Fault detection through physical modelling in an axial flow compressor of a combined-cycle power plant

Abstract

The methods that are most frequently used in fault detection are able to detect that some abnormal behaviour of a system is occurring, but are unable to explain directly its causes. The main goal of the new method proposed in this paper is to achieve the capability to explain incipient fault, detected using a physical interpretation. To this end a detailed physical model of an axial flow compressor has been developed using empirical off-design correlations and turbo machine fundamentals to determine flow conditions at the mid-span of each blade. Then the methodology has been applied using real faulty behaviour data from the compressor of a combined-cycle plant. An incipient fault could be detected using the method over a nonmeasurable variable of the system, the enthalpy at the outlet of first rotor cascade. These results suggest the possible improvements that the application of this methodology could bring.