Scholarlink

The integration of advanced sensorSensors technology into Alcoa’s R&D pots has enabled real-time monitoring of individual anodesAnode, replacing labor-intensive manual voltage dropVoltage drop measurementsMeasurements. The contactless system continuously captures detailed data, allowing for proactive anodeAnode fault detection. By applying machine learningMachine learning techniques, including pattern recognition and anomaly detection, we identified indicators of these various anodeAnode disturbances through different types of signals. By collecting data on multiple pots, time clusteringClustering revealed strong correlations between current distribution patterns and localized variations. Since direct anode–cathode distance (ACD) measurementMeasurements is impractical in industrial settings, we inferred ACD dynamics from individual anodeAnode signals, enabling localized tracking. To improve model robustness, we embedded constraints, combining domain expertise with data-driven insights. This hybrid approach significantly enhanced the precision and reliability to determine impactful anodic problems such as spikes, deformations, burn-offs, and other irregularities, enabling faster, more targeted, and better response time.

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

Inferring Localized Current Patterns and Anodic Incidents Using Hall Sensors for Individual Anode Monitoring

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Samuel Duplessis,
  • Marcel Guerreiro Neiro,
  • Patrice Doiron,
  • Julien Lauzon-Gauthier,
  • Dany Gauthier

摘要

The integration of advanced sensorSensors technology into Alcoa’s R&D pots has enabled real-time monitoring of individual anodesAnode, replacing labor-intensive manual voltage dropVoltage drop measurementsMeasurements. The contactless system continuously captures detailed data, allowing for proactive anodeAnode fault detection. By applying machine learningMachine learning techniques, including pattern recognition and anomaly detection, we identified indicators of these various anodeAnode disturbances through different types of signals. By collecting data on multiple pots, time clusteringClustering revealed strong correlations between current distribution patterns and localized variations. Since direct anode–cathode distance (ACD) measurementMeasurements is impractical in industrial settings, we inferred ACD dynamics from individual anodeAnode signals, enabling localized tracking. To improve model robustness, we embedded constraints, combining domain expertise with data-driven insights. This hybrid approach significantly enhanced the precision and reliability to determine impactful anodic problems such as spikes, deformations, burn-offs, and other irregularities, enabling faster, more targeted, and better response time.