Predicting the Rheological Properties of Super-Plasticized Concrete Using Modeling Techniques

Area of Science:

• Civil Engineering
• Materials Science
• Computational Science

Background:

• Concrete pumpability is critically dependent on rheological properties like yield stress (YS) and plastic viscosity (PV).
• Accurate prediction of these properties is essential for optimizing concrete mix design and construction processes.

Purpose of the Study:

• To forecast the rheological properties (YS and PV) of fresh concrete using predictive machine learning (PML) techniques.
• To compare the performance of Artificial Neural Network (NN) and Random Forest (R-F) models in predicting concrete rheology.

Main Methods:

• Application of Artificial Neural Network (NN) and Random Forest (R-F) algorithms for predicting plastic viscosity and yield stress.
• Validation of model accuracy using statistical checks, k-fold cross-validation, and analysis of error metrics (MAE, MSE, RMSE).
• Investigation of the influence of input parameters on the prediction of rheological properties.

Main Results:

• The Random Forest (R-F) model demonstrated superior performance, achieving R-squared values of 0.92 for PV and 0.96 for YS.
• R-F models exhibited lower error metrics (e.g., RMSE for YS: 33.79 Pa, PV: 4.06 Pa·s) compared to NN models.
• Statistical validation and k-fold cross-validation confirmed the high precision and reliability of the R-F model.

Conclusions:

• The Random Forest model is highly effective for accurately predicting the yield stress and plastic viscosity of fresh concrete.
• This PML approach offers significant benefits to the construction industry by saving time, effort, and project costs.
• The study provides a valuable tool for researchers and industry professionals in concrete technology and construction management.