Abstract : To take into account the behavior of soils subjected to very large numbers of cycles requires the use of specific models, often complex. Furthermore, the simulation of this behavior throughout the entire loading history is time-consuming. In order to overcome these difficulties, we developed two numerical techniques. The first technique aims at reducing substantially the computational time by applying the asymptotic time homogenization method. The effectiveness of this method is measured by comparing simulations with and without homogenization in the case of undrained triaxial tests under one-way cyclic loading. Two models are used: one is based on the principle of the bounding surface plasticity; the other is a two-surface model with isotropic and kinematic hardening (bubble model). The time homogenization method is implemented in the finite element software CESAR-LCPC. The second technique is part of the broader framework of constitutive parameter identification by inverse analysis. It consists of a multi-objective identification of parameters by genetic algorithms. This method is tested on monotonic pressuremeter tests in order to predict the settlement of a shallow foundation. Finally, the two numerical techniques are applied to experimental results obtained from undrained triaxial tests under one-way cyclic loading on a normally consolidated clay.