The MICS Project Approach for Fault Detection in CNC Machine Bearings

Defects in bearings constitute a major issue in Computer Numerical Control (CNC) machinery, affecting the process quality. In this work, we propose a fault detection strategy based on multivariate control charts. The proposed method employs a preliminary phase to check if the indicators collected for the three axes of motion of the machine are statistically uncorrelated. If so, the second phase consists in building multivariate control charts and the corresponding control limits for each axis, characterizing the indicators behavior in normal operating conditions. Finally, in the monitoring phase, new incoming data are compared with the corresponding control limits to detect faults. The proposed strategy is tested on two different CNC machines, located in the Leonardo S.p.A plant of Cisterna di Latina, Italy, yielding accurate detection of non-faulty and faulty bearing conditions.