A Probabilistic Corrosion Model for Copper-Coated Used Nuclear Fuel Containers

Abstract

Lifetime predictions of used nuclear fuel containers destined for permanent storage in Deep Geological Repositories (DGRs) are challenged by the uncertainty surrounding the environment and the performance of both containers and engineered barriers over repos- itory timescales. Much of the work to characterise the response of engineered barriers to postulated evolving environmental conditions and degradation mechanisms is limited to very short-term laboratory tests or at best in-situ large-scale experiments spanning less than a few decades. While much is learned from these test programmes, the fact remains that long-term performance of many tens of thousands of Used Fuel Containers (UFCs) across a timescale of 100,000 years or more cannot be estimated with a significant degree of confidence by extrapolating single point results of short-term experiments. This is par- ticularly true when there is a desire to understand the progression of container failures and the timing of contaminants subsequently released into the geosphere. Used Fuel Container (UFC) lifetime predictions require a probabilistic approach to address uncertainty. Accord- ingly, this thesis addresses three objectives. The first is to develop a probabilistic model to estimate the time to penetrate through the copper coating of a UFC, assuming sulphide- induced corrosion is the primary degradation mechanism of concern. Within this model, also develop a framework to account for the design of the Engineered Barrier System (EBS) and proposed repository layout. The second is to enhance the probabilistic corrosion model by integrating the potential effects of latent copper coating defects and the single temper- ature transient predicted for the repository. The third is to develop a stochastic process model for pitting corrosion, integrate the same into the sulphide-induced corrosion model, and estimate the time to penetrate through the copper coating based on both degradation mechanisms. To satisfy the first two objectives, this work presents a unique Monte Carlo probabilistic framework. With respect to the third objective, modelling pitting corrosion in copper under postulated repository environments poses a significant challenge since there is no relevant data and the likelihood of this mechanism remains a much debated topic. To overcome this challenge and facilitate demonstration of the approach to modelling pit growth, surrogate data is utilised. In addition to detailing various options for modelling pit growth, this work presents a novel and more transparent, self-contained approach to the estimation of the underlying process intensity when pit growth is modelled via a non- homogeneous Markov process. Finally, the combined effect of pitting and sulphide-induced corrosion on UFC copper-coating lifetimes is demonstrated. The modelling results are for the purpose of illustrating a potential methodology only.