基于LSTM神经网络的机载光纤陀螺温度冲击误差补偿技术OA

The measurement accuracy of the onboard fiber optic gyroscopes could be reduced by environmental temperature shocks,consequently impacting the flight accuracy of the aircraft.A temperature error compensation model based on long short-term memory(LSTM)neural networks was proposed in this paper to improve the measurement accuracy of fiber optic gyroscopes under temperature shock.The zero bias and scale factor of the fiber optic gyroscope were predicted and corrected in real-time using the LSTM network,improving its measurement accuracy.Experimental results showed that under temperature shock,the scale factor error was compensated by the LSTM prediction model,which was less than 30ppm.The zero bias stability was improved by 0.0034(°)/h compared with the conventional linear fitting compensation model.In dynamic experiments,when the input of the turntable was set to 20°/s,the gyroscope output was stabilized in the range of 19.999～20.001(°)/s after LSTM compensation,and the error of the gyroscope original output was reduced by 0.008(°)/s.The changes of the zero bias and scale factor of the airborne fiber optic gyroscope under temperature shock were more effectively compensated by the LSTM network.The stability of the inertial navigation of aircraft was enhanced.