The proliferation of numerical tools in engineering science enables a deeper understanding of mechanical systems. However, large-scale finite element models used within the design process of safety-relevant automotive structures merely pretend to provide a predictability of high relevance and accuracy. In reality, these models exhibit uncertainty of polymorphic nature, stemming from various sources, which needs to be appropriately assessed and effectively incorporated in the design process. An efficient and unified approach to incorporate uncertainty within the numerical treatment of complex systems is possibility theory. It allows for a description of partial knowledge and ignorance in a stringent mathematical way. Possibility theory can be used for statistical reasoning and the propagation of families of probability distributions. However, the solution of the possibilistic forward problem and, therefore, the propagation of possibility distributions through numerically complex models is computationally demanding. Simplified models are often used for uncertainty analysis, with inaccurate results being accepted to reduce the computation time. This issue motivates the methodical core of the present thesis. The idea is to combine the high-fidelity model's high accuracy and the low costs of a strongly simplified low-fidelity model while limiting their respective disadvantages. The proposed possibilistic multifidelity approach exploits the functional dependency between the high- and low-fidelity models in a possibilistic way and achieves sufficiently accurate results, given only a low number of expensive high-fidelity model evaluation.weiterlesen