Viscous interference torque analysis of liquid floated gyroscope based on multi-source data learning

Abstract

The liquid floated gyroscope (LFG) is a core instrument of the inertial navigation system, which is used to obtain the angular motion information of the carrier. Under the thermal effect of electronic components, the floating oil inside the instrument flows slowly, thereby introducing a viscous interference torque (VIT) acting on the floater surface, which will affect the output accuracy of the instrument. Since the magnitude of VIT is extremely small, there is currently no effective means to obtain its accurate value. Therefore, this work aims to combine the advantages of experiment and simulation and then proposes a feasible method to predict the VIT. First, a gas–liquid–solid three-phase coupled heat transfer model of the LFG was established, and the relative error between the calculated temperature and the test temperature of the calibration point is 3.5%. The computational fluid dynamics method was adopted to calculate the VIT under different oil temperature distributions; the backpropagation neural network algorithm was selected to build a network model between the temperature distribution and the VIT, and the model fitting accuracy was 0.99. Then, the actual temperature distribution of the gyro oil was obtained through experiments, which was taken as an input of the neural network to predict the VIT. The relative error between the predicted and simulation values under the same conditions was 4.18%. The proposed method provides a feasible scheme to predict the microscopic VIT that is difficult to measure, which provides a theoretical reference for the accuracy improvement of LFGs.