Research on carbon emission prediction methods for CNC machine tools based on robust support vector machines

Due to severe environmental issues and stringent carbon emission policies, carbon emission from machine tool processing have become a key area of focus in current research. To address the limitations of traditional prediction methods in handling nonlinear relationships and small-sample conditions, this study proposes a Robust Support Vector Machine (RSVM) method that integrates adaptive multi-scale kernel functions for accurate carbon emission prediction. Experimental data were collected from a CNC lathe, and three machine learning prediction models were established: Multi-gene Genetic Programming (MGGP), Robust Support Vector Machine (RSVM), and an Improved Backpropagation Neural Network (IBP). Model performance was evaluated and compared using the coefficient of determination (R²), mean absolute error (MAE), and root mean square error (RMSE). The results show that RSVM achieved an R² value of 99.27% on the test set, with RMSE and MAE values of 0.0351 and 0.0278, respectively. The R² value of RSVM is 8.54% higher than that of MGGP and 6.17% higher than that of IBP. Comparative evaluation indicates that RSVM demonstrated the best predictive performance, outperforming MGGP and IBP.