Modeling Study For Actuating Mechanism Of Twin-Spool Low By-Pass Turbofan Engine

The research of aviation turbofan engines is complicated system engineering, which have many characteristics such as long development cycle, high technical difficulties, high research budget, and big development risk. The reason for the difficulty of aviation turbofan engines research is that a lot of large-scale test validation must be carried out. With establishing the mathematical model of the aviation turbofan engine, the design of aircraft engine can be optimized and performance of the aircraft engine can be evaluated. So that the engine test costs and development time can be greatly reduced, and the risk of the actual test trialscan be avoid. At the mean while all the necessary parameters of the actual test can be getted, the problem of multi-parameter optimization can be solved and so on. The actuating mechanism is an important part of aviation turbofan engine; its main role is to control fuel flow, stator vane angle and the area of variable nozzle, and to make the aircraft engine to make the output thrust on request in the flight envelope. Actuating mechanism of the aviation turbofan engine as the object of study, this paper use the method of system identification to carry out modeling study of engine actuating mechanism.After a brief introduction to the basic theory of system identification, the mathematical model of the turbofan engine actuating mechanism is divided into two parts; one is linear part, another is nonlinear part, and two parts are respectively. For the identification of the linear part, the least squares method is used, for the identification of the nonlinear part, the BP neural network is applied; the sum of two parts is the mathematical model of the turbofan engine actuating mechanism. Because the BP learning algorithm has the disadvantages of slow rate of convergence and is easy to fall into local minimum, this paper apply wavelet neural network to identify the nonlinear part of the mathematical model of the turbofan engine actuating mechanism. Finally, this paper use the extended Kalman filter as learning algorithm of wavelet neural network, The problems of connection weights between neurons, stretching and offset in wavelet neural network are treated as the optimal estimation problem of the Kalman filter state vector, the nonlinear part in mathematical model of the turbofan engine actuating mechanism is identified by wavelet neural network, and simulated and analyzed in MATLAB environment.The results of simulation and analysis show that the wavelet neural network designed in this paper with extended Kalman filter as learning algorithm which is used in modeling of the turbofan engine actuating mechanism can quickly and accurately identify the mathematical model of the executing agency. In the three system identification methods, its training times are the least, and its root mean square error of the mathematical model is the least.