Compressive Strength Assessment of Concrete Containing Metakaolin Using ANN

Artificial neural networks (ANNs) as a powerful approach have been widely utilized to demonstrate some of the engineering problems. A three-layer ANN including three neurons in the hidden layer is considered to produce a verified pattern for assessing the compressive strength of concrete incorporating metakaolin (MK). For this purpose, an extensive database including ۴۶۹ experimental specimens was obtained from the literature. Next, novel equations utilizing the developed ANN approach were developed to measure the compressive strength of concrete mixtures incorporating MK. To examine the model accuracy a comparison between the proposed formulas based ANN and an empirical formula based nonlinear least-squares regression (NLSR) was carried out. The results show that the proposed formula based on the ANN yields a higher correlation coefficient and fewer errors compared to the NLSR method. An analysis based weights incorporating was utilized to show the significance of input variables. Accordingly, the ratio of fine aggregate to coarse aggregate exerts a dominant influence on the compressive strength of the concretes containing MK.

Keywords:

Artificial Neural Network , Compressive Strength of Concrete , Metakaolin , Garson’s algorithm , Nonlinear least squares regression