On the prediction of separation-induced transition by coupling delayed detached-eddy simulation with γ transition model

The computation of industrial turbomachinery applications is still a big challenge for LES methods due to their requirements in terms of mesh resolution, which lead to high computational costs. Therefore, hybrid RANS/LES methods, such as Detached Eddy-Simulation, are gaining more attention. There is potential in combining the strengths of LES (accuracy where required) and RANS (reduced computational costs where possible) within one modelling approach. Nevertheless, so far, hybrid methods have mostly been developed for fully turbulent flow configurations. The application in transitional flows is still not well considered yet, although the phenomenon laminar-to-turbulent flow transition has a noticeable impact on turbomachinery performance. To capture the transition accurately is a key for improving the predictive quality of hybrid RANS/LES methods. Therefore, we propose a coupling of DDES and the γ-transition model. In this paper, we first introduce the underlying turbulence and transition model. A detailed investigation of how these two models interact with each other resulted in a potential coupled DDES-γ model with a modified k-transport equation. We show the application of DDES-γ and discuss numerical results with two exemplary test cases, namely a flat plat boundary layer with adverse pressure gradient, experimentally investigated by Volino & Hultgren (2000) and the low-pressure turbine cascade T106C, experimentally considered by Michálek et al. (2012). Both test cases represent characteristic flow conditions in turbomachinery such as separation-induced transition under low free-stream turbulence intensity. Main focus is the assessment of the proposed DDES-γ with focus on the improvement of predictive quality, but also potential issues, coming up when coupling DDES and the γ-transition model. The flat plate case serves as a starting point to assess the general behavior of DDES in transitional flows and how the γ-model interacts with the DDES. Secondly, the T106C case revealed also predictive improvements when analyzing turbomachinery-relevant values such as wake losses. For a better assessment, we always put the results into context and compare them with RANS and LES results. This supports the need for more sophisticated approaches such as DDES compared to RANS and illustrates the competitiveness of DDES approaches compared to LES. The considered cases helped to understand the model coupling and yield promising results for the DDES-γ model predicting separation-induced transition, while we showed, that the fully-turbulent DDES failed to capture relevant features for this transition type. After initially assessing DDES- γ for separation-induced transition in this paper, future research needs to address bypass transition to get a better sense for the performance of DDES-γ in transitional flows.