ABSTRACT
Fast-growing traffic volumes and loads will increasingly become a safety hazard for existing bridges, especially in developing countries. One of the problems induced by heavy traffic loads is the fatigue damage accumulation in the welded joints of steel bridges. This study developed a computational framework for fatigue reliability evaluation of orthotropic steel bridge decks using site-specific traffic data. Long-term monitored traffic data from a highway was utilized to simulate a stochastic traffic load model, and the traffic growth ratio was also included to consider future traffic. In order to solve the time-consuming problem in the bridge finite element analysis using the traditional stress spectrum simulation approach, a novel and efficient computational framework was presented based on neural networks. The proposed computational framework was subsequently utilized to evaluate fatigue reliability of the welded joints in a steel bridge deck. Numerical results show the efficiency and accuracy of the framework. Considering the annual growth rate of the traffic volume and the gross vehicle weight (GVW) are both 0.5%, the fatigue reliability index of the bridge in the 100th year decreases from 5.94 to 2.87 and 0.92, respectively. The proposed computational framework is expected to be utilized for fatigue reliability evaluation of more general steel bridges.
