Redundant Graph to Improve Fault Diagnosis in a Gas Turbine

This paper deals with fault diagnosis issues for a Gas Turbine, GT, of a Combined Cycle Power Plant, CCPP. To analyze under which conditions faults in the turbogenerator can be detected and isolated, structural properties of the model are used. The structure redundancy is studied by graph tools considering the standard available measurements. A non-linear dynamic model given by 37 algebraic and differential equations is considered to identify the required redundancy degrees for diverse fault scenarios without numerical values. As result 10 relations are obtained which detect faults in all units of the turbine except one: the thermodynamic gas path. Moreover, using the redundant graph concept it is suggested to add a sensor to increase the redundance and consequently to have detectability of the mechanical faults in the gas path. This is the main contribution of the work. The implementation of redundant relations with specific simulated data of a GT validates this statement.