Identifying divertor detachment using a machine learning model trained on divertor camera images from DIII-D.

Area of Science:

• Fusion energy research
• Plasma physics
• Machine learning applications

Background:

• Divertor detachment is a critical plasma physics phenomenon in tokamak fusion devices like DIII-D.
• Accurate and timely detection of divertor detachment is essential for maintaining stable plasma conditions and device protection.
• Current methods for detachment detection may not provide the rapid assessment needed for real-time operational control.

Purpose of the Study:

• To apply a convolutional neural network (CNN) based machine learning (ML) algorithm for automated detection of divertor detachment in DIII-D.
• To evaluate the accuracy and effectiveness of the ML model in identifying detached plasma conditions from CIII emission imaging.
• To establish a rapid assessment tool for divertor detachment to support DIII-D operations.

Main Methods:

• Utilized a CNN ML algorithm, previously developed by Boyer et al., for image analysis.
• Trained and tested the ML model on CIII emission images (465 nm) from upper and lower filtered divertor cameras in DIII-D.
• Developed separate ML models for lower single null (LSN) and upper single null (USN) plasma configurations with closed divertor shapes.

Main Results:

• The ML model achieved 100% accuracy in detecting divertor detachment from upper divertor images due to stark CIII emission contrast.
• The ML model achieved 96% accuracy in detecting detachment from lower divertor images, where the CIII emission contrast is less pronounced.
• The developed models demonstrated high performance in classifying attached versus detached plasma conditions.

Conclusions:

• The ML algorithm effectively detects divertor detachment in DIII-D using CIII imaging, with high accuracy for both upper and lower divertor configurations.
• This automated detection provides a rapid assessment crucial for real-time operational feedback in DIII-D.
• The methodology shows potential for extension to other fusion devices, enhancing operational capabilities and plasma control.