Research on Temperature Drift Compensation of Electrostatic Suspended Accelerometer Based on BP Neural Network

The electrostatic suspended accelerometer is affected by temperature during operation which can couple with the measured acceleration signal in the low frequency band, resulting in output drift. To further improve its measurement accuracy, a temperature drift compensation method based on neural network is proposed. In order to simulate the actual on-orbit environmental conditions of an electrostatic suspended accelerometer, a temperature control experiment was conducted in a thermal vacuum tank to collect the temperature of the temperature measurement point, as well as the corresponding accelerometer displacement detection and feedback voltage data. By utilizing the nonlinear mapping ability of BP neural network, a compensation model was established for the temperature measurement point temperature of the electrostatic suspended accelerometer and the impact of the electrostatic suspended accelerometer temperature drift. The magnitude of the temperature drift effect of the six output channels is directly predicted from the temperature data of the temperature measurement points collected by the electrostatic suspended accelerometer. The experimental results show that the temperature drift of the electrostatic suspended accelerometer can be effectively suppressed after using the proposed neural network model for temperature drift compensation.