ESTIMATION OF COMPRESSIVE STRENGTH OFCONCRETE USING ARTIFICIAL NEURALNETWORK (ANN)

Abstract

The standards within the construction industry require formulation ofa strength classes’ concrete. In this paper, we study the qualityinfluence of concrete on the compressive strength. More than 1600cylindrical specimens 16x32 cm, using local materials, were tested inour laboratory. The mixtures are obtained using the Dreux-Gorissemethod and the cure of the specimens is done both out in the openand immersed in water. We show that the components’ intrinsicproperties of the studied concrete and particularly the broken upparticles, offer to the concrete complete satisfactory resistances. Wealso show that the cement proportioning for the selected class doesnot offer notable differences as regards to compressive strengths. Theresults we obtained are collected as a data bank. We estimate thecompressive strength of concrete using the multi layer perceptronartificial neural network method (ANN). The inputs are the slump,the content air, dosage of water and dosage of cement, while theoutput is the compressive strength of concrete at 28 days. We foundthat the concrete compressive strengths at 28 days could be readilyand accurately estimated from the established neural network.