The containment building is the third safety barrier in nuclear power plants. Its role is to protect the environment in the event of a hypothetical accident by limiting releases to the environment. Its function is therefore closely linked to its tightness, which it must maintain throughout its operating life. Traditionally, the estimation of the leakage rate is based on a good knowledge of the hydric state and potential mechanical disorders, associated with transfer laws (such as permeability) in a chained (thermo-)hygro-mechanical simulation approach. While the mechanical behaviour of the structure is now generally well known, using advanced simulation tools, progress is still needed to improve the understanding and quantification of flows. This is particularly the case in the presence of heterogeneities (cracks, honeycombs, reinforcement, cables, etc.), all of which can locally disrupt permeability. This is the context of the proposed thesis topic. The aim is to improve the understanding and representation of flows through a reinforced concrete structure using an approach that combines experimental tests and modelling. An initial analysis will be used to define an optimised experimental design based on several configurations (leak paths, type of flow, temperature, saturation, etc.), which will then be implemented during the thesis. The results will be analysed in order to characterise empirically the influence of the leakage path on the macroscopic laws classically used (Darcy's law). A more refined simulation approach will then be developed, based on the finite element method. The aim will be to reproduce the experimental results and extend them to the behaviour of containment vessels, thereby improving the modelling tools currently available.