A method for monitoring tool wear in varying operating conditions
By combining local feature extraction and feature transformation methods with one-dimensional convolutional neural networks and bidirectional long short-term memory neural networks, and utilizing an unsupervised domain adaptation strategy to align feature distributions, the accuracy problem of tool wear monitoring under varying working conditions is solved, thereby improving monitoring effectiveness and autonomous decision-making capabilities.
Patent Information
- Application Number
- CN202410775033.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-17
- Publication Date
- 2026-07-24
- Estimated Expiration
- 2044-06-17
AI Technical Summary
Existing technologies for tool wear monitoring under varying operating conditions have failed to effectively construct feature quantities that are insensitive to changes in process parameters, and rely on target domain sample labels, which limits the application of monitoring methods.
Local feature extraction and feature transformation methods are used to select insensitive features. A parallel hybrid neural network is constructed by fusing a one-dimensional convolutional neural network and a bidirectional long short-term memory neural network. An unsupervised domain adaptation strategy that represents subspace distance and orthogonal basis mismatch penalty is used to align feature distribution. A contrastive learning module is combined to retain cutting signal information under new working conditions.
It improves the accuracy of tool wear monitoring under varying working conditions and enhances the dynamic perception and autonomous decision-making capabilities of the CNC machining process.