Prediction of Efficiency Factor of Natural Pozzolan by the Use of an Artificial Neural Network

Abstract

Because of the abundance of the mineral additives and the great variations on their physical and chemical characteristics, the development of a general concept for their use out of concrete is required. In this study, the concept of the efficiency factor is applied like a measurement of the relative performance of these materials compared with Portland cement. The rapid growth of the artificial intelligence had a very important impact on the concrete technology.  It makes it possible to solve complex prediction problems of the properties of the concretes with cementitous materials (slag, fly-ashes, silica fume and natural pozzolan). The main aim of this study is to test the validity of the approach of Artificial Neural Networks (ANN) in developing a model for the prediction of the natural pozzolan efficiency factor. The most suitable model is the feed-forward multi-layer network. It is produced to implement the complexity of the nonlinear relation between the data network (the Water/Binder ratio “”, the percentage of pozzolan and the age of testing) and the produced result (the efficiency factor). It is also established by an incorporation of a large experimental database and by a suitable choice of architecture and the training process. The model was validated by experimental tests. The ANN Model developed provided effective means for the formulation of the concretes containing natural pozzolan for a given water binder ratio (W/B), an age of testing (t) and a rate of substitution of natural pozzolan (P).