An interpretable surrogate modeling framework for rice husk ash concrete using copula-based virtual sampling.

Area of Science:

• Materials Science
• Civil Engineering
• Computational Mechanics

Background:

• Compressive strength of concrete is crucial for structural integrity.
• Traditional sensitivity analysis often assumes input independence, which is unrealistic for concrete mix design.
• Rice husk ash concrete offers sustainable benefits but requires careful mix design optimization.

Purpose of the Study:

• To develop an interpretable and variance-aware sensitivity analysis framework for rice husk ash concrete.
• To analyze the influence of correlated input variables on compressive strength.
• To support sustainable and robust concrete mix design.

Main Methods:

• Utilized Gaussian copula for virtual design space construction, preserving data dependencies.
• Employed Extreme Gradient Boosting surrogate models optimized via Covariance Matrix Adaptation Evolution Strategy.
• Applied a permutation-based sensitivity measure for global variance attribution under correlated inputs.

Main Results:

• Achieved high predictive accuracy (mean R² = 0.974) for surrogate models.
• Identified Cement, Water, and Age as dominant factors influencing compressive strength.
• Revealed significant interaction effects, e.g., Cement-Water and Coarse aggregate-Superplasticizer.

Conclusions:

• The proposed framework provides reliable and interpretable sensitivity analysis for correlated inputs.
• Findings support the optimization of rice husk ash concrete mix design for sustainability.
• External validation confirms the framework's practical relevance and robustness.