Rabie, Mohamed (2025) Sustainable and durable Cementitious Composites: experimental development, structural application and Machine Learning-Driven optimisation. Doctoral thesis, University of West London.

PDF Mohamed Rabie - Final PhD Thesis (June 26)_Suistainable and durable cementitious composites.pdf - Submitted Version Restricted to Repository staff only until 10 December 2026. Download (25MB)

Abstract

The escalating demand for sustainable infrastructure necessitates a paradigm shift towards construction materials that offer both low embodied carbon and enhanced durability. This doctoral thesis presents a holistic framework that integrates rigorous experimental investigations with advanced Machine Learning (ML) predictive modelling to develop, validate, and optimise sustainable cementitious composites. The research is structured across three distinct experimental programmes and a comprehensive computational framework.

Firstly, the study optimised the mix design of glass and carbon fibre-reinforced concrete. Experimental results indicated that while fibre inclusion marginally reduced compressive strength, it significantly enhanced split tensile strength-by up to 70% for glass fibres. Furthermore, the research validated a novel bio-inspired durability solution: an external enzymatic self-healing technique. It was found that fibres act as effective nucleation sites for precipitation, with the optimal mix achieving a 30% crack closure rate after 56 days. Life Cycle Assessment (LCA) confirmed that while fibres increase initial embodied energy, the potential for service life extension offers a viable sustainability trade-off.

Secondly, the research developed a sustainable, Ambient-cured Alkali-Activated Mortar
(AAM) as a low-carbon alternative to Portland cement. Replacing fly ash with Ground Granulated Blast-furnace Slag (GGBS) resulted in substantial early-age strength gains, though workability constraints were identified. This material was subsequently applied as a bonding matrix for Steel-Reinforced Grout (SRG) in the structural strengthening of continuous reinforced concrete beams. The SRG system yielded load-carrying capacity enhancements ranging from 24% to 104%. Crucially, the study revealed that low-density fabric configurations prevented premature debonding, thereby ensuring superior structural ductility compared to high-density alternatives.

Finally, to overcome the limitations of traditional empirical design, a computational framework was established using ensemble machine learning algorithms. The study demonstrated that Gradient Tree Boosting and XGBoost models could predict material behaviours, including the mechanical properties of GFRC, the strength of AAMs, and geotechnical soil shear parameters with exceptional accuracy (training R2 values exceeding 98%). Crucially, this framework was extended to Composite Reduced Web Section (RWS) connections, addressing a significant gap in the literature regarding data-driven seismic design. By superseding computationally expensive numerical simulations, the study successfully applied Multi-Objective Optimisation (NSGA-II) to balance conflicting criteria, such as maximising seismic performance while minimising embodied carbon. These computational models were deployed via user-friendly web interfaces, providing accessible, data-driven design tools for practising engineers with key codes presented in the appendix.

Collectively, this research advances the theoretical understanding of sustainable composites and provides a validated, practical toolkit for the construction industry’s transition towards net zero infrastructure

Item Type:	Thesis (Doctoral)
Identifier:	10.36828/thesis/15308
Additional Information:	All images belong to the author.
Subjects:	Computing > Intelligent systems Construction and engineering > Civil and structural engineering
Date Deposited:	09 Sep 2026
Dates:	Date Publication status September 2025 Submitted
School, department or research centre:	School of Computing and Engineering
URI:	https://repository.uwl.ac.uk/id/eprint/15308

Actions (admin access)