Mechanical and durability performance of fiber-reinforced highstrength concrete having GGBS: experiments and machine learning

Recent efforts in sustainable construction focus on reducing cement consumption while improving the performance of fiber-reinforced high-strength concrete (FRHSC). This study investigates the combined influence of ground granulated blast furnace slag (GGBS) and recycled steel fibers (RSF) on the fresh, mechanical, and durability properties of FRHSC, supported by machine learning-based strength prediction. Cement was partially replaced with GGBS at levels of 15% and 30%, while RSF were incorporated at 0.25%, 0.50%, and 0.75% by volume. Fresh properties were evaluated through workability-related tests. Mechanical performance was assessed using compressive strength (CS), splitting tensile strength (STS), and flexural strength (FS) tests at 7, 28, and 90 days. Durability was examined using sorptivity, water absorption, rapid chloride penetration, and electrical resistivity tests. The random forest (RF) and artificial neural network (ANN) models were developed to predict strength behavior. The results showed that GGBS improved workability and long-term strength due to its delayed pozzolanic activity and formation of secondary C-S-H gel, leading to a denser microstructure. RSF enhanced tensile and FS by bridging cracks and limiting crack propagation. At 90 days, the paste comprising 30% GGBS and 0.75% RSF achieved increases of 13.61% in CS, 35.73% in STS, and 31.2% in FS compared to the reference mix. Durability performance was significantly enhanced, with up to a 66.38% reduction in chloride ion penetration, attributed to pore refinement and reduced permeability. SEM analysis confirmed a compact and homogeneous microstructure in GGBS-fibermodified concrete. Among the machine learning models, the RF approach demonstrated superior predictive accuracy compared to ANN, with higher R2 values and lower prediction errors. The findings highlight the synergistic role of GGBS and recycled RSF in producing durable and high-performance concrete. The integration of machine learning further provides an efficient tool for strength prediction, reducing experimental dependency. This study offers practical guidance for developing sustainable FRHSC with enhanced mechanical performance, durability, and predictive reliability.