Bold claim: AI is redefining metal 3D printing by predicting internal defects before they happen.
A team led by Dr. Jeong Min Park from Korea Institute of Materials Science (KIMS) and collaborating researchers Dr. Jaemin Wang and Prof. Dierk Raabe from the Max Planck Institute have created an artificial intelligence (AI) model that can evaluate the likelihood and characteristics of internal defects during the design stage of metal additive manufacturing. This breakthrough promises to boost the reliability of metal 3D-printed parts and pave the way for mass production in industrial settings.
Metal additive manufacturing, especially using laser powder bed fusion (LPBF), holds great promise for producing complex, high-value components. Yet industrial adoption has been held back by microscopic internal defects that can trigger failures and degrade performance. Traditional quality checks often focus on basic indicators like porosity, but real-world mechanical performance depends heavily on defect shape, size, location, and distribution—factors that simple metrics may miss.
To tackle this, the researchers developed an explainable artificial intelligence (Explainable AI) model that can systematically map the relationships between process conditions, defect morphology, and mechanical performance. In other words, the AI can predict not just if defects will occur, but how their form and distribution will influence how a part behaves under stress. This enables defect-aware process design and proactive quality management from the earliest design decisions.
A core strength of the model is its emphasis on defect morphology—shape, size, and spatial arrangement—using microstructural images to extract features like pore size, non-circularity, and distribution. The model then correlates these features with mechanical properties, providing a quantitative explanation of how defects translate into performance changes under different process conditions. Unlike many AI systems that resemble a black box, this model offers transparent reasoning for its predictions.
The team conducted a broad analysis spanning process conditions, powder characteristics, defect imagery, and mechanical data across several metals, including steel, aluminum alloys, and titanium alloys. They trained the AI on these datasets to build an integrated framework that can step-by-step predict how process variables and powder properties drive defect formation, and how those defects, via their morphology, impact mechanical performance.
This technology has the potential to significantly improve the quality reliability of metal 3D-printed components and accelerate their adoption in mass production for high-value parts. It can support process optimization and quality control across industries that require highly reliable metal components, such as aerospace, defense, and mobility. By reducing defect rates and minimizing material waste and rework, the approach could raise overall industrial efficiency.
Dr. Park emphasized that the work does more than cut defects; it creates a scientific framework that explains how particular defect types directly affect performance. The researchers anticipate that their approach will support broader industrial adoption of metal additive manufacturing, especially in high-performance domains like aerospace, space, and defense.
Funding came from multiple sources, including the KIMS Fundamental Research Program, the Materials and Components Technology Development Program funded by the Ministry of Trade, Industry and Energy, and the Energy Efficiency Innovation Technology Development Program. The findings were published online on January 1, 2026, in Acta Materialia, a leading metallurgy journal.
Looking ahead, the team plans follow-up studies to extend the technology into a digital-twin–based quality management system suitable for real-world industrial use.
Would you like this rewritten piece to lean more toward a technical briefing for engineers, or a layperson-friendly overview with more everyday analogies and examples?
