The next-generation satellite bus data has undergone a significant leap in terms of type, volume, and format. Traditional analysis methods rely on manual labour and single indicators, resulting in low efficiency, inadequate anomaly detection, and difficulty in addressing the challenges posed by massive data volumes and complex communication systems. To address these issues, this study focuses on innovative research into a three-dimensional framework analysis method for satellite bus data. The core objective is to establish an efficient and universal three-dimensional framework analysis system to achieve in-depth analysis of bus data and precise identification of anomalies. This paper constructs a ‘Time-Space-Type’ three-dimensional analysis framework, expanding traditional one-dimensional sequences into a ‘RT address-Message type-Timestamp’ data cube. It employs statistical threshold methods to rapidly identify anomalies in a single dimension; uses autocorrelation analysis to quantify sequence periodicity, detecting anomalies such as missing or unstable cycles; Density Peak Clustering (DPC) combined with Mahalanobis distance is used to identify composite faults with high-dimensional feature coupling; a closed-loop mechanism of ‘local detection-global verification-root cause identification’ is established for single-dimensional and multi-dimensional features to enhance anomaly identification accuracy. The software supported by this method significantly improves the efficiency of bus data processing for the entire satellite and has been applied to a specific satellite model, providing critical support for satellite fault diagnosis and health management.