Elsevier

Computers & Fluids

Volume 214, 15 January 2021, 104767
Computers & Fluids

Discontinuous Galerkin solution of the RANS and kLklog(ω) equations for natural and bypass transition

https://doi.org/10.1016/j.compfluid.2020.104767Get rights and content

Highlights

  • Novel transition model based on the laminar kinetic energy.

  • Implementation of the transition model in a Discontinuous Galerkin solver.

  • Validation of the model in the prediction of the flow field over flat plates.

  • Effect of y+ on transition model accuracy.

  • Computation of the flow fields through T106A and LS89 turbine nozzles.

Abstract

Transition modelling represents a key ingredient for improving the performance predictions of many industrial applications. Among transition models, local formulations seem to guarantee better robustness, accuracy and easiness of implementation in modern CFD solvers. These models have been proposed in the finite volume context to predict the laminar-turbulent transition, but only few attempts have been made in the high-order framework. In this paper a new phenomenological transition model based on the concept of laminar kinetic energy has been proposed and implemented in a high-order accurate Discontinuous Galerkin code, named MIGALE. The transition model is validated and assessed by computing the transitional flow around flat plates with zero/adverse pressure gradients and through different turbine nozzles (T106A and LS89) for different values of Reynolds number and turbulence intensity. The computed results have been compared with experimental data and reference numerical solutions.

Introduction

In the last decades, many reliable and robust turbulence models [1], [2], [3], [4] have been developed and coupled with Reynolds Average Navier-Stokes (RANS) equations to accurately simulate a wide range of engineering flows. This enhanced modelling capability made Computational Fluid Dynamics (CFD) a tool commonly adopted for industrial analysis and design, and considered complementary to experimental investigations.

However, these turbulence models are not well suited for low to moderate Reynolds numbers flows, which are characterized by strong transitional phenomena. Laminar-turbulent transition is common in aerospace, turbomachinery, maritime, automotive, and cooling applications; and a wrong prediction of this flow feature can dramatically affect their overall performance. As a consequence, numerical tools able to accurately predict transitional flows are mandatory for the efficient design of many industrial applications.

The laminar-turbulent transition can be predicted with different approaches: (i) Direct Numerical Simulation (DNS) [5], (ii) Large Eddy Simulation (LES) [6], [7], [8], [9], [10] and (iii) RANS approach coupled with a transition model. However, DNS and LES computational cost for many industrial flows characterized by low to moderate Reynolds numbers is still too large for a routine use in industry, and only the RANS approach can represent a viable solution. Moreover, in many applications where the transition is enforced within a limited area of the flow by geometric discontinuities, pressure gradient and/or flow separation, an accurate prediction of the flow features can be guaranteed also by a RANS approach supplemented with a transition model.

Next generation CFD solvers will be based on innovative numerical technology and turbulence models. In this context, the greater accuracy and geometrical flexibility guaranteed by Discontinuous Galerkin (DG) methods in solving the RANS and transition model equations could represent an appealing solution to enhance the predicting capabilities of standard industrial CFD codes, without resorting to intensive computational approaches, such as DNS and LES.

DG methods [11], [12], [13], [14], [15], [16] have emerged as a promising technique, as they provide a very high accuracy together with the geometrical flexibility required by complex industrial configurations, possibly adopting non-conforming grids made of arbitrarily shaped cells. Although this higher accuracy requires an increased computational cost compared to standard finite volume methods, the compact stencil of the DG discretization, involving only one cell and its neighbours, makes the method very well suited for massively parallel computer platforms.

Transition models can be classified into three families: (i) low-Reynolds, (ii) non-local and (iii) local models. Low Reynolds turbulence models [17], [18] have shown some limitations in the prediction of the transition for general flow conditions. Non-local models are based on correlations [19], [20], which relate the momentum thickness Reynolds number to local free-stream conditions, such as turbulence intensity and pressure gradient. These models can be easily calibrated and can often capture correctly the major physical effects. Several correlations have been developed for different transition mechanisms, i.e. natural, bypass, separation induced and cross-flow transition. The main drawback lies in their non-local formulation, i.e. the information on the integral thickness of the boundary layer and the state of the flow outside the boundary layer are required. Finally, the local models are built on transport equations, that require only local variables. Among these models, we can distinguish two different types: empirical correlation models [21], [22] and phenomenological transition models [23] based on the concept of laminar kinetic energy (LKE). The development of phenomenological models, that aim at incorporating the physics of boundary layer transition, is an attractive but challenging research field as many mechanisms influencing the flow phenomenon are not yet fully understood. Many authors [24], [25], [26], following the seminal work of Walters [23], [27], proposed different variants of the three equation phenomenological transition model based on the concept of the LKE proposed by Mayle [28].

In the higher-order context, only few attempts have been made to simulate transitional flows with a RANS approach. In the DG framework, the predicting capability of a low Reynolds variant of the kω turbulence model was assessed in [29], while Crivellini et al. [30] investigated the Spalart-Allmaras model ability to capture transition. Lorini et al. [31] presented the first implementation of a non-local transition model in a DG solver, while Ferrero et al. [32] presented the implementation of a non-local transition model based on laminar kinetic energy. To date, local transition models have been mainly developed in the low-order finite volume (FV) framework, and few attempts have been made to extend these models to high-order solvers [33], [34].

The objective of this work is to develop and implement a phenomenological transition model based on the LKE concept into the high-order DG code MIGALE. The prediction capabilities of the kLklog(ω) transition model are assessed by computing turbulent flows experiencing natural and bypass transition. The model was validated by computing the flow (i) over flat plates with zero/adverse pressure gradients and different values of turbulence intensity, (ii) through a low pressure turbine cascade (MTU T106A) at Reynolds number Re=1.1×106, and (iii) through the turbine cascade LS89 for different values of turbulence intensity.

The paper is organized as follows. Section 2 describes the transition model employed, while the descriptions of the DG spatial discretization and implicit time integration are given in Section 3. Section 4 is dedicated to the discussion of the numerical results. Finally, Section 5 gives concluding remarks.

Section snippets

Transition model

The transitional model presented in this section is based on the work of Walters and Cokljat [27], where the RANS system is closed with three additional equations for the turbulent kinetic energy k, the laminar kinetic energy kL, and the specific dissipation rate ω=ϵ/k, where ϵ is the dissipation.

The complete set of the RANS and kLkω˜ model equations for a compressible flow can be written asρt+xj(ρuj)=0,t(ρui)+xj(ρujui)=pxi+τ^jixj,t(ρE)+xj(ρujH)=xj[uiτ^ijq^j]ρ(PkDk)ρ(PkL

Space and time discretization

The governing equations for m variables in d=2,3 dimensions can be written in compact form asP(w)wt+·Fc(w)+·Fv(w,w)+s(w,w)=0,where w=[p,ui,T,k,kL,ω˜]TRm is the unknown solution vector, Fc,FvRmRd are the convective and viscous flux functions, sRm is the vector of source terms, and P(w)RmRm is the transformation matrix arising from the use of the primitive variables.

By multiplying Eq. (45) by an arbitrary test function v={v1,,vm}, and integrating by parts, we obtain the weak

Results

The proposed kLkω˜ model has been validated by computing the flow over the flat plates of the ERCOFTAC T3 series, with zero and non-zero pressure gradients, investigated experimentally in the early 1990’s [45], [46]. These test cases are mainly characterized by bypass transition, due to the levels of freestream turbulence intensity, i.e., Tu1 ≥ 1%. To assess the model accuracy when dealing with natural transition, the Schubauer and Klebanoff (SK) flat plate has been also computed [47].

Conclusions

The implementation of a novel local transition model based on the LKE concept into a high-order accurate DG code has been here presented. The work represents one of the first attempts to introduce this class of models in a high-order numerical context.

Validation has been performed by computing the flow over a flat plate with zero/adverse pressure gradient and different boundary conditions. The performance has been also assessed by computing complex transitional turbomachinery flows, i.e., the

CRediT authorship contribution statement

M. Lorini: Conceptualization, Methodology, Software. F. Bassi: Conceptualization, Methodology, Software. A. Colombo: Conceptualization, Writing - review & editing. A. Ghidoni: Conceptualization, Methodology, Software, Supervision, Writing - review & editing. G. Noventa: Software, Validation, Visualization, Writing - review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (53)

  • A. Crivellini et al.

    Spalart-allmaras model apparent transition and RANS simulations of laminar separation bubbles on airfoils

    Int J Heat Fluid Flow

    (2014)
  • M. Lorini et al.

    High-order implementation of a non-local transition model in a DG solver for turbomachinery applications

    Comput Fluids

    (2016)
  • F. Bassi et al.

    Discontinuous Galerkin solution of the Reynolds-averaged Navier-Stokes and kω turbulence model equations

    Comput Fluids

    (2005)
  • T.-H. Shih et al.

    A new kε eddy viscosity model for high reynolds number turbulent flows

    Comput Fluids

    (1995)
  • B. Launder et al.

    Application of the energy-dissipation model of turbulence to the calculation of flow near a spinning disc

    Lett Heat Mass Transf

    (1974)
  • D. Wilcox

    Formulation of the kω turbulence model revisited

    AIAA J

    (2008)
  • K.-Y. Chien

    Predictions of channel and boundary-layer flows with a low-Reynolds-number turbulence model

    AIAA J

    (1982)
  • Spalart P., Allmaras S.. A one-equation turbulence model for aerodynamic flows. In: 30th aerospace sciences meeting and...
  • F.R. Menter

    Two-equation eddy-viscosity turbulence models for engineering applications

    AIAA J

    (1994)
  • X. Mao et al.

    Transition induced by linear and nonlinear perturbation growth in flow past a compressor blade

    J Fluid Mech

    (2017)
  • A. Uranga et al.

    Implicit large eddy simulation of transition to turbulence at low Reynolds numbers using a discontinuous Galerkin method

    Int J Numer Methods Eng

    (2011)
  • Fernández P., Nguyen N.C., Peraire J.. On the ability of discontinuous Galerkin methods to simulate under-resolved...
  • F. Bassi et al.

    A p-adaptive matrix-free discontinuous Galerkin method for the implicit LES of incompressible transitional flows

    Flow Turbul Combust

    (2020)
  • F. Bassi et al.

    Time integration in the discontinuous Galerkin code MIGALE - steady problems

  • F. Bassi et al.

    Time integration in the discontinuous Galerkin code MIGALE - unsteady problems

  • S. Lardeau et al.

    Modelling bypass transition with low-Reynolds-number nonlinear eddy-viscosity closure

    Flow Turbul Combust

    (2004)
  • Cited by (0)

    View full text