Reacting-Flow Control PhD: Using Machine Learning to Anticipate Extreme Events

Reacting-Flow Control PhD: Using Machine Learning to Anticipate Extreme Events

Research cut-off: 6 November 2026
Assessment: Promising with material uncertainties

Quick Decision

AE0092 is a 3.5-year Imperial Aeronautics studentship on hybrid scientific machine learning and reacting-flow control. The current project page proposes combining a differentiable reacting-flow solver with a machine-learning controller that uses short-term state information to mitigate rare flow extremes. Hydrogen and sustainable aviation fuels are the motivating context; spray-water injection and electrically driven fuel-droplet control are listed as examples of possible actuation concepts, not fixed operating instructions or a promise that both will be used. Dr Nguyen Anh Khoa Doan is the named supervisor. Imperial’s current staff profile records his appointment in Aeronautics from January 2026, following a Delft role through 2025. (AE0092 project page; Nguyen Anh Khoa Doan profile)

The official page states 3.5 years, full tuition coverage for Home, EU and international students, and £23,805 annual tax-free stipend. It gives a 4 January 2027, 23:59 UK deadline and a start window from 1 August 2027 to 1 July 2028. Applicants need a first-class MEng/MSci or equivalent in Aeronautical/Mechanical Engineering, Applied Mathematics, Computer/Data Science, or Applied/Computational Physics; experience in fluid mechanics, turbulence, CFD, machine learning or reacting flows is sought. AE0092 project page

The strongest fit is an applicant who can connect fluid mechanics or combustion context with machine learning, differentiable simulation, optimization or control. Key points to clarify are the exact reacting-flow problem, solver and controller boundary, how rare-event performance will be validated, what compute and data access are confirmed, and the supervision/CONTEXT relationship. Similar Imperial-linked work offers strong research context, but several papers use lower-order or non-reacting model systems and cannot be treated as evidence that AE0092’s controller has been built or validated. (CONTEXT project; Luca Magri publications; Imperial data-driven engineering research; thermoacoustic digital-twin paper; thermoacoustic adjoint-sensitivity paper; data-assimilated model-informed reinforcement-learning paper; extreme-event control echo-state-network paper; quantum reservoir computing paper on chaotic dynamics)

Research Problem and Likely Doctoral Work

Scientific question

The project addresses rare, large changes in reacting-flow behaviour that may matter in sustainable propulsion systems. The advertised challenge is that turbulence, nonlinear reactions and multiscale interactions make the early evolution difficult to anticipate. AE0092 proposes to combine a differentiable reacting-flow solver and a machine-learning controller. The solver is intended to provide short-term state dependence to inform the controller’s mitigation decision. The project page names hydrogen flashback and thermoacoustic instability as examples motivating the work, and notes possible actuator concepts. It does not provide an exact combustor configuration, solver package, dataset, control objective, approved experimental activity or quantitative success criteria. AE0092 project page

Methods and research questions

At a high level, the work is likely to connect reacting-flow simulation, differentiable computation, data-driven state estimation, control optimization and validation on selected flow cases. A proposal or interview discussion can ask what is differentiated, which state information is available to the controller, how uncertainty and rare-event imbalance will be treated, and what independent checks will establish whether an apparent mitigation is robust. These are questions to clarify, not additional requirements or a prescribed implementation.

The sources support a research-scale computational project. Imperial says high-fidelity simulations and College HPC will be available as opportunities; facility pages describe central and departmental computing. Access policies require registration, and a public facility does not establish a dedicated GPU allocation, compute hours, wall-clock priority or particular software. (Imperial Research Computing Service; Aeronautics computing resources)

Recent Research and Field Direction

• Nóvoa et al. (2024), “A real-time digital twin of azimuthal thermoacoustic instabilities.” Integrates a low-order model with sparse experimental observations to infer state, parameters and model bias in real time. It is close group context for hybrid modelling in combustion, but is not AE0092’s controller or mitigation demonstration. thermoacoustic digital-twin paper
• Ozan and Magri (2024), “Data-driven computation of adjoint sensitivities…” Uses data-driven models to estimate sensitivities for thermoacoustic dynamics and explores optimization. It informs the differentiable/sensitivity theme but does not validate this project's reacting-flow solver. thermoacoustic adjoint-sensitivity paper
• Ozan et al. (2025), “Data-assimilated model-informed reinforcement learning.” Combines a low-order model, sequential data assimilation and reinforcement learning for control with partial observations. Its demonstration is a mathematical chaotic flow, not reacting combustion. data-assimilated model-informed reinforcement-learning paper
• Racca and Magri (2023), “Control-aware echo state networks…” Studies prediction plus control for extreme-event suppression in a chaotic-turbulent flow model. It is conceptually adjacent but a preprint/workshop paper, not a combustor result. extreme-event control echo-state-network paper
• Ahmed et al. (2024), “Prediction of chaotic dynamics and extreme events…” Explores reservoir computing for forecasting rare events and chaotic dynamics in mathematical flow systems. It studies prediction rather than reacting-flow control. quantum reservoir computing paper on chaotic dynamics

These papers illustrate active work on scientific ML for fluid prediction, data assimilation, sensitivities and control. The field challenge is translating methods from tractable models to high-dimensional, stiff, multiphysics simulations while retaining stable control, credible validation and clear uncertainty. The project's potential contribution is therefore plausible, but the project page does not establish that a real-time deployment, experimental combustor or safety claim is part of the PhD.

School, Programme and Research Environment

The doctorate is in Imperial Aeronautics, with an advertised 3.5-year full-time studentship. The department lists groups in scientific machine learning/data assimilation, flow control, turbulence and computational fluid dynamics. Imperial reports that Aeronautics had the UK's highest proportion of world-leading researchers in REF 2021; this is a department-reported research-assessment statement, not an individual project rating or a current global rank. (Aeronautics PhD application process; Aeronautics PhD programme; Aeronautics research groups; Aeronautics research and REF 2021)

The project page places the work in the CONTEXT research setting and names opportunities to interact with Luca Magri, George Rigas and Andrea Nóvoa. CONTEXT’s European Commission project description focuses on scientific machine learning for predicting and controlling extreme events in flows. The precise grant work-package, student reporting line and collaboration schedule are not public. (CONTEXT project; Luca Magri publications)

Imperial describes departmental Linux/GPU facilities, central HPC access and research-software support. The department's facilities include experimental aerodynamics resources, although AE0092 does not promise a particular experimental phase. Ask whether the work will use simulations only, share a laboratory test case, or have access to experimental data; do not assume an operational or industrial deployment. (Aeronautics research facilities; Imperial Research Computing Service; Aeronautics computing resources)

The saved position description also advertises Early Career Researcher Institute workshops on research communication, computing and data science, and professional progression, plus the opportunity to participate in international conferences. These are opportunities; specific attendance, timing and outcomes are not guaranteed. Early Career Researcher Institute professional-development opportunities

Supervisor and Lab Assessment

The named supervisor’s current Imperial profile lists an Aeronautics appointment from January 2026, so a historic Delft affiliation should not be treated as evidence that he is absent from Imperial. The official studentship page and current staff profile both identify the Imperial role. The advertised collaboration links the project to CONTEXT and to department experts in turbulence and machine learning. Public sources do not state how daily supervision will be divided across these researchers or whether any external team will co-supervise. (AE0092 project page; Nguyen Anh Khoa Doan profile; CONTEXT project; Luca Magri publications; Imperial data-driven engineering research)

Imperial-linked publications provide a coherent foundation in hybrid models, state estimation, sensitivities and extreme-event prediction. They are adjacent evidence: the 2024 digital twin concerns thermoacoustic estimation, and the 2025 reinforcement-learning paper uses a mathematical flow. They do not demonstrate success on the exact AE0092 reacting-flow cases.

Funding, Application and London Costs

The official page promises full tuition fees and an annual tax-free stipend of £23,805 for Home, EU and international students, for 3.5 years. Confirm the award wording and any stipend review for the selected 2027/28 entry term. Imperial's 2026 student estimate for average London living costs is £1,821–£1,893 per month; the listed lower and upper ranges are £1,264 and £2,118. Accommodation choices drive much of this spread. The figures exclude tuition and do not represent an employee budget. £23,805 divided by 12 is roughly £1,984 per month, a rough comparison only. (AE0092 project page; Imperial 2026 London living-cost guide; Imperial 2026–27 postgraduate tuition fees)

Apply using Imperial’s online system with a CV, personal statement, transcripts and contact details for two referees. Select “Aeronautics Research (PhD),” “Aero,” topic AE0092, Dr Nguyen Anh Khoa Doan and Autumn 2027/8 or Spring 2027/8. A maximum of two Imperial PhD project applications is allowed in the academic year; two applications require different entry terms. References are requested after submission and must arrive by 23:59 on 5 February 2027 UK time. Department guidance says applicants to advertised studentships do not need a separate research proposal; the current project form and checklist control. (AE0092 project page; Aeronautics PhD application process)

Applicant Fit and Preparation

Formal requirement versus useful evidence

Formal academic eligibility is a first-class MEng/MSci or higher, or international equivalent, in Aeronautical/Mechanical Engineering, Applied Mathematics, Computer/Data Science, or Applied/Computational Physics. The project seeks experience in fluid mechanics, turbulence, CFD, machine learning or reacting flows. The page does not require every listed experience area.

Useful evidence could include a CFD or scientific-ML project, reproducible code, numerical validation, control/optimization work, or a thesis involving nonlinear systems. Clearly distinguish a class project from research experience and identify what you personally implemented. A personal statement can explain why solver-plus-controller coupling matters and what kind of validation question interests you, without proposing unsafe operating details.

Career Context, Risks and Next Steps

EngineeringUK estimates 6.3 million people in UK engineering and technology roles. HESA reports 87% of 2023/24 graduates in work or further study 15 months after leaving, but neither source forecasts doctoral outcomes in combustion or scientific ML. Transferable skills include CFD, numerical methods, machine learning, optimization, simulation software and scientific computing. Potential sectors include aerospace, energy, industrial simulation, research labs and academia; no placement is promised. (EngineeringUK workforce briefing; HESA Graduate Outcomes 2023/24)

Before applying

• Confirm entry-term selection, application checklist, personal-statement expectations and two-reference timing.
• Ask whether the advertised research scope is tied to a particular CONTEXT work package and whether project data are already defined.

At interview

• Clarify the target reacting-flow case, the differentiable solver and the controller’s high-level inputs/outputs.
• Ask what constitutes a measurable precursor and how rare-event performance, robustness and model error will be assessed.
• Confirm simulation versus experimental data use, compute needs and the division of supervision with named collaborators.

Before accepting

• Obtain written confirmation of tuition coverage, stipend, duration, start date and any annual funding conditions.
• Confirm access to the required solver, HPC/GPU resources, data and supervision, and whether external collaboration or any actuator study is actually included.

Evidence and Sources

Official project, admissions and programme

• AE0092 project page
• Aeronautics PhD application process
• Aeronautics PhD programme

Research environment, supervisors and facilities

• Aeronautics research groups
• Aeronautics research facilities
• Imperial Research Computing Service
• Aeronautics computing resources
• Aeronautics research and REF 2021
• Nguyen Anh Khoa Doan profile
• CONTEXT project
• Imperial data-driven engineering research

Recent papers and field evidence

• Luca Magri publications
• thermoacoustic digital-twin paper
• thermoacoustic adjoint-sensitivity paper
• data-assimilated model-informed reinforcement-learning paper
• extreme-event control echo-state-network paper
• quantum reservoir computing paper on chaotic dynamics

Funding, costs and career context

• Imperial 2026 London living-cost guide
• Imperial 2026–27 postgraduate tuition fees
• EngineeringUK workforce briefing
• HESA Graduate Outcomes 2023/24
• Early Career Researcher Institute professional-development opportunities
• South Kensington campus