A damage identification method for cable-stayed bridge multi-source monitoring data fusion
By synchronously collecting acceleration and displacement sensor data in cable-stayed bridges, constructing Gram matrix and CNN-LSTM model, the fusion of multi-source data is realized, solving the problems of data singularity and insufficient spatial correlation in bridge damage identification, improving identification accuracy and reliability, and making it suitable for health monitoring of cable-stayed bridges.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HANGZHOU POLYTECHNIC
- Filing Date
- 2026-01-28
- Publication Date
- 2026-06-12
AI Technical Summary
Existing bridge damage identification technologies suffer from problems such as data uniformity, insufficient spatial correlation, and low identification accuracy, making it difficult to meet the high-precision engineering requirements.
A multi-source monitoring data fusion method for cable-stayed bridges is adopted. Through a spatial-temporal dual-channel fusion mechanism, data is collected synchronously using accelerometers and displacement sensors. A Gram matrix is constructed to extract global spatial features, and dynamic features are extracted by combining the CNN-LSTM temporal module. The resulting joint features are then input into a classification and recognition model for damage identification.
It improves the comprehensiveness, stability and accuracy of damage identification, enabling early identification of minor damage and providing reliable basis for health monitoring and maintenance decisions. It is applicable to cable-stayed bridges and other long-span bridge structures.
Smart Images

Figure CN122196723A_ABST