优化BP神经网络的光纤陀螺温度漂移建模与补偿  

Modeling and compensation algorithm of FOG temperature drift with optimized BP neural network

作　　者：郭士荦 许江宁[1] 李峰[1] 何泓洋[1] 

机构地区：[1]海军工程大学导航工程系,武汉430033

出　　处：《中国惯性技术学报》2016年第1期93-97,共5页Journal of Chinese Inertial Technology

基　　金：国家重大科学仪器开发专项(2011yq12004502);国家自然科学基金(61503404);航空科学基金(20150816002)

摘　　要：光纤陀螺(FOG)温度漂移误差是影响其输出精度的主要误差源之一。针对基于传统BP神经网络FOG温度误差补偿方案适用性较差的问题,提出了优化预测数据的BP神经网络补偿算法,利用最优线性平滑技术以及滑动平均技术对神经网络待补偿数据进行预处理,可以有效减小FOG输出白噪声对温度漂移网络模型补偿精度的干扰,优化神经网络模型的补偿效果。使用FOG温度漂移实测数据对所提出的优化算法进行验证,结果表明利用本文提出的两种建模及补偿方案进行补偿后的FOG温度漂移数据标准差相比传统BP神经网络补偿方法减少50%以上。In view of the poor applicability of traditional BP neural network＇s compensation algorithm, a new BP neural network with optimized prediction data was proposed by using an optimal liner smoothing technique and a moving average technique to pre-process the data for neural network compensation, which effectively reduces the white noise disturbance＇s influences on the compensation results of neural network model. The verification of the proposed optimization algorithm is made by using FOG temperature drift data, which shows that the standard drift of FOG temperature drift data is reduced to below 50% by using the proposed two modeling and compensation schemes, compared with traditional BP neural network compensation method.

关 键 词：光纤陀螺 温度漂移 补偿方案 BP神经网络 优化算法 

分 类 号：U666.1[交通运输工程—船舶及航道工程]