The prediction of crack initiation, propagation, and ductile fracture in a large scale metallic structure has posed a great challenge to the design and certification community. The complexity in component geometry, material heterogeneity, and the 3D stress distribution, will likely make the crack growth curvilinear. An accurate 3D stress prediction is essential to predict the ductile fracture that is controlled by void nucleation, growth and coalescence. To alleviate the computational burden associated with the use of a 3D solid finite element modeling coupled with a remeshing of the ductile failure prediction of a large scale structure, a novel approach based on the coupling of an extended finite element for shell elements and a plane strain core characterization of a cracked region is developed to efficiently simulate an arbitrary crack growth in a large scale thin-walled structure and its associated load deflection curve while retaining a sufficient level of computational efficiency. The methodology is implemented in Abaqus’ explicit solver via its VUEL where a kinematic description of a cracked shell is accomplished via the use of phantom-paired elements. A plane strain core that controls the mixture of plane stress and plane strain components is introduced for a rational representation of localized plasticity induced 3D stress state for the prediction of a ductile failure initiation. A cohesive injection is applied on the newly created crack surface to dissipate the energy during the crack propagation. The resulting XSHELL toolkit for Abaqus is applied for the ductile failure prediction of the 2012 Sandia fracture challenge problem. A blind prediction is performed first using geometry independent plane strain core parameters followed by a refined analysis based on a set of well-calibrated material and plane strain core parameters from Sandia’s tensile and compact testing data.