Hydro-mechanical modelling of underground excavations behaviour with an elastoviscoplastic model

The long term behaviour of underground excavations is a social and economic stake, in particular in the context of storage in deep geological formation of high activity and long life nuclear waste. Several experimental galleries have been dug in the underground research laboratory (URL) of Meuse/Haute-Marne located close to Bure in France, where studies are leaded in order to understand the global behaviour of the constitutive rock which is the Callovo-Oxfordian (COx) argillite. The purpose of this PhD Thesis is to establish a modelling with Code-Aster of underground excavations, and especially of a Meuse/Haute-Marne laboratory gallery, taking into account non local approach, creep effect and hydromechanical coupling in the framework of the mechanics of porous media, and then to compare numerical results with available experimental data. The specific elastoviscoplastic model used in this study is the L and K model: it offers a coupling between instantaneous and delayed behaviour, and it takes into account the dilation, parameter which governs the volume strains of the material during a solicitation, and its strong variation, a specificity of geomaterials and so of COx argillite. The fluid flowing through the material adds a hydraulic component to the modelling, which is coupled to mechanic component thanks to Biot's equations. Another novelty of this work concerns the coupling between such complex rheological behaviour and a non-local approach in an industrial way. Among methods of regularization available in Code-Aster, the second gradient of dilation is well fitted to geomaterials. Its aim is to correct mesh dependency and numerical localized solutions. After describing numeric tools and setting parameters of the L and K model on laboratory tests, a good general agreement was found between numeric results and in situ measures, without resetting parameters. Time effects experimentally measured on displacement and pore pressure evolution are observed in the same modelling, validating the followed predictive approach. (author)