Performance Degradation Diagnosis of Das Turbine Based on Improved FUKF

Owing to the difficulty in assessing the performance of a gas turbine during usage, because of sudden changes in its operation state, the use of an fading unscented Kalman filter with residual similarity (FUKF-RS) algorithm for the health parameter estimation of a gas turbine is proposed in this study. At first, under the common fading unscented Kalman filter (FUKF) framework, the health parameter estimation algorithm of the gas turbine was built. The weights before and after estimation were adjusted by multiplying the fading factor with the variance of measured value during the updating process of estimation; the fading factor was estimated by keeping the residual vectors to be orthogonal. Then, the similarity of the residual matrix was represented by the cosine value of the residual vector before and after the estimation, and the proportion of residual matrixes was determined according to the magnitude of the similarity. Finally, the fading factor of the algorithm was substituted by such proportion to calculate the residual matrix and obtain the quantitative parameter required for the calculation. The results show that the FUKF-RS algorithm can trace the sudden change of state of the gas turbine, and its accuracy in parameter estimation is higher by approximately 3% as compared to that of the FUKF algorithm. Additionally, as the component performance changes slowly, the parameter estimation curve will be smoother than that of the common FUKF, and the estimation accuracy will be increased by approximately 2%.