Theoretical study of GDM-SA-SVR algorithm on RAFM steel

Machine learning in nuclear materials research

“…There have been a number of studies using ML to model radiation effects in low-activation ferriticmartensitic (LAFM) steels with fitting to databases containing measured doses up to 90 or 100 dpa, relevant for fusion and next-generation fission applications. There are three primary studies, by Kemp et al 95 (in 2006) and Long et al 96 (in 2020) on the yield strength (sy), and Cottrell et al 97 on Charpy ductile-brittle transition temperature (DBTT) shifts (in 2007). Windsor et al, [98][99][100] followed up the Kemp and Cottrell studies with a series of tests using flux extrapolation of those models, and Kemp et al 101 and Windsor, et al 102 also applied these models in experimental and materials design for fusion reactor materials testing and development, respectively.…”

“…Finally, we discuss a very recent study from Long et al 96 that used what appears to be the same database from Yamamoto, as Kemp et al in their original work 95 , and modeled yield stress as a function of all 37 compositions, irradiation parameters, and processing parameters available in that database. The author performed extensive comparison to multiple ML methods, including backpropagation and general regression NNs, linear regression, random forest, and their new approach, a method of Support Vector Machine denoted GDM-SA-SVM.…”

Insights into Prismatic Loop Formation in Irradiated Fe–Cr Alloys from Hypothesis-Driven Active Learning and Causal Analysis

“…An alternative approach is to take advantage of the developments in machine learning (ML) techniques to construct models from experimental data and infer the underlying mechanisms . There have been studies using ML to model radiation effects in low-activation ferritic-martensitic (LAFM) steels at doses up to 90 or 100 dpa. − - Neural networks have been used to model the irradiation hardening of low-activation ferritic-martensitic steels with prediction target of yield stress . Bayesian neural network model has been predicted changes in the Charpy ductile-brittle transition temperature (ΔDBTT) of low-activation martensitic steels .…”

“…Bayesian neural network model has been predicted changes in the Charpy ductile-brittle transition temperature (ΔDBTT) of low-activation martensitic steels . Long et al in ref , used a support vector regression model to successfully predict the yield stress of RAFM steel when compared with other similar models, including the neural network, random forest, linear regression, and general regression neural network (GRNN).…”