一种在轨航天器技术支持系统智能运维方法OA

To address the challenges of low association across multi-source heterogeneous data,complex fault evolution mechanism,and the difficult root-cause localization in the technical sup-port system of on-orbit spacecraft,an intelligent operation and maintenance method that in-tegrates physical topology constraints with multi-modal data is proposed in this paper.Firstly,it designs a semantic alignment and cross-modal fusion mechanism to construct a multi-layer knowl-edge graph with spatiotemporal topology awareness.Secondly,it proposes a fault evolution mod-eling algorithm that fuses dynamic graph neural networks with causal inference to quantitatively characterize the spatiotemporal propagation of faults in complex topologies,thereby improving the accuracy of real-time diagnostics,anomaly early warnings,and complex root-cause localiza-tion of the system.The test validation of the navigation in-orbit technical support system has shown that the method achieves an accuracy rate of 93%for the top 3 root cause analysis of faults,significantly enhancing the level of intelligent spacecraft operation and maintenance.