A Fast Algorithm For Straightening Stroke Prediction Of Linear Guide Considering Fluctuation Of Material Properties

Linear guide is the core functional component of machine tool,its accuracy retention and straightness are important indexes to measure the performance of machine tool.Heat treatment and straightening are the key technologies in the machining process of linear guide rail.The hardened layer produced by heat treatment on the surface of guide changes the material properties of guide,which has an important influence on the prediction of subsequent straightening stroke.After heat treatment,the performance parameters of the guide are not exactly the same even in the same batch,there is a certain degree of fluctuation,which will directly affect the prediction accuracy of straightening stroke.In this paper,the linear guide rail with single hardened layer distribution after quenching is taken as the research object,and based on the prediction model of straightening stroke constructed based on elastic-plastic theory,the mathematical relationship between material property parameters and straightening stroke is explored.And for a kind of linear guide rail made of 45 steel,a fast straightening stroke algorithm which can adapt to the fluctuation of material property parameters is designed.In this algorithm,the straightening experiment and simulation data are deeply mined,and the neural network algorithm is used to accurately identify the material property parameters in the straightening process to predict the straightening stroke accurately and quickly,so as to improve the straightening efficiency.The main research contents and work are as follows:(1)Based on the elastic-plastic bending theory,the evolution law with the increase of bending curvature of section stress of the linear guide rail with different hardened layer distribution in the process of pressure straightening is analysed.And the moment-curvature model in the process of guide straightening is established.And then the theoretical model of straightening stroke prediction of the linear guide with hardened layer is formed.It is found that the yield strength ratio and relative thickness greatly affect the elastic-plastic distribution of the guide section,and then affect the straightening stroke of the guide.(2)The response surface method was used to design the simulation test with the initial deflection,yield strength ratio,relative thickness and straightening stroke as the influence factors and the residual deflection as the response value,and the finite element simulation was carried out according to the experimental group.The response surface model of the simulation data was analyzed by using Design-Expert software,and the regression equation between the material property parameters and the residual deflection was fitted to establish the mathematical model between the material property parameters and the straightening stroke.(3)Design and implement a fast algorithm for straightening stroke prediction of linear guide rail.The algorithm is implemented by Python programming language.The material property parameter identification module based on BP neural network is used as the core to adapt to the fluctuation of material property parameters.The regression equation fitted by response surface method is used to calculate the straightening stroke to improve the accuracy of stroke prediction.(4)The linear guide after single surface induction hardening is taken as the experimental object to measure the material property parameters and conduct the pressure straightening experiment.The relative thickness and yield strength ratio of linear guide after medium frequency and high frequency induction quenching are measured by hardness method,which can be used to determine the parameter values for subsequent straightening experiments.By comparing the straightening times required to reach the same straightness by using different straightening travel prediction methods,the speed and accuracy of the straightening travel prediction algorithm and the accuracy of the regression equation fitted by response surface method and BP neural network identification parameter module were verified.