Uncertainty quantification is an essential part of model-based evaluation processes such as the long-term safety assessment of radioactive waste repositories. Viewing simulation results obtained during such safety investigations in a probabilistic way as a consequence of the existing data, parameter, model and scenario uncertainties is expected to lead to more meaningful quantitative expressions of integrity. The aim of such an approach is to avoid common pitfalls of classical analyses which may yield non-conservative results or only insufficiently cover the actual parameter space.
The present subproject is intended to develop and test methods for probabilistically sound (geological) barrier-integrity assessment that are based on modern mathematical methods of uncertainty quantification. Mathematical modeling – tailored to the used THM models and measured parameter distributions – from the fields of probability calculation, statistics, numerical mathematics and data sciences enables an innovative, largely automated approach to quantify the propagation of uncertainties and the problem-specific engineering characteristics.
The project goals relate to the integration of workflows, methods and instruments for experimental-numerical analysis of parameter uncertainties with respect to the safety analysis of potential repository sites in a numerical environment. To realize and demonstrate the project goals, the following workflow is proposed using the above mentioned mathematical framework:
1. Initial assessment of parameter variability with little site-specific information
|
|
2. Improved characterization with site-specific information at tunnel scale
|
![]() |
3. Application at repository scale
|
The main outcome of this project is to propose methods and demonstrate their feasibility for propagating parameter uncertainty through THM model-based integrity analysis with a particular focus on questions related to how scale affects the parametrization; how the integrity criteria can be transferred to a probabilistic context; and finally, how efficient numerical mathematics can be used in all stages of this workflow.
Additionally, the planned illustration of simulation results with uncertainties in the context of virtual realities represents a significant asset for an improved communication of the difficult topic of “uncertainties in safety assessment”.