Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article

Degree

Master of Engineering Science

Program

Civil and Environmental Engineering

Supervisor

Nehdi, Moncef

Abstract

Whilst recycled aggregate (RA) can alleviate the environmental footprint of concrete production and the landfilling of colossal amounts of demolition waste, there need for robust predictive tools for its effects on mechanical and durability properties. In this thesis, state-of-the-art machine learning (ML) models were deployed to predict properties of recycled aggregate concrete (RAC). A systematic review was performed to analyze pertinent ML techniques previously applied in the concrete technology field. Accordingly, three different ML methods were selected to determine the compressive strength of RAC and perform mixture proportioning optimization. Furthermore, a gradient boosting regression tree was used to study the effects of RA and several types of binders on the carbonation depth of RAC. The ML models developed in this study demonstrated robust performance to predict diverse properties of RAC.

Summary for Lay Audience

Worldwide concerns regarding the environmental footprint of concrete production have imposed more rigorous requirements for construction and urban development. To enhance the sustainability of concrete, it is important to enhance its durability, lower the energy consumption in its production and placement processes, and promote the use of recycled materials in its mixture design. In the pursuit of such goals, this study explores the mechanical and durability properties of recycled aggregate concrete.

Recycled aggregate concrete (RAC) could contribute to mitigating the local shortages of natural aggregates, prevent the landfilling of massive amounts of construction and demolition waste, and reduce carbon emissions of concrete construction. Accordingly, this thesis presents state-of-the-art machine learning (ML) models to predict two main properties of RAC: compressive strength and resistance to carbonation. The development of these ML models ensured that the used datasets were diverse and comprehensive to capture the intrinsic principles involved in the properties of RAC. The carbonation depth of RAC was predicted for the first-time using ML. Furthermore, a hybrid ML model was developed to optimize the mixture design of RAC for various classes of compressive strength. The results demonstrated the superiority of ML techniques in the prediction of RAC properties. The models developed herein could be further harvested to achieve sustainable production of concrete with optimal recycled aggregate content, least cost, higher durability, and least environmental footprint.

Download

Share

COinS