Unscented Kalman-filter-based simultaneous diagnostic scheme for gas-turbine gas path and sensor faults
Liping Yan, Hualiang Zhang, Xuezhi Dong, Qiao Zhou, Haisheng Chen, Chunqing Tan
Abstract
Abstract Sensor faults can cause incorrect estimations of gas path fault amplitudes in gas turbines. In this paper, an unscented Kalman-filter (UKF)-based simultaneous diagnostic scheme for gas-turbine gas path and sensor faults is proposed. A fault detection and isolation (FDI) system based on the UKF method avoids the requirement to establish different hypothetical models for hierarchical multiple-model-based FDI. Moreover, a fault identification module based on the weighted sum of squared residuals from a bank of filters is proposed to confirm the actual fault. The corresponding fault amplitude is then estimated to adaptively update the related parameters of the fault-diagnosis system according to the actual determined fault. Finally, several simulation case studies are conducted, based on a three-shaft gas turbine. The simulation results show that when two faults coincide, the proposed scheme not only has diagnostic accuracies of 97.3% and 93% for sensor faults and gas path faults, respectively, but also estimates the fault magnitude to a high degree of accuracy.