Orbit Anomaly Detection Technology Based on Segmentation Optimization

In recent years, the large number of deployments of low-orbit giant constellation satellites have a significant impact on the operational safety of low-orbit satellites in orbit, and it is critical to detect orbital anomalies of constellation satellites in time. As a result, this study picks the Starlink satellite constellation as the research objects for detecting orbit anomaly and presents an improved orbit anomaly detection approach - segmentation optimization. Based on the concept of dynamic optimization, the approach improves the orbit anomaly detection method by using satellite orbit semi-major axis data as an analysis parameter, transforming it from threshold screening to optimization search. First, by assuming the segmentation points and randomly distributing them across the entire semi-major axis data series, the data in the time window are randomly segmented. Based on the variance of the segmented data, each segment's loss function is built. The iterative function are then designed using the loss function for optimization iterations. In order to determine the optimal segmentation method, the random segmentation is finally optimized with the goal of minimizing the sum of the total loss functions. This research finds that the anomaly detection effect is the best for the semi-major axis data by the differential processing after evaluating a range of data. The segmentation optimization approach has various sensitivities to different data after removing the noisy data. In conclusion, this work uses the TLE (Two-line element) data and the ephemeris data from the Starlink satellite for example verification, which was launched on 28 February 2023. The method's effectiveness in detecting orbital anomaly of constellation satellites is demonstrated, which are simple and efficient.