Research On Robust Modeling Technique Of Thermal Error Of NC Machine Tool Spindle Based On Thermal Image

The thermal error of the machine tool is one of the main factors affecting the machining accuracy of the CNC machine tool.The spindle as the core component of the CNC machine tool is one of the main heat sources of the machine tool,and its thermal deformation directly determines the metal removal rate and the machining accuracy of the parts.Therefore,accurate mathematical modeling of the thermal error of the machine tool spindle is the key to improve the machining accuracy.In order to more accurately model the thermal error of the machine tool spindle,this study proposes a thermal image-based robust modeling technique for the thermal error of the CNC machine tool spindle.The thermal image and the convolutional neural network are introduced to make the modeling research intelligent and Universality.In this paper,the spindles of Shenyang machine tool CAK3665 lathe and Hanchuan machine tool XK714 D milling machine are studied.Based on the thermal image of the spindle and the convolutional neural network,a modular model of spindle thermal error prediction is established to achieve high-precision and high robustness prediction of spindle thermal error.This study provides important technical support for improving machine tool machining accuracy.The main research contents of this article are arranged as follows:(1)Theoretical explanation of convolutional neural network.The basic theory of convolutional neural network is explained,the corresponding calculation of each basic module in convolutional neural network is explained in detail,and the operation calculation process of the thermal image of the front bearing of the spindle of the CNC machine tool in the convolutional neural network is visually displayed and explained.(2)Experimental data acquisition.An experimental platform is built,Infrared thermal imaging camera is used to take batch thermal image of the front bearing of the spindle,and at the same time,the thermal deformation data of the spindle is obtained based on the displacement sensor.In order to verify the robustness of the model,the thermal error data of the spindle at different speeds of the machine tool spindle are collected.(3)Data processing and modeling.Batch thermal pre-processing is performed on the thermal images,including digitization,median denoising,flipping and random cropping,etc.,to enhance and expand the data.Finally,the main axis thermal error data is used as the label to create and segment the data set.A deep learning platform is built,and a modular model is proposed to compare with the basic model and the traditional BP neural network model.(4)Comparative analysis of model prediction accuracy.The accuracy of thermal error prediction between the modular model,the basic model and the traditional BP neural network model is compared,which verifies the high accuracy and high robustness of the modular model.The results confirm that the thermal error model of the numerical control machine tool spindle based on the thermal image can effectively predict the spindle thermal error.It has been verified by experiments that the model predicts the thermal error of the spindle of the lathe and milling machine at a temperature of more than 80%,and at a temperature of less than84%.After changing the spindle speed,the prediction accuracy of the model reaches more than80% under the temperature rising state.Under the cooling state,they all reach more than 82.2%.The verification results confirm that the thermal error prediction model based on thermal image has high precision and high robustness.It has been verified by cutting experiments that the size error of the workpiece after lathe compensation and cutting is reduced by about 45% on average,and the size error of the workpiece after milling and compensation is reduced by about 54% on average.The cutting experiment results confirmed the practical feasibility of the model.