This work presents the development of the Proper Generalized Decomposition (PGD) method for solving couple transient multiphysics problems with different characteristic times. This method consists in approximating solutions ( Partial Differentiai Equations with separated representations. The 2D transient heat equation is initially considered. A automatic adaptive mesh technique is proposed in order to make the discretization fit the different transient domains. Tw different couplings between the PGD method and the adaptive mesh refinement technique are discussed: the frrst on consists in computing the PGD solution for each new mesh from the null solution; the second one consists in enrichin the PGD solution for each new mesh from the basis functions generated on the previous meshes. The frrst coupling . more efficient since fewer modes are required to accurately describe the solution on the final mesh. Nevertheless, th second one decreases the number of enrichments cumulated tbrough the mesh refmement pro cess. Regardless of th coupling used, the adaptive mesh technique is able to automatically describe the localized transient zones. The II transient heat equation with a non linear source term is also studied. A new approach combining the PGD method and th Asymptotic Numerical Method (ANM) is tested, which allows to efficiently solve sorne families of non linear transiel problems. Finally, two muItitime and multiphysics problems are considered. It consists of a partially coupled he diffusion problem and a strongly coupled thermoviscoelastic problem. The PGD method gives an accurate prediction c the response of these muItiphysics problems for which the coupling terms lead to specific transient zones. Combined wit the PGD method, the adaptive mesh technique is particularly suitable for these situations of strongly coupled tim multiscale. This combination brings to the same conclusions as in the case of a single physical phenomenon. Th discussion focuses on two strategies of mesh construction: concatenating the time meshes of each physical phenomeno or refme each mesh independently. The concatenation of two meshes allows a convergence with fewer steps of mes refmement but with a much bigher mesh density.
