Researchers at POSTECH and the Korea Institute of Materials Science (KIMS) have developed an AI-based analytical framework that can predict the mechanical strength of metal 3D printed components in seconds, even when internal defects are present.

The work was led by Professor Kim Hyeong-seop of POSTECH’s Graduate Institute of Ferrous & Eco Materials Technology and Department of Materials Science and Engineering, and was conducted in collaboration with Senior Researcher Park Jung-min of KIMS. Findings were published in the international materials science journal Acta Materialia.

The defect problem in metal additive manufacturing

The layering process in LPBF frequently produces small, bubble-like internal voids that can become critical weaknesses in parts intended for demanding applications such as aircraft engines or automotive assemblies. But to quantify the effect of these voids on structural strength requires extensive and costly repetitive testing.

AI framework predicts metal 3D printed part strength in seconds

Instead of attempting to eliminate defects, the research team trained the AI to work with them. The model ingested a broad dataset covering manufacturing parameters — including laser power and scanning speed — alongside data on internal microstructure, as well as the size and spatial distribution of voids.

A technique described as “data-selective learning” was then applied to identify the variables with the greatest influence on strength, improving prediction accuracy.

Explainable results and verified accuracy

A distinguishing feature of the framework is its capacity to produce human-readable equations alongside its predictions, rather than functioning as a black box.

Validation tests conducted on an aluminum-silicon-magnesium (Al-Si-Mg) alloy, widely used across the aerospace and automotive sectors, returned strength predictions with a mean error of 9.51 megapascals (MPa), a result the team stated was more than four times more accurate than existing methods.

The researchers indicated the framework could be extended into a defect-aware design map, giving engineers advanced visibility into how component performance varies with changes in manufacturing conditions.

“This technology will enhance the reliability of metal 3D printed parts, greatly accelerating their commercialization in fields like aerospace and automotive,” said Kim Hyeong-seop.