Implementation of Artificial Neural Networks for Prediction of Chloride Penetration in Concrete

Abstract

Artificial Neural Networks (ANN) has received a great attention from researchers in previous decade to predict different aspect of engineering problems. The aim of this research is to present an implementation of ANN to predict the Chloride penetration of self-consolidating concrete (SCC), containing various amounts of cement replacement minerals including fly ash, silica fume, and slag. The ability of concrete to resist chloride penetration is measured using Rapid Chloride Penetration (RCP) test through an experimental program. One- and two-layer ANN models were developed by controlling the critical parameters affecting chloride penetration to predict the results of RCP test. The ANN models were developed using various parameters including ratio of water-to-binder (w/b), course aggregate, fine aggregate, fly ash, and silica fume. It was shown that the prediction accuracy of ANN models was sensitive to combinations of learning rate and momentum. Data used to train and test the ANN were obtained through an experimental program conducted by the authors.