In-process Quality Control of Drilled and Reamed Bores using NC-Internal Signals and Machine Learning Method

Abstract

The increasing amount of available process data from the machining processes and machine learning methods lead to new approaches to the workpiece quality control in an early machining stage. We gather process data directly from the numerical control (NC) of a milling machine during drilling and reaming of hydraulic valves in a serial production. Different feature extraction strategies are applied to obtain features from the time series data and the machine learning method of extra tree regressor is selected to predict the quality characteristics of the bores. A comparison of the feature extraction strategies is conducted and very precise prediction results are obtained.