A Multiagent System for Monitoring Industrial Gas Turbine Startup Sequences

This paper introduces a novel technique for condition monitoring of gas turbine start-up sequences. The vast amount of data and the complex processes behind on-line fault detection indicate the need for an automated solution. A multi-agent system that views the problem as the interaction of simple independent software entities for effective use of the available data is presented. The overall solution is derived from the combination of partial solutions provided by the components of the multiagent system. As a consequence, data interpretation is achieved by converting the data into appropriate information and combining individual agents´ information, resulting in an automatic fault diagnosis for the engineers. This multiagent system can employ various intelligent system techniques and has been implemented using the ZEUS agent building toolkit.