Evaluation of high performance concrete hardness properties using fuzzy logic based modeling

For construction engineering to be most effective, it's essential to understand the hardness properties of high-performance concrete (HPC), such as compressive strength and slump. However, predicting these properties is challenging due to complex, non-linear relationships and the inherent variability of factors like the water-cement ratio and mineral additives. Traditional modeling methods often struggle to handle these uncertainties effectively. This paper introduces a novel Hybrid Fuzzy Artificial Neural Network (HFANN) architecture that integrates fuzzy logic with advanced machine learning models, including Support Vector Regression (SVR), Multilayer Perceptron (MLP), and Gradient Boosting Machine (GBM), all optimized by the Chaos Game Optimization (CGO). The fuzzy logic system employs Gaussian membership functions and a Sugeno-type rule base to preprocess inputs, generating 18 fuzzy features per sample to capture data uncertainties. The CGO enhances model performance by optimizing hyperparameters and ensemble weights through a chaotic search based on the Sierpinski triangle. The HFANN framework demonstrates superior predictive accuracy compared to traditional models. It achieves a 25% reduction in RMSE for compressive strength and R² values of 0.98 for compressive strength and 0.97-0.99 for slump across training, validation, and testing phases. These results, validated through MANOVA and Tukey's HSD tests, indicate significant improvements over conventional methods. The HFANN framework offers a robust tool for enhancing HPC mix design and advancing sustainable construction practices.