Research On Thermal Error Of CNC Machine Tool Based On DBSCAN Clustering And BP Neural Network Algorithm

The thermal error of machine tool is caused by the thermal expansion of the structure affected by the heat source,which occupies an important part in the total error.In the study of thermal error compensation,the establishment of thermal error model is mainly divided into finite element method and mathematical modeling method.Because the machine tool is subject to many internal and external heat sources,and the heat exchange mode can not be idealized,and the contact parts are complex and many,the analysis of the whole machine is extremely difficult,so it is difficult to establish accurate temperature field and thermal deformation model by general finite element method.The mathematical modeling method is to analyze the relationship between the temperature data of the key points of the machine tool and the thermal error data of the spindle to get the thermal error model.According to the model,the compensation parameters of the controller are calculated to compensate the thermal error,which is the most popular method at present.The key point temperature of machine tool has a nonlinear relationship with the thermal error of spindle,and it has certain dynamic and periodicity.How to establish an effective and accurate model has become a research hotspot.In order to reduce the influence of thermal error on the machining accuracy of machine tools,the methods of data acquisition system,selection of temperature measurement points,data processing and optimization of temperature measurement points,sample enhancement and modeling are studied.Combined with the spatial and thermal characteristics of the machine tool,the laser displacement sensor and infrared temperature sensor are determined to obtain the temperature of the machine tool and the displacement data of the main shaft.RJ45 and USB are used as the communication schemes for the temperature and displacement collectors respectively.The upper computer is designed to align the data according to the time and develop the storage and export functions.According to the principle of heating,the position of main heat source of machine tool is qualitatively analyzed.Determine the main heat exchange mode of the heat source part and the influence on the error of the machine tool.According to the correlation analysis of the influence of the ambient temperature on the thermal error,the temperature distribution and the ambient temperature distribution under the static condition of the machine tool are obtained.The temperature measured at the initial time of the machine tool can be taken as the ambient temperature,which shows that the influence of the ambient temperature on the thermal error is taken into account in the process of the thermal error data collection of the machine tool.The location of temperature and displacement measurement was determined,and the experimental scheme was designed quarterly according to the environmental impact.In order to ensure that the collected data are more close to the actual machining conditions,four machining conditions are designed according to the cutting elements.Data collection according to the scheme.K-S algorithm is used to analyze the temperature data which does not obey the normal distribution,so the box chart method is used to filter the outliers and delete them.The missing data is processed by Lagrange interpolation.Density clustering and Pearson correlation coefficient method are used to optimize the temperature measurement points.Through density clustering,temperature data can be divided into several categories.In each category,density clustering can effectively classify the vectors with linear correlation or similar density space into the same category,calculate the correlation coefficient between the temperature vector and the spindle displacement in each category,and select the temperature vector corresponding to the maximum correlation coefficient in the category as the optimized temperature vector data.GAM method is used to test the non-parametric regression of the optimized temperature data,and the regression fitting degree under single factor and multi factor is analyzed.It is found that the optimized temperature data have a great influence on the change of thermal error,indicating that the optimization effect is good,and the analysis shows that there is a non-linear relationship between them.The dynamic characteristics of thermal error caused by vibration,machining and other factors are studied,and the method of random dynamic error is proposed to enhance the data set to obtain the enhanced data set.The thermal error model of machine tool is established by BP neural network.Considering that the external force produced by the tool in the actual machining process has a certain impact on the thermal error of the machine tool,combined with the multi feature fusion method,the three features of cutting speed,the amount of cut and the feed speed are fused,and the thermal error model is established.It is found that the network prediction is more accurate after multi feature fusion.