Deep learning-based framework for time-dependent reliability analysis of a cable-stayed bridge with corroded PSC box...

Area of Science:

• Civil Engineering
• Structural Engineering
• Computational Mechanics

Background:

• Civil infrastructure maintenance is crucial for transportation safety.
• Prestressed concrete (PSC) box girders in bridges face risks from corrosion-induced deterioration.
• Current reliability assessment methods are often computationally intensive or neglect system-level behavior.

Purpose of the Study:

• To develop an efficient deep learning framework for evaluating the time-dependent reliability of cable-stayed bridges with corroded PSC box girders.
• To accurately model structural degradation and system-level responses under corrosion.
• To provide a robust tool for assessing the reliability of aging bridge infrastructure.

Main Methods:

• Development of a finite element model for simulating progressive structural degradation.
• Training a deep neural network (DNN) to predict flexural characteristics of PSC box girders under corrosion.
• Integration of DNN predictions into a global bridge model for Monte Carlo-based system reliability analysis.
• Incorporation of uncertainties related to environmental factors, geometry, and load redistribution.

Main Results:

• The proposed framework efficiently evaluates the time-variant reliability of bridges with corroded PSC box girders.
• Numerical investigations highlight the significant impact of corrosion on structural performance and load redistribution.
• The deep learning approach provides accurate predictions of structural behavior under varying corrosion levels.

Conclusions:

• The deep learning-based framework offers a computationally efficient and robust solution for assessing the reliability of aging cable-stayed bridges.
• This approach enhances structural health monitoring and maintenance strategies for critical infrastructure.
• Accurate reliability assessment is vital for ensuring the long-term safety and serviceability of bridges.