Presentation Information
[1G12]Study on Machine Learning Methods for NNP Development for Nuclear Materials
*Baopu WANG1, Hiroto OSAKI1, Yuting CHEN1, Liangfan ZHU1, Kazunori MORISHITA1, Kenichi FUKUMOTO2 (1. Kyoto Univ., 2. Univ. of Fukui)
Keywords:
Nuclear materials,Radiation damage,Neural network potential,Machine learning method
To simulate the irradiation damage behavior of nuclear materials, interatomic potential functions are used in molecular dynamics simulations, and their accuracy has a significant impact. However, conventional empirical potentials have difficulty in adequately describing the complex phenomena of irradiation damage, and improvements in this area are needed. Therefore, to improve the accuracy of potential functions, neural network potentials (NNP), which are based on machine learning, are being studied and developed. However, it is necessary to evaluate the learning capability of machine learning and its effectiveness in improving the accuracy of the potential functions. In this study, we aim to develop and evaluate NNPs for simulating irradiation damage in nuclear materials by exploring machine learning methods to reproduce existing empirical potential functions, with the goal of eventually applying these methods to the development of potential functions for nuclear materials.