Predictive Models of Fatigue Failure in Steel Bridges under Severe Dynamic Loading: An Advanced Fracture-Mechanics, Probabilistic, and Digital-Twin Framework

Abstract

Sudden failures of steel bridges are frequently associated with fatigue cracks that initiate at welded joints, geometric discontinuities, bolted connections, orthotropic decks, diaphragm attachments, and corrosion-damaged regions. Although fatigue failure is usually a progressive phenomenon, its final stage can occur abruptly when a subcritical crack reaches a critical size and the stress-intensity factor exceeds the fracture toughness of the material. Under high dynamic loading, the conventional static stress-range approach may be insufficient because structural inertia, resonance, traffic-induced vibration, impact effects, mixed-mode fracture, corrosion, residual stress, and uncertainty in the actual load spectrum interact simultaneously.This paper develops an advanced predictive framework for steel-bridge fatigue failure based on the coupling of structural dynamics, elastic and elastic–plastic fracture mechanics, stochastic traffic loading, probabilistic reliability, sensor-based condition assessment, and digital-twin technology. The central mathematical model combines the dynamic equation of motion, rainflow-derived stress spectra, nonlinear fatigue damage accumulation, crack-growth laws based on the Paris–Erdogan and Forman equations, and a limit-state reliability function. The proposed framework further incorporates Bayesian updating and physics-informed machine learning to reduce uncertainty using strain gauges, accelerometers, acoustic-emission sensors, weigh-in-motion systems, and machine-vision inspection. The most reliable fatigue-prediction technology is not a single predictive algorithm but a multi-layered system combining mechanics-based analysis, continuous monitoring, probabilistic inference, and conservative maintenance decision-making.