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

Journal Article · Thu Aug 01 00:00:00 EDT 2024 · Review of Scientific Instruments

DOI:https://doi.org/10.1063/5.0218724· OSTI ID:2439941

Victor, B. S. ^[1]; Scotti, F. ^[2]

---

1. Lawrence Livermore National Laboratory (LLNL), Livermore, CA (United States); General Atomics
2. Lawrence Livermore National Laboratory (LLNL), Livermore, CA (United States)

This paper describes the application of a machine learning (ML) algorithm using a convolution neural network, first developed in Boyer et al. [“Classification and prediction of detachment in DIII-D using neural networks trained on C III imaging,” Nucl. Fusion (submitted) (2024)], to detect divertor detachment in DIII-D. Detachment detection is based on images from tangentially viewing upper and lower filtered divertor cameras that measure CIII emission at 465 nm. Separate ML models are developed for lower single null and upper single null configurations with mostly closed divertor shapes. Due to the viewing angle and divertor geometry, camera images of the upper divertor show a stark contrast in CIII emission between attached and detached conditions and the model identified detachment with 100% accuracy in the test dataset. For the lower divertor images, the contrast between attached and detached conditions is lower and the model identifies detachment with 96% accuracy. Furthermore, this ML model will be applied to the image data after each shot to provide a rapid assessment of divertor detachment to aid operation of DIII-D with the potential extension to other devices in the future.

• Research Organization: General Atomics, San Diego, CA (United States)
• Sponsoring Organization: USDOE Office of Science (SC) Fusion Energy Sciences (FES); USDOE Office of Science (SC), Basic Energy Sciences (BES). Scientific User Facilities (SUF)
• Grant/Contract Number: FC02-04ER54698; AC52-07NA27344
• OSTI ID: 2439941
• Alternate ID(s): OSTI ID: 2439364
• Journal Information: Review of Scientific Instruments, Journal Name: Review of Scientific Instruments Journal Issue: 8 Vol. 95; ISSN 0034-6748
• Publisher: American Institute of Physics (AIP)
• Country of Publication: United States
• Language: English

Related Subjects

70 PLASMA PHYSICS AND FUSION TECHNOLOGY
Artificial neural networks
Fusion energy
Imaging spectroscopy
Linear filters
Machine learning
Optical spectroscopy
Plasma confinement
Plasma temperature
TV cameras
Tokamaks