In this work a fuzzy based fault diagnosis system for an industrial gas turbine engine is presented. The system compares measured parameters and those calculated by a computer model, which sets the new and clean reference of the equipment. A fuzzy system classifies and quantifies the faults, which includes part load operation and takes into consideration the compressor variable geometry, which implies in a modification of the deviation pattern between the reference and the model. A commercial software, the GSP (Gas turbine Simulation Program), has been used for the process simulation. The reference software used is called NGGT (Natural Gas and Gas Turbine). Both were calibrated using a real industrial gas turbine operation data. Two compressor faults were simulated using GSP, the resulting variables were compared with those generated by NGGT. The fuzzy sets of the entry variables and the inference rules implemented in the fuzzy systems are dedicated to each fault. Thus, it is possible to add new faults and correct existing fault patterns without disturbing the other faults implemented. The system was tested for the two implemented faults by generating faulty test cases to several geometries and levels of damage. The results showed the robustness of the system and that it is possible to clearly identify the faults, for the several cases tested.
Multi-mode diagnosis of a gas turbine engine using an adaptive neuro-fuzzy system