A reconstructed variable regression method for thermal error modeling of machine tools

作者：Li, Yang; Zhao, Ji^*; Ji, Shijun

来源：International Journal of Advanced Manufacturing Technology, 2017, 90(9-12): 3673-3684.

DOI：10.1007/s00170-016-9648-3

• Thermal error
• Empirical modeling method
• CNC machine tools
• Reconstructed variable regression
• Fuzzy C means clustering

For the thermal error model of computer numerical control (CNC) machine tools based on empirical modeling method, selecting the appropriate temperature-sensitive points is very important. Owing to the existence of multi-collinearity among temperature variables, the accuracy and robustness of the error model are degraded. Therefore, a reconstructed variable regression (RVR) algorithm is presented firstly in this paper. The reconstructed variable is derived from representative temperature points which are selected from the raw temperature data through fuzzy C means (FCM) clustering method. Then, the reconstructed variable is utilized to build the thermal error model based on regression analysis. Meanwhile, a criterion is proposed to find out the optimal cluster result from the whole results of clustering. RVR algorithm was validated through the thermal error experiments of positioning error on a 3-axis machine tool. The number of temperature measuring points could be reduced from 9 to 2. The results of experiments showed that the fitting accuracy of RVR model could reach 92 % and the prediction accuracy could reach about 90 %. It is verified that the RVR model is of high prediction accuracy and strong robustness which is a great choice for thermal error modeling.

• 出版日期2017-6
• 单位吉林大学