The goal of this thesis is to propose a stress state dependent ductile fracture criterion for shell elements that can be used for crash simulations for example collisions of ships with offshore structures. Fracture has been extensively researched for solid elements for finite element analysis in Bai & Wierzbicki [2009], among others. Shells have also been investigated as a special case of the findings of solid elements. These elements are mostly used in large, thin-walled structures such as ships, as they reduce the minimum required elements in a model, which reduces computation time. Dunand & Mohr [2010] reported that these elements are unable to capture through-thickness necking. This necking localizes strain and creates stress concentrations that decrease the validity of plane stress assumption of shells. In contrast to prior research, this thesis focuses on shell elements that are in the range of 0.5 to 5 times the material thickness by proposing a phenomenological correction method. This method estimates the effective strain perpendicular to the neck at the instance of fracture based on the assumption that the stress state does not change at the onset of necking and a simplified analytical model of the neck itself. The preference for a correction method is validated from a numerical plate model based on analysis proposed by Marciniak et al. [1973]. From comparison with the numerical results, it can be concluded that the proposed method performed better than existing methods for smaller shell elements lengths and similar for larger shell elements. The improvement is most dependent on the accuracy of the necking strain. A complete simplified approach for testing and effective fracture strain prediction is proposed and validated with calibration and raking experiments from Haag et al. [2017] and Bijleveld [2018].