Progressive Solutions to Improve Corrosion Resistance of Nuclear Waste Storage Materials: Atomistic Simulations, Machine Learning and Artificial Intelligence for Glass Corrosion, Glass Design and Canisters Lifetime
Sponsored by: TMS Corrosion and Environmental Effects Committee, TMS: Nuclear Materials Committee
Program Organizers: Madeleine Jordache, Stevens Institute of Technology; Gary Pickrell, Virginia Tech; Bai Cui, University of Nebraska Lincoln
Wednesday 9:45 AM
October 12, 2022
Room: 333
Location: David L. Lawrence Convention Center
Session Chair: Madeleine Jordache, Stevens Institute of Technology; Gary Pickrell, Virginia Tech; Bai Cui, University of Nebraska -Lincoln
9:45 AM Introductory Comments
9:50 AM
Environmental Cracking Lifetime Prediction through the Development of Pitting and SCC Models for Nuclear Waste Storage Casks: Sarah Blust1; James Burns1; 1University of Virginia
Used nuclear fuel (UNF) is currently stored across the US in passively cooled stainless steel dry storage canisters (DSC). Due to the design of the DSC, aerosols from the outside environment are able to deposit on the stainless-steel canisters. Over time the deposited aerosols will deliquesce on canisters to form concentrated salt brines resulting in localized corrosion, which when coupled with the high residual stress around welds can lead to stress corrosion cracking (SCC). The objective of this study is to create a model for the life-management of DSC, which will inform a framework to quantify and manage a risk-based ranking of storage sites. The specific goals of this work are to: (a) validate the maximum pit size model for DSC-relevant corrosion conditions (b) coupling the limiting pit size/Kondo approaches, (c) generate da/dt vs. K data and perform probabilistic FM predictions of SCC growth, and (d) validate the model predictions.
10:10 AM
Diminished Diffusion in the Aged Hydrated Gels of Irradiated Borosilicate Glasses: Amreen Jan1; N M Anoop Krishnan1; 1Indian Institute of Technology Delhi
Under aqueous conditions, borosilicate glass, which is used as a matrix to store nuclear waste, forms a hydrated gel layer which plays a significant role in the durability of the glass matrix. However, the properties of this gel layer remain poorly understood. Here, we study the effect of irradiation on the structure and transport properties of the hydrated gels in a series of borosilicate glasses. By analyzing the gel structure of pristine and irradiated glasses, we observe that, indeed, there is a difference in the structure (connectivity and short- and medium-range order) of aged gels formed from pristine and irradiated glass. Particularly, aging has a pronounced effect on the irradiated gels as compared to gels obtained from pristine glasses. Furthermore, we observe that irradiated gels exhibit larger ring sizes in comparison to pristine gels. In contrast to the conventional hypothesis, we observe that diffusion of H decreases in gels from irradiated glasses in comparison to their pristine counterparts. We suggest that the topology of the ring plays a major role, in addition to their size, in governing the dynamics.
10:30 AM Cancelled
Characterization of Hydrated Aluminosilicate Gel from Glass Corrosion: Reaction Mechanism, Structure and Properties from Reactive Molecular Dynamics: Jincheng Du1; 1University of North Texas
The long-term chemical durability of nuclear waste glasses is critical for the geological storage of medium and high level nuclear waste materials. The gel layer formed as a result of corrosion of nuclear waste glasses was found to play a critical role on their long-term chemical durability. In this talk, I will present results on the gel layer structure, stability and mechanical properties by using reactive potential based molecular dynamics simulations. The overall water-silicate gel structure was found to be qualitatively similar to that of real gel layer observed by Atom Probe Tomography. The simulated glass was then characterized to understand the atomistic and nano-structures, diffusive transport of water and dissolved species, as well as the vibrational spectrum and mechanical properties. The results from the reactive MD, combined with experimental results were shown to provide a good understanding of the atomistic phenomena in the interfacial layer.
11:00 AM Invited
Predicting the Long-term Durability of Nuclear Waste Immobilization Glasses using Machine Learning: Mathieu Bauchy1; 1University of California, Los Angeles
The long-term durability of glasses is a key performance metric for nuclear waste immobilization applications. Although some models have been proposed to predict the short-term forward dissolution kinetics of glasses, long-term dissolution is a more complex behavior that is influenced by the glass structure, the feedback from the solution, and the precipitation of secondary phases. This complexity has thus far limited our ability to robustly predict the long-term dissolution rate of nuclear waste immobilization glasses. Here, based on the analysis of a large dataset of vapor hydration tests (VHT), we develop using machine learning a Gaussian Process Regression (GPR) model to predict the long-term durability of glasses. Importantly, GPR models are non-parametric and intrinsically capture the uncertainty of the prediction. We demonstrate that our GPR model features excellent accuracy. This model allows us to decipher the propensity for each oxide to accelerate or slow down the dissolution kinetics of glasses.
11:30 AM Invited
Designing Glasses for Nuclear Waste Immobilization with AI and ML: N M Anoop Krishnan1; 1Indian Institute of Technology Delhi
Despite the use of glasses for more than 2000 years, the composition-structure-property relationships in these materials remain elusive. Here, we discuss how the improved glasses can be designed for nuclear waste immobilization. Specifically, we focus on three aspects namely, (i) ML-based property prediction, (ii) physics-informed ML for viscosity prediction, and (iii) information extraction from the literature. First, we discuss how ML models can be used for predicting electrical, and mechanical properties. Second, we show the use of physics-informed ML for predicting viscosity, which combines the MYEGA equation with ML to predict the viscosity of unknown glasses. Further, the ML models are used as surrogates along with constrained optimization to identify new potential glass compositions. Finally, we discuss how natural language processing can be used to extract information from the literature regarding nuclear waste immobilization, which can be effectively used for accelerating the discovery of new glasses.