关键词: 时变自回归滑动平均模型, 增广最小二乘估计, 组合目标函数, 遗传算法

Abstract: To solve the problem of spectral deviation for Time-Varied Auto-Regressive Moving Average (TVARMA) parametric spectrum estimation, a TVARMA parameter estimation method based on Genetic Algorithm (GA) and combined objective function was proposed and then used for estimating the spectrum of flight vibration time series. Firstly, an initial estimation of the model parameters was acquired by the long time-varied auto-regressive and augmented Least-Square (LS) method; secondly, through extremum condition of continuous function, a constraint equation of model parameters was derived and a combined objective function was constructed based on the punishment function idea; lastly, the GA was used to optimize the initial parameters and the optimum parameters were the ones that made the combined objective function be the least. The experimental results demonstrate that the new method can overcome the spectrum deviation and improve the model precision both in time domain and frequency domain.

Key words: TVARMA Model, Augmented LS Estimate, Combined Objective Function, GA