Abstract: To address the performance bottleneck caused by frequent network requests when the Cesium framework loads large-volume BIM models in a highway BIM platform, this paper proposes a non-intrusive caching optimization scheme. By intercepting XMLHttpRequest requests through the ajax-hook library and combining IndexedDB with LocalStorage, efficient persistent storage and retrieval of model data are achieved. Experiments show that this scheme significantly improves model loading efficiency, reduces network request time consumption by over 90%, and lowers memory occupancy to within 5%, providing effective technical support for the visualization of large-volume BIM models.