Bridge data reconstruction method based on multi-scale spatio-temporal fusion and uncertainty perception

CN121959459BActive Publication Date: 2026-06-09JILIN UNIVERSITY
View PDF 8 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
JILIN UNIVERSITY
Filing Date
2026-03-30
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Existing bridge monitoring technologies, when faced with complex environments, lack in-depth exploration of spatiotemporal features, dynamic adaptability, and sufficient integration of multi-scale features with physical laws. Furthermore, they lack the ability to quantify the uncertainty of reconstruction results, which may lead to reconstruction results that violate physical laws and pose a risk of safety misjudgment.

Method used

Multi-scale convolution is used to capture temporal patterns at different frequencies. A dual-path temporal coding architecture is used to fuse long short-term memory. A dynamic attention mechanism is introduced to mine spatial correlations of sensors. Based on a probabilistic model, the reconstruction results with confidence intervals are output.

Benefits of technology

It significantly improves the accuracy and reliability of bridge monitoring data reconstruction, can adapt to dynamic sensor correlation, provides a basis for engineering decision-making, and improves the interpretability and security of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121959459B_ABST
    Figure CN121959459B_ABST
Patent Text Reader

Abstract

This invention relates to a bridge data reconstruction method based on multi-scale spatiotemporal fusion and uncertainty perception, belonging to the field of civil engineering structural health monitoring and artificial intelligence data processing technology. It aims to address the problems of insufficient dynamic adaptability in spatiotemporal feature mining and inadequate multi-scale feature fusion in existing technologies. The method includes: collecting multi-dimensional sensor time-series data and generating a structured dataset; extracting multi-scale temporal features using a parallel causal convolutional network; constructing a dual-path temporal coding architecture including an LSTM main path and a GRU auxiliary path; after hierarchical attention and adaptive temporal fusion, introducing sensor spatial correlation dynamic modeling to generate high-dimensional enhanced features; using a dual-branch network to predict mean and heteroscedasticity uncertainties, and performing end-to-end optimization training through a hybrid loss function. The trained model is used to output reconstructed bridge data with uncertainty confidence intervals. This invention can significantly improve the accuracy and reliability of bridge monitoring data reconstruction.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Structural health monitoring data reconstruction method and system

    CN118133105A

  • Bridge sensor time sequence missing data reconstruction method based on EMD and GRU

    CN118193504A

  • Bridge temperature monitoring data reconstruction method and system

    CN119025788A

  • Mass fatigue damage monitoring data anomaly identification method based on time series data reconstruction and adaptive threshold

    CN119357854A

  • Fault detection method of multi-dimensional data reconstruction mode based on transformer

    CN120030312A