频域内参数估计

Parameter estimation in the frequency domain

研究了在频域中利用方程误差法结合递推的傅里叶变换技术进行线性系统实时参数估计的方法。分别从频域中的方程误差法、递推的傅里叶变换和计算频率范围选择三个方面对参数估计进行了较详细地分析。通过对某型战斗机进行模拟计算 ,结果表明 :该方法能自动过滤噪声 ;占用计算机内存固定 ;计算量小 ;不需要输入任何初值并在计算中调整任何参数 ;所估计的参数值能够在系统模态周期 1 .0～ 1 .5倍的时间内逼近真值 ,即使在信号水平较低的情况下也能够比较精确地估计参数。

A technique is developed for estimating linear model parameters in real time, using an equation error formulation in the frequency domain with a recursive Fourier transform. Simulation examples of a fighter demonstrated that the method gives accurate real time estimates of model parameters and standard errors. The examples also show that the advantages of this approach include automatic noise filtering, fixed memory requirements regardless of the time record length, low computational requirements, no starting values were required for the parameter estimates, no tuning parameters had to be adjusted during the parameter estimation, typical convergence times are less than 1~1 5 cycle time of the dominant dynamic mode, and good performance even for low information content in the data.