The accumulated damage in aging Steel Structures, especially due to fatigue, is considered as a critical phenomenon that affects safety and serviceability of civil engineering structures. Although, fatigue damage is influenced by various parameters such as the frequency of loading, sequence of load application, material properties, geometry, etc, in practice simplified SN curve is typically used for condition assessment. In order to mitigate risks of catastrophic failure resulting from fatigue brittle nature, even in normally ductile materials, researchers have generated several non-linear damage models to predict the remaining service life of the structure considered. These models are mainly based on the S-N curve, material dependent parameters and loading conditions. However, due to the complexity of the fatigue phenomenon and expensive-long term full scale experimental testing, the models presented in literature have shown high degree of uncertainty due to simplifications of mathematical models, parametric uncertainties and varying loading conditions. Furthermore, the usage of S-N curve generated from experimental work is limited to identical loading mechanism and constant boundary conditions. Therefore, this study presents a structural health monitoring approach to overcome the limitation and inaccurate estimation of damage quantification models. The suggested framework relies on fatigue damage prediction models incorporated with real-time damage records. All sources of uncertainty are incorporated in the health monitoring scheme to guarantee an optimal statistical identification of the state damage. The accuracy and robustness of the presented scheme will be assessed through a set of controlled experiments and numerical simulation of real case scenario.