Scholarlink

The monitoring of data quality in high-energy physics experiments is essential during both data acquisition and offline analyses to ensure the reliability of datasets. The Compact Muon Solenoid (CMS) experiment at the Large Hadron Collider (LHC) has recently implemented Data Quality Monitoring (DQM) at the granularity of individual “luminosity sections” (LSs), each corresponding to approximately 23 seconds of data collection. This paper presents a novel application of AutoEncoders for anomaly detection in DQM, specifically targeting quantities associated with jets and missing transverse energy (MET). The developed method allows for the detection of anomalies at the LS level, which might be missed when examining integrated quantities. By automating the identification of anomalies, this approach enhances the efficiency and precision of the DQM process, ultimately improving the quality of the datasets used for analysis.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

AutoEncoder-Based Anomaly Detection for CMS Data Quality Monitoring

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Alkis Papanastassiou,
  • Valentina Gori,
  • Piergiulio Lenzi

摘要

The monitoring of data quality in high-energy physics experiments is essential during both data acquisition and offline analyses to ensure the reliability of datasets. The Compact Muon Solenoid (CMS) experiment at the Large Hadron Collider (LHC) has recently implemented Data Quality Monitoring (DQM) at the granularity of individual “luminosity sections” (LSs), each corresponding to approximately 23 seconds of data collection. This paper presents a novel application of AutoEncoders for anomaly detection in DQM, specifically targeting quantities associated with jets and missing transverse energy (MET). The developed method allows for the detection of anomalies at the LS level, which might be missed when examining integrated quantities. By automating the identification of anomalies, this approach enhances the efficiency and precision of the DQM process, ultimately improving the quality of the datasets used for analysis.