The reliability of the gas path components (compressor, burners and turbines) of a gas turbine (GT) is usually high when compared with other GT systems such as fuel and control. However, their availability could be relatively low as high downtimes are normally associated with these components when subjected to forced outages. One way of improving availability is by improved maintenance practices that involve applying such approaches as condition based monitoring (CBM). Unfortunately, this cannot be achieved without the existence of a proper instrumentation set that can adequately and repeatedly track down the levels of deterioration in these components, thereby allowing for optimally scheduled maintenance. Different engine handles (operating point or parameter that is held constant with respect to other parameters) would require different instrumentation sets for proper gas path fault diagnosis. Sometimes, the instrumentation sets used makes the required diagnostic analysis impossible. Furthermore, allowing redundancy in instrumentation, unless specified with knowledge of the diagnostic technique to be used, is not only unnecessary but also cost ineffective. The central theme of this paper is to present a means of attaining an optimum instrumentation set using a non-linear gas path analysis (NLGPA) programme. Firstly, some of the common gas path faults are considered, some theoretical backing is given to the principles involved in this work, the implications of unoptimised instrumentation set as viewed from the users’ perspective is examined and finally, results presented for the NLGPA approach when applied to a two-shaft and a three-shaft industrial gas turbine. Also, we show how the engine handle can affect the choice of the instrumentation set.
FAULT DIAGNOSIS OF AN INDUSTRIAL GAS TURBINE USING NEURO-FUZZY METHODS
