Sandro Lic, Reseracher in Area Orchestration, has successfully defended his master’s thesis, “Sensorless Classification of Severe Tool Wear in CNC Milling using ResNet,” completed at Johannes Kepler University Linz.
His research addresses the high costs and complexities of traditional tool condition monitoring by proposing a fully sensorless approach that leverages existing internal machine data, specifically spindle torque, to predict when a cutting tool is nearing failure. By transforming these high-frequency machine signals into visual spectrograms and analyzing them with deep learning models like ResNet-50, the developed system achieves accuracy comparable to setups requiring expensive external sensors. Ultimately, this work provides the manufacturing industry with a highly scalable, cost-effective, and software-driven solution to predict tool wear, optimize machining processes, and significantly reduce unplanned downtime.
