一种基于物理引导神经网络的缓进给磨削温度预测方法
By introducing a physical-guided neural network into the temperature prediction of creep-feed grinding, and combining physical laws and loss functions, the problems of large error and low accuracy of traditional neural network models are solved, and more accurate temperature prediction and workpiece surface burn prevention are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NORTHWESTERN POLYTECHNICAL UNIV
- Filing Date
- 2023-08-16
- Publication Date
- 2026-07-17
AI Technical Summary
In existing methods for predicting grinding temperature using creep feed, traditional neural network models suffer from large errors and low accuracy. Furthermore, the black-box nature of these models leads to inaccurate predictions, making them unsuitable for effectively guiding actual production.
A physical-guided neural network approach is adopted, which transforms the physical laws of the grinding process into physical constraints, constructs a neural network model, and uses a physical-guided loss function and an empirical loss function to construct the final loss function. The model is then trained with an augmented dataset to improve the accuracy of temperature prediction.
It improves the accuracy of grinding temperature prediction, reduces errors, provides a basis for preventing workpiece surface burns, reduces data acquisition costs, and shows better prediction performance in small sample and noisy environments.
Smart Images
Figure CN117195695B_ABST