基于阈值去噪和Elman神经网络的激光陀螺误差补偿  

Laser gyro error compensation based on threshold denoising and Elman neural network

作　　者：刘颖 蔡月艳 LIU Ying;CAI Yueyan(School of Automation,Xi’an University of Posts&Telecommunications,Xi’an 710121,China)

机构地区：[1]西安邮电大学自动化学院,陕西西安710121

出　　处：《传感器与微系统》2023年第3期19-21,26,共4页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(51405387);陕西省自然科学基金资助项目(2016JQ5051);陕西省教育厅科学研究计划资助项目(12JK0791)。

摘　　要：为了降低激光陀螺信号中噪声对导航解算精度的影响,提出了一种基于阈值去噪和Elman神经网络相结合的误差补偿方法。首先,采用指数型阈值函数对原始数据进行阈值去噪;然后,以激光陀螺的3轴原始数据作为Elman神经网络的输入,经小波阈值处理后的单轴信号作为神经网络的输出,建立一个3输入1输出的Elman神经网络模型,再用神经网络训练后的预测误差作为补偿信号对阈值信号进行补偿处理;最后,用Allan方差对补偿前后的信号进行误差分析。结果表明:所提方法可以降低激光陀螺的随机误差,相对原始信号的随机误差降低了38%~89.8%,方差降低了2个数量级。In order to reduce the influence of noise in laser gyro signal on precision of navigation solution, an error compensation method based on threshold denoising and the Elman neural network is proposed.Firstly, the exponential threshold function is applied to denoise the original data.Then, the three-axis original data of laser gyro is applied as the input of Elman neural network, and the single axis signal processed by wavelet threshold is used as the output of neural network.Elman neural network model of three-input one-output is established.Furthermore, the prediction error of neural network training is applied as compensation signal to compensate the threshold signal.Finally, the Allan variance is used to analyze the error of the signal before and after compensation.The result shows that this method can reduce the random error of the laser gyro, compared to the random error of the original signal, it is reduced by 38 %~89.8 %,and the variance is reduced by two orders of magnitude.

关 键 词：激光陀螺 阈值去噪 ELMAN神经网络 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置] TH703[自动化与计算机技术—控制科学与工程]