Generalisable, Physics-Aware Machine Learning for Gas Turbine Engines
About the Project
Gas turbine engines are among the most complex engineered systems in operation today, generating vast quantities of multi-modal data across global fleets. Modern machine learning methodologies are transforming how these engines are monitored and understood. Much like the human body, each engine is a unique asset with its own operational history and characteristics. In medicine, patient-specific modelling has revolutionised healthcare by moving beyond a “one-size-fits-all” approach towards tailored, data-driven precision care. This project applies the same philosophy to the aerospace sector, developing machine learning models that capture the underlying physics of gas turbine engines while being tailored to individual assets, enabling more accurate predictions and more meaningful operational insight.
This PhD will develop generalisable machine learning models for gas turbine engines that can predict system outputs and health across multiple engine types by capturing the underlying physical principles and fine tuning for a specific asset. The project will exploit recent advances in generative modelling and transfer learning, using multi-modal fleet data from Rolls-Royce combined with data that quantifies the environmental and operational context. The aim is to capture underlying gas-turbine physics while enabling rapid specialisation to a new engine type using limited data. The models will be used to monitor degradation and detect faults across the fleet of in-service engines and will be available for new engines as the enter into service.
The PhD programme will address research challenges including:
- Generalisation vs specialisation: How to learn common physical behaviour across different gas turbine engine types while adapting efficiently to individual engine variants without retraining from scratch.
- Adaptation and transfer learning: How to transfer knowledge from mature engine fleets to new or low-data engines using state of the art machine learning techniques such as generative, transfer-learning, meta-learning or hybrid physics–ML approaches.
- Sparse multi-modal data: How to integrate time-series sensor data, flight summary data, maintenance records and environmental data within a unified modelling framework, despite uneven data availability across engines.
The research will be guided by a supervisory team with expertise in both machine learning and industrial applications. This 4-year PhD offers the opportunity to pioneer original ML research with a focus on physics-aware AI and transfer learning. You will be mentored throughout the process, ensuring your research is backed by technical resources and expertise. You will develop machine learning models for gas turbine engines using data gathered from across the Rolls Royce civil aerospace fleet, containing data from hundreds of engines over many years of operation. Models will be extended by investigating generative approaches suited to complex, multi-modal engineering data. You will validate model forecasts against real Rolls-Royce fleet data and develop and test methods for rapid adaptation to new engine types using limited historical data.
The PhD is based at the Rolls-Royce University Technology Centre (UTC) at the University of Sheffield, providing a unique opportunity to work closely with leading academics and industrial partners. Graduates from the Sheffield UTC are well placed for careers in academia or industry, including research and advanced engineering roles in aerospace, energy and AI.
Industrial Sponsorship
The student will be based in the University Technology Centre in Controls, Monitoring and Systems Engineering – part of the school of Electrical and Electronic Engineering at the University of Sheffield. Access to industrial datasets as well as appropriate computing infrastructure will be provided. There will be regular contact with industrial collaborators at Rolls Royce throughout the project, including an on-site placement and engagement with the Rolls Royce Doctoral Network.
Applicant qualifications:
- An Undergraduate or Master’s degree in Engineering, Mathematics, Computer Science, Physics, or a closely related discipline.
- Proficiency in programming using Python, MATLAB, or a similar language.
- A demonstrable interest in Machine Learning (ML), particularly its application to real-world physical systems.
- Prior knowledge of gas turbine engines is not required.
Funding Notes
The PhD is fully funded for Home students through the EPSRC IDLA scheme and supported by Rolls-Royce, offering a unique opportunity to gain experience in both academia and industry. The funding includes (i) a tax-free annual stipend (£20,780 in 2025/26) plus a stipend enhancement of £8,000, totalling £28,780 per annum, (ii) Tuition fees at the home rate, (iii) generous funding for consumables and travel.
Find Your Best Opportunity
Tell them AcademicJobs.com sent you!
