Tidal turbines could provide over 10% of the UK’s electricity, but issues with reliability and cost are hampering their commercial viability. Maintenance at sea is expensive, and so remote monitoring is a critical part of lowering the cost of tidal power. Data on turbine performance and mechanical loading is being gathered from full-scale devices currently deployed in test locations around the world, but there is a need to understand the data and to develop methods for diagnosing the early signs of faults and scheduling pre-emptive maintenance to avoid unplanned shutdowns or catastrophic failures.
This PhD will explore the use of artificial intelligence in remote monitoring and fault detection for tidal turbines. While data is available for turbines operating normally in various sea states, failures are rare, and this means that data on failures is sparse or non-existent. AI schemes will therefore be developed for normality modelling. Furthermore, it is difficult to measure turbine behaviour (blade stresses, shaft loads etc) directly; engineers are instead fed secondary information such as nacelle vibration, which is not directly related to the health of any particular component. This ambiguity between measurement and cause motivates the use of neural networks in understanding the data. Training will be achieved using data from a turbine in one site, and validation will be undertaken on a turbine deployed in another location.
Ironically, the sensors themselves cost money and require maintenance, and erroneous data can trigger unwarranted maintenance. The project will therefore establish the most efficient use of sensors and use artificial intelligence to interpret sparse data sets such that condition monitoring can be achieved with minimal measurements and/or redundancy can be achieved.
As more tidal turbines are deployed, their reliability will become critical for grid stability, and the use of these novel condition monitoring algorithms will therefore place great responsibility into the hands of the AI. This PhD will therefore examine the risks of using AI for turbine maintenance scheduling and quantify these risks against the potential benefits – can the AI reduce the risk of unplanned outages by enabling pre-emptive maintenance, or might it cause complacency and lead to catastrophic failures? Can AI allow for longer gaps between scheduled maintenance without compromising turbine availability? Further to this, when the AI detects a fault, it is important that this fault can be traced back to its root cause so that future turbine designs can be improved. This PhD will therefore create transparent AI schemes that can be interrogated by engineers: in parallel with the AI work, low-order models of the hydrodynamics and structural dynamics of the turbine will be developed to link measured outputs with their physical causes.
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, a First or Upper Second Class Honours degree in a relevant subject. You will also need to have taken a mathematics course or a quantitative methods course at university or have at least grade B in A level maths or international equivalent. A Master’s level qualification and programming experience would also be desirable.
Informal enquiries about the project should be directed to Dr Anna Young.
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].
Start date: 2 October 2023.