A problem of modeling fracture and failure preceded by large-scale yielding of ductile shells is consideredfrom the point of view of large-scale structural analysis. Special emphasis is placed on the computational efficiency of the constitutive formulation. In this context, a formulation is sought that is embedded in the shell mechanics framework and can be easily implemented into a large-scale explicit dynamic finite-element code without precluding vectorization or parallelization. This is achieved through the elasto-plastic damage constitutive model for finite-element analysis of plates and shells. The proposed damage model is purely phenomenological, with a scalar damage parameter, which has no physical interpretation, except that it represents on a global scale the micromechanical changes undergone bythe material during the process of necking and fracture. The localization leading to softening and fracture is represented by the damage calibration function, with exponential damage growth after the onset of necking. The proposed phenomenological damage model uses a general plasticity and shell mechanics framework, which makes it general and easily implementable into existing finite-element codes. The proposed formulation has been implemented into the explicit dynamic finite-element software code EPSA (Atkatsh et al., 1980, 1983).