摘要
为减小光纤陀螺随机漂移,采用时间序列分析法对其进行ARMA模型辨识.提出一种全局最优的模型阶次搜索算法,将模型适用性检验方法中的BIC(Bayesian information criterion)用于模型阶次搜索,并采用Pandit-Wu建模思想,把二维搜索化为一维搜索,得到了模型阶次的一致性估计.提出了一种改进的U-C算法,并与长自回归模型计算残差法相结合共同估计模型参数.它将非线性参数估计过程转化为线性过程,使用了正置与逆置漂移时序参与估计,以前向和后向模型的滤波误差平方和最小为参数估计的指标,在p+1维空间中求极小值.采用上述方法确定的模型其残差标准差为0.0024°,最大预报误差为0.0112°,能准确预报光纤陀螺随机漂移趋势.
To reduce random drift in fiber optic gyros, an autoregressive moving average （ARMA） model for measured data drift was developed using time series analysis. A global optimum algorithm for selecting model order was proposed, based on Bayesian information criterion （BIC） and the Pandit-Wu modeling method. It changed a twodimensional search program into a one-dimensional one, giving a consistent method for estimation of model order. A parameter estimation algorithm was put forward that improves the unit commitment （U-C） method, which was then combined with the long autoregressive residual method to estimate parameters. In this way the non-linear parameter estimation process was transformed into a linear one. Both measured sequence and its reverse were used to estimate parameters, so drift data information was fully utilized and the accuracy of parameter estimation improved. The parameter estimation goal was the minimum sum of forward and backward filtered errors squared, which was found in p ＋ 1 dimensional space. Residual standard deviation of the model was established as 0. 002 4°, and maximum predictive error was 0. 011 2°. This model predicts gyro drift with high accuracy.
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
北大核心
2009年第11期1251-1255,共5页
Journal of Harbin Engineering University
基金
2008国际科技合作计划基金资助项目(2008DFR20420)
关键词
光纤陀螺
随机漂移
ARMA模型
模型阶次辨识
参数估计
fiber optic gyro （FOG）
random drift
ARMA model
model order identification
parameter estimation
