About the project
Physics-Informed Neural Networks (PINNs) provide an innovative, efficient alternative to traditional methods employed in fluid dynamics simulations, delivering Computational Fluid Dynamics (CFD)-level of accuracy at significantly lower computational costs, thus reducing energy use and benefiting the environment.
Artificial Intelligence (AI) has become deeply integrated into modern society, offering immense benefits that often eclipse its limitations. Neural Networks (NNs), the most prevalent form of AI, typically demand vast training datasets and substantial computational power, resulting in high energy consumption and a considerable carbon footprint. Additionally, collecting and preparing data remains a major challenge that limits AI accessibility.
However, many problems in physics and engineering are governed by well-established laws, often expressed as partial differential equations (PDEs). By incorporating this physical knowledge into neural networks, the need for large datasets can be greatly reduced, or even eliminated, along with much of the computational cost. These specialized networks are known as Physics-Informed Neural Networks (PINNs).
In this project, building on previous research by the host team, you will work towards the development and optimisation of PINNs for applications in various problems associated with fluid mechanics, hydraulics and general thermofluids. Examples of applications include:
- fast modelling of floods
- alternative Navier-Stokes solvers
- simulations of various fluid-structure interactions
You will join a diverse, vibrant and growing PINNs research group, with access to one of the UK’s most powerful supercomputers, Iridis, and support from a community of PhD students working in various aspects of Computational Fluid Dynamics and AI. You will also be part of a growing network of international collaborators.
You will receive comprehensive training to ensure the successful completion of your PhD. This includes access to dedicated support at Southampton for High Performance Computing and Computational Fluid Dynamics. Additionally, the host research team, comprising several research students working on PINNs, will provide further guidance and collaboration opportunities.
