Application of Soft-Computing Methods to Evaluate the Compressive Strength of Self-Compacting Concrete

Area of Science:

• Civil Engineering
• Materials Science
• Computational Intelligence

Background:

• Self-compacting concrete (SCC) is a vital material in modern construction.
• Accurate prediction of SCC compressive strength (CS) is crucial for structural integrity and design.
• Existing prediction methods may lack precision or efficiency.

Purpose of the Study:

• To evaluate and compare the efficacy of machine learning (ML) techniques for predicting the compressive strength (CS) of self-compacting concrete (SCC).
• To identify the most accurate ML model among multilayer perceptron (MLP), bagging regressor (BR), and support vector machine (SVM) for SCC CS prediction.
• To analyze the influence of various constituent materials on SCC CS using sensitivity analysis.

Main Methods:

• Utilized three machine learning models: multilayer perceptron (MLP), bagging regressor (BR), and support vector machine (SVM).
• Compiled a dataset of 169 data points from published literature, including 11 input parameters (e.g., cement, fly ash, aggregates, admixtures) and SCC compressive strength as the output.
• Employed k-fold cross-validation and statistical measures (MAE, RMSE, MAPE) to validate model accuracy and reliability.

Main Results:

• The bagging regressor (BR) model demonstrated superior performance in predicting SCC compressive strength, achieving a coefficient of determination (R²) of 0.95.
• Support vector machine (SVM) and multilayer perceptron (MLP) models yielded R² values of 0.90 and 0.86, respectively.
• Sensitivity analysis indicated that cement was the most influential input parameter (16.2%), while rice husk ash had the least impact (4.25%).

Conclusions:

• Machine learning, particularly the bagging regressor, offers a robust and accurate approach for predicting the compressive strength of self-compacting concrete.
• The findings provide valuable insights for optimizing SCC mix design by understanding the contribution of individual components.
• This research supports the use of data-driven methods for enhancing the predictability and performance of construction materials.