@misc{MarinkovićMedinaMendezSchmidtetal., author = {Marinković, Pavle and Medina M{\´e}ndez, Juan Ali and Schmidt, Heiko and Klein, Marten}, title = {Ongoing development of a hybrid reduced order stochastic/LES solver for turbulent flows}, address = {Cottbus}, pages = {1}, language = {en} } @misc{MarinkovićMedinaSchoepsetal., author = {Marinković, Pavle and Medina, Juan A. and Sch{\"o}ps, Mark Simon and Klein, Marten and Schmidt, Heiko}, title = {Experiences from the bottom-up development of an object-oriented CFD solver with prospective hybrid turbulence model applications}, series = {Proceedings in Applied Mathematics and Mechanics}, volume = {25}, journal = {Proceedings in Applied Mathematics and Mechanics}, number = {1}, publisher = {Wiley}, issn = {1617-7061}, doi = {10.1002/pamm.202400190}, abstract = {In this work, we discuss recent experiences related to the development and enhancement of a hybrid stochastic computational fluid dynamics (CFD) solver, the C++ version of the Implicit/Explicit (IMEX) time-advancement algorithm used in the one-dimensional turbulence-based (ODT) large eddy simulation (LES) model, abbreviated as ODTLES. After being ported from Fortran 90, the current capabilities of the C++ code are restricted to the reproducibility of turbulent channel flow simulations with respect to the former Fortran code version that was able to achieve reasonable agreement with available reference direct numerical simulation (DNS) for low to moderate Reynolds number turbulent channel flows. This is far from satisfactory so that current efforts are centered on improving the solver code structure through comprehensive refactoring, robust unit testing, and strict adherence to code style guides, following the principles of Clean Code. We focus the discussion on a methodology to balance unit, regression, and integration testing, here for the LES component of the code. The objective is to frame a starting point that is relevant also for other CFD codes, irrespective of whether they utilize conventional or novel discretization or flow modeling approaches.}, language = {en} } @misc{MarinkovicMedinaKleinetal., author = {Marinkovic, Pavle and Medina, Juan and Klein, Marten and Schmidt, Heiko}, title = {Application of extended large-eddy simualtion (XLES) to turbulent channel flow}, series = {Proceedings in applied mathematics and mechanics : PAMM}, volume = {25 : special issue : 95th Annual Meeting of the International Association of Applied Mathematics and Mechanics (GAMM)}, journal = {Proceedings in applied mathematics and mechanics : PAMM}, number = {2}, publisher = {Wiley}, address = {Weinheim}, issn = {1617-7061}, doi = {10.1002/pamm.70010}, pages = {1 -- 7}, abstract = {Extended large-eddy simulation (XLES) offers a promising middle ground between computational efficiency and predictive accuracy in turbulent flow simulations. By integrating the one-dimensional turbulence model as a stochastic subgrid-scale approach, XLES autonomously resolves turbulent microscales in a dimensionally reduced setting. Our implementation employs a coarse grid solely for pressure calculations, while three specialized high-resolution grids handle the turbulence dynamics. This multi-scale architecture reduces computational requirements by approximately two orders of magnitude compared to direct numerical simulation while maintaining comparable accuracy. Our results for turbulent channel flow at Reynolds number 395 demonstrate that XLES with a base resolution of just cells closely matches DNS reference data, vastly outperforming implicit LES using identical resolution and discretization schemes. This updated implementation of XLES, though still under active development, demonstrates significant potential for accurately simulating wall-bounded turbulent flows while substantially reducing computational requirements compared to traditional high-fidelity approaches.}, language = {en} } @misc{MarinkovicMedinaMendezKleinetal., author = {Marinkovic, Pavle and Medina M{\´e}ndez, Juan A. and Klein, Marten and Schmidt, Heiko}, title = {ODTLES : large-eddy simulation with autonomous stochastic subgrid-scale modeling applied to turbulent duct flow}, series = {Proceedings of the Conference on Modelling Fluid Flow CMFF'25}, journal = {Proceedings of the Conference on Modelling Fluid Flow CMFF'25}, publisher = {Budapest University of Technology and Economics}, address = {Budapest}, isbn = {978-615-112-002-6}, pages = {8}, abstract = {In this work, we discuss the application of the One-Dimensional Turbulence-based (very) Large-Eddy Simulation model, abbreviated as ODTLES, to turbulent duct flow. ODTLES is a multi-scale flow model in which an autonomous stochastic One-Dimensional Turbulence (ODT) model, capable of simulating the full bandwidth of time and length-scales in a 1-D domain, is supplemented with large-scale 3-D information coming from a very large eddy simulation (VLES) grid. ODTLES is more expensive than any other VLES, but could be cheaper than highly resolved LES or, naturally, than Direct Numerical Simulation (DNS). Unlike Reynolds-Averaged Navier-Stokes (RANS) and VLES, ODTLES does neither need a wall model, nor a damping function. The correct near-wall behavior is naturally obtained from one SGS ODT domain that is locally wall-normal. The proposed hybrid (3-D/1-D) approach allows the resolution of all relevant scales, modeling certain aspects of 3-D turbulence on the SGS scale. Here, turbulent duct flow is considered as an example, which poses a moderate challenge for traditional LES due to emerging secondary flows that manifest themselves by corner vortices that crucially depend on the accurate capturing of small and large scale motions. Preliminary results indicate a reasonable match with DNS for mean velocity profiles, although capturing secondary flow structures remains a challenge at this stage. Further refinements of the solver and modeling approach are ongoing to improve accuracy and predictive capabilities.}, language = {en} }