Predicting Compressive Strength of Self-Repairing Concrete Using Artificial Neural Networks

Abstract

In this study, an extensive simulation program was employed to determine the optimal artificial neural network (ANN) model for effectively forecasting the compressive strength of concrete. The prediction accuracy was based on the mineral admixture variation and crystalline admixture variation. To achieve this, a comprehensive experimental database consisting of 200 samples was compiled from existing literature and utilized to identify the most suitable ANN architecture. The primary objective of this research paper was to utilize an ANN to forecast the strength properties of self-repairing concrete (SRC) incorporating different mineral admixtures and a crystalline admixture. A total of 200 samples with various concrete mixtures were analyzed after a 28-day curing period. The ANN model was trained using the experimental data, which included four input parameters: exposure type (ET), crystalline admixture variation (CAV), mineral admixture variation (MAV) and mineral admixture (MA). The output parameter of interest was the concrete strength (Fc). Notably, the investigational data exhibited a significant correlation with the values forecast by the ANN model. In conclusion, the findings indicate that an ANN can accurately assess the strength characteristics of SRC. The utilization of ANN in this context holds promise for evaluating SRC strength properties with precision.