This paper compares simple constitutive models that are widely used in engineering practice with more sophisticated methods in the context of a case study. In particular, four constitutive modelling approaches have been considered: a simple elasto-plastic constitutive model (modified Cam-clay), with and without Rayleigh damping; the same model coupled with a cyclic non-linear model that can simulate pre-yield hysteresis; and finally an advanced kinematic hardening model, which is an improved version of the Al-Tabbaa & Wood two-surface model. These four approaches are used to analyse the seismic response of a section of the Bolu tunnels during the 1999 Duzce earthquake. To shed light on the performance of the constitutive models, simple site response finite-element analyses were first undertaken for the studied site, paying particular attention to the calibration of the Rayleigh damping parameters. The results of these analyses, in terms of maximum shear strain, were then used as input to an analytical elastic method (extended Hoeg method) for calculating the thrust and bending moment acting in the tunnel lining. Finally the results of dynamic time domain plane-strain analyses, employing the four adopted constitutive modelling approaches, are compared against field observations and results obtained by the extended Hoeg method, to investigate the ability of the models, of ranging complexity, to mimic soil response under seismic excitation. On effectue enfin une comparaison des résultats des analyses dans le domaine temporel de la déformation sur un plan, faisant usage des quatre méthodes de modélisation constitutives adoptées, avec les résultats obtenus à l'aide de la méth-ode étendue de Hoeg, pour examiner la capacité pour des modèles de complexité variable de simuler la réaction des sols à une excitation sismique.
INTRODUCTIONUnderstanding the behaviour of geotechnical structures such as tunnels, earthfill dams and retaining walls during earthquakes is recognised as a necessary condition for their rigorous and safe seismic design. In order to predict the seismic response of these structures correctly, it is essential to develop and use constitutive models that can simulate soil behaviour appropriately under dynamic loading. Numerical analysis has developed significantly over the last decades, and nowadays a wide range of constitutive models is available. Equivalent linear models and simple non-linear models are used with acceptable accuracy for situations involving small to medium strains, whereas for higher strain levels coupling of a non-linear model with an elasto-plastic model, or the use of more advanced models, is needed. However, not all of the above-mentioned types of models are available in commercial codes. Therefore in engineering practice very simple elasto-plastic models (e.g. Mohr-Coulomb and/or