Uncertainty Quantification

Most numerical simulations use deterministic input data or boundary conditions, in contrast to realistic values. In reality, all physical values are non-deterministic data and usually follow some kind of statistical distribution. Realistic data can be obtained by measurements for example. Multiple mathematical approaches exist covering realistic input data for numerical simulations. Our chair focuses on spectral methods, like the generalized Polynomial Chaos expansion. Therefore, the input data is represented by spectral expansions. This method is advantageous considering the accuracy and computational effort.

Uncertainty quantification is a main research topic of the Chair of Vibroacoustics of Vehicles and Machines. From measuring physical variables, over distribution identification techniques and up to numerical modelling of stochastic variables, field, vectors or processes.

Current research topics are the modelling of uncertainties in additive manufacturing processes, the influence of imperfections of ideally periodic structures and the definition of an uncertainty model for unit cells in the reciprocal lattice.