Abstract: Based on the random vibration analysis of a coupled train-bridge system, an approach for fatigue reliability assessment of railway steel bridges is presented. A coupled train-bridge system model is established. Train speed and track irregularities are selected as the basic random variables to perform the random vibration analysis and to identify probabilistic models of the equivalent fatigue stress range and its cyclic number of bridge components. On this basis, a fatigue limit state function is constructed to conduct fatigue reliability analysis by using the S-N curve approach. As an illustrative example, the fatigue reliability assessment is performed for a through steel truss bridge on railroad. The effects of train speed and track irregularities on the fatigue reliability of components are discussed. The results show that the presented approach can be utilized to effectively assess the fatigue reliability of railway steel bridges. Because of the randomness in train speed and track irregularities, both the train-induced equivalent fatigue stress range and its number of cycles of bridge components are indeterminate, which should be considered as random variables and can be represented by a lognormal distribution. Train speed and track irregularities can significantly affect the fatigue reliability of bridge members. The reliability indexes of fatigue-critical members increase with the enhancement in track smoothness.