A team led by Prof. ZHENG Mingjie from Hefei Institutes of Physical Science (HFIPS) of the Chinese Academy of Sciences (CAS) recently proposed a property-oriented design strategy of high-strength ductile reduced activation ferritic-martensitic (RAFM) steels based on machine learning.
This work was published in Materials Science and Engineering: A.
RAFM steels have been regarded as promising candidate structural materials for fusion reactors, owing to their good thermo-physical, thermomechanical and irradiation-resistant properties. However, it is a big challenge for the traditional trial-and-error experiments to raise the upper limit of its operation temperature. Machine learning models, which were developed in recent years, are mainly forward models built from compositions and processing parameters to properties. It is more attractive to build a property-oriented design model which can quickly find out the reasonable combination in the search space of compositions and processing parameters according to the targeted properties.
Therefore, the team proposed novel strategy based on machine learning. They also designed and prepared a new type of RAFM steel with excellent high-temperature tensile properties.
Prof. ZHENG explained the whole process. In this research, they chose gradient boosting regression algorithm to construct the forward model which has high precision and reliability for predicting the tensile properties of RAFM steels. Then they established reverse model using artificial neural network algorithm to provide the possible combinations of compositions and heat treatments for given tensile properties.
In the end, the intelligent design model, combining the forward model and the reverse model, was developed to implement the property-oriented compositional and processing design.
Scientists verified the validity of this model with the experimental data reported in the relevant literatures. They even designed a new RAFM steel with the help of the model and prepared experimentally.
"The ultimate tensile strength (UTS) and total elongation (TE) reached 539 MPa and 20.6% at 600 °C, respectively," said Prof. ZHENG, "it's consistent with the targeted values. Especially when maintaining the comparable elongation, the UTS at 600 °C is higher than the conventional RAFM steels."
It proved that the strategy proposed was suitable for the property-oriented design of RAFM steels and could also be used as a promising approach to develop other kinds of high-performance structural materials.
This work was supported by the National Natural Science Foundation of China and the National Magnetic Confinement Fusion Science Program of China.
The schematic diagram of the intelligent design model. (Image by LI Xiaochen)