论文部分内容阅读
动态Allan方差是分析光纤陀螺随机误差的一种新方法。针对其采用固定窗长的窗函数截断信号,导致动态跟踪效果与方差估计值的置信度不能兼顾的问题,提出了一种基于峭度自动调节窗长的改进算法。算法引入峭度参数表征陀螺输出信号的平稳程度,以滑动窗内数据的峭度值为变量构造窗长截取函数,应用窗长函数根据短时信号的稳定程度自动确定截断窗窗长,并用其来截取随机信号进而进行Allan方差估计及误差系数实时辨识和提取,同时将方差估计值和误差系数时间曲线分别绘制在三维或二维图上。对仿真和实测数据分析结果表明:新算法既能在平稳性好的数据段保持较高的置信度,又能在发生动态变化时及时跟踪信号,可以更好地对光纤陀螺的随机误差系数进行实时在线提取和分析。
Dynamic Allan variance is a new method of analyzing the random error of fiber optic gyro. Aiming at the problem that the window function with fixed window length truncates the signal, which leads to the fact that the dynamic tracking effect and the variance estimation value can not be balanced, an improved algorithm based on kurtosis to automatically adjust the window length is proposed. The algorithm introduces the kurtosis parameter to characterize the smoothness of the gyro output signal, constructs the window length interception function with the kurtosis of the data in the sliding window as a variable, and uses the window length function to automatically determine the truncated window length according to the stability of the short-time signal To intercept the random signal and then Allan variance estimation and error coefficient real-time identification and extraction, while the variance estimate and the error coefficient time curve were drawn in three-dimensional or two-dimensional map. The results of simulation and measured data analysis show that the new algorithm can not only maintain high confidence in the stationary data segment but also track the signal in time when the dynamic changes occur, so that the random error coefficient of the FOG can be better Real-time online extraction and analysis.