One of the most common approaches used in CFD is Reynolds-Averaged Navier-Stokes (RANS) modeling, which we know is computationally efficient but suffers from several limitations, such as the inability to accurately predict complex turbulent flows with secondary motions and swirling effects. On the other hand, Large Eddy Simulation (LES) modeling is capable of capturing (by resolving) most of the turbulent length scales and provides more accurate predictions of turbulent flows in a time-dependent fashion. However, LES requires significantly more computational resources than RANS, making it computationally expensive for many practical engineering applications.
To overcome the limitations of both RANS and LES, researchers have developed a hybrid approach called Hybrid RANS/LES (or HRL) modeling. This approach has been combining the best of both worlds for both accuracy and efficiency motives. The first hybrid RANS/LES model was proposed by Spalart et al. in 1997, known as the detached eddy simulation (DES) method (or DES97) [1]. This DES method is a model that combines RANS in the near-wall regions where the flow is likely to be attached, and activated LES in the far-field regions where the separation is more likely and where large-scale unsteady structures dominate the flow field. The transition from RANS to LES is controlled by a blending function that determines the fraction of the flow that is simulated using RANS and LES. The aim of this approach has been to provide an accurate and computationally efficient simulation of both attached and separated flows. This method has since become one of the most widely used hybrid RANS/LES models in the CFD community.
N.B.: DES97 method was originally developed for turbulent flows over smooth walls, but it has been extended to rough wall flows, combustion flows, and other complex flow configurations. Since its introduction, many other hybrid RANS/LES models have been proposed, each with their own advantages and limitations.
Following the introduction of the DES97, several studies have been conducted to evaluate its accuracy and applicability to various flow configurations. For example, Kim et al. (1999) compared DES with RANS and LES methods for a backward-facing step flow and found that DES provided better predictions of the mean and fluctuating flow fields compared to pure RANS and LES [2].
Later, Menter and Egorov (2005) proposed an improved DES method, known as Delayed DES (DDES), which delays the transition from RANS to LES until a certain level of turbulence is present in the flow [3]. The DDES method has been shown to improve the accuracy of DES for turbulent flows with low levels of turbulence intensity.
In recent years, several other variations of DES have been proposed, such as the Improved Delayed DES (IDDES) method by Spalart and Shur (2009) [4], which further improves the transition from RANS to LES by introducing a blending function that accounts for the turbulent length scales in the flow. Furthermore, hybrid RANS/LES methods based on DES have been extended to more complex flow configurations, such as unsteady flows, rotating flows, and multiphase flows. For instance, the Unsteady DES (UDES) method was proposed by Shur et al. (2008) for unsteady flows, which employs a time filter to smooth out the turbulent scales [5].
More recently, several studies have focused on improving the computational efficiency of DES, such as the Adaptive DES (ADES) method by Manceau and Leschziner (2010), which adaptively switches between RANS and DES regions based on the local turbulence intensity [6].
There are two main types of DES: the Spalart-Allmaras DES (S-A DES) and the k-ω based DES.
The Spalart-Allmaras DES (S-A DES) is based on the one-equation Spalart-Allmaras turbulence model. In S-A DES, the RANS equations are solved for the entire flow field, and the unresolved turbulent stresses are modeled using the Spalart-Allmaras turbulence model [7]. In regions where the flow is attached, the model behaves like a RANS model, while in separated regions, the model switches to an LES mode and solves the filtered Navier-Stokes equations.
The k-ω based DES is based on the k-ω turbulence model. In this approach, the RANS equations are solved for the entire flow field, and the unresolved turbulent stresses are modeled using the k-ω turbulence model. The k-ω based DES method also switches from RANS to LES mode in separated flow regions.
There are also several variants of DES, such as Improved Delayed DES (IDDES), which is an extension of S-A DES that uses a delayed DES model in regions of attached flow and a full LES model in regions of separated flow. Another variant is the Scale-Adaptive Simulation (SAS), which uses a scale-similarity model to smoothly transition between RANS and LES modes.
The modification done to DES into DDES by Spalart in 2000 is that the DDES approach blends the RANS equations and the LES equations in the near-wall region using the Smagorinsky model to model subgrid-scale stresses. In contrast, the original DES method divided the simulation domain into two regions: the RANS region near the wall and the LES region in the outer flow region. However, DES had limitations in capturing the near-wall turbulence, especially in highly separated flows. To overcome this limitation, the DDES method was proposed to combine DES and LES in a more seamless manner. The dynamic Smagorinsky model is used in the LES region to model subgrid-scale stresses.
As we know, Unsteady RANS (URANS) models are typically used to simulate attached turbulent flows, but they often struggle to accurately predict separated flows where large-scale unsteady structures dominate the flow field. On the other hand, Large Eddy Simulation (LES) models are better suited for capturing these unsteady structures in separated flows, but they can be computationally expensive as we mentioned earlier, especially for industrial applications.
DES seeks to combine the strengths of RANS and LES by using a RANS model in attached flow regions and transitioning to an LES approach in separated flow regions. This allows DES to capture both attached and separated flows in a computationally efficient manner, while still providing accurate predictions of flow separation, vortex shedding, and other unsteady phenomena (e.g. see Fig. 1 below).
For this reason, DES has been shown to be effective in a wide range of engineering applications, including aerospace, wind energy, and automotive industries, where accurate predictions of complex flow phenomena are crucial for designing efficient and safe systems.
As you could probably tell, hybrid RANS/LES represents a successful and efficient approach for simulating turbulent flows in CFD. By combining the advantages of both RANS and LES methods, hybrid RANS/LES can provide accurate and efficient solutions for a wide range of engineering problems. Despite the challenges associated with implementing and validating hybrid RANS/LES models, recent advances in CPU resources and modeling techniques have made it possible to apply this approach to practical engineering problems.
As the field of CFD continues to evolve, hybrid RANS/LES is expected to play an increasingly important role in the design and optimization of complex systems. Whether it is in the automotive, aerospace, or energy industries, hybrid RANS/LES models have the potential to provide valuable insights into flow physics and improve the performance of engineering designs. As such, it is important for CFD practitioners to be aware of the theoretical foundations, implementation challenges, and practical applications of hybrid RANS/LES methods.
In my view, the development and refinement of hybrid RANS/LES models is one of the most exciting areas of research in CFD and it holds a great promise for the future of Fluid Dynamics (ah yes alongside the exponential evolution of AI 😁)!
[1] Spalart, P. R., Jou, W. H., Strelets, M., & Allmaras, S. R. (1997). Comments on the feasibility of LES for wings and on a hybrid RANS/LES approach. In Advances in DNS/LES (pp. 137-147). Springer, Dordrecht.
[2] Kim, S. E., Kwon, O. C., & Ahn, J. H. (1999). A comparison of RANS, DES, and LES for the computation of turbulent flows over a backward-facing step. Computers & Fluids, 28(4-5), 421-443.
[3] Menter, F. R., & Egorov, Y. (2005). The scale-adaptive simulation method for unsteady turbulent flow predictions. Part 1: Theory and model description. Flow, Turbulence and Combustion, 74(3), 183-204.
[4] Spalart, P. R., & Shur, M. L. (2009). On the role of wall-distance in zonal approaches to LES. International Journal of Heat and Fluid Flow, 30(5), 849-857.
[5] Shur, M. L., Spalart, P. R., Strelets, M. K., & Travin, A. K. (2008). A hybrid RANS-LES approach with delayed-DES and wall-modelled LES capabilities. International Journal of Heat and Fluid Flow, 29(6), 1638-1649.
[6] Manceau, R., & Leschziner, M. A. (2010). An adaptive version of the DES and its application to separated flows. International Journal of Heat and Fluid Flow, 31(6), 1059-1071.
[7] Spalart, P.R. and Allmaras, S.R. (1994) A one-equation turbulence model for aerodynamic flows. La Recherche Aerospatiale, 1, 5–21.
