Prediction of rapid chloride permeability of ground granulated blast furnace slag and <scp>silica fume</scp> based ultra‐high performance concrete using hybrid gradient boosting algorithms

Abstract Reducing concrete permeability to chloride ions is crucial for extending the lifespan of concrete structures. This study investigates the enhancement of chloride ion penetration resistance in ultra‐high‐performance concrete (UHPC) through the incorporation of ground granulated blast furnace slag (GGBS) and silica fume (SF). To overcome the labor‐intensive and costly nature of conventional evaluation methods like the rapid chloride permeability test (RCPT), machine learning models were developed using gradient boosting (GB) combined with Bayesian optimization (BO), differential evolution (DE), and particle swarm optimization (PSO). A dataset comprising 264 experimental records, with variables such as age, silica fume, cement, GGBS, superplasticizer, aggregates, and temperature, was utilized. The GB‐DE and GB‐BO models achieved high predictive accuracy with R 2 values of 0.9536 and 0.9404, respectively. SHAP analysis of the best‐performing GB‐DE model provided insights into the influence of input variables, enhancing model interpretability. Results confirm that UHPC mixtures containing GGBS and SF exhibit significantly reduced chloride permeability at 28 days due to denser microstructure and accelerated C‐S‐H formation.