Guido Mazzà *, Antonella Frigerio* Laboratoire Central des Points et Chaussées (LCPC) – RILEM TC-ACS-M Parigi, 17-18 Novembre 2008 PRESENTAZIONE POWER POINT * CESI RICERCA The AAR problem in concrete concerns different actors, such as structures builder, structures owners, concrete provider, chemists who consider concrete as a background for complex chemo-physic processes and civil engineers dealing with structural problems due to concrete expansion. All of them may be interested in using numerical models to study this pathology. But are they thinking to the same tool when they speak about “model”? The numerical models developed in the last few decades study the AAR phenomena from different points of view, but they can be gathered in two main groups. The first one consists in modelling the chemical reaction itself; the second one focus on modelling the overall behaviour of concrete affected by AAR. Since the global AAR phenomenon in concrete is a multi-physics problem, involving chemistry, thermodynamics, diffusion in porous media and mechanics in strongly heterogeneous material, most of the numerical models devoted to this topic include both chemical and mechanical aspects, hence the term “chemo-mechanical” used to qualify them. The scope of the RILEM Technical Committee ACS-M is to outline a guide to classify and present different AAR-models in order to point out their potentiality or their drawbacks. As the coexistence of several models relating to different aspects of AAR could be a source of confusion, the first aim of the present guide is to help potential users to find out which numerical model is more suitable to their particular purposes. Obviously, the criteria that will be developed and outlined in the guide to discriminate different models shall not be considered as strict borders between incompatible objects, but rather as a few milestones aimed at clarifying a complex landscape. The second motivation in writing this guide is to propose (or recall) some advices in the use of different models. Numerical models are powerful, interesting and cost effective tools, but they present two major dangers. First, whatever the quality and the complexity of a model is, it can never attain good results if it has been fed with poor data. Secondly, a numerical tool give always results, often in a shiny, good-looking form of colourful 3D-graphics, but these results must be compared and validated with experimental data to avoid wrong conclusions. The use of numerical results and their comparison to reality may be trickier than using numerical models, especially in the case of a complex multi-physic phenomenon such as AAR. Last, this guide would like to be considered as a state-of-the-art of existing tools in the field of AARmodelling. Of course, this is a vivid research field, and new models are constantly under development, but it can be useful to present, in a unique document, the different models existing at a given date. It may help potential users in the choice of the mostly appropriate tools and, hopefully, suggest some new researches and developments.