Journal bearing lubrication monitoring based on spectrum cluster analysis of vibration signals

Journal bearings are critical components for many important machines. Lubrication analysis techniques are often not timely and cost effective for monitoring journal bearings. This research investigates into vibration responses of such bearings using a clustering technique for identifying different lubrication regimes, and consequently for assessing bearing lubrication conditions. It firstly understands that the vibration sources are mainly due to the nonlinear effects including micro asperity collisions and fluid shearing interactions. These excitations together with complicated vibration paths are difficult to be characterized in a linear way for the purpose of condition monitoring. Therefore, a clustering analysis technique is adopted to classify the vibration spectrum in high frequency ranges around 10kHz into different representative responses that corresponds to different bearing modulus values and lubrication characteristics. In particular, the analysis allows sensitive signal components and sensor positions to be determined for monitoring the journal bearing effectively. Test results from self-aligning spherical journal bearings show that it allows different lubricant oils and different lubrication regimes to be identified appropriately, providing feasible ways to online monitoring bearing conditions.