Fully Funded UKRI & Swansea PhD Scholarship: Deep-learning Accelerated Computational Fluid Dynamic Model for Pollutant Dispersion in Urban Environment

Funding providers: UKRI and Swansea University

Subject areas: Computational Engineering/ Machine Learning

Project start date: 1^st July 2023 (Enrolment open from mid–June)

Project description: 

Exposure to hazardous substances from manufacturing, storage and transport poses significant risk to public health and to the environment. Planning control of hazardous substances has been one of the main strategies at different national agencies (e.g. Environmental Agency) to understand the associated risks during the planning phase as well as the operational stage of both new and existing facilities. Apart from extensive networks of physical sensors being deployed island-wide to monitor air quality and pollutants, numerical models of pollutant dispersion for planning and operational purposes have been used. The current industrial numerical approaches are often fast to provide predictions; they are, however, much less accurate due to inherent assumptions and simplifications embedded in those models. There exists a class of higher fidelity methodologies for the prediction of pollutant dispersion using computational fluid dynamic (CFD) approaches. Unfortunately, the CFD-based models are resource intensive and time-consuming to execute; thus, rendering them impractical for industrial usage. The current proposal aims at addressing this gap by developing a fast CFD-based approach for prediction of pollutant dispersion.

Computational methods for predicting complex unsteady flows are computationally demanding, while full-field experimental methods are very expensive. This project is at the forefront of the emerging field of machine learning and data-driven computational modelling. The project aims at developing unique new models to predict fluid mechanic characteristics by integrating advances in high order CFD techniques for simulations of urban flows with accuracy and robustness The aim is to build a surrogate model for wind wake prediction using high fidelity CFD data - thus taking into account the effects of the urban built environment (i.e. buildings and greenery). This development will form a fundamental building block for the hybrid platform on pollutant dispersion monitoring and response.

The resulting learning-based models will vastly reduce the computational cost of running CFD simulations. This reduction in cost will expand the potential applications of CFD into new areas such as generative design, digital twins, life-cycle forecasting for engineering structures and inverse problems.

Eligibility

Candidates must normally hold an undergraduate degree at 2.1 level (or Non-UK equivalent as defined by Swansea University) in Engineering or similar relevant science discipline.

English Language requirements: If applicable – IELTS 6.5 overall (with at least 5.5 in each individual component) or Swansea recognised equivalent.

Due to funding restrictions, this scholarship is open to applicants eligible to pay tuition fees at the UK rate only, as defined by UKCISA regulations.

Advert information