基于神经网络代数算法的电子罗盘的标定  

Calibration of electronic compass based on neural network algebra algorithm

作　　者：马斌良[1,2] 黄玉美[1] 史恩秀[1] 李引魁[1] 韩旭炤[1] 

机构地区：[1]西安理工大学机械与精密仪器工程学院,西安710048 [2]独山子石化公司,克拉玛依833600

出　　处：《仪器仪表学报》2008年第11期2304-2309,共6页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(50075069);陕西省教育厅产业化培育计划(03JC17)资助项目

摘　　要：为了标定KVH-C100型电子罗盘,分析了电子罗盘产生误差的原因,设计了电子罗盘信号采集的软硬件系统;为了减少随机因素的干扰,对采样数据进行了中值滤波;在此基础上,采用神经网络代数算法建立了电子罗盘的误差补偿模型,该模型能实现在样本空间的精确映射,具有较高的非线性逼近精度。通过实验证明该补偿模型比中值正弦误差补偿精度高,使得误差范围由标定前的-5.468 15°～1.457 96°减小到-0.492 8°～0.517 8°,重复性实验也证明该补偿模型一致性好,满足控制要求。To calibrate KVH-C100 model electronic compass, the error source of electronic compass was analyzed and the software and hardware systems for signal sampling of electronic compass was designed. In order to decrease the interference of random factors, a median filter was used to deal with the sampled data, and then the error compensation model of the electronic compass was established based on neural network algebra algorithm. The model has excellent characteristics of nonlinear approximation and can realize accurate mapping in sample space. Experiment result proves that the compensation model is more precise than the median sine error compensation and reduces the error range from -5. 468 15° -1. 457 96° to -0. 492 8° -0.517 8°. Repeat experiments also prove that the compensation model has better consistency and satisfies the requirement of the control task.

关 键 词：电子罗盘 标定 神经网络代数算法 误差补偿 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]