基于神经网络的捷联惯导系统温度补偿

Temperature Compensation of SINS Based on Neural Network

基于神经网络的捷联惯导系统温度补偿算法,以观测温度为输入变量,以数字滤波器输出信号为输出。采用BP算法训练网络。其对输出信号进行数字滤波,消除系统随机误差和采样电路噪声的影响。再与初始零位比较,得到只与温度变化有关的信号,以此作为网络学习的目标信号,得到陀螺随温度变化信号的估计值。实验结果表明,该算法不仅能获得较高的精度,还能提高系统实时性。

The temperature compensation technology of SINS was based on neural network. The observation temperature was taken as the input variable, and the output signal of digital filter was taken as the output, The BP algorithm was adopted to train network. The digital filter was used in output signal; the influence of system random error and sampling circuit noise was eliminated, Then, compared with the initial null, the signal which only related with the temperature change were acquired and the signal was taken as the target signal of network studying; at last, the evaluation value of top which varied with temperature change signal was acquired. The test results showed that this algorithm was not only improved the higher accuracy, but also improved the system real-time efficiency.