- Promotor: Prof. Geert Verdoolaege
- Supervisor: Jerome Alhage
- Study programs: Master of Science in Engineering Physics, Master of Science in Physics and Astronomy, Master of Science in Teaching in Science and Technology (Physics and Astronomy), European Master of Science in Nuclear Fusion and Engineering Physics, Master of Science in Information Engineering Technology
- Location: Technicum, at home
Problem setting
The worldwide research on controlled thermonuclear fusion aims to realize a source of energy on earth that is limitless, clean and safe. It is a large-scale enterprise that mimics the power source of the stars. Ghent University is involved in fusion R&D by developing data science methods and numerical models for understanding and predicting the complex science and technology of fusion devices based on magnetic plasma confinement.
One of many challenges on the road to fusion energy is the prevention or mitigation of plasma instabilities, threatening the efficient and safe operation of the device. Among those instabilities, edge-localized modes (ELMs) are of special importance. These are repetitive magnetohydrodynamic instabilities that occur near the edge of the plasma. When an ELM occurs, particles and energy expelled from the plasma edge may contribute to degradation of the wall materials. Moreover, occasional large ELMs pose a significant risk to machine operation, since transients heat loads can exceed material limits. Understanding the plasma conditions influencing ELM behavior is therefore crucial for safe plasma operation.
Video frames during an ELM at the ASDEX Upgrade tokamak (Germany).
Recently, UGent students conducted an extensive study of ELM behavior using data from the Joint European Torus (JET, UK). More than 10 000 of these events were manually annotated, and then used to train and test several detection algorithms as a proof-of-concept. Afterwards, the best method was selected to automatically process nearly a tenfold of timeseries data. However, examining these new samples shows inconsistency in detection: false positives, partially predicted events, overlapping event regions, skipping very close events, ...
Objectives
The goal of this thesis is to turn the exploratory algorithms developed before into production-ready tools, ready for use by researchers in fusion laboratories all over the world. You will benefit from existing software routines for data loading and ELM detection, refining the detection models and tuning their hyperparameters.
Milestones
- Review current literature on event detection.
- Become familiar with the existing workflow to build, tune, and test an ELM detection method.
- Improve an existing method or propose a new one. Evaluate on the extended dataset.
- Create a minimal detect_elms function and package the library.
- [optional] Evaluate performance on data from a different machine (ex: ASDEX Upgrade).
Signals pertaining to a series of ELMs at JET, detected with one of our techniques (Laplacian of Gaussian filter with a custom kernel). Top: Light emission from the plasma edge that accompanies an ELM, serving as its main indicator. Bottom: Plasma stored energy with red markers indicating energy drops caused by ELMs.
Examples of ELM detection errors in JET discharges.