Dam deformation prediction system fusing graph neural networks and transfer learning

The dam deformation prediction system, which integrates graph neural networks and transfer learning, solves the problems of insufficient spatial correlation modeling and cross-engineering knowledge reuse in existing technologies. It achieves stable and reliable prediction of dam deformation trends and is suitable for high-precision prediction under complex engineering conditions.

CN121834620BActive Publication Date: 2026-05-29CHANGJIANG SPATIAL INFORMATION TECH ENG CO LTD (WUHAN) +1
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
CHANGJIANG SPATIAL INFORMATION TECH ENG CO LTD (WUHAN)
Filing Date
2026-03-09
Publication Date
2026-05-29

AI Technical Summary

Technical Problem

Existing dam deformation prediction technologies suffer from insufficient spatial correlation modeling capabilities, unstable prediction accuracy under limited sample conditions, and difficulty in reusing cross-engineering knowledge, making it difficult to meet the demand for high-precision and strong generalization prediction under complex engineering conditions.

Method used

The dam deformation prediction system, which integrates graph neural networks and transfer learning, constructs a spatial structure model by uniformly processing dam monitoring data, learns the spatial correlation characteristics between monitoring points, and introduces a cross-engineering transfer learning mechanism to achieve stable and reliable prediction of dam deformation trends.

Benefits of technology

It improves the spatial correlation modeling capability for dam deformation prediction, enhances the prediction accuracy and stability under limited monitoring data or changing engineering conditions, is applicable to dam projects with different structural forms and operating conditions, reduces reliance on long-term historical monitoring data, and lowers model building costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121834620B_ABST
    Figure CN121834620B_ABST
Patent Text Reader

Abstract

The application relates to a dam deformation prediction system combining a graph neural network and transfer learning, which comprises a data processing unit, a spatial structure modeling unit, a spatial correlation feature processing unit, a time evolution feature processing unit, a cross-project model transfer unit and a prediction result output unit. Through preprocessing of dam multi-monitoring point deformation data and combining with a spatial layout relationship of monitoring points to construct a spatial structure model, the system learns the cooperative deformation features among the monitoring points under the constraint of the spatial structure, simultaneously models the evolution law of deformation with time, realizes dam deformation prediction, further introduces a transfer learning mechanism, transfers the learned deformation features and model parameters in a source project to a target project, and completes prediction modeling under the condition of limited monitoring data. The system can consider spatial correlation features, time evolution laws and cross-project applicability, and improves the stability and engineering applicability of dam deformation prediction.
Need to check novelty before this filing date? Find Prior Art