Probabilistic Modelling of Compressive Strength of Concrete Using Response Surface Methodology and Neural Networks

In this paper, we aim to achieve a probabilistic modelling of the compressive strength of concrete using three response surface models (RSM) and the artificial neural network (ANN) method. The input random variables for the three RSM and for the ANN are cement content, water content, measure of slump and air content, while the output for all the models is the compressive strength of concrete at 28 days. More than 800 cylindrical specimens 16×\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times$$\end{document}32 cm were tested. The experimental data are used to check the reliability of the suggested probabilistic models and their prediction capability. It is shown that the use of these new RSM is as simple as that of any of the basic formulas, yet they provide an improved tool for the prediction of concrete strength and for concrete proportioning. It is also shown that the concrete compressive strength can be readily and accurately estimated from the established ANN.