Research On Fault Diagnosis Of Gas Turbine Based On Kalman Filter

Gas turbine gas path fault has extremely obvious influence to the security and the economy of gas turbine, meanwhile gas path fault diagnosis is always as the key research content of maintenance based on condition both at home and abroad. In the process of gas turbine gas path fault diagnosis, the reliability of the sensor measurement data is the prerequisite to make sure state evaluation rationality of gas turbine. It has important engineering application value to detect gas turbine sensor to ensure the reliability of the sensor measurement parameters, then gas turbine condition monitoring can be performed.This paper studies the application of kalman filter in gas turbine sensor and gas path fault diagnosis technology. The detailed contents are shown in the following:(1) Established the state space model of dual shaft gas turbine. To solve the low precision problem of state space model in steady state, "Step-by-step fitting method" is adopted to establish the state space model of gas turbine and analyze the effects of different step on the state space model. The results show that the state space model of the average matrix which is corresponding to the ±0.5% step amount has high precision. The steady state error of the state space model is less than 0.21% and the linear range of the steady working condition is -1% to 1%. At the same time, analyzing the controllability, observability and stability of the state space model. The results show that the controllability, observability and stability of the state space model are all available. In order to characterize the performance deterioration of gas turbine, the health parameters were added to the state space model and the simplex method is adopted to establish the extended state space model of gas turbine. The results show that the steady output error between the extended state space model and the nonlinear model is less than 0.02% and the extended state space model can characterize the performance deterioration of gas turbine.(2) The state space model of gas turbine was discreted to build kalman filter. In view of the filtering effects of different kalman filter parameters, the control variable method is adopted. The kalman filter was used to estimate the measurement parameters of gas turbine in different system noise covariance and measurement noise covariance. Then the estimate value was compared with the measurement. By analyzing the parameters of the kalman filter, the appropriate system noise covariance and measurement noise covariance were determined.(3) The application of kalman filter in sensor fault diagnosis of gas turbine. In view of the problem of extracting the fault features of gas turbine sensor. The statistical features of the residual signal were analyzed, meanwhile, the weighted sum of squares residual was used to normalize the residual signal. The fault characteristics of the gas turbine sensor was successfully extracted. In view of the fault indicator signal irregular disturbance problem, the method of moving average is used to eliminate the influence of random disturbance. The results show that the method of moving average can effectively eliminate the influence of random disturbance. The fault of constant deviation, gradient type and mutant type always appear in gas turbine sensor. This paper respectively design kalman filter group to detect and isolate the single sensor fault and dual sensor fault of gas turbine. The results show that kalman filter group can achieve the detection and isolation of gas turbine sensor fault.(4) The application of kalman filter in gas path fault diagnosis of gas turbine. The measurement noise lead to the uncertainty of the measured data. Adding the health parameters in extended state space model of gas turbine to state variables, the augmented state space model can be achieved, which can be used to construct the kalman filter. The results show that kalman filter can effectively reduce the influence of measurement noise and improve the diagnostic accuracy by filtering process. The fault of compressor and turbine always appear in gas turbine. The kalman filter is used to detect and isolate the common faults. Results show that kalman filter can achieved the estimation of health parameter, no matter the amount of parameter is single or more and the absolute error is less than 0.12%.