On the use of ensemble averaging techniques to accelerate the Uncertainty Quantification of CFD predictions in wind engineering

In this work we focus on reducing the wall clock time required to compute statistical estimators of highly chaotic incompressible flows on high performance computing systems. Our approach consists of replacing a single long-term simulation by an ensemble of multiple independent realizations, which are run in parallel. A failure probability convergence criteria must be satisfied by the statistical estimator of interest to assess convergence. The error analysis leads to the identification of two error contributions: the initialization bias and the statistical error. We propose an approach to systematically detect the burn-in time needed in order to minimize the initialization bias, as well as techniques to choose the effective time needed to keep the statistical error under control. This is accompanied by strategies to reduce simulation cost. The proposed method is assessed in application to the prediction of the drag force over high rise buildings and specifically in application to the CAARC building, a relevant benchmark for the wind engineering community.

Automated summary

This work focuses on reducing the computation time for statistical estimators of chaotic incompressible flows by using parallelization and applying convergence criteria. The error analysis identifies initialization bias and statistical error and proposes methods to minimize them. The method is evaluated in predicting the drag force on high rise buildings and specifically applied to the CAARC building.