基于GSA-LSTM的光纤陀螺标度因数迟滞误差补偿研究

Research on FOG Scale Factor Hysteresis Error Compensation Based on GSA-LSTM

摘要:

光纤陀螺是光纤捷联惯导系统的核心器件，已经广泛应用于航空、航天、航海等领域。标度因数是影响光纤陀螺动态性能的主要因素。由于光纤陀螺内部的光电器件对温度变化十分敏感，受温度影响会产生标度因数误差，进而影响了光纤陀螺的精度。在变温环境下，各光电器件受热不均匀，使得标度因数误差具有迟滞性。本文针对光纤陀螺标度因数迟滞误差进行了研究，搭建了多温度传感器测量系统，通过实验结果明确了标度因数迟滞误差的来源及特性。基于此，提出了一种基于引力搜索算法（Gravitational Search Algorithm，GSA）和长短期记忆（Long Short-Term Memory，LSTM）网络的误差补偿算法，通过GSA对LSTM网络的参数进行寻优，并将LSTM模型用于标度因数迟滞误差补偿。实验结果表明：本文提出的算法将全温范围内的标度因数误差的峰值由835.1×10^-6减小至38.02×10^-6。通过与多层感知机（Multilayer Perceptron，MLP）和传统LSTM等算法的补偿结果进行对比，进一步验证了本文提出的算法在标度因数迟滞误差补偿方面的有益效果。

Abstract:

Fiber Optic Gyroscope (FOG) is the core component of the fiber optic strapdown inertial navigation system, which has been widely used in aviation, aerospace, navigation and other fields. The scale factor is the main factor affecting the dynamic performance of FOG. Because the photoelectric devices inside FOG are highly sensitive to temperature vaviations, the scale factor error will be produced under the influence of temperature, which will affect the precision of FOG. In the variable temperature environment, each photoelectric device is heated unevenly, which leads to the hysteresis of the scale factor error. In this paper, the scale factor hysteresis error of FOG is studied, and a multi-temperature sensor measuring system is built. The source and characteristics of the scale factor hysteresis error are determined by the experimental results. Based on the above analysis, an error compensation algorithm based on gravitational search algorithm (GSA) and long short-term memory (LSTM) network is proposed. The parameters of LSTM network are optimized by GSA, and the LSTM model is used to compensate the scale factor hysteresis error. The experimental results show that the peak-to-peak value of scale factor error in the whole temperature range is reduced from 835.1×10^-6 to 38.02×10^-6 by the proposed algorithm. By comparing with the compensation results of multilayer perceptron (MLP) and traditional LSTM algorithm, the effectiveness of the proposed algorithm in scale factor hysteresis error compensation is further verified.