Optimization of Mix Design for Geopolymer Concrete Using Artificial Neural Network

Engineers’ experience and empirical relationships serve as the foundation for concrete mix design investigation and optimisation. Numerous trial mixes must be made and examined for different design factors in order to achieve the required mix proportion. This obviously takes much time. To overcome, the prediction of the concrete mix proportions with experimental data can help an artificial neural network (ANN). Using a large number of trial mixes and resources, the ANN approach predicts the concrete mix proportions that can be both economical and effective for producing concrete. ANNs are trained to forecast mix proportions that are determined by a variety of input characteristics by using experimental mix design data from the past. A dataset of mix proportions was employed to train the ANN model. In order to anticipate the output, the following input parameters were set: the specific gravity of cement, the W/B ratio, the workability, the target mean strength, and the grading zone of fine aggregate. Five mix proportions obtained from ANN was used to validate the trained network. The error rate observed as 0.193%. The developed ANN model has been found to be highly efficient than linear regression analysis.