Prediction of thermal errors in dual spindle CNC machine tools considering outlier data repair

In order to better model the thermal deformation of the spindle for error compensation and solve the influence of unavoidable outliers in experimental data caused by different measurement errors and environmental interference on modeling, this paper uses the quartile method to detect and remove outliers, and uses the Adaboosting-bp method for data repair to reduce the influence of outliers on modeling. Firstly, the experimental data collected is subjected to outlier detection and outlier data repair. Then, the processed data is screened for important temperature measurement points using hierarchical clustering and Pearson correlation coefficient method. For dual spindle data, this paper considers the influence of another spindle on predicting the temperature of the spindle, establishes feature matrices separately, and studies the nonlinear prediction of thermal deformation through slime mold algorithm optimized gated recurrent unit. Finally, the results of temperature prediction for single and double spindles with and without outliers were compared and analyzed. The results show that the thermal deformation prediction of removing outliers with dual spindle is basically consistent with experimental testing, Compared with the thermal deformation prediction of single spindle with outlier removal and single spindle without outlier removal, for spindle 1, the root mean square error (RMSE) decreased by 0.0921 μm and 0.1164 μm, the mean absolute percentage error (MAPE) decreased by 0.0425 and 0.0803, R-squared (R 2 ) increased by 0.0187 and 0.0389; for spindle 2 , the Rmse decreased by 0.0195 μm and 0.0307 μm, and MAPE decreased by 0.001 and 0.014, R 2 increased by 0.0043 and 0.007 respectively. This indicates that the dual spindle thermal error prediction model, which removes outliers and performs data repair, is effective and of great significance for precision control in high-precision manufacturing.