Intelligent distributed condition monitoring for nuclear power plants

This paper describes the use of a Multi Agent System (MAS), to perform the collection, analysis and storage of nuclear reactor Condition Monitoring (CM) data such that a distributed CM framework is created. It is shown that various styles of CM techniques can be deployed, with varying degrees of decentralisation, to meet the CM requirements of nuclear reactors and related systems. Indeed the sociability of agents and their ability to share expertise allow them to solve complex CM problems that may require sequences of actions involving access to multiple data sources. The creation of an off-line, prototype framework based on these principles is described, using as a case study data from Advance Gas-cooled Reactors (AGR) in the UK. It is shown that by virtue of the requirement to create a common agent language, or ontology, the expertise previously applied by engineers can be encoded and retained in software rules. The automation introduced through the use of a MAS also allows for more rapid analysis of data and a more structured method of introducing new analyses and data sources. Examples of existing and newly developed CM analyses are shown, highlighting the desire to maximize the use of the data by correlating events wherever possible and providing distributed access to the CM capabilities of the system.