Ultra-Low Frequency Micro-Thrust Drifting Compensation Using Autoencoder Prediction Based Auto-Zero Mechanism

Micro-thrust measurement technology plays a crucial role in detecting space gravitational waves. However, the precision of micro-thrust measurements is significantly challenged by ultra-low frequency stochastic drift. To address this issue, this paper proposes a multiple scale drifting compensation method for ultra-low-frequency micro-thrust measurement based on auto encoder prediction. The method extracts a short-time micro-scale drift skeleton of the raw signal using a bidirectional Butter worth low-pass filtering approach. An optimized auto encoder model is then employed to predict a long-time sequence of the low-frequency drift skeleton. The predicted skeleton sequence is removed from the raw signal to obtain a micro-thrust measurement signal with drifting compensation. The effectiveness of this auto encoder predictionbased multiple scale compensation method for low-frequency drift in micro-thrust measurement is validated by reducing the noise level of the measured micro-thrust data to 0.30 μm/Hz1/2 at a frequency of 8 × 10-5 Hz, demonstrating its effectiveness in suppressing low-frequency noise.