A novel maximum entropy-based Bayesian network for multi-scale corrosion-fatigue damage prognosis for slender coastal bridges is planned. The framework fuses the information and knowledge from the material level, the structural level, and the system level for the probabilistic prognosis and reliability assessment. The inter-correlations among different levels of nodes in the network are developed by using coupled dynamic analysis and corrosion-fatigue damage analysis. Advanced experimental testing for fatigue and simulation methods will be combined together for the physics-based prediction of remaining useful life of costal bridges. Uncertainties will be propagated through the Bayesian network and the system level reliability will be updated and reassessed. The Bayesian network can update itself with information from experimental measurements, field observation, and historical experiences. In this methodology, coupled structure dynamics model will capture the realistic service and environmental loads during the lifetime span of the bridges.