Temperature Drift Modeling and Compensation of a Quartz Differential Resonant Accelerometer Based on a Bi-LSTM Network

Thermal stability constitutes a critical performance characterization for accelerometers, particularly in high-precision applications such as autonomous navigation systems, satellite orbital control, and IMU-integrated platforms. Quartz resonant accelerometers have been paid more attention due to their exceptional stability and repeatability, while their operational precision remains vulnerable to ambient temperature fluctuations. This study theoretically and experimentally models the frequency–temperature characteristic of a quartz differential resonant accelerometer under varying temperatures based on a standard test procedure. Then, the bidirectional long short-term memory (Bi-LSTM) network is used to model and compensate for the temperature drift of the accelerometer. In addition, several compensation methods are compared, such as variable coefficients regression (VCR) and support vector machine (SVM). Results suggest that the temperature repeatability compensated by Bi-LSTM is improved by more than 70% in environments with a temperature ramp rate, and the compensation effect of Bi-LSTM is better than that of VCR and SVM methods.