Research On The Gas Path Fault Diagnosis Method For Three-Shaft Gas Turbine

Gas path fault diagnosis of gas turbine provides early fault warning,and reasonably arranges maintenance plan to ensure safety and efficient operation of gas turbine.The gas path fault diagnosis of the three-shaft gas turbine is very challenging due to its complex structure and remarkable nonlinear characteristics.At present,the accuracy of gas path fault diagnosis method is limited,and there is no effective technical way for coupling fault diagnosis of gas path and sensor.This paper focuses on the research work of gas path fault identification,gas path fault degree evaluation,gas path and sensor coupling fault diagnosis for three-shaft gas turbine,so as to improve the accuracy,stability and fault tolerance of gas path fault diagnosis.The main contents of this paper are as follows:(1)Simulation research of typical gas path faults based on the three-shaft gas turbine component model.Based on the simulation model of three-shaft gas turbine established by component method,the influence of fouling and erosion on gas turbine performance was studied.It was found that the gas path fault will reduce the thermal efficiency of gas turbine,and the thermal parameters change linearly with the fault severity.(2)Optimise the data-driven diagnosis method for gas path fault.(a)Aiming at the problem that the shallow learning algorithm——kernel extreme learning machine(KELM)hadn’t considered the different feature contribution to fault classification,leading to poor diagnostic performance,a improved extreme learning machine method FWKELM-RF for gas turbine fault diagnosis was proposed.Fault simulation results showed that FWKELM-RF had better fault identification accuracy and stability than other traditional KELM algorithms under four different working conditions.(b)Aiming at the problem that a large number of initial parameters in deep learning algorithm——deep belief network(DBN)affected the diagnosis accuracy,this study proposed a gas path fault diagnosis method based on genetic algorithm(GA)optimizing DBN.Fault simulation results showed that GA-DBN had a higher accuracy(up to 98.4%)of fault identification than other diagnostic algorithms,and GA optimization had stronger global search performance than PSO and SA.(3)Establish the model-based diagnosis method for gas path fault.In order to further estimate the gas path fault degree,unscented kalman filter(UKF)was applied to the gas turbine model to diagnose the gas path fault;the combination of measurement parameters with high evaluation accuracy was selected based on the observability analysis method;through the simulation experiment validation,it was found that the diagnosis method had a high evaluation accuracy for gas path fault of the three-shaft gas turbine.(4)Propose the simultaneous fault diagnosis method of gas path and sensor based on Unscented Kalman filter.In view of the poor performance of single fault diagnosis method under the coupling fault of gas path and sensor,this paper adopts a sensor fault diagnosis method,and verified the reliability of this method through simulation experiments.On this basis,a diagnosis method for fault coexistence of gas path and sensor was proposed,which eliminated the sensor fault influence through the operation data under multiple working conditions;a diagnosis method considering fault time sequence for gas path and sensor was proposed,in which a fault identification module of gas path and sensor was established,and the parameters of the fault diagnosis system were updated adaptively.Finally,simulation experiments verified the coupling fault diagnosis ability of the two methods.