ABSTRACT
Orthotropic bridge decks are fatigue critical components since they directly suffer from cyclic traffic loads. This study utilized a stochastic fatigue truck load model to simulate the fatigue stress spectra and to evaluate fatigue reliability of orthotropic steel bridge decks. A computational framework associated with deep learning technique was presented to deal with the uncertainty-induced computational complexity. In the deep learning approach, initially, several uniformly designed training samples was generated accounting for traffic load parameters including vehicle type, axle weight, etc. Subsequently, these training samples were in- putted to the finite element model of the orthotropic steel bridge deck to simulate the equivalent stress range under each truck load as the output samples. Finally, the training and output samples were connected to the deep learning machine with a Gaussian Kernel function, and the accuracy of the learning machine was checked in a parametric study. Based on the deep learning technology, the fatigue stress spectra of a prototype bridge under actual traffic load was simulated It is concluded that the machine learning approach provides an intelligent and efficient approach for probabilistic fatigue analysis of orthotropic bridge decks.
