Prediction of mechanical properties of concrete by novel method of mixing using Artificial Neural Network and Support Vector regression

This study investigates the effect of a novel pre-dissolution mixing method, wherein Ground Granulated Blast Furnace Slag (GGBS) is dissolved in water prior to blending with other concrete constituents, on the mechanical and durability properties of M30 grade concrete. The hypothesis is that pre-dissolution promotes early ionic release from slag particles, lowering the Ca/Si ratio and accelerating denser C–S–H gel formation relative to conventional direct addition. Five slag replacement levels (0%, 2.5%, 5%, 10%, and 20% by cement weight) were evaluated under both mixing methods. The novel method consistently outperformed the conventional approach, with 2.5% slag at one hour immersion identified as optimum based on compressive strength, carbonation resistance, and sustainability criteria. To mitigate accelerated carbonation at higher replacement levels, fly ash was introduced at levels of 0–40% by cement weight, forming a ternary blended system. The mix GBS2.5FA25 achieved compressive strength improvements of 17.4% and 10.18% at 28 and 90 days respectively over the equivalent traditional mix. Mechanical properties were predicted using ANN and SVR models. ANN (4-18-1 architecture, R2 = 0.97) outperformed all SVR kernel variants, confirming its superior generalisation. These findings demonstrate that pre-dissolution mixing is an effective low-carbon strategy for simultaneously enhancing mechanical performance and carbonation resistance.