Prediction of compressive strength of concrete by neural networks

In this paper, a method to predict 28-day compressive strength of concrete by using multi-layer feed-forward neural networks (MFNNs) was proposed base d on the inadequacy of present methods dealing with multiple variable and n onlinear problems. A MFNN model was built to implement the complex nonlinea r relationship between the inputs (many factors that influence concrete str ength) and the output (concrete strength). The neural network (NN) models g ive high prediction accuracy, and the research results conform to some rule s of mix proportion of concrete. These demonstrate that using NNs to predic t concrete strength is practical and beneficial. (C) 2000 Elsevier Science Ltd. All rights reserved.