|(VKI) 3rd LECTURE SERIES ON UNCERTAINTY QUANTIFICATION IN COMPUTATIONAL FLUID DYNAMICS|
|Applied Vehicle Technology|
Calibration, Computational Fluid Dynamics, Data assimilation, Modelform uncertainties, Robust design and optimization, Stochastic propagation methods, Uncertainty Quantification, Validation and Verification
The availability of powerful computational resources and general-purpose numerical algorithms creates increasing opportunities to attempt flow simulations in complex systems such as military vehicle, space platforms, urban canopies, heat exchangers, and wind farms. How accurate are the resulting predictions? Are the mathematical and physical models correct? Do we have sufficient information to define relevant operating conditions? In general, how can we establish “error bars” on the results? At the interface between physics, mathematics, probability and optimization, there have been significant advances in Uncertainty Quantification (UQ) efforts in computational science over the past decade. The 2011 AVT-193 RTO-VKI LS on theory, applications and numerical tools for UQ was the opportunity to introduce UQ to the larger computational fluid dynamics (CFD) community and focused in particular on the difficulties stemming from the strong non-linearity and multiscale nature of flow dynamics, in particular in hypersonics. The 2013-2014 AVT-235 STO-VKI LS presented new trends in UQ in CFD focusing on model-form uncertainties and data assimilation. Both the theoretical and algorithmic aspects of stochastic CFD computations, and applications and results were discussed.
Uncertainty Quantification aims at developing rigorous methods to characterize the impact of “limited knowledge” on quantities of interest. After demonstrating the importance of uncertainty quantification for improving the predictive capabilities for complex fluid dynamics systems, the present course will review the new trends in UQ in CFD, focusing on the application of UQ for robust design and optimization.
Probabilistic analyses are at the core of current UQ approaches, and therefore, the challenges offered by complex flow simulations are multiplied when uncertainty characterization is required. The additional requirement for optimization under uncertainty poses a tremendous burden on computational algorithms but also offers great opportunities to introduce engineering reasoning and a new paradigm for modeling physical phenomena under uncertainty. The current plan is to divide the lecture series in two parts, the first more focused on the theoretical and algorithmic aspects of stochastic CFD computations, the second devoted to applications and results.