Overview on condition monitoring for bearings in variable speedconditions: the example of a packaging machine

This paper proposes a review of some techniques for the diagnostics of ball bearingbased on the experience of the authors on an industrial application. Frequently engineering industrysuggests non-trivial problems and new fields of research for the academy. This is the case of bearingdiagnostics in direct-drive motors, for example. Direct-drive are brushless motors fully controlledby the drive system. Thanks to an encoder or a resolver mounted on the shaft they could performcomplex motion profiles such as polynomial or splines, including reverse rotation of the shaft. Themain advantage of direct-drive motors is the removal of cams or gearboxes afterwards the motorwith a consequent reduction of economic and maintaining costs. Indeed the main drawback is thedifficulty to make diagnostics on their bearings. Regarding the bearing diagnostics, most of theliterature techniques are based on the search of fault-characteristic frequencies in the vibration spectrumof the motor. These fault frequencies are linearly dependent on the rotational frequency of theshaft supposing it is constant. In direct-drive motors the rotational speed changes continuously andconsequently the fault frequencies are meaningless. Diagnostics of machineries in non-stationaryconditions is attractive and promising field and recently different papers have been proposed in literature[1] and thematic conference [2] organized, covering a wide range of applications, e.g. frommanufacturing [3] to mining industry [4]. Focusing on a specific industrial case the authors runthrough their experience on bearing diagnostics for a packaging machine working under variablespeed condition. The developed techniques include an improved version of the order tracking, theuse of correlation function, wavelet analysis and two supervised learning approaches: the artificialneural networks and the support vector machines. Moreover the closing part of the paper covers thebearing diagnostics by mean of the stator current signal analysis of the motor