Title: Machine Learning Models for Predicting Rheological Properties of Self-Consolidating Concrete (SCC)

Author(s): Abdelhamid Hafidi, Ilhame Harbouz, Benoit Hilloulin, Ahmed Loukili, and Ammar Yahia

Publication: Symposium Paper

Volume: 362

Issue:

Appears on pages(s): 249-267

Keywords: machine learning (ML); rheology; self-consolidating concrete (SCC); supplementary cementitious materials (SCM); viscosity; yield stress

DOI: 10.14359/51740887

Date: 6/6/2024

Abstract:
This study investigates the potential of machine learning (ML) models to predict the rheological properties of self-consolidating concrete (SCC), with a focus on yield stress and viscosity. The significance of this research arises from the environmental impact of cement production and the pressing need to explore low-carbon alternatives. Supplementary cementitious materials (SCM), such as slag and fly ash, offer promise for reducing carbon emissions in the cement industry. However, their incorporation can alter the rheological properties of concrete, impacting its mechanical and durability characteristics. Predicting these properties is complex due to the multifaceted interplay of various factors. To address this challenge, ML models were employed, including Random Forest (RF) and Gradient Boosting (GB). A comprehensive database comprising 12 input parameters, such as mixture proportions, aggregate characteristics, and rheological attributes, was meticulously compiled from existing literature. Training and testing these ML models revealed GB as a standout performer for predicting yield stress, while RF excelled in forecasting viscosity. Furthermore, a comprehensive SHapley Additive exPlanations (SHAP) analysis was conducted to unravel the most influential parameters impacting yield stress and viscosity. These findings can contribute in advancing our understanding of SCC behavior and the development of sustainable construction materials that align with environmental objectives.