Electrical Engineering and AI PhD: Machine Learning Digital Twins for High-Torque Electric Propulsion

Electrical Engineering and AI PhD: Machine Learning Digital Twins for High-Torque Electric Propulsion

Quick Decision

This doctorate combines multi-scale motor modelling, scientific machine learning and controlled experiments. Its central task is to predict temperature, torque capability and efficiency when electric propulsion motors operate at low speed and high torque under changing, sometimes multi-hour duty cycles. The plan includes safety-constrained dynamometer experiments at Exeter and work with an industrial collaborator, so applicants should be ready to move between model design, measurement and engineering interpretation.

The project is based at the University of Exeter’s Department of Engineering, Streatham. Dr Devika Koonthalakadu Baby is lead supervisor and Dr Prathyush P Menon is additional supervisor. The EPSRC Doctoral Landscape Award offers up to 3.5 years of full-time support, tuition fees and £5,000 in Research Training Support Grant (RTSG). The detailed funding section gives £21,805 per year as the current 2026/27 stipend and says the University covers the international-fee difference for successful international awardees. The award summary calls £21,805 a 2027/28 rate, however; that future amount is not confirmed. Applications are due at midday on 8 December 2026. Before applying, confirm which reference deadline applies, how the project-specific degree requirement will be assessed, and what data and facility access the student will receive.

Research Problem and Doctoral Work

What Is Unsolved and Why It Matters

Marine vessels, mining equipment and off-highway machinery can require sustained high torque at low rotational speed, for example during bollard pull, winching or heavy excavation. Fast electromagnetic response and saturation interact with load-history effects and slower heating. Temperature then changes resistance, efficiency and available torque. Repeatedly testing a high-power motor near its thermal limits is expensive and may damage insulation or permanent magnets. Detailed physics models can be too slow across long, variable cycles, while conventional ML may not generalise beyond its training conditions, especially for long-horizon temperature prediction.

Work, Methods, and Potential Outputs

The advertised work investigates a multiscale, physics-informed scientific machine-learning digital twin that predicts thermal trajectories, torque capability and efficiency over time. Three candidate model families map to distinct time scales:

• Deep neural networks for nonlinear electromagnetic relationships and saturation.
• Recurrent or neural state-space models for transients and load-history dependence.
• Neural ordinary differential equations for continuous-time thermal accumulation over multi-hour horizons.

The doctoral researcher is expected to design safety-constrained dynamometer experiments at Exeter’s Centre for Future Clean Mobility (CFCM), collect high-frequency electrical, mechanical and thermal measurements, and use the data to train and validate the twin. A central question is whether carefully chosen, non-destructive tests can support reliable predictions for conditions the motor has not directly experienced. Physics-based constraints, data-efficient experiment design and uncertainty quantification are part of the stated approach. The intended outputs are fast, uncertainty-aware operating-profile predictions to inform motor sizing and duty-cycle decisions; commercial deployment is not guaranteed.

Recent motor-twin studies offer useful comparators. A 2026 study combines finite-element EM/thermal snapshots, proper orthogonal decomposition (POD) and a neural surrogate, then updates the model from measurements. A 2025 thermal-neural-network study uses measured traction-machine data and reports a best validation MSE of 0.41 °C², although the machine and oil-spray cooling differ. Another 2025 paper uses a multiphysics-informed DeepONet for a permanent-magnet synchronous motor. These methods may inform comparisons, but the project does not state that POD, DeepONet or any particular baseline will be adopted.

One possible evaluation, to discuss with the supervisors rather than assume, is to compare physics-based or reduced-order, temporal data-driven and hybrid models on held-out duty cycles. Relevant measures could include prediction error, calibration, safe-envelope coverage and compute time. The public materials do not specify a motor model or rating, dataset volume, acceptance thresholds, software stack, publication targets or detailed test schedule.

School, Programme and Field Context

The degree-awarding institution is the University of Exeter. The listed route is a PhD in Engineering in the Department of Engineering, at Harrison Building on the Streatham campus. No joint degree or second degree-awarding partner is stated. Exeter Engineering’s research themes include computational modelling and data science, clean-energy engineering, engineering AI, robotics/control and digital twins, a relevant environment for a project crossing physical modelling, data methods and experimental engineering.

The most specific research-quality evidence is REF 2021 Engineering (UoA 12): 32% of submitted activity was rated 4-star and 60% 3-star. Exeter’s QS 2027 rank of 136 and THE 2027 rank of 157 are overall university rankings, not subject rankings or project-quality ratings. The project names CFCM as its experimental setting and advertises interdisciplinary training in electric machines, scientific machine learning, thermal modelling and digital-twin validation. It does not specify formal modules, cohort coursework or a detailed training schedule. Facility pages describe test capability, but a student’s access to a specific motor, instruments, test time or technical support still needs confirmation.

The UK’s 2025 maritime decarbonisation strategy targets zero fuel-lifecycle greenhouse-gas emissions from domestic maritime activity by 2050, with at least 30% reduction by 2030 and 80% by 2040 relative to 2008. This gives the application domain policy relevance. The research itself concerns motor modelling and testing; it does not establish that a particular vessel or propulsion technology will be commercialised.

Supervisor and Lab Assessment

Supervisor Research Directions

The project lists Dr Devika Koonthalakadu Baby as lead and Dr Prathyush P Menon as additional supervisor. The University directory identifies Baby as a Lecturer in Mechanical Engineering. Menon’s profile describes control and autonomy research and includes earlier applied work on DC motor design for a marine drivetrain. Recent team publications are adjacent rather than identical to the proposed motor twin: a 2023 study models heavy-goods-vehicle electrification strategies; a 2024 co-authored paper examines observer-based power allocation for a simulated shipboard microgrid; and a 2025 paper develops Gaussian-process classification and active sampling to map safe/unsafe regions in uncertain systems. The latter connects conceptually to uncertainty-aware operating-envelope estimation, but its examples are not motor dynamometer tests.

Recent Important Publications

• Rohith et al. (2023), “Sustainable Heavy Goods Vehicle Electrification Strategies for Long-Haul Road Freight Transportation.” Compares electrification strategies through road-freight modelling, supporting experience with heavy-duty applications. It is not a marine motor thermal twin.
• Rinaldi, Baby and Menon (2024), “Optimal Observer-Based Power Imbalance Allocation for Frequency Regulation in Shipboard Microgrids.” Combines a super-twisting observer, online optimisation and control for a simulated shipboard microgrid. It is numerical simulation rather than vessel or motor testing.
• Wang et al. (2025), “Safety exploration using Gaussian process classification for uncertain systems.” Uses GP classification and active sampling to estimate safe/unsafe parameter regions, including a spacecraft example. Transfer to motor tests needs validation.

Lab, Resources, and Collaboration Environment

The project promises safety-constrained dynamometer work at CFCM and onsite collaboration with Ecomar Innovations, a marine power-systems company. The candidate is expected to test the framework against industrial hardware and contribute to thermal-management research. The facility and company pages establish an applied clean-mobility setting; the project PDF provides the clearest evidence for promised student activities. It does not specify the Ecomar site, placement length, travel support, exact motor access, data ownership, confidentiality or publication approval. Ask about those conditions, the supervision schedule and available technical support.

Funding, Application, and Living Cost

The EPSRC DLA offers up to 16 fully funded doctoral places for September 2027. The stated package is up to 3.5 years full-time, an index-linked stipend, tuition fees and £5,000 RTSG over 3.5 years. The detailed award text and UKRI’s table identify £21,805 as the 2026/27 minimum stipend; the award page summary labels it as the 2027/28 rate. The 2027/28 amount is therefore unconfirmed until written into an offer or confirmed by admissions.

International candidates may compete for a limited number of places, up to 30% of the cohort. Exeter says successful international awardees’ full tuition is covered through the studentship and a University fee waiver. Visa, healthcare surcharge and relocation expenses are not covered. The RTSG is research support, not personal maintenance.

The application deadline is midday Tuesday 8 December 2026. Applications are project-led; submit a separate application for each selected project, up to three. Scheme materials request a personal statement naming the project, CV, transcript(s), two references and English-proficiency evidence when applicable. Documents must be in English and academic qualifications require certified translations. Use the project-specific application route in the relevant PDF so the application reaches the correct panel.

Reference timing conflicts across the official Exeter pages. The project-specific Ref 6014 award page says references must arrive by 8 January 2027, while the central DLA page says 15 February 2027. The 6014 project PDF itself does not list a reference date. Treat 8 January as the safer planning deadline and ask PGR Admissions to confirm which date applies and whether references may follow the 8 December application. The central DLA page is also internally inconsistent about interviews: its overview says February 2027 interviews and early March offers, while a later paragraph gives February and March 2026; the Ref 6014 page lists likely general interviews between 15 and 24 February 2027. Confirm the current schedule.

Exeter estimates basic living costs at £1,300–£2,000 per month. At the current reference stipend, £21,805 averages about £1,817 per month before considering payment timing or deductions, so a gap is possible near the upper end, before visa/IHS and arrival costs. Current postgraduate accommodation examples range from £124.74 per week for a standard self-catered room to £257.04 per week for a studio; these are examples, not guaranteed 2027/28 prices or availability.

Applicant Fit and Preparation

Required Background and Skills

The project-specific PDF asks for a First or Upper Second Class UK Honours degree, or international equivalent, in electrical/electronic engineering, mechanical engineering, control engineering, mechatronics, applied mathematics, computer science or a closely related discipline. It also asks for a strong quantitative background and interest in electric propulsion, modelling, control, machine learning or data-driven engineering. MATLAB, Python or similar computational-tool experience is expected.

The central DLA page says a Lower Second may be considered at scheme level if accompanied by a master’s degree or significant relevant non-academic experience, while warning that projects may set specific requirements. Since this project states its own degree-class criterion, ask whether the general exception applies instead of assuming it does.

Preferred Experience and Suggested Preparation

The project does not name a required programming language beyond MATLAB, Python or similar, nor require prior marine-industry experience. Applicants can strengthen their case with a reproducible modelling/control project, evidence of quantitative reasoning and experimental/data analysis, and a concise explanation of how physics-informed learning could help predict coupled motor behaviour. Awareness of distribution shift, calibration, sparse thermal coverage and unsafe extrapolation is useful preparation. These are suggestions, not extra eligibility requirements.

Career Market Context

Adjacent career routes include electric-drive and powertrain R&D, controls and scientific-ML engineering, thermal systems, simulation and digital twins, clean propulsion, and industrial decarbonisation. Skills England projects 47,000 additional workers (13%) in priority advanced-manufacturing occupations over 2025–35, alongside replacement demand. Maritime UK and IET identify relevant skills in propulsion efficiency, autonomous control, AI and automation, simulation, data and software, and sustainability; the UK maritime decarbonisation strategy provides further policy context. These are broad sector signals, not a placement or salary forecast for this PhD and not a job guarantee. Demand will depend on investment, infrastructure, policy and technology costs. (Skills England advanced-manufacturing assessment, Maritime UK skills report, IET skills survey, UK maritime decarbonisation strategy)

Next Actions and Questions to Verify

Before Applying

1. Ask the lead supervisor whether the target motor and test conditions are selected, and how the first year is divided between modelling, experiments and Ecomar work.
2. Ask PGR Admissions whether references for project 6014 are due 8 January or 15 February 2027, and whether they can follow the 8 December application.
3. Confirm the 2027/28 stipend, indexation, international fee-gap coverage and permitted RTSG expenses.
4. Ask whether the scheme-wide Lower Second exception applies to this project’s stated academic requirement.
5. Confirm the correct project-specific application link and current English-evidence/certified-translation rules.

At Interview

• Which motor model, rated power, cooling system and load profiles will be tested? Which work is at Exeter and which at Ecomar?
• What are the safety limits, sensor channels, sample rates, planned tests and data-access conditions?
• Which physics-only, reduced-order and ML baselines or held-out duty cycles will be used? How will calibration and safe-envelope coverage be judged?
• What placement schedule, onsite supervision, confidentiality rules and publication terms apply?

Before Accepting an Offer

Obtain written confirmation of stipend year/rate, duration, fees and international waiver, RTSG, dynamometer and industrial-hardware access, data/publication rights, supervision and any travel or visa support. The applied setting is attractive, but these details determine whether the training and financial package match your expectations.

Evidence and Sources

Project and funding

• Project 6014 description
• EPSRC DLA award page, Ref 6014
• EPSRC DLA eligibility and application information
• UKRI stipend and fee guidance
• Another 2027 DLA project page also listing 8 January references

University, field and facilities

• Exeter living-cost estimates
• Exeter postgraduate accommodation examples
• University of Exeter rankings
• REF 2021 results
• Engineering research themes
• Computational Modelling and Data Science
• Engineering academic staff
• Prathyush P. Menon profile
• Centre for Future Clean Mobility team
• Clean propulsion dynamometer facility
• Ecomar Innovations
• UK Maritime Decarbonisation Strategy

Careers and recent research

• Skills England advanced-manufacturing assessment 2026
• Maritime UK Skills for Green Jobs Report 2024
• IET 2025 engineering and technology skills survey
• Sustainable HGV electrification strategies (2023)
• Shipboard microgrid frequency regulation (2024)
• Safety exploration with Gaussian-process classification (2025)
• Thermal neural networks for traction machines (2025)
• Physics-driven digital twin for transient thermal systems (2024)
• Real-time electromagnetic-thermal motor digital twin (2026)
• Motor twin temperature prediction over a drive cycle (2025)
• Multiphysics-informed DeepONet for a PMSM (2025)