Machine Learning for sensing and control of dynamic stall in realistic conditions

Dynamic stall is an aerodynamic phenomenon that can occur on lifting surfaces in extreme conditions, such as gust encounters and rapid manoeuvres, and is characterised by the formation and convection of a distinct, rotating flow structure - the Dynamic Stall Vortex (DSV). This DSV can create excessive excursions in aerodynamic loads which severely degrades vehicle performance and control. Robust detection and disruption of this event will present unique opportunities for the next generation of aerodynamic vehicles to push beyond traditional operational envelopes and attain new levels of performance. These opportunities however present significant challenges in prediction and control, as dynamic stall exhibits strong non-linearities and stochastic properties which have proven extremely difficult to predict despite the significant progress made to date.

This PhD will explore the use of Machine-Learning (ML) approaches for dynamic stall sensing and control to leverage their strengths and capabilities for use in complex systems. Large sets of wind tunnel training data will be obtained by the successful candidate in the University of Bath’s novel Test Rig for Unsteady aeroDYnamics (TRUDY) to explore dynamic stall through a vast range of representative conditions. ML will then be applied to these data to search for new mechanisms, precursory parameters and underlying dynamics that govern the dynamic stall process. This project asks the questions: Can ML accurately detect dynamic stall events with sufficient warning? Is ML able to mitigate these events better than traditional control methods? Are ML methods robust enough to operate in realistic conditions?

The final question is crucial, as robustness implies a certain level of trust in methods and models operating in uncertain conditions, and this project aims to build that trust through the application of Accountable, Reliable and Transparent Artificial Intelligence (ART-AI). Machine Learning methods demonstrate promising performance in many aspects of engineering, but these methods can often lack transparency, offer no performance guarantees, and fail to extrapolate or generalise beyond the data set they were trained on. The accountability, transparency and robustness of ML methods will form a critical aspect of certification for the next generation of aerodynamic vehicles and have a significant impact in shaping public perception and confidence as these enter the market. This project will prioritise these aspects through:

·        extensive validation using TRUDY’s capability of simulating a wide range of extreme, unsteady environments.

·        open-source data sets, methods, and code to facilitate the development of benchmarking procedures and encourage external validation

·        a focus on methods that avoid “black box” solutions for more interpretability and transparency.

This is an exciting opportunity to become an expert in the use of Machine Learning techniques for unsteady aerodynamics, as well as gaining hands-on experience with novel wind tunnel testing equipment. The successful student will also need to consider the implications of their work for certification purposes. Certification is a rigorous process designed to uphold principles of safety and transparency, and it is imperative to pursue methods that adhere to these principles. This project is associated with the UKRI Centre for Doctoral Training (CDT) in Accountable, Responsible and Transparent AI (ART-AI). We value people from different life experiences with a passion for research. The CDT's mission is to graduate diverse specialists with perspectives who can go out in the world and make a difference.

Applicants should hold, or expect to receive, either a First or Upper Second Class Honours degree in a relevant subject or a relevant master’s level qualification. Applications will close when a suitable candidate is found; therefore, early application is recommended.

Formal applications should be accompanied by a research proposal and made via the University of Bath’s online application form. Enquiries about the application process should be sent to [Email Address Removed]. Enquiries about the research project should be sent to Dr Bull.

Start date: 2 October 2023.