Computer scientists at the University of the West of England are developing software that uses AI and machine learning to help construction companies reduce the amount of embodied carbon in their building and infrastructure projects. Dr Lukman Akanbi explains how the project could help the UK achieve its net zero targets
In 2019, the UK pledged to achieve net zero emissions by 2050, which may just have provided the construction industry with the impetus to push the boundaries of technology when looking for ways to reduce its carbon emissions.
As we know, a large proportion of construction projects create a carbon footprint, once completed, whose size depends on the project’s energy efficiency. However, we should not forget about embodied carbon. This pertains to the energy (and therefore CO2) consumed during the construction phase, including the extraction of building materials, manufacturing, transport of materials, assembly, maintenance, deconstruction and disposal.
Reducing embodied carbon at the planning stages
I work at the University of the West of England (UWE Bristol) and am part of a team of computer scientists in the UK who are developing software that uses artificial intelligence (AI) and machine learning to radically speed up the process of determining how to reduce embodied carbon at the planning stages.
The two-year, £800,000 project, funded by Innovate UK, is led by UWE Bristol’s Big Data Laboratory in collaboration with Winvic Construction and Costain, as well as Edgetrix, a start-up that specialises in cloud and AI solutions.
Tackling embodied carbon in construction projects has become an increasingly important topic around the world in the last decade and construction companies have begun looking at the whole-life carbon of a building. In the UK, although operational carbon emissions are currently regulated for buildings, there are no equivalent regulations for embodied carbon and it is hoped that the government will bring in such regulations in line with its carbon neutral strategy.
However, regulation or not, buildings are increasingly constructed in ways that make them more energy efficient and in the near future, the energy sources to run these buildings will likely rely on locally generated low or zero carbon heat and power sources. As a result, the proportion of the building’s lifecycle CO2 that comes from the embodied carbon is set to become more significant.
Assessing this needs to be done early on in a project: at the design stage and particularly in the planning of the building’s structure if carbon savings are to be made for the lifetime of the project. Existing methods of determining which construction materials to use to reduce embodied carbon can be extremely time-consuming when done by a human, and are costly. This is where our software comes in, as it makes the calculations for you and its machine learning capability means that it will gradually speed up through time as it gathers more data.
Developing machine learning models
ASPEC (AI System for Predicting Embodied Carbon) begins by collecting embodied carbon data from previous projects and the more data we have, the better. We are already developing the machine learning models, whose algorithms will then analyse this data to spot any patterns and learn from them. As the system receives and sifts through millions of items of data, it will gradually become quicker and quicker and will eventually be able to come up with material suggestions in a fraction of the time.
For a large-scale project, for example, we estimate that instead of taking five to 10 hours to work out alternative materials, it could initially take only one to two hours and further down the road, once more data is gathered, just a few minutes.
One of the challenges is that there is still a lack of available data about embodied carbon and existing datasets are often fragmented and in silos. But with the collaboration and support of construction companies on the project, this will change in time.
Our work, which started in November, will initially use Winvic and Costain’s commercial premises and over the next two years, we will use them as test sites for the software.
Once we are satisfied that the system works as efficiently as possible and once we commercialise it, the objective is to make the system available to building designers, who will be able to buy and use it as a plugin as part of their existing design systems, such as Autodesk Revit. This way they can implement embodied carbon findings incrementally throughout a construction project’s delivery.
Based on UWE Bristol’s Frenchay campus, the Big Data Lab is an exciting place where our academic researchers and systems analysts aim to find innovative solutions to a variety of challenges. We have already developed a system that uses Internet of Things (IoT) sensors to determine the presence and level of pollutants on selected motorways in the UK. We are also testing a software that uses similar sensors to forecast energy requirements of commercial buildings to enable effective on-demand energy purchase to save cost.
UWE Bristol is a university that’s environmentally aware and is aiming to be carbon neutral by 2030, so it’s great that we are taking a lead in a project that is looking to help achieve similar aims and could help save valuable time for the industry as it sets out ways of reducing its carbon footprint.
Dr Lukman Akanbi
Big Data Enterprise & Artificial Intelligence Laboratory (Big-DEAL), University of the West of England
LinkedIn: University of the West of England