SBIR-STTR Award

When nuclear reactor performance drops below desired energy levels, the radioactive waste must be properly disposed. Until deep geological disposal sites are identified for very long-term storage of this waste (1000?s of years), interim storage methods must be used for much longer than initially intended (~100 years). With hundreds of dry storage canisters approaching or already exceeding their intended design life, the Department of Energy needs an aging management program to assess the long-term integrity of these canisters. The primary objective of this work is to develop a web-based machine learning software tool for optimizing life-cycle decisions and ensuring long-term performance of welded stainless-steel dry storage canisters for the storage of spent nuclear fuel. Incorporating this tool into daily operations will promote pro-active decision-making and improved risk management to minimize the likelihood of a potentially catastrophic, failure event (loss of the canister?s containment boundary). To ground the outcome of Phase I, and demonstrate the feasibility of the approach, working prototypes for each component of proposed tool will be developed, including: (i) the artificial intelligence Bayesian decision network engine, (ii) the chloride-induced stress corrosion cracking environ-mechanical model with supporting experimental data, and (iii) the web-based delivery interface. The outcome will be a fully functional life-cycle management tool that can make more informed decisions and guide operations without disregarding the need for simplicity and computational efficiency. Further refinement of the computational engine through subsequent phases of this project will further improve the predictive capability of proposed tool. The computational framework developed here, is directly applicable to other failure modes, across all energy industry sectors, making the commercialization of this technology extremely feasible. Not only is there an economic benefit to making better decisions that lower the risk of failure, the public benefits as well.

When nuclear reactor performance drops below desired energy levels, the radioactive waste must be properly disposed. Until deep geological disposal sites are identified for very long-term storage of this waste (1000’s of years), interim storage methods must be used for much longer than initially intended (~100 years). With 1000’s of dry storage canisters presently loaded with spent nuclear fuel and more loaded each year, reactor sites and canister vendors must properly address aging effects in their aging management programs to ensure canister longevity. The primary objective of this work is to develop and deliver a web-based machine learning software tool that addresses aging effects for optimizing life-cycle decisions and ensuring long-term performance of welded stainless-steel dry storage canisters for the storage of spent nuclear fuel. Incorporating this tool into daily operations will promote pro-active decision-making and improved risk management to minimize the likelihood of a potentially catastrophic, failure event (i.e. loss of the canister’s containment boundary). The outcome of Phase I was a fully-functioning prototype software tool that successfully demonstrated the feasibility of the approach, including: (i) a web-based user-interface and online database to satisfy the workflow requirements of the end-users, (ii) the artificial intelligence Bayesian decision network engine, and (ii) the chloride-induced stress corrosion cracking model with supporting experimental data. Even though the Phase I prototype software is all encompassing and predictive, assumptions were made in populating the engine, and configuring the user-interface. During Phase II efforts, these assumptions will be relaxed to improve predictability and reduce uncertainty. Efforts will also focus on further anchoring the engine into the fundamentals of the damage mechanisms, as well as refinement, verification, validation, and commercialization of both the engine and the web-based user-interface and online database, to ensure market-readiness.