Predicting polymer milling accuracy via an artificial neural network integrating UTS and tool diameter

Achieving high dimensional accuracy in CNC machining of polymers is challenging due to their low stiffness and viscoelasticity. This study introduces an artificial neural network (ANN) framework to predict multi-axial dimensional deviations in the milling of polyamide (PA12) and low-density polyethylene (LDPE). The model innovatively integrates a fundamental mechanical property, ultimate tensile strength (UTS), with a key geometric parameter, tool diameter. A comprehensive experimental campaign was conducted under dry machining conditions with constant cutting parameters, systematically varying the tool diameter (3, 6, and 8 mm). The results demonstrated that larger tool diameters and higher material UTS values consistently led to a significant reduction in dimensional errors across the length, width, and thickness of the machined components. The developed feedforward ANN successfully captured the complex, non-linear relationship between these inputs and the outputs, demonstrating exceptional predictive performance with coefficients of determination (R 2 ) exceeding 0.997 for all deviations. This work conclusively shows that a simplified model based on intrinsic material and tool properties can effectively forecast machining outcomes, providing a powerful proof-of-concept tool for virtual process optimization to enhance precision and promote resource-efficient manufacturing of high-tolerance polymer components.