An integrated combined methodology for the outline gas turbines performance-based diagnostics and signal failure isolation.

The target of this research is the performance-based diagnostics of a gas turbine for the online automated early detection of components malfunctions with the presence of measurements malfunctions. The research proposes a new combination of multiple methodologies for the performance-based diagnostics of single and multiple failures on a two-spool engine. The aim of this technique is to combine the strength of each methodology and provide a high rate of success for single and multiple failures with the presence of measurement malfunctions – measurement noise. A combination of Kalman Filter, Artificial Neural Network, Neuro-Fuzzy Logic and Fuzzy Logic is used in this research in order to improve the success rate, to increase the flexibility and the number of failures detected and to combine the strength of multiple methods to have a more robust solution. The Kalman Filter has in his strength the measurement failure treatment, the artificial neural network the simulation and prediction of reference and deteriorated performance profile, the neuro-fuzzy logic the estimation precision, used for the quantification and the fuzzy logic the categorization flexibility, which are used to classify the components failure. All contributors are also a valid technique for online diagnostics, which is a key objective of the methodology. In the area of gas turbine diagnostics, the multiple failures in combination with measurement issues and the utilization of multiple methods for a 2-spool industrial gas turbine engine has not been investigated extensively. This research investigates the key contribution of each component of the methodology and reaches a success rate for the component health estimation above 92.0% and a success rate for the failure type classification above 95.1%. The results are obtained with the first configuration, running with the reference random simulation of 203 points with different level of deterioration magnitude and different combinations of failures type. If a measurement noise 5 times higher than the nominal is considered, the component health estimation drop to a minimum of 70.1% (reference scheme 1) while the classification success rate remains above 88.9% (reference scheme 1). Moreover, the speed of the data processing – minimum 0.23 s / maximum 1.7 s per every single sample – proves the suitability of this methodology for online diagnostics. The methodology is extensively tested against components failure and measurement issues. The tests are repeated with constant simulations, random simulation and a deterioration schedule that is reproducing several months of engine operations.