Prediction of Compressive Strength of Concrete using Artificial Neural Network

Concrete cube strength determination tests are usually performed at three days to one year afterpouring the concrete. The waiting period required to perform such test may delay the construction progress,decision making and neglecting such test would limit the quality control checks in large constructionprojects. Therefore it becomes necessary that the rapid and reliable prediction of concrete strength isessential for pre-design or quality control of construction. It is possible to facilitate the modification of themix proportion if the concrete does not meet the required design stage, which may save time andconstruction costs. The early prediction of concrete strength is essential for estimating the desirable time forconcrete form removal, project scheduling, quality control and estimating delay if any. Artificial NeuralNetwork (ANN) is used to predict the compressive strength of concrete. Standard back propagation andJordan–Elman algorithms are used to train the networks. Networks are trained and tested at various learningrate and momentum factor and after many trials these were kept constant for this study. Performance ofnetworks were checked with statistical error criteria of correlation coefficient, root mean squared error andmean absolute error. It is observed that artificial neural networks can predict compressive strength ofconcrete with 91 to 98 % accuracy.