Abstract Miniaturization is the future of the industrial revolution, and technologies like micromilling are crucial in ensuring that products meet market quality standards. Therefore, micromilling must be efficient and free from anomalies such as tool failures, vibrations, and chatter. The avoidance and detection of the anomalies can be achieved via force prediction due to their correlations. This work investigates the applicability of machine learning (ML) ensemble regressors for force prediction in micromilling. A systematic series of experiments was performed on hardened AISI H13 (50 ± 1 HRC), with force signals recorded, processed, and used to develop the regressors. Models including Random Forest (RF), stacked generalization, extreme gradient boosting (XGBoost), voting, adaptive boosting (AdaBoost), and gradient boosting (GB) were developed, evaluated, and compared. XGBoost achieved the best force prediction performance, with average RMSE , MAPE , and R 2 values of 0.41, 7.83%, and 0.98, respectively, on validation data. The XGBoost model was then utilized to develop the XGBoost-Grey Wolf Optimization algorithm, which optimized cutting parameters and forces. The optimization results indicate that a feed in the range 2.68–3.75 (µm/rev) combined with an axial depth of cut in the range 25-47.97 (µm) will yield optimal cutting ( F c = 1.55 N) and axial ( F z = 1.74 N) forces during micromilling of hardened steels using a TiAlN-coated carbide end mill. It is also found that the axial depth of cut is the most significant parameter in generating cutting forces, whereas feed has the highest impact in generating axial forces. These findings can enhance the precision of micromilling of hardened steels.
Isaya et al. (Mon,) studied this question.