Bridge steel structure corrosion prediction method based on ensemble learning algorithm

By integrating learning algorithms with sensors and hyperspectral cameras, accurate corrosion rate prediction and dynamic early warning of bridge steel structures have been achieved, solving the problems of low prediction accuracy and delayed early warning in existing technologies and improving the corrosion monitoring effect of bridge steel structures.

CN122153767APending Publication Date: 2026-06-05CHONGQING JIAOTONG UNIV
View PDF 1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN Β· China
Patent Type
Applications(China)
Current Assignee / Owner
CHONGQING JIAOTONG UNIV
Filing Date
2026-01-27
Publication Date
2026-06-05

AI Technical Summary

Technical Problem

Existing methods for monitoring corrosion of bridge steel structures are insufficient to fully characterize the corrosion evolution process under complex environments, have low prediction accuracy, and the early warning system is at risk of lag or false alarms.

Method used

An ensemble learning algorithm is used to combine temperature, humidity and concentration sensor data to acquire corrosion images through a hyperspectral camera. U-Net is used for image segmentation to calculate the corrosion area ratio, a GRU time series prediction model is constructed to generate a corrosion threat index and set dynamic early warning levels.

Benefits of technology

This improved the accuracy of corrosion prediction and the flexibility of early warning strategies, enabling precise prediction and timely early warning of corrosion rates for bridge steel structures.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122153767A_ABST
    Figure CN122153767A_ABST
Patent Text Reader

Abstract

The application provides a bridge steel structure corrosion prediction method based on an ensemble learning algorithm, relates to the technical field of material corrosion monitoring, and adopts a temperature, humidity, Cl ‑ The environmental parameter sensor collects environmental parameters, corrosion area detection is carried out in combination with a hyperspectral camera, image segmentation is carried out by using U-Net, the corrosion area is accurately extracted and the corrosion area ratio is calculated, so that the integrity and accuracy of data are improved, the mass loss method is used to measure the mass loss per unit area under different environments, the corrosion rate is calculated in combination with historical data, the mapping relationship from the environmental parameter to the corrosion rate is constructed by using a GRU prediction model, the prediction accuracy is improved, the future corrosion rate is predicted by using the trained GRU model, and a corrosion threat index is generated, the warning level is set based on a dynamic proportional interval in combination with the corrosion area ratio, so that the warning strategy is more flexible and adaptive.
Need to check novelty before this filing date? Find Prior Art