Evaluation of the Compressive Strength of Generic and Geopolymer Concrete by Artificial Intelligence

In this study, the compressive strength of generic concrete and fly ash-based geopolymer concrete (GPC), made of cement by-products, has been compared using appropriate laboratory experiments and fully connected multi-layer artificial neural network (ANN) models. The main objective is to predict the compressive strength by the models and compare the models based on accuracy. ANN is one of the well-known supervised learning algorithms used in the field of artificial intelligence and it can effectively replace the current conventional labor-intensive time-consuming process of laboratory experiments. To prepare the experimental dataset, cylindrical specimens were prepared for generic concrete and GPC. The ANN model takes in fine aggregates, coarse aggregate, and sample size as the input for generic concrete and fly ash, slag, and sodium silicate solution for GPC to predict the output compressive strength based on the dataset used to train the network. All the experimental results and prediction models divulged that the ANN model trained for generic concrete had better accuracy with less error than the GPC one.