高精度时变参数模型谱估计及应用

High Precision Time-Varying Parametric Model Spectrum Estimation and Its Application

现有时变自回归(TVAR)模型参数谱估计容易导致谱峰漂移。针对该问题,提出一种基于组合目标函数和遗传算法的TVAR参数估计方法,并将之应用于飞行器结构响应序列的建模及谱估计。通过U-C算法获得TVAR模型参数的初始估计;依据现代谱估计理论结合连续函数极值存在的必要条件,推导模型参数的频域约束条件并构造组合目标函数;采用遗传算法对模型参数初始估计值进行优化。应用结果证明了该方法的有效性。

To solve the problem of spectral deviation for Time-Varying Auto-Regressive（TVAR） model parametric spectrum estimation, this paper proposes a TVAR parameter estimation method based on Genetic Algorithm（GA） and combined objective function and the method is used to model the flight vibration time series and its spectrum estimation. An initial estimation of the parameters is acquired by the U-C algorithm; through modem spectrum estimation theory and the necessary condition of extreme value for continuous function, a constraint equation of model parameters is derived and a combined adaptability function is constructed; the GA is used to optimize the initial parameter estimation. Application results verifies the efficiency of the method.