Hybrid Modeling of Cutting Forces in Turning of Engineering Polymers Using Kienzle Approach and Data Augmentation-Enhanced Random Forest Algorithm

In this study, the turning processes of Polyamide 6 (PA 6) and High-Density Polyethylene (HDPE) engineering polymers were analyzed by integrating experimental, mathematical, and machine learning-based methods. Within the framework of a Taguchi L8 experimental design, the effects of cutting speed (Vc), cutting depth (ap), and feed rate (f) on cutting forces (Fc) were investigated. ANOVA results revealed that cutting depth was the most dominant factor for both materials (44.49% for PA 6 and 64.49% for HDPE), while PA 6 was found to be twice as sensitive to feed rate variations compared to HDPE.Through the Kienzle mathematical approach, specific cutting force coefficients (kc1.1) were calculated as 152.45 N/mm^2 for PA 6 and 74.89 N/mm^2 for HDPE. However, to overcome the predictive limitations of traditional models caused by the viscoelastic and non-linear behavior of polymers, the Random Forest (RF) regression algorithm was incorporated into the process. To enhance the model's generalization capability under limited data conditions, a Data Augmentation technique was employed.The findings demonstrated that the RF model, supported by data augmentation, achieved remarkably high average accuracy rates of 99.09% for HDPE and 91.53% for PA 6. This research proves that hybrid approaches, where traditional physical models are supported by machine learning, provide a high-reliability force prediction infrastructure for the precision manufacturing of polymer-based components and digital twin applications.