ON THE CONFIDENCE OF LEARNED CONSTRAINTS ATTAINMENT OPTIMIZATION THROUGH CONFORMAL PREDICTION IN 3D PRINTING

3D printing is a relatively recent manufacturing method with significant potential due to its speed, cost-effectiveness, and geometric freedom. However, high variability in quality and mechanical strength limits its widespread industrial application. This study analyzes surface rough ness, cylindricity, and compressive strength across various printing parameters. Supervised learning models were trained for conformal prediction of these responses relative to a defined parameter region and coverage probability, with the objective to compare the confidence of the results. Cons trained optimization was performed on surface quality to maximize compressive strength. Using conformal prediction, the optimization was evaluated at 90 % and 50 % confidence levels. These levels identified two distinct viable regions for optimization for the analyzed settings, with the highest confidence levels presenting lower compressive strength than the lower confidence levels, highlighting the impact of confidence analysis.