Predicting the compressive strength of silica fume concrete using hybrid artificial neural network with multi-objective grey wolves

The use of silica fume as a partial replacement for Ordinary Portland Cement provides a wide variety of benefits such as reduced pressure on natural resources, reduced CO2 footprint, and improved mechanical and durability properties. Compressive strength of concrete is one of the most important parameters in the design of concrete structures. Artificial Neural Network (ANN), as one of the powerful tools in the artificial intelligence field, has been widely used to predict various properties of concrete mixtures. One of the major drawbacks of the developed ANN models in predicting the concrete properties is their high complexity. In this study, estimating the compressive strength of silica fume concrete using the ANN method was considered as a two-objective optimization problem. These two objectives were the accuracy and the complexity of the developed ANN models. In this regard, a new multi-objective optimization method called Multi-Objective Grey Wolves Optimization (MOGWO) method was served to find a simple ANN model with acceptable error. After solving the optimization problem, a total number of 31 optimized ANN models were achieved. An ANN model with just one hidden layer with five neurons and the Pearson correlation coefficient of 0.9617 for all data was chosen as the final ANN model. Moreover, a sensitivity analysis was carried out to investigate the capability of the final ANN model in predicting the trend of the compressive strength of silica fume concrete with changing the effective variables on the compressive strength. (C) 2018 Elsevier Ltd. All rights reserved.