Design of low-carbon cement blends for thermal well conditions using machine learning and industrial wastes

High-temperature well cementing presents critical challenges due to the thermal degradation of traditional Portland cement systems, often resulting in mechanical strength loss and compromised well integrity. Despite advances, there remains a lack of sustainable, thermally stable cement blends designed for demanding conditions such as steam injection and geothermal wells. This study aimed to develop and validate a low-carbon, high-performance cementitious system incorporating industrial wastes to replace silica flour, enhancing thermal stability while reducing environmental impact. A systematic experimental workflow was implemented, including raw material characterization, fluid and hardened-state testing of 24 binary and ternary formulations, and high-temperature curing simulations representative of bottomhole conditions (280 °C, 1500 psi). A support vector regression (SVR) model guided the blend optimization, followed by experimental validation, microstructural analysis (XRD and SEM), and thermal behavior. The optimized ternary blend (B31.3RHA45), composed of Portland cement, rice husk ash (RHA), and mortar sand residue (MSR), achieved compressive strength of 40.6 ± 0.6 MPa after thermal cycling and exhibited low permeability (0.12 mD). Experimental values closely matched SVR predictions (deviation