Uncertainty quantification in LES of channel flow

Abstract

Summary In this paper, we present a Bayesian framework for estimating joint densities for large eddy simulation (LES) sub‐grid scale model parameters based on canonical forced isotropic turbulence direct numerical simulation (DNS) data. The framework accounts for noise in the independent variables, and we present alternative formulations for accounting for discrepancies between model and data. To generate probability densities for flow characteristics, posterior densities for sub‐grid scale model parameters are propagated forward through LES of channel flow and compared with DNS data. Synthesis of the calibration and prediction results demonstrates that model parameters have an explicit filter width dependence and are highly correlated. Discrepancies between DNS and calibrated LES results point to additional model form inadequacies that need to be accounted for. Copyright © 2016 John Wiley & Sons, Ltd.

Optimal Compressed Sensing and Reconstruction of Unstructured Mesh Datasets

“…We have also obtained data from the simulation of threedimensional high-speed internal flow in a flat channel [42]. The data involves several time step, each consisting of 8 primitive variables.…”

Calibration and Forward Uncertainty Propagation for Large-eddy Simulations of Engineering Flows

“…In a previous study, Safta et al [65], used a Bayesian framework to incorporate filtered DNS turbulence information to estimate uncertainties in the parameters for k sgs turbulence model. By varying the parameters C e and C µ e found in equations 2 and 5 above for 25 channel flow cases we were able to use Polynomial Chaos Expansion (PCE) to build a response surface.…”

“…These surrogate models are constructed based on LES simulations corresponding to select sample values for q = h wall , h center , shown in Table 3.1. Originally these samples were optimally placed to construct surrogate models based on Polynomial Chaos concepts [19,65]. As some of the model runs failed to produce valid turbulent results, resulting in an unstructured valid region, we turned our attention to Radial Basis Functions (RBFs) [6] for the construction of surrogate models for the Quantities of Interest resulted from LES computations.…”

“…In a previous paper [65], we used a Bayesian framework to incorporate directly filtered DNS turbulence information to estimate uncertainties in the parameters for a sub-grid scale turbulence model. This was followed by a forward UQ study on the predictive capabilities of this approach.…”

Eigensensitivity analysis of subgrid-scale stresses in large-eddy simulation of a turbulent axisymmetric jet

“…A more sophisticated approach is to consider the closure parameters uncertain and estimate their effects on the QoIs by forwardpropagating them as probability distributions. This strategy has been applied to Reynolds-averaged Navier-Stokes (RANS) [20] and LES [21] models and extended to incorporate simulation data from DNS [22] and utilize Bayesian inference techniques [10,23,24]. In the case of complex flows, some methodologies predict on the basis of an ensemble of solutions obtained using different models, such as in earth sciences for weather and ocean forecasting [25,26,27].…”