A fuzzy-based gas turbine fault detection and identification system for full and part-load performance deterioration

Considering the fact that the signs of performance deterioration in the main components of gas turbines vary in different operating points, thus designing a Fault Detection and Identification (FDI) system using a single fault pattern for all engine operating ranges can reduce the ability of the diagnosis system to identify the intensity or even the type of any potential degradation. So, in this article, with the aid of the “load” parameter as an augmented input and using the fault patterns obtained at different part-load conditions, a fuzzy-based FDI system is proposed for an industrial two-shaft gas turbine with the ability to use in both the full and part-load operations. In the proposed FDI system, fuzzy rule base is generated by a table look-up scheme and by employing the available input/output data extracted from fault signature table. Moreover, a global optimization technique is used to determine some database parameters, such as the number of membership functions of input variables and their standard deviations. The optimization is carried out to make a compromise between the diagnosis accuracy and robustness against measurement noise. In the present work, the performance of the proposed FDI system is evaluated against the most common cause of gas turbine performance deterioration i.e. fouling and erosion in terms of the success rate and the estimation accuracy at different levels of sensor noise. The results obtained indicate that the proposed FDI system can considerably reduce the average estimation error by 0.1-0.75% and increase the success rate by 10-20% compared with the diagnosis systems designed for a specific operating point. The results also demonstrate that the proposed FDI system has robust performance against measurement uncertainties, and moreover smearing effects are rarely seen in the results.