Predictive maintenance of ball bearings for machines rotating with arbitrary velocity profiles

Recent research and development on direct–drive motor technology and on their control system push forwardthe application of these devices as electric cams in a large numbers of industrial applications, such as examplemotion control system for packaging machineries. Performances of these devices are much higher than themechanical solutions for machine motion, in terms of precision and operational speed, and, of course, intime required to reconfigure the motion profile.However, classical approaches to the ball bearing diagnosis that use analysis of vibration signals are notlonger applicable, as they have been developed on the hypothesis of drive constant speed. In fact, in mostof the current industrial application, the control system drives the motor to follow a complex and cyclic (i.e.motor shaft reverses its motion direction at each operational cycle) non constant speed motion profile.In those application, even Computed Order Tracking [1] (COT), which is the main fault bearing diagnosistechnique used in non-constant velocity applications, fails to detect incipient faults, as highlighted by Fyfeand Munck in [2].This paper presents a new procedure that modifies the COT to be successfully applicable to the diagnosisproblem of ball bearings in variable speed motion applications.