Institut Français des Sciences et Technologies des Transports, de l'Aménagement et des Réseaux

Type de document

COMMUNICATION AVEC ACTES INTERNATIONAL (ACTI)

Langue

anglais

Auteur

PRADELLE ; THIERY ; BAROGHEL BOUNY

Résumé / Abstract

Numerous models can be used to predict chloride ingress into immersed concretes. A comparative study is necessary to point out the efficiency of these models to predict total and free chloride profiles. Moreover, testing costs become prohibitive and it is necessary to limit models input data which have to be obtained experimentally. The objective of this paper is to contribute to the development of a more reliable engineering tool for predicting chloride ingress into concrete. This paper presents a review of four models: one mono-ionic and three multi-species models. Four different chloride binding isotherms (CBIs) are tested to describe ionic interactions with the cementitious matrix: linear, Freundlich, Langmuir and ion-exchange isotherms. Input data of these four models are limited to a minimum number: accessible-to-water porosity (φ), effective chloride diffusion coefficient (DCl− ) and coefficients of CBI (1 or 2 parameters). φ and DCl− are determined experimentally. The coefficients of CBI are fitted by a numerical inverse analysis performed on one experimental total chloride content (tcc) profile at a given time of exposure. This approach has been performed for five concretes in laboratory (three concretes) and in-situ (two concretes) conditions . The accuracy of this fitting is determined using two statistical criteria: the adjusted coefficient of determination (Radj ) and the normal mean error (NME). Then, profiles after other exposure times can be predicted. Comparisons between numerical predictions and available experimental data corresponding to several exposure times are led. The results of this study allow to conclude about the most relevant chloride ingress model. The authors recommend the use of a multi-species model coupled with a Langmuir or Freundlich isotherm.