Research On The Key Technologies Of Thermal Error Modeling For CNC Machine Tool Spindle

With the improvement of automatic production,CNC machine tools have become an indispensable force for industrial development and technological progress.In the case of high quality requirements,the machining accuracy of CNC machine tools directly reflects the level of processing and manufacturing in a country.Among the many error sources that affect the machining accuracy of machine tools,thermal error can account for 40%-70% of the total error.Therefore,the modeling and compensation technology of spindle thermal error of CNC machine tools can effectively improve the machining accuracy.The measurement of the thermal error of the spindle,selection of the temperature-sensitive points and establishment of thermal error model are the preconditions for the implementation of thermal error compensation.This paper takes the vertical machining center as the analysis object,and studies the related technology of thermal error modeling for the spindle of CNC machine tools.Based on the active reference and absorption of related theories and research results,the research work of spindle thermal error measurement is carried out.The temperature sensor,displacement sensor and corresponding data acquisition board are selected reasonably according to the actual situation.Arrange the installation position of the temperature sensor according to the main factor strategy and the minimum distribution strategy,and the installation arrangement of the displacement sensor is carried out based on the "five-point method" in ISO230-3.Different experimental conditions are designed and the measurement experiments of the temperature field and thermal error of the spindle are implemented,which provided the data support for the subsequent research of the thermal error modeling technology.In view of the different influence of different heat source areas on the spindle thermal error,a multi theory synthesis method considering the whole heat source area is proposed to select the combination of temperature sensitive points.Based on the different influence on the thermal deformation of the spindle,the machine tool is divided into five heat source areas.The K-means clustering algorithm based on correlation analysis is used to cluster and filter the temperature variables to obtain the key temperature variable combination corresponding to different K-values.Setting the value range of K can effectively solve the problem that K-value is difficult to determine.Based on the key temperature variable combinations in each heat source area,the multi theory synthesis method is used again to select the global optimization temperature-sensitive point combination.BP neural network model is used to establish the internal relationship between temperature variables and thermal deformation.Residual mean value(RMV)and root mean square error(RMSE)are used to evaluate the results of the model to more intuitively select the global optimal temperature-sensitive point combination that are most conducive to thermal error modeling.According to the different influence degree of different temperature variables on the spindle thermal error,a method of temperature-sensitive point combination selection and thermal error modeling based on weighted integration is proposed.Set the value range of K,analyze the data based on correlation analysis and K-means clustering algorithm,establish BP neural network models respectively for different key temperature variable combinations,assign weights to the prediction output of different models,and obtain the optimal combination of weights.Each weight in the optimal weight combination and its corresponding temperaturesensitive point combination and thermal error model can be weighted integrated to obtain the required weighted integrate temperature-sensitive point combination and weighted integrated thermal error model.It should be noted that the weighted integrated temperature-sensitive point combination is used as the input of thermal error model,while the weighted integrated thermal error model can be used for the final thermal error modeling.The results obtained by the same method should be used differently.The weighted integrated temperature method can not only avoid the problem of local extremum,but also reduce or even eliminate the adverse effect of the collinearity of temperature variables.The test results under different experimental conditions also reflect the effectiveness of the weighted integrated temperature method.Based on the selected temperature-sensitive point combination,the spindle thermal error modeling of CNC machine tool is studied.Considering the nonlinear relationship between temperature variable and thermal deformation,the excellent nonlinear mapping ability and no local extremum problem of radial basis function(RBF)neural network,one RBF neural network model based on CSO algorithm optimization is established.RBF neural network is responsible for thermal error modeling,and CSO algorithm is used to optimize the initial structural parameters of RBF neural network to improve the prediction performance of the model.The feasibility of CSO-RBF neural network in the thermal error modeling of machine tool spindle and its good prediction accuracy and robustness are proved by the verification experiments under different working conditions.