In-process detection of grinding burn using machine learning

Abstract The improvement of industrial grinding processes is driven by the objective to reduce process time and costs while maintaining required workpiece quality characteristics. One several limiting factors burn. Usually applied techniques for burn are conducted often only selected parts can be consuming. This study presents a new approach detection realized each ground part under near-production conditions. Based on in-process measurement acoustic emission, spindle electric current, power signals, time-frequency transforms derive almost 900 statistical features as an input machine learning algorithms. Using genetic programming, optimized combination between feature selector classifier determined detect application results in high classification accuracy about 99% binary problem more than 98% multi-classdetection case, respectively.