A Frequency Domain Identification with Maximum Likelihood Method for Aircraft Engine

Due to the complicated non-linearity of aircraft engine, it is very hard to design the control system directly. Now, linear model which could take place of real engine is necessary during the design process of aircraft engine control system. Normally, partial derivative method [1] and fitting technology [2, 3] usually are used to the linear model establishment of aircraft engine. Partial derivative method could be used for the steady state point identification, and fitting could extend the state variable model to transient process. The accuracy of fitting generally better than partial derivative method. However, when the noise exists, the accuracy of previous two method is not satisfactory.In this paper, the statistical characteristic of input, output and noise signal is analyzed, and a maximum likelihood identification method in frequency domain is proposed. This method establishes an optimized parameter estimation criterion function in frequency domain. By designing a multi-sinusoidal excitation signal, the Levenberg-Marquardt algorithm is applied to solve the criterion function to establish the control system model. The proposed method is applied to identify the shaft speed system of aircraft engine and the linear models at different points are established. From the simulation results, it can be seen that the aircraft engine models established have high accuracy.