Computation of High-Performance Concrete Compressive Strength Using Standalone and Ensembled Machine Learning...

Area of Science:

• Materials Science
• Civil Engineering
• Computational Science

Background:

• Modern research seeks cost-effective methods for predicting material properties.
• Machine learning (ML) is increasingly applied to material science for property prediction.
• Predicting concrete strength is crucial for construction quality and safety.

Purpose of the Study:

• To forecast the 28-day compressive strength of high-performance concrete (HPC).
• To evaluate standalone and ensemble ML techniques for concrete strength prediction.
• To compare the predictive accuracy and performance of different ML models.

Main Methods:

• Applied support vector regression (SVR) as a standalone ML technique.
• Utilized AdaBoost and random forest as ensemble ML techniques.
• Validated model performance using coefficient of determination (R²), statistical analysis, and k-fold cross-validation.
• Performed sensitivity analysis to determine input parameter contributions.

Main Results:

• All employed ML techniques demonstrated improved prediction performance.
• Random forest achieved the highest accuracy with an R² of 0.93.
• Support vector regression (R²=0.83) and AdaBoost (R²=0.90) also showed acceptable predictive performance.
• Cross-validation confirmed the random forest model's superior performance based on lower error metrics.

Conclusions:

• Machine learning techniques are effective for predicting the mechanical properties of high-performance concrete.
• The random forest model is a highly accurate tool for forecasting concrete compressive strength.
• ML offers a promising avenue for reducing experimental costs and time in material characterization.