Professor Kim Hyun-seop and doctoral student Lee Jeong-ah from the Graduate School of Ferrous and Eco-Materials Technology and Department of Materials Science and Engineering at Pohang University of Technology (POSTECH), in collaboration with Professor Figueiredo from the Department of Metallurgy and Materials Engineering at the Federal University of Minas Gerais in Brazil, have developed an advanced artificial intelligence model that accurately predicts the yield strength of various metals, effectively overcoming traditional time and cost constraints.
Yield strength indicates when a material, such as a metal, begins to change shape due to an external stress. In materials engineering, accurately predicting yield strength is essential for creating high-performance materials and improving structural stability.
However, predicting this property requires taking into account many factors, such as the material’s grain size and the type of impurities, and typically requires extensive experiments over a long period of time to collect the data.
To address this challenge, the Hall-Petch equation, which establishes the relationship between the yield strength of a material and grain size, is commonly used, but this equation has limitations in accurately predicting the yield strength of new materials, taking into account their specific properties and various environmental conditions such as temperature and strain rate.
This research project integrated physical principles and AI techniques to improve prediction accuracy and reduce the time and cost required to predict yield strength. By understanding how particles move within a material, a machine learning model was created along with a machine learning algorithm that utilizes the “grain boundary sliding” mechanism in predicting yield strength.
Initially, the team used a black-box model to evaluate the effect of different material properties on yield strength. Later, a white-box model with clearly defined inputs and outputs was developed to improve the accuracy of yield strength predictions.
Model validation included testing a range of ferrous alloys that were not included in the training data for the yield strength prediction model, and showed that even when predictions were made on untrained data, the model showed high accuracy with a mean absolute error of 7.79 MPa compared to the actual yield strength.
Professor Kim Hyun-seop of POSTECH expressed his aspirations as follows: “We have developed a generic AI model that can accurately predict yield strength for many different metals and under a variety of experimental conditions.” He added, “We will continue to actively utilize AI technology to make great advances in materials engineering research.”
Journal References:
- Jeong Ah Lee, Roberto B. Figueiredo, Hyojin Park, Jae Hoon Kim, Hyoung Seop Kim. Uncovering the Yield Strength of Metallic Materials Under Diverse Experimental Conditions Using Physics-Enhanced Machine Learning. Acta Materialia, 2024; DOI: 10.1016/j.actamat.2024.120046
