Concrete strength prediction by means of neural network

Abstract

In this article a model is developed, based on neurocomputing, for predicting, with sufficient approximation, the compressive strength of cement conglomerates. First, the principles of connectionism are briefly recalled. Some neural networks are then constructed in which the different mix-design parameters of a variety of cement conglomerates i.e. the compressive strength, are associated. The experimental data obtained during construction of the ‘Alto Sulcis Thermal Power Station’ at Portovesme, Italy, were used in the tests. The availability of a substantial amount of data enabled the method to be suitably calibrated and satisfactory results were obtained for evaluating the mechanical properties of different concrete mixes.