Measurement And Analysis Of Thermal Error In CNC Machine Tool

The CNC machine tool is the main machinery equipment in production and it’s machining accuracy determines the quality of products. To improve the machining accuracy of machine tool can greatly promote the process of industrialization. Through the measurement and research of thermal error from CNC machine tool to build the thermal error model with high accuracy is the main purpose of this paper and the thermal error compensation techniques can improve the machining accuracy of machine tool.This paper uses the finite element analysis method to proceed the thermal structure coupling analysis of the spindle box assembly for CNC machine tool CK6142 and concludes that the contact area of bearing with spindle and spindle box is the heat sensitive area, by combining with the type of sensors to determine the installation position of temperature and displacement sensors. Through the C++Builder software the multi-channel data acquisition system has been programmed and finally the acquisition of experimental data has been completed. By the utilization of correlation analysis and fuzzy clustering method to optimize the classification of multiple temperature data from measuring points, and two better classification results are obtained. Referred to the correlation coefficient with thermal error the respective optimization points are determined and finally by comparing the corresponding model’s precision to conclude that the plan with the correlation analysis method to proceed optimizing measuring points is better.Taking the method of multiple linear regression analysis and BP neural network to build the thermal error models and combing different measuring points’ plan to build more models, by comparing the accuracy of models it is concluded that with more temperature measuring points the precision of model form the method of neural network is higher; and the accuracy of model built by multivariate linear regression is higher when less measuring points are used, this is because the neural network needs a large number of training samples. Finally, by verifying the optimal model in experiment, it is known the prediction accuracy of thermal error model is higher and the research purpose is achieved.The thermal error model built has high accuracy and reaches the requirement of compensation for thermal error, which provides the basis for the realization of the compensation technique. Meanwhile, it is shown the correctness of the plan with optimized measuring points and provides an economic and convenient way for thermal error modeling and compensation to speed up the development of compensation technology.