Dimensional Stability and Mechanical Performance of Thermally Conditioned High‐Heat Polylactic Acid Parts Produced by Fused Filament Fabrication

Additively manufactured high‐heat PLA parts must balance dimensional fidelity and tensile performance. However, the joint effects of printing temperature and post‐printing annealing on geometry, mechanics, and model predictability remain under‐quantified. This study integrated controlled experiments with machine learning to clarify these links. Standardized Ingeo 3D850 specimens were printed at 190°C–260°C, and half were annealed at 100°C for 4 h. The dimensional accuracy (height, width, and length) and tensile metrics (ultimate tensile strength, stress at break, and deformation) were measured. Predictive models were trained for each thermal state using the printing temperature as the input. Annealing consistently tightened dimensional spread and centering, with the most stable outcomes typically at 210°C–240°C. Nonlinear learners outperformed linear regression for geometric targets, achieving high test‐set accuracy (width and length with R 2 up to ≈0.96), whereas mechanical targets were less learnable. UTS showed a near‐linear dependence on printing temperature in the as‐printed condition ( R 2 = 0.87), but predictability dropped after annealing ( R 2 = 0.59), consistent with competing crystallization, diffusion, and early degradation effects. The integrated analysis quantifies printing temperature‐ and condition‐dependent trade‐offs and provides process guidance. Annealing within 210°C–240°C printing histories improves dimensional reliability, and nonlinear machine learning is preferred for geometry, whereas linear models suffice for UTS in as‐printed parts.