Drilling is a crucial process in industrial production and the quality of the machined hole has a decisive impact on the final part quality. However, there are various disturbances in the manufacturing process, which makes the non-value-adding quality inspection unavoidable. In this paper, high-frequently recorded internal NC-signal data and the vibration sensor data from the machine protection control unit (MPC) are used to predict hole quality for the drilling process. The analysis of the preprocessed data reveals a linear association between the hole quality characteristics and extracted features. For inline quality monitoring, interpretable models for the quality characteristics of straightness and roundness are developed. The proposed approach showcases the potential as an economical alternative to quality inspection by random sampling in mass production.
Science Journal 