Overview
This project aims to improve the efficiency of fire safety design by combining machine learning with CFD fire modelling. It will reduce the time and computing power needed for simulations, making detailed fire modelling more practical for everyday use. Working with real-world data from our industry partner SOCOTEC (https://www.socotec.co.uk/), the project offers a realistic step toward faster and more reliable fire safety solutions.
About this opportunity
Fires in urban environments remain a critical safety challenge as cities become denser and more architecturally complex. The consequences of fire, ranging from structural damage and economic loss to injuries and fatalities, are considerable. In the UK, the Home Office estimated the marginal cost of fire in England at approximately £3.2 billion in 2020. Addressing these risks requires robust fire safety design tools that support estimation of fire behaviour within buildings. Computational Fluid Dynamics (CFD) is currently one of the most popular tools available to fire safety engineers, used for modelling fire growth, smoke spread, and heat transfer in complex buildings.
However, detailed CFD simulations are highly resource-intensive and time-consuming, often taking weeks to run on supercomputers. This creates a significant barrier for practical use in iterative design, regulatory review, or emergency response planning. This PhD project aims to address that challenge by developing a machine learning (ML)-based surrogate modelling tool to enhance the efficiency of CFD simulations. By training ML models on a database of CFD outputs from real-world fire scenarios (in close collaboration between SOCOTEC and University of Liverpool), the project seeks to build an efficient design tool that significantly reduces the computational time while maintaining acceptable accuracy.
The candidate will benefit from a multidisciplinary supervisory team with expertise in fire dynamics, fire modelling, machine learning, and AI. They will receive training in advanced CFD modelling, ML algorithm development, and software implementation. In addition, the project involves close collaboration with SOCOTEC (https://www.socotec.co.uk/), who will provide industry guidance, case studies, summer interns, and opportunities for site visits and further knowledge exchange. The candidate will have an industry mentor, Karla Sandoval, who is the Research Leader at the fire division at SOCOTEC. The candidate will also engage with an academic specialised in ML and AI in fire safety engineering from Delft University of Technology (TU Delft) in Netherlands.
University of Liverpool is established in 1881, we are an internationally renowned Russell Group university recognised for our high-quality teaching and research. We are consistently ranked as one of the best universities both nationally and globally, and the majority of our research is rated world-leading or internationally excellent.