Abstract:
Today, hydraulic materials (concrete) are the most used in the world for civil and industrial
constructions, without being able to control or predict as always their specific behavior
according to various conditions.
During the introduction of hydraulic materials, they are subjected to physical and chemical
changes, the variation depends on the cement content, the water content, the quality of the
aggregates, cement quality and other influences .
This study to show the application of a nonparametric approach called ""Artificial Neural
Networks"" aims. to forecast the evolution of dimensional changes of chemical and
mineralogical composition of clinker strength, expansion and consistency.
The use of this approach allows the development of models for their prediction using a
multilayer network backpropagation. They are also based on a large database of experimental
results collected in the literature and on an appropriate choice of architectures and processes
of training utilized.
