Optimizing the Mix Design of Relatively Strength Lightweight Concrete Using the Artificial Neural Network Method

Today, concrete is used as the main material in the construction of structures in our beloved country of Iran. Therefore, various types of concrete have been researched by different researchers in terms of behavior, weight, and resistance, and have been an important concern of researchers for a long time. Lightweight concrete, high-strength concrete, porous concrete, self-compacting concrete, etc. are examples of this research in the field of civil engineering, which has been held in various competitions in recent years. The present research has been discussed in the field of mixing design of a specific type of relatively light and relatively resistant concrete. The main goal of this article is to optimize the cubic concrete mixing design (5x5 cm) using artificial neural network models. The experiment program in this research started with 200 mixing designs in the field of cubic concrete and preparing a database and it was modeled by using various perceptron multilayer neural network structures. The evaluation indices used in this research include the regression coefficient (R) and the mean square error (MSE), which have been used to introduce the optimal structure of the artificial neural network. The results of the modeling based on the laboratory tests showed that the correlation coefficient for most of the models is more than 85%, which indicates the appropriate efficiency of the artificial neural network based on the Schmidt criterion. Also, pseudo-neuron with double hidden layer of 8 neurons with an average correlation coefficient of 93% in three sets of learning, training and evaluation and an average error index of 0.182 has been selected as the optimal structure in the set of structures used.