A tool chatter detection method based on signal graphical representation and light weight deep network
By acquiring and graphically processing multi-channel signals in real time using intelligent cutting tools, and combining this with the EfficientNetV2-S network for chatter detection, the problems of signal non-stationarity, noise interference, and poor adaptability across working conditions in existing technologies are solved, achieving high-precision, low-latency online chatter detection and hierarchical identification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHONGQING UNIV
- Filing Date
- 2026-04-07
- Publication Date
- 2026-07-17
AI Technical Summary
Existing tool chatter detection technologies struggle to achieve high-precision, low-latency online detection and hierarchical identification when faced with non-stationary signals, noise interference, poor adaptability across operating conditions, and insufficient model interpretability.
By integrating a triaxial vibration sensor and a triaxial PVDF force sensor into a smart tool, multi-channel signals are acquired in real time, preprocessed, and then graphically mapped. Combined with the EfficientNetV2-S lightweight deep network for end-to-end supervised learning, real-time identification of flutter categories is achieved.
The method improves robustness to noise and transient disturbances, enhances cross-condition adaptability and interpretability, reduces computational overhead, and meets the real-time and engineering feasibility requirements of actual processing sites.
Smart Images
Figure CN122401166A_ABST