Abstract
The main reason that restricts the increase of fast reactor burn up is the radiation swelling of the cladding. The reported research shows that the main elements and trace alloy/impurity elements have a great influence on the radiation swelling. However, the behavior of the above elements in the process of coupling with radiation defects is complex, and it is difficult to directly measure the relationship between these elements, radiation defects and microstructure evolution in experiments. The emergence of machine learning and big data mining technology will help to gain new understanding of the impact of radiation swelling on austenitic stainless steel, so as to find a new type of austenitic stainless steel cladding material resistant to radiation swelling. Therefore, in this work, about 1000 groups of data such as composition, irradiation conditions and radiation swelling of austenitic stainless steel are collected, and the data are cleaned and screened for modelling by machine learning. The deep neural network with back propagation is used in this work, and the correlation between alloy composition such as Cr, Ni, Ti and C, irradiation dose and temperature and radiation swelling of austenitic stainless steel is established. The results show that the addition of a certain amount of Ti and Si can effectively inhibit the radiation swelling of austenitic stainless steel, but the addition of Ni will aggravate the swelling effect. The addition of Cr, Ni, Ti and Cr will increase the swelling inflection point dose, while the addition of C and P will reduce the swelling inflection point dose. Besides, the influence of multi factor coupling such as composition on radiation swelling of austenitic stainless steel will help to promote the material optimization design of austenitic stainless steel cladding material.
