Predicting Density and Compressive Strength of Concrete Cement Paste Containing Silica Fume Using Arti cial Neural Networks

Abstract

Abstract. Arti cial Neural Networks (ANNs) have recently been introduced as an e cient arti cial intelligence modeling technique for applications involving a large number of variables, especially with highly nonlinear and complex interactions among input/output variables in a system without any prior knowledge about the nature of these interactions. Various types of ANN models are developed and used for di erent problems. In this paper, an arti cial neural network of the feed-forward back-propagation type has been applied for the prediction of density and compressive strength properties of the cement paste portion of concrete mixtures. The mechanical properties of concrete are highly in uenced by the density and compressive strength of concrete cement paste. Due to the complex non-linear e ect of silica fume on concrete cement paste, the ANN model is used to predict density and compressive strength parameters. The density and compressive strength of concrete cement paste are a ected by several parameters, viz, watercementitious materials ratio, silica fume unit contents, percentage of super-plasticizer, curing, cement type, etc. The 28-day compressive strength and Saturated Surface Dry (SSD) density values are considered as the aim of the prediction. A total of 600 specimens were selected. The system was trained and validated using 350 training pairs chosen randomly from the data set and tested using the remaining 250 pairs. Results indicate that the density and compressive strength of concrete cement paste can be predicted much more accurately using the ANN method compared to existing conventional methods, such as traditional regression analysis, statistical methods, etc.