挠性陀螺刻度系数标定算法  

Calibration Algorithm for Flexible Gyroscope Scale Factor

作　　者：刘育浩[1] 黄新生[1] 徐婉莹[1] 

机构地区：[1]国防科学技术大学机电工程与自动化学院,长沙410073

出　　处：《数据采集与处理》2010年第1期117-121,共5页Journal of Data Acquisition and Processing

基　　金：兵器预研基金(2020203)资助项目

摘　　要：为减小刻度系数非线性误差,建立了陀螺刻度系数分段非线性插值标定方法。针对分段非线性插值方法的不足,建立刻度系数神经网络模型,用扩展卡尔曼滤波方法训练;并在此基础上对神经网络进行再训练以提高其精度。实验结果表明,分段非线性插值方法有效地减小了刻度系数非线性误差;用训练好的神经网络可以代替分段非线性插值算法,并可简化实现过程;对刻度系数神经网络再训练,可提高神经网络对实际刻度系数的逼近程度,减小刻度系数的非线性误差。In a strapdown inertial navigation system（SINS）, the calibration methods for piecewise nonlinear interpolating （PNIP） and multilayer feedforward neural network（MLFN） are presented to compensate the nonlinear errors of flexible gyroscope scale factor. To reduce the nonlinear errors of the scale factor, the PNIP method is used to calibrate the scale factor. Aimed at the flaw of PNIP method, a MLFN is built and trained with extended Kalman filter （EKF） to calibrate the scale factor. Based on the trained MLFN, a retrained method is put for- ward to improve its performance. Experimental results show that the PNIP method reduces the nonlinear errors of the scale factor. And the trained MLFN can replace the PNIP method and simplify the realization process. Results also show that the retrained MLFN can approach the actual scale factor and reduce the nonlinear error.

关 键 词：捷联惯导系统 刻度系数 分段非线性插值 神经网络 

分 类 号：V241.5[航空宇航科学与技术—飞行器设计]