AI Model Developed to Predict Internal Defects in Metal Additive Manufacturing
A research team, led by Dr. Jeong Min Park of the Nano Materials Research Division at the Korea Institute of Materials Science (KIMS) in collaboration with Dr. Jaemin Wang and Prof. Dierk Raabe of the Max Planck Institute in Germany, has developed an artificial intelligence (AI)-based model. This model assesses the likelihood and characteristics of internal defects during the process design phase of metal additive manufacturing, aiming to enhance the reliability of metal additive manufacturing parts and expand their applicability for industrial mass production.
Overcoming Limitations in Metal Additive Manufacturing
Metal additive manufacturing faces limitations in industrial application due to microscopic internal defects that can cause component failure and performance degradation. Traditional quality evaluation methods have primarily focused on simple indicators like porosity. However, the impact on mechanical performance varies significantly based on the shape, size, location, and distribution of these defects.
Introducing an Explainable AI for Defect-Aware Design
To address these issues, the research team created an explainable artificial intelligence (Explainable AI) model. This model systematically analyzes and predicts the relationships between metal additive manufacturing process conditions, defect morphology, and mechanical performance. The approach allows for the prediction of potential internal defects and their performance impact from the process design stage, establishing a new framework for defect-aware process design and quality management.
Advanced Defect Analysis and Prediction
Key features of the AI model include its ability to analyze and predict internal defects generated during the laser powder bed fusion (LPBF) process based on morphological characteristics such as shape and distribution, rather than solely on the number or fraction of defects. Utilizing microstructural images, the model automatically analyzes pore size, non-circularity, and spatial distribution, correlating these factors directly with mechanical properties. This enables a quantitative explanation of how defects influence performance. The model is also designed to explain the reasons for increased defects and performance deterioration under specific process conditions.
For training, the research team comprehensively analyzed process conditions, powder characteristics, defect images, and mechanical property data across various metal additive manufacturing materials, including steel, aluminum alloys, and titanium alloys. This established an integrated framework that predicts how process variables and powder characteristics influence defect formation, and how defect morphology subsequently affects mechanical performance.
Driving Industrial Adoption and Efficiency
This technology is expected to improve the quality reliability of metal 3D-printed components and accelerate their mass production for high-value parts. It is applicable for process optimization and quality control in industries requiring highly reliable metal components, such as aerospace, defense, and mobility. Potential benefits include reducing defect rates, material waste, and rework costs, thereby enhancing overall industrial production efficiency.
Dr. Jeong Min Park stated that the research establishes a scientific framework explaining how specific types of defects directly influence performance. This work is expected to contribute to the broader industrial adoption of metal additive manufacturing in high-performance sectors.
The research was supported by the KIMS Fundamental Research Program, the Materials and Components Technology Development Program, and the Energy Efficiency Innovation Technology Development Program. The findings were published online on January 1, 2026, in Acta Materialia. The team plans to develop this technology into a digital twin–based quality management system for industrial application.