Agent-based technology for data management, diagnostics and learning within condition monitoring applications

Research output: Contribution to conferencePaper

Abstract

Online condition monitoring systems are used to prolong the life of electrical power equipment by continually monitoring for any signs of faults. To be of most use, a condition monitoring system should be flexible enough to accommodate various sensors and different data interpretation techniques. To provide such flexibility this paper proposes an agent-based architecture, where autonomous modules (agents) perform separate parts of the data management and interpretation tasks. This means that only the agents associated with required tasks need to be deployed. This paper presents an example of a flexible agent-based system that can be used to diagnose defects in a power transformer using data from various sensors.
The agent-based architecture also provides an extensible framework to integrate different types of data interpretation. This paper shows this by detailing the addition of further interpretation agents for pattern recognition, diagnosis and learning. One employs a knowledge-based approach to diagnose defects in transformers, based on fundamental partial discharge behaviours. Other agents provide on-line learning of the plant behaviour, automatically identifying normal and abnormal modes, leading to advanced anomaly detection capabilities.
Original languageEnglish
Number of pages10
Publication statusPublished - 2007
Event2nd World Congress on Engineering Asset Management (EAM) and The 4th International Conference on Condition Monitoring - Harrogate, United Kingdom
Duration: 4 Jan 20074 Jan 2007

Conference

Conference2nd World Congress on Engineering Asset Management (EAM) and The 4th International Conference on Condition Monitoring
CountryUnited Kingdom
CityHarrogate
Period4/01/074/01/07

Keywords

  • condition monitoring
  • power transformers
  • multi-agent systems

Fingerprint Dive into the research topics of 'Agent-based technology for data management, diagnostics and learning within condition monitoring applications'. Together they form a unique fingerprint.

Cite this