基于FLANN的三轴磁强计误差校正研究  

Research on correction of tri-axial magnetometer based on FLANN

作　　者：吴德会[1] 黄松岭[1] 赵伟[1] 

机构地区：[1]清华大学电机系电力系统国家重点实验室,北京100084

出　　处：《仪器仪表学报》2009年第3期449-453,共5页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(50305017);中国博士后基金(20070420358)资助项目

摘　　要：提出一种基于函数链接型神经网络(FLANN)的三轴磁强计误差修正方法。由于三轴非正交、灵敏度不一致及零点漂移所引起的误差降低了三轴磁强计的测量精度,因此有必要进行校正。本文先对与三轴磁强计系统参数有关的测量进行详细分析和理论计算;然后,设计矩阵形式的数学模型对该误差进行修正。通过构造相应的FLANN网络结构,实现对模型参数矩阵的辨识。用实际地磁场测量数据进行测试,结果表明,三轴磁强计的转向误差由800 nT修正到12 nT以下。因此,该研究为提高三轴磁强计性能提供了一种可行方法。An error correction method for tri-axial magnetometer based on functional link artificial neural network （FLANN） was proposed. Because the errors caused by nonorthogonality, different sensitivities, and zero-shifts among three axes reduce the accuracy of tri-axial magnetometer, therefore calibration is needed. Firstly, analysis and calculation of the measurement errors related to the tri-axial magnetometer system parameters were made. Then, a math- ematic model in matrix form was designed to correct these errors. Through corresponding FLANN structure, the parameter matrix of this model was identified successfully. Practical measurement data are used to test the system, and results show that the diversionary error of the tri-axial magnetometer can be rectified from 800 nT to 12 nT. So, the study provides a better way to improve the performance of the tri-axial magnetometer.

关 键 词：函数链接型神经网络 三轴磁强计 误差校正 辨识 

分 类 号：TH762.3[机械工程—仪器科学与技术]