Thermal Error Model of Linear Motor Feed System Based on Bayesian Neural Network

The linear motor feed system has been in service complex working conditions for a long time, thus causing the nonuniform distribution of temperature field distribution. Thus, thermal error become key factor affecting motion accuracy. In order to maximize accuracy and efficiency compensation systems, an improved modeling method based on Bayesian neural networks is proposed combination with strong generalization performance avoidance overfitting networks. And specific ideas are as follows: Firstly, X-Y cross-type two-axis taken test object. Due traditional network’s slow convergence, overfitting, underfitting problems, network used model system. Secondly, avoid influence multicollinearity data final results, grey relation analysis screen measuring points. large degree selected ensure prediction network. Thirdly, variables sensitive points positioning errors input samples. Fourthly, established. Fifthly, hyperparameters determined by calculating Hessian matrix Gauss-Newton approximation. finally, comparison constructed ordinary Levenberg-Marquardt algorithm after series experimental demonstrations see that can be enhanced up 10%. It also shows advantages high precision, ability, anti-disturbance solid robustness, etc. Therefore, expected widely predicting compensating high-speed CNC machine tools practical machining occasions.