Real-time monitoring of drilling operations in Computer Numerical Control (CNC) machine tools is crucial to ensure process reliability and minimize downtime. This is enabled by the detection of anomalies such as missing workpieces, incorrect workpiece positioning and drill bit breakages. The aforementioned anomalies can be detected by segmenting the drilling process signals into distinct phases and identifying unexpected transitions. In addition, the segmented data can be utilized for a fine-grained, phase-specific monitoring and to support further analytical processes, enabling the development of more sophisticated algorithms for fault diagnosis and process optimization. This paper presents TEPS (Tool Engagement based Phase Segmentation), which is a novel, computationally lightweight unsupervised machine learning algorithm - inspired by k-means clustering - that efficiently determines a binary state indicating whether the drill is engaged or not. To ensure reliability, the algorithm continuously adapts to handle complex scenarios such as weak or noisy signals. By combining the engagement state with other machine signals, a second segmentation into finer phases is performed. The presented approach is specifically designed for streaming data directly from a CNC control unit. This eliminates the need for additional sensors or other costly hardware investments and keeps the complexity to a minimum. The algorithm adapts quickly to new process parameters and requires minimal memory. Experimental results demonstrate reliable engagement-state detection, with the derived phases enabling anomaly detection and opening new possibilities for monitoring in single-part and small-series settings.
Denkena et al. (Thu,) studied this question.