| Gas turbine engine, as the heart of the aircraft, has an important significance to insure the performance and flight safety of the aircraft. However, the existence of complex structure and strong nonlinearity results in the difficulty in modeling, controller design and fault diagnosis based on physical analysis. This dissertation presents a method for modeling and fault diagnosis in gas turbine engine by applying system identification approach. This algorithm has practical significance because it can effectively avoid the complicated structure and computational burden of the system.The proposed multi-stage least squares identification algorithm for closed-loop system is useful for gas turbine engine systems’parameter identification, by assuming that the input-output data is measurable by sensing from SISO systems and MIMO system. It turns out that the time delay parameter of process, system model parameters and parameters of feedback channel can be exactly estimated.The process of this algorithm is divided into three stages. Firstly, method of determinant ratio is used to estimate the value of time delay by calculating the order of the system. Then, applying recursive extended least squares (RELS) algorithm, the values of the parameters of noise model and auxiliary models are obtained. Finally, the parameters of the system model as well as unknown feedback channels are calculated from the auxiliary models.Based on the above mentioned multi-stage least squares identification algorithm for closed-loop system, this dissertation further studies the fault diagnosis problem for fuel mechanism of gas turbine engine. Three types of faults are considered here:fouling and wear of compressor; blocking or structure changes of last vents; atomized instability or fuel flow changes in the injection process. The detailed illustrations about the principles and steps of multi-stage least squares identification algorithm for closed-loop system for the SISO and MIMO systems of biaxial gas turbine engine. Specially, for the MIMO case, a sub-model decomposition method is considered to simplify the identification model.Finally, some discussions about the choice of forgetting factor, the difference between the faults and disturbances are given. By numerical examples, the fault diagnosis for the gas turbine engine is achieved, which verifies the effectiveness of the proposed multi-stage least squares identification algorithm for closed-loop system. |
