Sheet metal forming is typically simulated using shell finite elements. These elements are computationally very efficient, however, their kinematics are not sufficient to represent complex three-dimensional stress states that may arise when the thickness of the metal sheet is comparable to the curvature radius. Solid finite elements provide a better approximation, however, at a higher computational cost. This thesis presents a computational framework, in which the benefits of both element classes are combined by using solid elements only locally where the shell elements are insufficiently precise. A mechanically-motivated model error indicator is introduced to assist the choice of the elements. The results of the adaptive computation show good agreement with the purely solid discretization, however, at a lower computational cost.The work presented in the thesis is a fully automated model-adaptive finite element simulation methodology. The methodology is aimed at precise simulation of complex deformation of thin-walled structures, such as sheet metal. The key components of the method are readily exchangeable with the necessary interfaces described in this work. While most of the core ingredients for the model-adaptive method exist within the applied finite element codes, the potential for automatic model-adaptive techniques is yet to be realized by simulation engineers. The use of commercial finite element software as "black-boxes", for the algorithms implemented in this work, allows the implementation of the presented methodology using, in principle, any general-purpose finite element software.