Operation and maintenance method and device of equipment, nonvolatile storage medium and electronic equipment
By constructing a target graph and a multi-level attention framework, and combining equipment operation characteristics and multimodal data, the problem of low operation and maintenance efficiency in existing technologies is solved, and efficient anomaly root cause localization and fault diagnosis are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA TELECOM CORP LTD
- Filing Date
- 2026-02-11
- Publication Date
- 2026-05-29
AI Technical Summary
Existing technologies cannot comprehensively consider the operational data of different modes of various devices in the system, resulting in low operation and maintenance efficiency and an inability to accurately locate the root cause of abnormal areas and devices.
By constructing a target graph, nodes and devices are matched one-to-one. Node features include the device's operating characteristics and time and location information. Abnormal areas are identified using correlation indicators, and the causes of abnormal device failures are determined through multimodal operating data analysis. A multi-level attention framework is used for deep fusion and diagnosis.
It enables efficient and accurate location of the root cause of anomalies, improves operation and maintenance efficiency, and can comprehensively consider multimodal data for accurate diagnosis and fault analysis.
Smart Images

Figure CN122120109A_ABST