飞行过程中航空发动机闭环辨识研究

On Closed-Loop Identification of a Certain Aeroengine under Flight Conditions

介绍了飞行试验时,利用在随机扰动作用下的发动机输入—输出信号对发动机数学模型的辨识.分析和研究了发动机的闭环可辨识性;在飞行过程中发动机动态信号的采集和处理方法;信号的相关性与功率谱.最后利用两种飞行状态下不同形式的信号应用递推广义最小二乘方法进行参数辨识,得到了较满意的结果.

Past research on indentification of aeroengine under stochastic disturbance condition appears to be inadequate. The authors now present the results of their comprehensive study of a certain Chinese made aeroengine (hereinafter referred to as aeroengine C) on all four following aspects of identification: system identifiability, model category, test conditions and methods of identification. In this paper, the authors analyze the conditions aeroengine C must satisfy to ensure its identifiability. Idnetifiability requires that the burning delay in the engine burner must be considered. As there does actually exist such burning delay, the authors can prove the consistency and uniqueness of parameter estimation. Theoretical investigation shows that the model construction of aeroengine C is a two-order one. Under different flight conditions, flight tests have been implemented for aeroengine C. Under stochastic disturbance excitation, the engine input (fuel flow) and output response (engine speed) have been recorded on-line by special equipment. An example of fuel flow record is shown in Fig.2; an example of engine speed record is shown in Fig.3. Because the test signals are contaminated by noises, the authors have given the signals the treatment of weighted moving smoothing and that of FIRDF filtering. For aeroengine C, for the first time, the correlation functions and power spectrums of the random signals have been analyzed. The authors' research indicates that the output is obviously correlated to input, the frequency spectrum band of input is larger than that of the output and all the features of information in the engine are included in the output. Signal analysis shows that the input and output satisfy identification requirements. The recursive generalized least squares method is applied to parameter identification of two different types of test signals of aeroengine C. The deviations of parameter estimation are very small and the results of parameter estimation may be considered satisfactory.