ML-Based Prediction of Surface Roughness in CNC Turning Using Cutting Forces and Machining Parameters

Achieving optimal surface finish in CNC turning operations remains a critical challenge in precision manufacturing, directly influencing component longevity, assembly precision, and operational performance. Conventional approaches relying on empirical optimization and post-process inspection demonstrate significant limitations in real-time quality assurance and predictive capability. We present a computational modeling framework employing machine learning algorithms to forecast surface roughness (Ra) during turning operations on AISI H13 steel under fresh tool conditions. Our investigation utilized cutting parameters (speed, feed, depth) combined with three-axis force sensor measurements from a publicly available experimental dataset. We evaluated three regression architectures - linear baseline, Random Forest ensemble, and Gradient Boosting to establish predictive relationships. Experimental results revealed that ensemble methods substantially outperformed linear approaches, with Gradient Boosting achieving R² = 0.92, MAE = 0.09 μm, and RMSE = 0.15 μm. Feature analysis identified feed rate and resultant force magnitude as dominant predictors. These findings validate computational intelligence as a viable tool for surface quality forecasting in smart manufacturing environments.