Hierarchical Dynamic Bayesian Network-Based Fatigue Crack Propagation Modeling Considering Initial Defects.

Area of Science:

• Structural Engineering
• Mechanical Engineering
• Reliability Engineering

Background:

• Orthotropic steel decks (OSDs) in long-span bridges are prone to fatigue damage from repeated vehicle loads.
• Accurate modeling of fatigue crack propagation is crucial for bridge safety and maintenance.

Purpose of the Study:

• To develop a probabilistic framework for modeling fatigue crack propagation in OSDs.
• To integrate dynamic Bayesian networks (DBNs) with fracture mechanics to account for uncertainties and correlations in fatigue details.
• To predict component-level fatigue reliability.

Main Methods:

• Fracture mechanics-based uncertainty analysis of crack propagation, material properties, and crack length observations.
• Hierarchical dynamic Bayesian network (DBN) model to capture dependencies among uncertainties and multiple crack propagations.
• Stochastic traffic model using weight-in-motion (WIM) data and finite element analysis.
• Refined exact inference algorithm for posterior distribution estimation of equivalent initial crack length.

Main Results:

• The proposed method successfully models fatigue crack propagation and predicts component-level fatigue reliability.
• The probability distribution of equivalent initial crack size was derived with a larger mean value compared to conventional methods.
• Mutual correlations among multiple fatigue details were effectively considered.

Conclusions:

• The integrated DBN and fracture mechanics framework provides a robust approach for probabilistic fatigue analysis of OSDs.
• The method enhances the accuracy of fatigue crack propagation modeling and reliability prediction in bridges.
• The study highlights the importance of considering uncertainties and correlations in fatigue analysis for critical infrastructure.