基于BP神经网络的GFSINS角速度预测  

Prediction of the angular velocity of GFSINS by BP neural network

作　　者：韩庆楠[1] 郝燕玲[1] 刘志平[1] 王瑞[2] 

机构地区：[1]哈尔滨工程大学自动化学院,黑龙江哈尔滨150001 [2]海军飞行学院教研部,辽宁葫芦岛125001

出　　处：《华中科技大学学报（自然科学版）》2011年第3期115-119,共5页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(60604019)

摘　　要：针对无陀螺捷联惯导系统(GFSINS)中传统角速度算法解算精度不高的问题,提出一种可避免复杂代数运算的反向传播(BP)神经网络算法来求解角速度.基于一种十加速度计构型方案,选择10个加速度计输出、采样周期和臂杆距离等12个已知量作为网络输入,以对数法得到的角速度值作为期望输出,针对5 000个样本在不同的隐含层层数、单层神经元个数以及学习步数等情况下进行网络训练,构建了一个含有30个隐含层神经元的3层BP网络模型.采用此模型对角速度进行实时预测,结果表明:网络具有很好的适应能力和实时性,角速度实时预测时间与对数法相当,且其预测精度比对数法提高大约3倍.Aimed at low precision for traditional angular velocity algorithms in gyro-free strapdown inertial navigation system （GFSINS）, a BP （back-propagation）neural network algorithm without complex mathematic computation was put forward to calculate angular velocity. Based on a ten-accelerometer configuration scheme, the accelerometer output, sample interval and fixed position were chosen as input, angular velocity got by lognormal algorithm was chosen as output, and 5 000 samples were trained in several conditions with different hiding layers, neural cells and training steps. Then a three-layer BP network model with 30 hiding layer neural cells was built. Finally, the angular velocity was predicted in real time by the model. Results demonstrate that network has strong adaptive capability and instantaneity, and compared with lognormal algorithm, prediction time is almost the same, but the prediction precision of angular velocity is nearly improved by 3 times.

关 键 词：无陀螺捷联惯导系统 角速度预测 反向传播神经网络 对数法 十加速度计 

分 类 号：U666.1[交通运输工程—船舶及航道工程]