AI-Driven Mix Design Optimization for Hybrid Fiber-Reinforced High-Strength Concrete with Cement Replacement

Abstract

Because of its exceptional durability and compressive strength, high-strength concrete (HSC) is frequently used in contemporary construction. However, structural performance is seriously hampered by its brittleness and vulnerability to micro- and macro-cracking. The combined effects of hybrid steel and polypropylene fiber reinforcement and partial cement substitution with ground granulated blast furnace slag (GGBS) on the mechanical characteristics and crack resistance of M60-grade concrete are examined in this study. Specimens with different fiber content, aspect ratios, and cement replacement levels were subjected to an experimental program that included compressive, split tensile, and flexural strength tests as well as digital image analysis measures of crack width.

A dataset generated from these experiments was utilized to develop machine learning models — To forecast crack resistance and improve the concrete mix, use Artificial Neural Networks (ANN), Random Forests (RF), and Extreme Gradient Boosting (XGBoost). Model performance was evaluated using R², RMSE, and MAE. The XG Boost model demonstrated the highest predictive accuracy (R² = 0.965), followed by RF (R² = 0.948) and ANN (R² = 0.936). Optimization results identified an optimal configuration: steel fibers with aspect ratio 70–75 partially aligned, polypropylene fibers with aspect ratio 550–600 randomly distributed, and 30% GGBS replacement. This combination reduced crack width by approximately 62% and improved flexural strength by 28% compared to control HSC.The findings highlight the potential of AI-driven mix design for accelerating high-performance, sustainable HSC development, minimizing trial-and-error experimentation, and improving durability and structural performance.