Application of inverse modeling methods to thermal and diffusion experiments at Mont Terri Rock laboratory

The Opalinus clay parameter identification at Mont Terri Underground Rock Laboratory (URL) is performed with inverse modeling. This article focuses on a comparison between the global and local inverse approaches, in a deterministic framework, applied to the thermal (HE-C) and diffusion (DI) experiments respectively. Each experiment presents a similar diffusion process described by the same equation. This simple forward model makes easier the comparison of each inverse approach and allows to easily apply different methods such as the neural networks and the adjoint state method. The synthesis yields to valuable information about their merits and flaws, and highlights the importance of the parametrization. The singular value decomposition of the Jacobian matrix is presented to identify the best parametrization. In each experiment, the numerically found parameter values are in good agreement with the experimental methods. In DI experiment a spatial variability description of the medium is a hypothesis to explain the rapid decrease of the tracer in the injection chamber. (authors)