Embodied carbon analysis and benchmarking emissions of high and ultra-high strength concrete using machine learning algorithms. (20th July 2020)

• Record Type: Journal Article
• Authors: Thilakarathna, P.S.M.; Seo, S.; Baduge, K.S. Kristombu; Lee, H.; Mendis, P.; Foliente, G.
• Is Part Of: Journal of cleaner production. Volume 262(2020)
• Publication Date: 2020-07-20
• Subjects: High strength concrete -- Machine learning -- Artificial neural networks -- Gaussian process regression -- Embodied carbon -- Sustainable concrete; Factory and trade waste -- Management -- Periodicals; Manufactures -- Environmental aspects -- Periodicals; Déchets industriels -- Gestion -- Périodiques; Usines -- Aspect de l'environnement -- Périodiques; 628.5
• DOI: 10.1016/j.jclepro.2020.121281
• Languages: English
• ISSNs: 0959-6526

Abstract

Abstract: High strength concrete (HSC) (50–100 MPa) and ultra-high strength concrete (UHSC) (>100 MPa) have been increasingly used in the construction industry due to its inherent performance characteristics. However, these concrete mixes have a higher carbon footprint and it is vital to consider the embodied carbon of the HSC and UHSC due to the massive consumption throughout the world. In this study, embodied carbon analysis, using machine learning algorithms has been carried out to minimize the carbon footprint of concrete without jeopardizing the mechanical properties of the concrete. Machine learning models are developed using experimental results in the literature and used to predict the compressive strength of concrete using the constituent materials. Using the experimental data and machine-learned models for mix designs, embodied carbon emissions were calculated. It is shown that there can be many mix compositions which have the same compressive strength while having significantly different embodied carbon values. Based on experimental and machine learned mix designs, an equation to predict the average embodied carbon value for concrete mixes is proposed. The study suggested proposed intervals for the benchmark function in order to propose a region where the embodied carbon value of a concrete mix should lie while achieving the desired compressive strength. Finally, it is shown that machine learning can be used successfully to identify the high strength concrete mixes while minimizing the embodied carbon value of that mix composition. Finally, guidelines are presented to produce a concrete mix within proposed benchmark limits while achieving the desirable strength grade. Highlights: Machine learning algorithms can be used to predict the embodied carbon emissions. Dependent on the mix design, different embodied emissions exist for the same compressive strength. An expression proposed for embodied carbon emission of high and very-high strength concrete. A benchmark function and limits are proposed for embodied energy. Sustainable mix designs and methods are proposed using ma 3chine learning algorithms.