Anomaly classification by inserting prior knowledge into a max-tree based method for divertor hot spot...

Area of Science:

• Nuclear Fusion Engineering
• Plasma Physics
• Materials Science

Background:

• The WEST tokamak divertor manages high heat fluxes, crucial for plasma control and exhaust.
• Exceeding heat flux limits can damage plasma-facing tungsten components.
• Ensuring divertor operation safety is vital for future fusion reactors like ITER.

Purpose of the Study:

• To develop a robust method for detecting and classifying hot spots on the WEST tokamak divertor surface.
• To enhance operational safety by accurately monitoring divertor thermal conditions.
• To provide a divertor monitoring solution applicable to ITER.

Main Methods:

• Utilized infrared (IR) thermography for real-time surface temperature monitoring.
• Developed a novel classification approach based on max-tree representation and image attributes.
• Applied the method to analyze IR images of the WEST tokamak divertor strikelines.

Main Results:

• The max-tree classifier accurately identified 88% of abnormal hot spots on the divertor.
• The method demonstrated efficient computation, suitable for execution between tokamak pulses.
• Achieved classification without requiring labeled training data, unlike SVM or CNN methods.

Conclusions:

• The max-tree based approach offers a reliable and efficient solution for divertor hot spot classification.
• This method enhances operational safety for actively cooled tungsten divertors in tokamaks.
• The technique shows promise as a monitoring solution for ITER divertor systems.