Design, implementation and evaluation of an expert system for operating regime detection in industrial gas turbine

Abstract

Operating regime detection plays a pivotal role not only in modeling and control, but also in performance monitoring of gas turbines. In this paper, an expert system is designed and implemented to discriminate the operating regimes of an industrial gas turbine. The proposed expert system identifies start-up regimes (successful/unsuccessful ignition, flame, starts, purge, and manual cooling), stand-still regimes (steady-state and the type of load), and turn-off regimes (normal/emergency shutdown, barring, and off), along with the type of active controller (three types) of the gas turbine. A comprehensive set of expert rules is extracted from the control system’s logic, the knowledge of technicians, and several examples of operating regimes in different environmental conditions. The rules of the expert system are designed so that the minimum number of input sensors and parameters are required to detect the operating regimes. The devised expert system works both online and offline. A concise implementation of the rules is achieved by reducing memory and processing resources on the hardware. To this end, the rules are designed to have the minimum order of complexity, while the processing and memory are confined to the detected beginning and end of each operating regime. The proposed expert system is implemented on an embedded hardware, and its performance is evaluated by using real gas turbine data. Evaluation data is gathered from a different turbine with a different manufacturer. Obtained results show a promising discrimination performance with a hint at the generalization capability of the proposed expert system across different turbines.