Numerical determination of Callovo-Oxfordian clay-stone thermal characteristics: TED experiment

Document available in extended abstract form only. In the framework of the research program on nuclear waste repository in deep geological formation, TED in-situ experiment was performed by ANDRA in Meuse/Haute Marne underground research laboratory to determine THM behaviour of Callovo-Oxfordian clay, and more precisely to identify thermal characteristics of the host rock: thermal conductivity and specific heat capacity. clay-stone thermal characteristics are obtained by an inverse method : comparison numerical process between calculated temperature from physical modelling and experiment temperature from in situ sensors. First of all a thermal modelling of TED experiment is performed. Here the main physical process taken into account is transient heat transfer by conduction from hot packer to host rock. A single diffusion partial differential equation is used. This equation is solved by linear Finite Element Method implemented in CAST3M code with unstructured three-dimensional mesh and Crank-Nicolson time discretization. The diffusion tensor is function of two thermal conductivity; one (λl) characterizes thermal diffusion parallel to the bedding plane, the other (λt) perpendicularly. Packers heating is represented by source term r. A first sensitivity analysis completes the implementation of direct method. The inverse method modelling is then applied: - three input thermal parameters are identified: longitudinal (λl), transversal (λt) thermal conductivity, and specific heat capacity (Cp); experimental variables are represented by 108 sensors temperatures in host rock, during more than two years of packers heating; - a meta-model is built with the help of neural network, to replace CAST3M modelling, too time consuming; the methods employed, like LHS experiment plan performance, come from URANIE platform; - a minimization of objective function, depending on experimental and calculated temperatures, is realized to determine thermal parameters; - solution parameters confidence domain is obtained by taking into account experimental temperature uncertainties