CMS探测器中基于异常检测的自动数据质量监控
高能物理 - 实验
2026-03-27 v1 数据分析、统计与概率
仪器与探测器
摘要
大型粒子探测器如巴黎联合实验室(CERN)大型强子对撞机(Compact Muon Solenoid, CMS)的成功运行需要快速、深入的评估数据质量。我们引入了``AutoDQM''系统，用于基于先进统计技术和无监督机器学习的自动化数据质量监控。基于beta-binomial概率函数、主成分分析和神经网络自编码器图像评估的异常检测算法在CMS在2022年收集的全套质子-质子碰撞数据上进行测试。AutoDQM以4--6倍的速率识别受显著探测器故障影响的异常``坏''数据，凸显其作为通用数据质量监控工具的有效性。
引用
@article{arxiv.2501.13789,
title = {Anomaly Detection for Automated Data Quality Monitoring in the CMS Detector},
author = {Andrew Brinkerhoff and Chosila Sutantawibul and Robert White and Caio Daumann and Chad Freer and Indara Suarez and Samuel May and Vivan Nguyen and Jonathan Guiang and Bennett Marsh and Darin Acosta and Alex Aubuchon and Emanuela Barberis and Aaron Bundock and Evan Collins and Preston Epps and Johannes Erdmann and Henning Flaecher and Junshen Huang and Ryan Nie and Sudarshan Paramesvaran and John Rotter and Kaitlin Salyer and Siddhesh Sawant and Tanvi Sheokand and Darien Wood},
journal= {arXiv preprint arXiv:2501.13789},
year = {2026}
}
备注
16 pages, 14 figures