subsequently used to compare a series of RDEs to analyse changes in BB performance characteristics. Using only shaft position and rotating speed data, this investigation showed that delevitation severity indicators change as the bearing degrades. A distinctive linear pattern of degradation is identified by calculating AVVAL for the duration when rotor whirl and bouncing occurs. A threshold value when BB failure occurs is also identified. Using the linear degradation pattern and identified threshold failure value, two life prediction methods are formulated: the safe envelope method (SEM) and the linear extrapolation method (LEM). The SEM and LEM were validated with successful life predictions at various initial conditions and provided an average prediction accuracy of 91%. The two methods were found to be applicable only when BB life exceeded that of the bearing’s run-in phase. Large and sudden changes in rolling friction were detected by calculating the values of DVAL and AVVAL for the duration when a rolling motion is induced. The changes serve as an early warning for possible catastrophic failure of the bearing and enable a form of BB failure detection. The failure detection capability was verified by uncovering the linear relationship between rolling friction and AVVAL. This linear relationship further shows that AVVAL is indicative of bearing degradation. A novel method for quantifying rotor movement is obtained from ΔDVAL. This method enables critical frequency analysis of the BB system, identification of rotor delevitation severity, and forward or backward whirl detection capabilities. Different rotordynamic motions were found to depend on the rotor traversing specific critical frequencies of the AMB system. The magnitude of transverse movement was also found to be independent of the delevitation speed. Application of this method would be the comparison of rotor delevitation quality by various BB manufacturers for design and implementation purposes. This method also provides accurate verification of delevitation modelling by comparing simulated transverse movement to actual transverse movement Recommended future work includes the integration of delevitation severity indicators in RDE modelling. The effect of BB support stiffness and damping on life prediction methods should further be studied. An investigation of the effect of cage-less ceramic or lubricated bearings on life prediction methods is also recommended. A method for determining the identified failure thresholds from basic system variables is also required