Novel gas turbine fault diagnosis method based on performance deviation model

Effective fault detection and identification methods are crucial in gas turbine maintenance. To express the gas turbine performance of the fault symptom state precisely and to reduce the individual differences of different gas turbines, a novel performance deviation model based on real-life operation data of gas turbines is proposed in this paper. A backpropagation neural network is adopted to establish the performance deviation model. Performance deviation values calculated by the model are regarded as fault signatures of the gas turbines. To enhance the accuracy of the fault diagnosis, a multikernel support vector machine is employed in the fault classification experiment. A contrast experiment showed the accuracy of the fault diagnosis method based on the performance deviation model and multikernel support vector machine.