Advanced analysis of defect clusters in nuclear reactors using machine learning techniques.

Area of Science:

• Materials Science
• Nuclear Engineering
• Computational Physics

Background:

• Understanding material degradation under irradiation is crucial for nuclear reactor safety and longevity.
• Defect formation and evolution significantly impact the performance of reactor materials.

Purpose of the Study:

• To investigate point defects and their clusters in reactor pressure vessel materials using large-scale molecular dynamics data and machine learning.
• To develop novel methods for efficient and accurate classification and characterization of defect clusters.
• To provide insights into irradiation mechanisms and enhance the radiation resistance of reactor materials.

Main Methods:

• Integration of large-scale molecular dynamics (MD) datasets from cascade collisions.
• Application of machine learning (ML) techniques for defect analysis.
• Development of a novel physical characteristic-based clustering method for defect classification and noise reduction.
• Implementation of a component-based defect cluster configuration recognition method using a dual-pointer lattice-filling technique.
• Scalable algorithm demonstrated on systems with millions of atomic coordinates.

Main Results:

• Accurate classification of defect clusters and identification of experimental cluster morphologies.
• Demonstrated scalability and robustness of the developed algorithm for large atomic systems.
• Visualization of 3D spatial distribution and 2D spatial density maps of vacancy and interstitial clusters.
• Precise characterization of spatial relationships within defect clusters.

Conclusions:

• The study provides critical insights into material degradation mechanisms under irradiation.
• The developed methods enhance the understanding and characterization of defects in reactor materials.
• This work contributes to improving the radiation resistance and extending the service life of nuclear reactor components.