An Intelligent Model to Predict the Compactness of Granular Mixtures Used in Conventional (CC) and Roller-Compacted Concrete (RCC)

The granular compactness is the first-order geotechnical parameter. Its optimization is very important from a technical and economic perspective. It depends on several geometrical and physical parameters. It is experimentally measured or predicted using theoretical models that are not always applicable for particular granular mixtures such as roller-compacted concrete (RCC), which contain large diameter aggregates and fillers. This work aims to develop an artificial intelligence model to predict the compactness of granular mixtures used in conventional and roller-compacted concrete. The modeling is carried out through a database resulting from an experimental study conducted in the lab on 140 granular mixtures. The compactness of these mixtures was measured by an experimental system designed and optimized in the lab. Thus, three artificial neural network (ANN) models were developed to predict the compactness as a function of some basic geometrical parameters (Dmax, intermediate diameters Di, curvature Hazen’s coefficients, and the percentage of fillers). After validation, the global prediction model of the compactness showed a good performance with a good correlation (R > 0.91 and low root mean squared error).