Project Summary
The objective of this activity is to develop a methodology and toolsets for integrating complex, coupled models (such as THMC or THC) into the GDSA model for evaluating disposal system performance for nuclear waste. Because the GDSA model is designed for the entire repository with a thousand emplacement tunnels, a systematic methodology is needed on how to simplify some coupled processes/parameters. In this search activity, we are seeking a reduced-order model (ROM) that allows for integrating coupled processes into the GDSA model. The uncertainty quantification and sensitivity analysis
Research highlight: Through a THC model, we simulated time-varying Kd values representing the buffer degradation. We performed global sensitivity analysis to find that key parameters evolve over time, although the most dominant parameter (the adsorption sites density on smectite) stayed the same. We then constructed ROMs based on neural networks and Random Forest to represent Kd as a function of key parameters.
Selected publications:
Ermacova, D. et al. Global Sensitivity Analysis for U(VI) Transport with coupled Thermal, Hydrological, and Chemical (THC) Modeling, AGU 2020
Lu, H. et al., Data-driven Reduced Order Modeling for ReactiveTransport in Nuclear Waste Repository Assessments, AGU 2020
Contact: Haruko Wainwright