Advanced analysis of defect clusters in nuclear reactors using machine learning techniques
Authors:
Open Access
- 1 July 2025
- journal article
- research article
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-31
Abstract
Studying defects and defect clusters in reactor materials is essential for understanding the degradation mechanisms of materials under irradiation. By uncovering the formation and evolution of defects, this paper provides critical insights for enhancing the radiation resistance of reactor materials and extending their service life. The contributions of this study are as follows: (1) Integration of a large-scale molecular dynamics (MD) dataset generated by cascade collisions and the application of machine learning (ML) techniques to investigate point defects and their clusters in reactor pressure vessel (RPV) materials; (2) Proposal of a novel clustering method based on physical characteristics, enabling efficient classification of defect clusters while effectively reducing data noise; (3) Development of a component-based defect cluster configuration recognition method using a dual-pointer lattice-filling technique, accurately capturing all defects and identifying several cluster morphologies observed in experiments; (4) Demonstration of the algorithm’s outstanding performance in handling systems containing millions of atomic coordinates, showcasing its scalability and robustness; (5) Visualization of the three-dimensional spatial distribution of defect clusters and provision of two-dimensional spatial density distribution maps for vacancy clusters and interstitial clusters, offering precise characterization of spatial relationships within defect clusters and new insights into the irradiation mechanisms of materials.Keywords
Funding Information
- National Key Research and Development Program of China (2023YFB3003003)
This publication has 52 references indexed in Scilit:
- From process to property: multi-physics modeling of dislocation dynamics and microscale damage in metal additive manufacturingComputational Mechanics, 2024
- Estimating the Optimal Number of Clusters in Categorical Data Clustering by Silhouette Coefficient2019
- Rearrangement of interstitial defects in alpha-Fe under extreme conditionActa Materialia, 2019
- Performance Prediction in Nuclear Materials by Using a Collaborative Framework of Supercomputing, Big Data and Artificial Intelligence2019
- Improved shape matching and retrieval using robust histograms of spatially distributed points and angular radial transformOptik, 2017
- Kernel optimization for short-range molecular dynamicsComputer Physics Communications, 2017
- Multiscale simulations on the coarsening of Cu-rich precipitates in α-Fe using kinetic Monte Carlo, molecular dynamics and phase-field simulationsActa Materialia, 2012
- Compressed Histogram of Gradients: A Low-Bitrate DescriptorInternational Journal of Computer Vision, 2011
- Radiation damage cascades: Liquid droplet treatment of subcascade interactionsJournal of Nuclear Materials, 1991
- A proposed method of calculating displacement dose ratesNuclear Engineering and Design, 1975