Effects of mix design parameters on workability and compressive strength of teff straw ash-based geopolymer mortars: Experimental evaluation and machine learning prediction
Tajebe Bezabih, David Sinkhonde, Derrick Mirindi, Samson Kiprop, Frédéric Mirindi, Oluwapelumi Abiodun
Abstract
Despite its high carbon footprint, Portland cement (PC) remains the dominant construction binder, prompting the search for sustainable alternatives. Geopolymers, formed by alkali activation of aluminosilicate-rich precursors, offer a low-carbon option with enhanced durability. While conventional precursors (like fly ash) are well studied, teff straw ash (TSA), a silica- and alumina-rich agricultural by-product, remains underexplored. Moreover, no previous study has systematically investigated the effects of key mix parameters on the fresh and hardened properties of TSA-based geopolymer mortar (TSA-GM) or applied machine learning (ML) for predictive modelling in this context. This study addresses these gaps through an experimental program that varied NaOH concentration (8–14 M), alkaline activator solution-to-TSA (AAS/TSA) ratio (0.6–0.9), Na₂SiO₃/NaOH ratio (1.5–3.0), and curing regime (ambient and heat curing at 65–80 °C), to evaluate workability and compressive strength. Four ML models (Random Forest (RF), Artificial Neural Network, Decision Tree, and Support Vector Regression) were trained to predict compressive strength and analyse parameter influence. Results show that an optimal mix (12 M NaOH, AAS/TSA = 0.7, Na₂SiO₃/NaOH = 2.5) achieved a compressive strength of 36.02 MPa at 90 days under heat curing, confirming its viability as a high-performance binder. Furthermore, ambient curing achieved a comparable long-term strength of 34.68 MPa, highlighting its potential as an energy-efficient alternative for non-time-sensitive applications. Among the models, RF demonstrated the best predictive performance (R² = 0.95). SHapley Additive exPlanations (SHAP) analysis identified water, NaOH content, curing temperature, and age as the most influential factors. This study is among the first to establish TSA as a viable precursor in geopolymer technology. It introduces an integrated experimental-ML framework to accelerate the development of sustainable construction materials.