This thesis introduces a numerical model for describing damage processes and fracture in a severely strain-hardened aluminum alloy. The macroscopic phenomenology and fracture mechanisms of sheet-metal specimens as well as thin-walled structures have been studied experimentally and numerically for various loading conditions. The model’s essential feature is capturing different fracture mechanisms under various stress states. This has been accomplished by introducing a stress state dependent damage variable in the elasto-plastic constitutive behavior. A custom material model with nonlocal regularization has been implemented into commercial finite element software. The model has not only been applied to modeling uncracked and pre-cracked alloy specimens, but has also been put at test in a complex loading scenario occurring in a real world mechanical component. The model’s true predictive ability has been assessed thereby. The model predictions for various geometrical as well as loading scenarios are found to be in very good agreement with experimental findings. This has been accomplished even for fundamentally different damage and fracture processes without any changes of the material model. Furthermore, a simplified method for determination of material constants has been introduced that assists traditional inverse identification of parameters by artificial intelligence based on machine learning methods. All research findings together encourage the general use of the constitutive model for the design of aluminum and other sheet-metal structures.