PhD student in nuclear fuel performance modelling using machine learning Full-time NEW

PhD Student In Nuclear Fuel Performance Modelling Using Machine Learning

Are you interested in working with nuclear fuel modelling, machine learning and statistical methods as part of a close collaboration between Uppsala University, Westinghouse, and Vattenfall? Would you like to contribute to the development of fast, robust and reliable computational methods for current and future nuclear energy systems, with the support of competent and friendly colleagues in an international environment? Are you looking for an employer that promotes a sustainable working environment and offers secure, favourable working conditions? We welcome you to apply for a PhD student position at Uppsala University.

The Department of Physics and Astronomy, Division of Applied Nuclear Physics at Uppsala University conducts research and education in applied nuclear physics and nuclear engineering. The research includes modelling and simulation of nuclear fuel behaviour in reactors, with particular emphasis on statistical methods, uncertainty quantification and machine learning to improve the accuracy and computational efficiency of fuel-performance analyses. The work contributes to the development of safe, reliable and competitive nuclear energy systems for both existing and future reactors, including small modular reactors.

As a PhD student, you will be part of a research group working on nuclear fuel performance, uncertainty quantification and advanced computational methods for nuclear engineering applications. The group combines physics-based fuel-performance codes with statistical model calibration, machine learning and data analysis. The research environment is international and interdisciplinary, with close links between fundamental method development and technically relevant applications.

The project is carried out within the competence centre ANItA (Academic-industrial Nuclear technology Initiative to Achieve a sustainable energy future). You will be part of a joint academic and industrial research environment, with supervision from both Uppsala University and industrial partners. Through close collaboration with industry, you will have the opportunity to develop research methods with clear industrial relevance and contribute to knowledge exchange between academia and industry.

Reliable and fast fuel-performance calculations are important for both current reactors and future designs, including small modular reactors (SMRs). The project builds on research conducted within CaNel – Calibration of Fuel Performance Codes, where methods based on statistical calibration, uncertainty quantification and machine-learning-based surrogate modelling have been developed for nuclear fuel-performance simulations. You will further develop these methods to enable faster and more accurate predictions with well-quantified uncertainties.

The project includes calibration against time-dependent and axially resolved measurement data, propagation of uncertainties between coupled sub-models, and the development of temporal machine-learning-based surrogate models. The models will accommodate variations in, for example, fuel type, enrichment and gadolinia content, be evaluated outside their training domain, and be applied to a broad range of fuel designs and operating conditions. They will also be demonstrated in industrially relevant applications, such as predicting cladding hoop stress and assessing the risk of PCI-related fuel damage.

Duties The duties mainly consist of doctoral studies, where you will conduct research within the project and take courses within the doctoral education programme. The work includes development, implementation, and evaluation of statistical and machine-learning-based methods for calibration and fast simulation of nuclear fuel performance.

The duties include:

• developing calibration methods that incorporate time-dependent and axially resolved measurement data, • developing methods for propagating uncertainties between calibrated and coupled sub-models, thereby avoiding double counting of effects, • developing and evaluating temporal machine-learning models for sequence-to-sequence prediction of fuel behaviour, • extending surrogate models to accommodate multiple fuel types, enrichment levels and gadolinia concentrations and, where possible, to predict uncertainties, • generating and analysing training and validation data covering a broad range of power histories and operating conditions, and assessing model generalisation beyond the training domain, • demonstrating the surrogate models in industrially relevant applications, for example prediction of cladding hoop stress and assessment of PCI-failure risk, • implementing, testing and documenting computational tools, for example in Python, and contributing to reproducible computational workflows, • compiling, publishing and presenting research results in international journals and conferences and participating in the research group's seminars, project meetings and other scientific activities.

Teaching and other departmental duties may be included, up to a maximum of 20 percent of full-time employment.

Requirements To meet the entry requirements for doctoral studies, you must

• hold a Master's degree in engineering physics, nuclear engineering, applied physics, energy engineering, computational science, applied mathematics, statistics, machine learning or another area relevant to the project, or
• have completed at least 240 credits in higher education, with at least 60 credits at Master's level, including an independent project worth at least 15 credits, or
• have acquired substantially equivalent knowledge in some other way.

The position also requires:

• good knowledge of physics, numerical methods, statistics and/or machine learning,
• good programming skills, for example in Python, Julia, C++ or equivalent,
• good ability to work independently and in a structured manner,
• good collaboration skills,
• good ability to express yourself in spoken and written English.

Additional qualifications Experience in one or more of the following areas is considered a merit:

• nuclear fuel, fuel-performance modelling, reactor physics or nuclear engineering,
• numerical modelling of heat transfer, material behaviour or solid mechanics,
• Bayesian inference, model calibration, and Markov Chain Monte Carlo methods,
• uncertainty quantification, statistical modelling, and Gaussian processes,
• machine learning for time series, sequence-to-sequence models or surrogate modelling,
• scientific computing, large datasets from simulations or measurements, and high-performance computing
• version control and reproducible computational workflows.

Great emphasis will be placed on personal qualities such as analytical ability, initiative, accuracy and motivation to pursue doctoral studies in a field involving close collaboration between academia and industry.