Predicting and interpreting the irradiation-induced tensile behavior of fast reactor cladding using explainable machine learning

HT9 and Gr.92 (Mod) steels are promising candidates for advanced nuclear fuel cladding. In this study, 1,590 tensile test data points of 9–12Cr ferritic/martensitic (FM) steels were collected to develop machine learning (ML) models for predicting irradiation-induced changes in the tensile properties of HT9 and Gr.92 (Mod). Regression models were evaluated after Optuna-based hyperparameter optimization using test-set evaluation and k -fold cross-validation (CV). Extreme gradient boosting (XGB) exhibited the best predictive performance; however, an artificial neural network (ANN) was selected for its continuous response, physical consistency, and stable representation of irradiation-dependent trends. Using independent experimental data, the ANN yielded mean absolute percentage errors (MAPEs) of 6.26 ± 0.47% for ultimate tensile strength (UTS) and 10.69 ± 0.46% for yield strength (YS). The ANN model predicted the YS across irradiation conditions and showed that the irradiation-induced increase in YS for HT9 and Gr.92 (Mod) steels saturated at 5–10 displacements per atom (dpa). SHapley Additive exPlanations (SHAP), an explainable artificial intelligence (XAI) method, quantified the contributions of dpa and irradiation temperature, revealing that irradiation temperature has a greater overall influence on the predicted YS. These results provide both reliable property predictions and physical insights into the effects of irradiation conditions.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-19
DOI
https://doi.org/10.1038/s41598-026-81129-1
Primary Topic
Fusion materials and technologies
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Predicting and interpreting the irradiation-induced tensile behavior of fast reactor cladding using explainable machine learning

Jeong Mok Oh, Sunghwan Yeo, Hongin Kim, Junhwan Kim et al.
Scientific Reports
Fusion materials and technologies
article

Predicting and interpreting the irradiation-induced tensile behavior of fast reactor cladding using explainable machine learning

Jeong Mok Oh, Sunghwan Yeo, Hongin Kim, Junhwan Kim, Dongha Kim, Young-Kook Lee, Wooksang Jeon
article en

Abstract

HT9 and Gr.92 (Mod) steels are promising candidates for advanced nuclear fuel cladding. In this study, 1,590 tensile test data points of 9–12Cr ferritic/martensitic (FM) steels were collected to develop machine learning (ML) models for predicting irradiation-induced changes in the tensile properties of HT9 and Gr.92 (Mod). Regression models were evaluated after Optuna-based hyperparameter optimization using test-set evaluation and k -fold cross-validation (CV). Extreme gradient boosting (XGB) exhibited the best predictive performance; however, an artificial neural network (ANN) was selected for its continuous response, physical consistency, and stable representation of irradiation-dependent trends. Using independent experimental data, the ANN yielded mean absolute percentage errors (MAPEs) of 6.26 ± 0.47% for ultimate tensile strength (UTS) and 10.69 ± 0.46% for yield strength (YS). The ANN model predicted the YS across irradiation conditions and showed that the irradiation-induced increase in YS for HT9 and Gr.92 (Mod) steels saturated at 5–10 displacements per atom (dpa). SHapley Additive exPlanations (SHAP), an explainable artificial intelligence (XAI) method, quantified the contributions of dpa and irradiation temperature, revealing that irradiation temperature has a greater overall influence on the predicted YS. These results provide both reliable property predictions and physical insights into the effects of irradiation conditions.

Scientific Reports
Korea Atomic Energy Research Institute (KR), Yonsei University (KR)
Fusion materials and technologies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon