Abstract
Unsteady turbulent wall bounded flows can include complex flow physics such as temporally varying mean pressure gradients, intermittent regions of high turbulence intensity, and interaction of different scales of motion. As a representative example, pulsating channel flow presents significant challenges for newly developed and existing turbulence models in computational fluid dynamics (CFD) simulations. The present study investigates the performance of the dynamic hybrid Reynolds‐averaged Navier‐Stokes-large eddy simulation RANS-LES (DHRL) modeling framework for nonstationary turbulent flows using two variants of an exponential temporal filter for calculating statistics of the resolved turbulent flow. The first adopts a static filter size (static exponential time filtering-SETF) based on the characteristic time scale of imposed mean flow unsteadiness. The second uses a dynamic filter size (dynamic exponential time filtering-DETF) to vary the filter size based on local statistics of the resolved turbulent flow. Both of the time-filtered variants and the baseline (stationary) form of DHRL are compared against an industry-standard RANS model, monotonically integrated large eddy simulation (MILES), and a conventional hybrid RANS-LES (HRL) models. Model performance is evaluated based on comparison with previously documented direct numerical simulation (DNS) and LES results. Simulations are performed for a fully developed flow in a channel with time-periodic driving pressure gradient. Results highlight the relative merits of each model type and indicate that the use of a time-filtering technique improves the accuracy of the DHRL model for nonstationary flow, and that a dynamic filter offers clear advantages over a static filter.
