Geosynchronous Orbit Satellite Maneuver Detection Method Based on Dual-Model Parallel GLRT

Ground-based radar can obtain high-precision radial velocity measurements through coherent integration Doppler processing, providing an important data source for geostationary orbit (GEO) satellite maneuver detection. An orbital maneuver causes a persistent mean shift in radial velocity residuals, whereas attitude motion or tumbling due to malfunction induces velocity scintillation, which manifests as a sudden variance amplification. Since both anomalies cause velocity residuals to deviate significantly from normal levels, a single detection model can easily misclassify one as the other. To address this problem, a maneuver detection method is proposed that combines precise orbit determination with a dual-model parallel Generalized Likelihood Ratio Test (GLRT). A high-accuracy initial state is first obtained through the Gibbs method and weighted least squares. Two detection models are then constructed based on distinct residual characteristics: mean shift for maneuver detection and variance amplification for scintillation detection. An ε-contaminated Gaussian mixture GLRT detector is introduced to suppress the influence of outliers. Combined with Monte Carlo threshold calibration and a maneuver-priority competitive decision strategy, the method effectively distinguishes between the two anomaly types. Simulation results demonstrate that under strict false alarm constraints, the detection probability for both maneuvers and velocity scintillation exceeds 0.9, effectively eliminating mutual misclassification. The method achieves stable detection of weak maneuvers lasting tens of seconds with radial velocity increments of approximately 0.03 m/s, demonstrating strong engineering application potential.