Physics-Informed Machine Learning with Applications in Hydrogen Fuel Propulsion Systems

What are you going to do?

Background

The Airbus A380 ZEROe demonstrator aircraft used for testing hydrogen-powered propulsion (source: https://simpleflying.com/rolls-royce-patent-hydrogen-electric-engine-systems-explained/)

Hydrogen fuel systems are pivotal to the future of aviation, offering a pathway to net-zero carbon emissions. However, their adoption introduces substantial maintenance challenges, as operating with hydrogen involves harsher conditions, stricter safety requirements, and more intricate system behaviour than conventional fuels. Prognostics and Health Management (PHM) can address these challenges by enabling a shift towards condition-based, predictive, and prescriptive maintenance strategies, reducing unplanned downtime, extending component life, and lowering operational costs while improving safety. Realising these benefits, however, depends on the quality of the underlying predictive models, and current approaches face important limitations: purely data-driven models struggle to generalise across operating regimes and require large volumes of representative data, while traditional physics-based models are computationally costly and constrained by simplifying assumptions. Physics-Informed Machine Learning (PIML) offers a promising alternative by embedding physical laws into data-driven frameworks, enabling more accurate, generalisable, and efficient prediction of component behaviour and degradation. This project therefore aims to explore the application of PIML to hydrogen fuel propulsion systems

The assignment will include the following tasks:

• Preliminary assessment and identification of which sub-systems and components of hydrogen fuel propulsion systems are most suitable for PIML-based modelling, given the current state of the technology and data availability. For this, we have identified several possible components and corresponding datasets (see below).

• A literature study on PIML and maintenance of hydrogen fuel propulsion systems, including an analysis on how the PIML approach can be used within a PHM context to support the monitoring and health management of hydrogen propulsion systems.

• Designing and implementing a PIML framework to model key phenomena in the selected sub-systems or components.

• Validating the framework on real-world or high-fidelity simulation datasets and benchmarking it against state-of-the-art methods.

Result

The final outcomes of this assignment will be:

• A structured assessment, as part of a literature study, identifying the hydrogen fuel propulsion sub-systems or components most suitable for PIML-based modelling, together with a justified selection of the target application.

• A PIML-based method and/or workflow capable of predicting the behaviour and/or degradation of the selected hydrogen fuel propulsion system, sub-system, or component, accounting for interpretability and uncertainty quantification.

• A technical thesis report, following the guidelines of your faculty, describing the approach, results, and conclusions of the work.

• Optionally a conference and/or journal publication.

Available datasets and models

Relevant datasets to this research problem are:

• Proton Exchange Membrane fuel cell (IEEE PHM Data Challenge 2014)

• Hamidi S, Haghighi S, Askari K. Dataset of Standard Tests of Nafion 112 Membrane and Membrane Electrode Assembly (MEA) Activation Tests of Proton Exchange Membrane (PEM) Fuel Cell. ChemRxiv. 2020; doi:10.26434/chemrxiv.11902023

• Zuo, Jian, Hong Lv, Daming Zhou, Qiong Xue, Liming Jin, Wei Zhou, Daijun Yang, and Cunman Zhang. "Long-term dynamic durability test datasets for single proton exchange membrane fuel cell." Data in Brief 35 (2021): 106775.

Relevant physics-based models:

• McKay, Denise A., Jason B. Siegel, William Ott, and Anna G. Stefanopoulou. "Parameterization and prediction of temporal fuel cell voltage behavior during flooding and drying conditions." Journal of Power Sources 178, no. 1 (2008): 207-222.

Notice that additional data can also be obtained through lab tests at NLR facilities.

Duration

Standard duration of a thesis at your faculty.

What do we expect from you?

• You are an MSc student in Aerospace Engineering, Physics, Computer Science or a similar master.

• You have experience with programming in Python and packages for machine learning within Python, for instance PyTorch.

• You have completed a Machine Learning or AI course (e.g., DSAIT4005, DSAIT4115, AE2224-II)

• You have a solid physics/engineering background that allows you to understand the physics of failure models.

What do we offer you?

• You are an MSc student in Aerospace Engineering, Physics, Computer Science or a similar master.

• You have experience with programming in Python and packages for machine learning within Python, for instance PyTorch.

• You have completed a Machine Learning or AI course (e.g., DSAIT4005, DSAIT4115, AE2224-II).

• You have a solid physics/engineering background that allows you to understand the physics of failure models.

Your new working environment?

Royal NLR has been the ambitious research organisation with the will to keep innovating for over 100 years. With that drive, we make the world of transportation safer, more sustainable, more efficient and more effective. We are on the threshold of breakthrough innovations. Plans and ideas start to move when these are fed with the right energy. Over 800 driven professionals work on research and innovation. From aircraft engineers to psychologists and from mathematicians to application experts.

Our colleagues are happy to tell you what it’s like to work at NLR.

Recommended Literature

On Physics-Informed Machine Learning:

§ Keith, Brendan, Thomas O’Leary-Roseberry, Benjamin Sanderse, Robert Scheichl, and Bart van Bloemen Waanders. "Scientific machine learning: a symbiosis." Foundations of Data Science 7, no. 1 (2025): i-x.

§ Hao, Zhongkai, Songming Liu, Yichi Zhang, Chengyang Ying, Yao Feng, Hang Su, and Jun Zhu. "Physics-informed machine learning: A survey on problems, methods and applications." arXiv preprint arXiv:2211.08064 (2022).

§ Fink, Olga, Vinay Sharma, Ismail Nejjar, Leandro Von Krannichfeldt, Sergei Garmaev, Zepeng Zhang, Amaury Wei et al. "From Physics to Machine Learning and Back: Part I-Learning with Inductive Biases in Prognostics and Health Management." Reliability Engineering & System Safety (2026): 112213.

§ Thiyagalingam, Jeyan, Mallikarjun Shankar, Geoffrey Fox, and Tony Hey. "Scientific machine learning benchmarks." Nature Reviews Physics 4, no. 6 (2022): 413-420.

§ Psaros, Apostolos F., Xuhui Meng, Zongren Zou, Ling Guo, and George Em Karniadakis. "Uncertainty quantification in scientific machine learning: Methods, metrics, and comparisons." Journal of Computational Physics 477 (2023): 111902.

On Hydrogen Fuel Propulsion Systems:

§ Tiwari, Saurav, Michael J. Pekris, and John J. Doherty. "A review of liquid hydrogen aircraft and propulsion technologies." International journal of hydrogen energy 57 (2024): 1174-1196.

§ Fard, Majid T., JiangBiao He, Hao Huang, and Yue Cao. "Aircraft distributed electric propulsion technologies—A review." IEEE Transactions on Transportation Electrification 8, no. 4 (2022): 4067-4090.

§ Massaro, Maria Chiara, Roberta Biga, Artem Kolisnichenko, Paolo Marocco, Alessandro Hugo Antonio Monteverde, and Massimo Santarelli. "Potential and technical challenges of on-board hydrogen storage technologies coupled with fuel cell systems for aircraft electrification." Journal of Power Sources 555 (2023): 232397.

Are you interested?

Send your application, together with your Motivation letter and CV and we will contact you as soon as possible. For more information about the assignment contact Lisandro Jimenez ([email protected]) or Paul Stuiver ([email protected]).