BP神经网络在半球谐振陀螺仪零偏温度补偿中的应用

Application of BP Neural Network in Hemispherical Resonant Gyroscope Zero Bias Temperature Compensation

半球谐振陀螺仪(HRG)的核心部件(半球谐振子)容易受到温度变化的影响,从而导致陀螺仪输出产生漂移,很大程度上限制了半球谐振陀螺仪精度的提升,因此需要对陀螺仪零偏进行补偿。首先结合半球谐振陀螺仪的工作原理对其零偏漂移产生的机理进行分析,利用最小二乘法和反向传播(BP)神经网络建立了半球谐振陀螺仪零偏漂移模型,同时利用建立的模型进行了零偏补偿,补偿效果显著。实验结果表明:补偿后零偏稳定性相较于补偿前和最小二乘法分别提高了80%和60%以上,BP神经网络有着比最小二乘法更优的非线性拟合能力。

The core component(hemispherical resonator)of the hemispherical resonant gyroscope(HRG)is susceptible to temperature changes,which leads to output drift in the gyroscope,greatly limiting the improvement of the accuracy of the hemispherical resonant gyroscope,thus the gyroscope zero bias needs to be compensated.Firstly,combined with the working principle of the hemispherical resonant gyroscope,the generated mechanism of the zero bias drift was analyzed.The zero bias drift model of the hemispherical resonant gyroscope was established by the least square method and back propagation(BP)neural network.Meanwhile,the established model was used for zero bias compensation,and the compensation effect was remarkable.The experimental results show that compared with those before the compensation and by the least square method,the zero bias stability after the compensation is improved more than 80%and60%,respectively,showing the BP neural network has a better non-linear fitting ability than the least square method.