环形激光陀螺人工神经网络温度漂移建模

Ring laser gyro temperature drift model based on artificial neural network

研究了系统在常温条件下开机后,壳体密闭条件下环形激光陀螺(RLG)漂移的温度特性建模问题。一系列不同环境温度条件下的RLG自升温实验表明,由于系统壳体的保温和隔温功能,系统内部温度场变化缓慢、均匀,此时漂移的变化主要与内部温度场的温度值变化相关。利用一组25℃-55℃范围内静态漂移测试数据作为学习样本,建立起基于BP和RBF神经网络的温度漂移补偿模型,并利用四组不同测试条件下测得的静态数据对文中模型进行检验。结果表明,若采用均方根误差(RMS)指标进行评价,则得到的温度漂移模型可以有效补偿RLG的漂移输出趋势项,使陀螺的稳定性指标提高20%-40%,且BP网络建模补偿精度优于常规最小二乘中的一阶线性分段拟合,RBF网络建模优于二阶抛物线分段拟合。

The paper studies the modeling of temperature characteristics for the drift of ring laser gyro（RLG） with enclosed case after Strapdown Inertial Navigation System（SINS） being powered on under normal temperature.A set of self-rising temperature tests under different external temperature environments are made which show that the SINS box can keep the inside temperature rising gently and uniformly by its isolating capability,in which the drift is mainly related to the temperature changes of temperature field.Take the static test data from temperature 25℃ to 55℃ as training samples,we set up the drift temperature models based on BP and RBF neural network respectively,and use another four experiment data to check the validity of this method.The experiment results show that if taking the MSE（mean square error） as criteria,the stability of gyro can be improved by 20%-40%,the accuracy of BP network modeling is better than that of first-order linear piecewise fitting with Least Square（LS）,and the accuracy of RBF network modeling is better than that of second-order LS Fitting.