Dynamic Two-Way Parallelization of Non-Intrusive Methods for Uncertainty Quantification

摘要

This work deals with a two-way parallelization method of numerical flow simulations under uncertainties with sampling methods for the uncertainty quantification. In general, for such simulations, there are two possibilities to distribute the computation on multiple processors in order to reduce the overall computing time. One approach is to divide the flow domain into several blocks which can be calculated by different processors. The other is to solve the various independent deterministic problems that arise from the sampling method for uncertainty quantification in parallel. Both methods have advantages and disadvantages and can be applied simultaneously, depending on the provided number of processors for the simulation. The presented method assigns the available processors dynamically to both parallelization methods during the simulation of the uncertain flow problem. The aim is to reduce the overall computing time compared to a static parallelization strategy.