Harnessing explainable Artificial Intelligence (XAI) for enhanced geopolymer concrete mix optimization

Geopolymer concrete (GC) emerges as a sustainable alternative yet faces challenges in achieving optimal resource utilization for strength development. Balancing these aspects is crucial for its large-scale adoption as a sustainable material. The type and dosage of precursors, activator, curing, and mixing conditions influence compressive strength, setting time, and workability. Moreover, multiple experimental trials are required for a desirable geopolymer blend. Even the experimental parameters alone do not meet the design principles concerning sustainable construction. This paper presents a study on the mix design and interpretation of machine learning techniques (MLT) with XAI. To train the model, extensive experimental databases using the shapley additive explanations (SHAP) technique rank input factors that impact the strength aspect. The prediction models' performance was compared using coefficient of determination (R 2 ) and root mean square error (RMSE). SHAP interpretations reveal that temperature, Na to Al ratio, and NaOH molarity are the main factors influencing the compressive strength of GC. Further, these parameters were crucial in developing the dense geopolymer matrix. By integrating XAI into the MLT approach, we have also opened new criteria for understanding the complex relationships between geopolymer concrete potential parameters and their compressive strength. • Sixteen parameters were analysed to assess their impact on geopolymer strength. • SHAP was utilized to elucidate the machine learning model of geopolymer concrete. • SHAP clarified the decision-making process of the machine learning model. • Explainable machine learning models enhance viability of geopolymer mix design.