With the surge in the number of satellites in mega-constellations， the computational resources required for constellation orbit simulation and visualization have increased dramatically. To address the issue of inefficient visualization of mega-constellations on mobile devices with limited storage and computational power， this paper proposes an orbit model simplification method based on the HDBSCAN algorithm. This method employs a two-step clustering strategy： first， categorizing the constellation orbits， and then further subdividing the orbital planes within each category， using the cluster centers as simplified representations. Using the Starlink constellation as an example， experiments show that the orbit data is compressed from 6 778 to 279 tracks. Under the experimental hardware environment， the visualization frame rate is significantly improved， and the simplified orbit model effectively retains the structural characteristics of the mega-constellation.