Experimental and machine learning prediction of compressive strength of chemically activated RHA based RAC using SHAP...

Area of Science:

• Materials Science
• Civil Engineering
• Sustainable Construction

Background:

• Growing demand for sustainable construction materials.
• Need for effective reuse of industrial and agricultural waste in high-performance concrete (HPC).
• Challenges include strength loss with recycled concrete aggregates (RCA) and variable supplementary cementitious material performance.

Purpose of the Study:

• Investigate the synergistic effects of chemically activated rice husk ash (RHA), RCA, and foundry sand (FS) on HPC.
• Evaluate compressive strength and durability of HPC incorporating these waste materials.
• Develop predictive machine learning models for HPC performance.

Main Methods:

• Prepared six experimental HPC groups with varying RCA (0-100%) and activated RHA using sodium sulfate (Na2SO4).
• Included 20% FS as fine aggregate replacement and constant silica fume in all mixes.
• Assessed compressive strength over 120 days and durability via acid exposure tests; applied ML models (XGBoost, RF, ANN, KNN) for strength prediction.

Main Results:

• Chemically activated RHA with up to 40% RCA and 20% FS showed no compromise in long-term strength and acid resistance.
• Extreme Gradient Boosting (XGBoost) model achieved high accuracy (R²=0.951) in predicting compressive strength.
• SHAP and PDP analyses identified curing age, RCA, and Na2SO4 content as critical factors influencing HPC performance.

Conclusions:

• Up to 40% RCA can be effectively incorporated into HPC using activated RHA and FS without sacrificing performance.
• The study provides a robust framework for sustainable concrete design by integrating experimental data with interpretable machine learning.
• This research facilitates the wider adoption of waste materials in high-performance concrete applications.