Multi-fidelity approach for transitional boundary layer

• Published in: The International journal of heat and fluid flow 2023-08, Vol.102, p.109163, Article 109163
• Main Authors: Kim, Minwoo, Lim, Jiseop, Jee, Solkeun, Park, Donghun
• Format: Article
• Language: English
• Subjects: Computational fluid dynamics; Multi-fidelity method; Turbulent transition
• ISSN: 0142-727X; 1879-2278
• DOI: 10.1016/j.ijheatfluidflow.2023.109163

A multi-fidelity computational approach is proposed for transitional boundary layer flow in order to obtain both high fidelity and high efficiency in flow simulation. Because we can categorize the transitional flow into three regions, i.e., laminar region with instabilities, transition region, and turbulent region, three models are judiciously selected. A boundary layer stability method, here nonlinear parabolized stability equations (NPSE), is chosen for the laminar region incorporating nonlinear interactions between instabilities. Large-eddy simulation (LES) is applied for the transition region including the beginning of the turbulent flow. Reynolds-averaged Navier–Stokes (RANS) simulation is used for the downstream turbulent flow. Since the NPSE-coupled LES method has been validated for transitional boundary layers including both incompressible (Kim et al., 2018, 2019, 2020, 2021) and compressible (Lim et al., 2021) flows, appropriate treatments for the interface between LES and RANS are mainly investigated in this study. The current multi-fidelity approach is tested for a transitional boundary layer on a flat plate. Computational fidelity and cost are compared with previous high-fidelity computations. We successfully demonstrate that the proposed multi-fidelity approach provides both high fidelity and high efficiency in transitional flow computations. [Display omitted] •Multi-fidelity method for turbulent transition simulation.•Judiciously combined three methods: NPSE, LES, and RANS method.•Drastically reduced computational cost while maintaining reasonable fidelity.•Appropriate treatment at an interface between low- and high-fidelity methods.