Predictive tool path compensation using ANN to minimize slenderness-induced errors in dry turning of UNS A97075-T6

Abstract

This study investigates dimensional deviations in dry turning of UNS A97075-T6 aluminium alloy, focusing on the impact of part slenderness and machining parameters. Experimental tests were conducted on specimens with varying diameters (10–18 mm), cutting speeds (40–80 m/min), and feed rates (0.05–0.15 mm/rev). Results reveal nonlinear deviation patterns, with maximum deviations up to 0.40 mm in slender parts. An iterative tool path compensation process reduced average deviations by 87%, achieving final errors below 0.10 mm. Based on these experiments, a feedforward Artificial Neural Network (ANN) was trained using diameter, cutting speed, and feed rate as inputs, and the compensated tool path as output. The ANN showed excellent predictive performance (R2 = 0.98, RMSE = 0.0012 mm), enabling first-pass trajectory corrections without iteration. This approach improves part accuracy while reducing cost and time. The novelty lies in considering part slenderness as a key factor affecting final part dimensions to meet tolerance requirements. Additionally, the use of Artificial Neural Networks enables the inclusion of slenderness effects in tool path prediction, allowing for the correction of potential geometrical deviations. These results offer practical insights for enhancing dimensional control in manufacturing processes.