Prediction of cement paste mechanical behaviour from chemical composition using genetic algorithms and artificial neural networks

Computational intelligence (CI) techniques have attracted the interest of some engineers as valid tools for the representation of complex systems. A growing number of works are showing that they are also effective in optimisation. Building materials, such as concrete and mortar, usually display a complex behaviour hard to model and seem to be an interesting area to explore the application of CI as a modelling technique. This paper describes how CI can be used to model the performance of cement paste. The specific objective was to develop models able to predict the mechanical behaviour of this material using only data available from chemical composition of cement. The developed models showed the advantage of CI with respect to conventional techniques leading rapidly to useful results with reasonable precision and accuracy.