A Fourier Series-Neural Network Based Real-Time Compensation Approach for Geometric and Thermal Errors of CNC Milling Machines

Abstract

In order to improve the accuracy and efficiency of the real-time error compensation, a Fourier Series-Neural Network (FS-NN) based compensation methodology is developed to reduce the thermally induced geometric errors of CNC milling machines under varying conditions. Firstly, an error model is presented based on the principle of Fourier series with fast calculation and higher precision, which is regarded as the error base of original geometric errors. Secondly, the relationships between the slopes related to different error curves and the temperatures of key thermal points are figured out by using neural networks when concerning thermal effects. Then, a combined error model is established which is suitable for reducing axis positioning errors at any thermal status. Finally, a real-time dynamic error compensation system is developed featuring automatic error modeling and multiaxis synchronous compensation. The experimental results prove that the proposed methodology has satisfactory modeling accuracy and robustness to frequently changing working conditions.