Enhancing compressive and flexural strength prediction in high-performance concrete through integrated histogram gradient boosting and multiple optimization algorithms via an ensemble approach

Compressive and flexural strengths (CS and FS) constitute a vital parameter in designing various concrete structures, including rigid pavements, beams, and bridges, holding significant importance in ensuring their structural integrity and performance. The prevailing industry standard for concrete evaluation, the compressive strength test, is favored for its procedural simplicity. However, the estimation of CS and FS, especially for High-Performance Concrete (HPC), remains challenging due to material variability, mix complexity, curing conditions, and testing variations, necessitating advanced modeling approaches for accuracy. In response to this, the present research advocates the integration of Histogram Gradient Boosting (HGB) with Reptile Search Optimization (RSO), Arithmetic Optimization Algorithm (AOA), Sooty Tern Optimization Algorithm (STOA), Leader Harris hawks optimization (LHHO) and an ensemble of 4 optimizers to elevate the precision of such assessments. The results of this study reveal that among the various prediction models under examination, the HGB+RSA (HGRS) outperforms all other models, boasting the highest R2 value of 0.9914, and HGB+AOA (HGAO) is the second-best model with an R2 of 0.9846. When shifting the focus to FS estimation, the hybrid HGRS model shines with exceptional performance, displaying optimized R2 and RMSE values of 0.9922 and 0.2447, respectively. This underscores the effectiveness of the optimized Histogram Gradient Boosting with RSO in predicting CS and FS values for HPC. Interestingly, the study suggests that employing an ensemble of 4 selected optimizers developed reliable models with R2 of higher than 98% compatible with data exchange.