Search

One of the pillars of the smart factory concept within the Industry 4.0 paradigm is the capability to monitor the health conditions of production systems and their critical components in a continuous and effective way. This could be enabled through the implementation of innovative diagnosis, prognosis and predictive maintenance actions. A wide literature has been devoted to methodologies to monitor the manufacturing process and the tool wear. A parallel research field is dedicated to isolate the health condition of the machine tool from the production process and external source of noise. This study presents a novel solution for machine health condition monitoring based on the so-called “fingerprint” cycle approach. A fingerprint cycle is a pre-defined test cycle in no-load conditions, where the axes and the spindle are activated in a sequential order. Several signals are extracted from the machine controller to characterize the current health state of the machine. The method is suitable to separate drifts, trends and shifts in CNC signals caused by a change in machine tool health condition from any variation related to the cutting process and external factors. A machine learning method that combines Principal Component Analysis and statistical process monitoring allows one to quickly detect degraded conditions affecting one or multiple critical components. A real case study is presented to highlight the potentials and benefits provided by the proposed approach.