An incremental spatiotemporal joint analysis framework for real-time tool breakage monitoring

Tool breakage monitoring technology remains a bottleneck, constraining the reliability of high-precision manufacturing systems. The real-time detection of tool breakage through sensor signals still faces dual challenges of complex working condition recognition and subtle damage feature extraction. To address these challenges, this study proposes an incremental spatiotemporal joint analysis framework for real-time tool breakage detection. First, a novel spatiotemporal joint distance is proposed. By introducing a correlation-gated spatiotemporal joint distance, subtle breakage-induced transients are selectively amplified while process fluctuations are suppressed. Second, an incremental dynamic tracking mechanism is introduced, employing sliding time windows to perform matching analysis between real-time monitoring signals and historical machining cycle data. This strategy reduces signal traversal time and meets the response requirements for real-time breakage monitoring. Finally, the kurtosis-based adaptive threshold decision method dynamically determines the current threshold by integrating historical information and fluctuation characteristics of current signals, thereby superseding conventional fixed threshold methodologies. The effectiveness and advantages of the proposed framework are demonstrated through both milling simulation signals and data collected from real machining processes. Comparative analysis demonstrates that the proposed framework achieves higher monitoring accuracy while effectively eliminating false alarms caused by non-machining segments and machining fluctuations. Furthermore, the proposed method demonstrates strong anomaly amplification capability, as reflected by a maximum deviation ratio of 14.9, while maintaining high computational efficiency with an average processing latency of less than 160 milliseconds. These results indicate its suitability for real-time monitoring applications in automated manufacturing environments.