Enhancing durability and sustainability in fly ash-slag concrete using advanced metaheuristic algorithms and explainable ML for compressive strength prediction

Abstract. Fly ash slag concrete (FASC), a supplementary cementitious material, has transformed construction by lowering the carbon footprint, minimizing waste, reducing labor costs, and improving durability and precision. Predicting compressive strength (CS), a key mechanical property, is essential for optimal performance. Due to the nonlinear nature of FASC mixtures, researchers now utilize machine learning tools. This study evaluates three machine learning models by combining traditional AI algorithms, such as artificial neural networks (ANN), with nature-inspired optimization techniques, such as chicken swarm optimization (CSO), moth flame optimization algorithm (MOFA), and whale optimization algorithm (WOA). By addressing the gaps in mechanical property variation, dataset scope, and model comparison, this study demonstrated high accuracy in CS prediction for all three models. The ANN optimized by WOA consistently excelled across multiple metrics. Visual evidence supports the models' effectiveness, suggesting benefits like better quality control, cost savings, increased safety, and a cleaner environment.