Skip to content

Quench detection

← Back to all benchmarks

Date: 2024-10-15

Name: Quench detection

Domain: High Energy Physics

Focus: Real-time detection of superconducting magnet quenches using ML

Task Types: Anomaly Detection, Quench localization

Metrics: ROC-AUC, Detection latency

Models: Autoencoder, RL agents (in development)

AI/ML Motif: Anomaly Detection

Resources

Benchmark: Visit

Keywords

Citation

  • Maira Khan, Steve Krave, Vittorio Marinozzi, Jennifer Ngadiuba, Stoyan Stoynev, and Nhan Tran. Benchmarking and interpreting real time quench detection algorithms. In Fast Machine Learning for Science Conference 2024. Purdue University, IN, October 2024. indico.cern.ch. URL: https://indico.cern.ch/event/1387540/contributions/6153618/attachments/2948441/5182077/fast_ml_magnets_2024_final.pdf.
@inproceedings{quench2024,
  author = {Maira Khan and Steve Krave and Vittorio Marinozzi and Jennifer Ngadiuba and Stoyan Stoynev and Nhan Tran},
  title = {Benchmarking and Interpreting Real Time Quench Detection Algorithms},
  booktitle = {Fast Machine Learning for Science Conference 2024},
  year = {2024},
  month = oct,
  address = {Purdue University, IN},
  publisher = {indico.cern.ch},
  url = {https://indico.cern.ch/event/1387540/contributions/6153618/attachments/2948441/5182077/fast_ml_magnets_2024_final.pdf}
}

Ratings

CategoryRating
Software
1.00
Code not provided; no evidence of documentation or containerization
Specification
4.00
Real-time detection task is clearly described, but exact constraints, inputs/outputs, and evaluation protocol are only partially specified
Dataset
2.00
Dataset URL is missing; FAIR principles largely unmet
Metrics
3.00
ROC-AUC and latency are mentioned, but metric definitions and formal evaluation setup are missing
Reference Solution
1.00
No baseline or reproducible model implementation available
Documentation
2.00
Only a conference slide deck is available; lacks detailed instructions or repository for reproduction
Average rating: 2.17/5

Radar plot

Quench detection radar

Edit: edit this entry