Prediction of geopolymer concrete compressive strength using artificial neural network and genetic algorithm

Abstract

• A hybrid predictive model for geopolymer concrete was developed using ANNs with GA, PSO, and LM algorithms. • The ANN-LM model achieved the highest prediction accuracy ( R = 0.96701 for training, R = 0.9868 for testing). • The ANN-PSO model exhibited strong generalization capability with balanced error distribution. • The proposed AI-based approach optimizes mix design and reduces experimental costs. The environmental concerns associated with Portland cement production have led to the development of geopolymer concrete as a sustainable alternative with reduced carbon emissions. In this study, an advanced hybrid predictive model based on Artificial Neural Networks (ANNs) combined with three optimization algorithms—Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Levenberg-Marquardt (LM) algorithm—was developed to estimate the compressive strength of geopolymer concrete (GPC). A dataset of 162 mix designs, obtained from experimental studies, was used for model training and validation. The performance of the proposed models was evaluated based on statistical accuracy metrics, demonstrating that the ANN-LM model outperformed the other two in prediction precision. Notably, this study also considers the variations in the chemical composition of fly ash and sodium silicate solution, which significantly influence the strength behavior of GPC. The findings highlight the effectiveness of AI-based models in optimizing mix design parameters and predicting mechanical properties of sustainable concrete materials. The proposed approach can contribute to reducing experimental costs and accelerating the adoption of eco-friendly construction materials. This study uniquely integrates three optimization algorithms to enhance prediction accuracy, with test results confirming the superiority of the ANN-LM hybrid model.

Topics & Concepts