Durabilité des éléments en béton armé en milieu marin

Abstract

Corrosion of steel reinforcement is one of the main phenomena determining the life of the structure. Methods based on several indicators of the likelihood of corrosion can be followed. Some of these measures are more or less long and require very specific equipment. In recent years, several non-destructive tests have been developed to be relatively fast and less expensive, based on the measurement of corrosion potential. Several test benches have been developed to evaluate the risk of corrosion by nondestructive "CANIN +" tests carried out on ordinary concretes and hardened mortars. The results obtained using these measurements reveal a decrease in the corrosion potential of the specimens in an aggressive medium (containing 3% NaCl) relative to the reference medium. The progression of the tests in the different environments confirms the aggressiveness of the environment on the structures and the interest of the coating. The thesis was divided in two parts. The first one is aimed at a probabilistic study to predict the corrosion initiation time of concrete structures in chloride-containing environments. It is expressed as a mathematical model using the second Fick law and the statistical distribution properties of their parameters are included in this model. Dispersion under environmental exposure conditions and structural properties were considered random fields in the mathematical model with a probabilistic design. This probabilistic study is developed using a Monte Carlo simulation to determine the contribution of each input parameter and the statistical parameters of the random variables on the probability distribution functions of the initial corrosion time. In addition, a comparative study was conducted to analyze the impact of the probability distribution on the response (the initial corrosion time). The second part of the thesis was the neural analysis of the data obtained by the experimental measurement of the corrosion potential. Statistical analysis is performed using multiple linear regression to test the reliability of the data. Artificial neural networks (ANNs) are then used to develop a model for predicting the corrosion potential of reinforcement in a concrete or mortar. The results indicate that the artificial neural network is able to predict the corrosion potential with an acceptable degree of accuracy.