Modeling of fatigue crack growth plays a key role in risk informed inspection and maintenance planning for fatigue sensitive structural details. Probabilistic models must be available for observable fatigue performances such as crack length and depth, as a function of time. To this end, probabilistic fracture mechanical models are generally formulated and calibrated to provide the same probabilistic characteristics of the fatigue life as the relevant SN fatigue life model. Despite this calibration, it is recognized that the rather complex fracture mechanical models suffer from the fact that several of their parameters are assessed experimentally on an individual basis. Thus, the probabilistic models derived for these parameters in general omit possible mutual dependencies, and this in turn is likely to increase the uncertainty associated with modeled fatigue lives. Motivated by the possibility to reduce the uncertainty associated with complex multi-parameter probabilistic fracture mechanical models, a so-called normalized fatigue crack growth model was suggested by Tychsen (2017). In this model, the main uncertainty associated with the fatigue crack growth is captured in only one parameter. In the present contribution, we address this new approach for the modeling of fatigue crack growth from the perspective of how to best estimate its parameters based on experimental evidence. To this end, parametric Bayesian hierarchical models are formulated taking basis in modern big data analysis techniques. The proposed probabilistic modeling scheme is presented and discussed through an example considering fatigue crack growth of welds in K-joints. Finally, it is shown how the developed probabilistic crack growth model may be applied as basis for risk-based inspection and maintenance planning.