Application des Métaheuristiques dans l'optimisation de la formulation de béton compacté au rouleau (BCR) de barrage réservoir.

Abstract

The formulation of roller compacted concrete (RCC) is a fundamental operation in the design of a dam at the best technical-economic conditions. Although the current empirical or semiempirical formulation methods are used with success, the development of a more efficient approach is of major interest. The Artificial Neural Networks (ANN) and Particle Swarm. Optimization (PSO) have been proposed as optimization metaheuristics. The research here presented aims using artificial intelligence techniques to optimize the RCC dam’s formulation. The first part of study concerned the granular mixture of RCC’s compactness optimization. An experimental program was carried in the laboratory on 174 granular mixtures, with different maximum sizes and aggregates’ proportions. The second part deals with the prediction of the compressive strength "Rc", the water/cement ratio "E/C" and the cement rate "Dc". For this study a 500 vectors database of RCC formulations given by laboratory activity reports of 04 dam projects. After the database normalized, the most important parameters of formulation were selected for the development of five (05) prediction models. The compressive strength is predicted as a function of a maximum size of the aggregates (Dmax), cement rate (DC), limestone filler rate (Df), compactness (C), and water/cement ratio (W/C). For the prediction of E/C ratio and Dc rate, two modelling approaches were considered resulting in 04 model outputs, as a function of Dmax, Df, C, Rc and VeBe time. The models were experimentally validated in the laboratory and compared with the Laboratoire Central des Ponts et Chaussees (LCPC) method. Even with few differences in materials and concrete placement, the developed ANN and/or ANN-PSO systems provided very good predictions. Finally, a global methodology for RCC formulation was developed, based on the designed prediction models and a graphical interface created to simplify its use and allows to better explore the proposed methodology. a practical exploitation of the methodology.