R. Manceau

Current trends in modelling research for turbulent aerodynamic flows

The engineering tools of choice for the computation of practical engineering flows have begun to migrate from those based on the traditional Reynolds-averaged Navier–Stokes approach to methodologies capable, in theory if not in practice, of accurately predicting some instantaneous scales of motion in the flow. The migration has largely been driven by both the success of Reynolds-averaged methods over a wide variety of flows and the inherent limitations of the method itself. Practitioners, emboldened by their ability to predict a wide variety of statistically steady equilibrium turbulent flows, have now turned their attention to flow control and non-equilibrium flows, i.e. separation control. This review gives some current priorities in traditional Reynolds-averaged modelling research as well as some methodologies being applied to a new class of turbulent flow control problem.

A seamless hybrid RANS-LES model based on transport equations for the subgrid stresses and elliptic blending

The aim of the present work is to develop a seamless hybrid Reynolds-averaged Navier–Stokes (RANS) large-eddy simulation (LES) model based on transport equations for the subgrid stresses, using the elliptic-blending method to account for the nonlocal kinematic blocking effect of the wall. It is shown that the elliptic relaxation strategy of Durbin is valid in a RANS (steady) as well as a LES context (unsteady). In order to reproduce the complex production and redistribution mechanisms when the cutoff wavenumber is located in the productive zone of the turbulent energy spectrum, the model is based on transport equations for the subgrid-stress tensor. The partially integrated transport model (PITM) methodology offers a consistent theoretical framework for such a model, enabling to control the cutoff wavenumber κc, and thus the transition from RANS to LES, by making the Ce2 coefficient in the dissipation equation of a RANS model a function of κc. The equivalence between the PITM and the Smagorinsky model is ...

Turbulent inflow conditions for large-eddy simulation based on low-order empirical model

Generation of turbulent inflow boundary conditions is performed by interfacing an experimental database acquired by particle image velocimetry to a computational code. The proposed method ensures that the velocity fields introduced as inlet conditions in the computational code present correct one- and two-point spatial statistics and a realistic temporal dynamics. This approach is based on the use of the proper orthogonal decomposition (POD) to interpolate and extrapolate the experimental data onto the numerical mesh and to model both the temporal dynamics and the spatial organization of the flow in the inlet section. Realistic representation of the flow is achieved by extracting and modeling independently its coherent and incoherent parts. A low-order dynamical model is derived from the experimental database in order to provide the temporal evolution of the most energetic structures. The incoherent motion is modeled by employing time series of Gaussian random numbers to mimic the temporal evolution of hi...

A rescaled elliptic relaxation approach: Neutralizing the effect on the log layer

An alternative scaling for the relaxation function describing the velocity pressure–gradient correlation used in the elliptic relaxation procedure for both eddy-viscosity and Reynolds stress models is presented. While other alternatives have been proposed to neutralize the adverse effect on log-layer dynamics, they have relied on altering the original differential formulation. A simpler alternative is presented here that involves a rescaling of the relaxation function with the isotropic dissipation rate as well as the turbulent kinetic energy. Various comparative tests are made and the new rescaled formulation is shown to provide improved and accurate predictions for both the eddy-viscosity and Reynolds stress models.

Anisotropic linear forcing for synthetic turbulence generation in large eddy simulation and hybrid RANS/LES modeling

A general forcing method for Large Eddy Simulation (LES) is proposed for the purpose of providing the flow with fluctuations that satisfy a desired statistical state. This method, the Anisotropic Linear Forcing (ALF) introduces an unsteady linear tensor function of the resolved velocity which acts as a restoring force in the mean velocity and resolved stress budgets. The ALF generalizes and extends several forcing previously proposed in the literature. In order to make it possible to impose the integral length scale of the turbulence generated by the forcing term, an alternative formulation of the ALF, using a differential spatial filter, is proposed and analyzed. The anisotropic forcing of the Reynolds stresses is particularly attractive, since unsteady turbulent fluctuations can be locally enhanced or damped, depending on the target stresses. As such, it is shown that the ALF is an effective method to promote turbulent fluctuations downstream of the LES inlet or at the interface between RANS and LES in zonal hybrid RANS/LES modeling. The detailed analysis of the influence of the ALF parameters in spatially developing channel flows and hybrid computations where the ALF target statistics are given by a RANS second-moment closure show that this original approach performs as well as the synthetic eddy method. However, since the ALF method is more flexible and significant computational savings are obtained, the method appears a promising all-in-one solution for general embedded LES simulations.

LES, Zonal and Seamless Hybrid LES/RANS: Rationale and Application to Free and Wall-Bounded Flows Involving Separation and Swirl

An overview is given of the activities in the framework of the German-French Research Group on ”LES of Complex Flows” (DFG-CNRS FOR 507) with respect to the development of zonal and seamless hybrid LES/RANS computational methods based on a near-wall Eddy-Viscosity Model (EVM) and a near-wall Second-Moment Closure (SMC) respectively. The zonal scheme represents a two layer model with a two-equation EVM-RANS model covering the near-wall layer and the true LES employing the zero-equation subgrid-scale (SGS) model of Smagorinsky resolving the core flow. Due attention was payed to the exchange of the variables between the ensemble-averaged RANS field and the spatially-filtered LES field across the discrete interface separating the two sub-regions. A procedure for controlling the interface position in the flow domain was also in focus of the present investigations. After considering a few introductory test cases (fully-developed channel flow, flows separating from sharp-edged surfaces) the feasibility of the method was validated against the available experiments in a single tubo-annular, swirl combustor configuration (Exp.: Palm et al., [39]) and in the separated flows in a 3-D diffuser (Exp. Cherry et al., [10]) and over a 2-D hump including the case with the separation control by steady suction (Exp.Greenblatt et al., [23]). The seamless LES/RANS method employs the so-called Elliptic-Blending Reynolds-Stress Model (EB-RSM, Manceau, [33]; Manceau and Hanjalic, [34]) being active in the entire flow field. This RANS-based SGS model represents a near-wall Second-Moment Closure model relying on the elliptic relaxation method. The model coefficient multiplying the destruction term in the transport equation for the scale-supplying variable e (dissipation rate of the turbulence kinetic energy) was made filter-width (corresponding to the grid spacing) dependent, i.e. dependent on the location of the spectral cutoff, by applying a multiscale modelling procedure originating from spectral splitting of filtered turbulence in line with the Partially Integrated Transport Model (PITM) proposed by Dejoan and Schiestel, [48] and Chaouat and Schiestel, [8]. Herewith, the dissipation rate level was obtained, which suppresses the turbulence intensity towards the subgrid (i.e. subscale) level in the regions where large coherent structures dominate the flow. The resulting model was validated by computing some free flows (a temporal mixing layer) and wall-bounded flows (a fully-developed channel flow). Finally, the PITM method applied to the high-Reynolds number RSM model due to Speziale et al., [53] was used to compute the flow separated from a 2-D hill (with reference LES by Frohlich et al., [19] and Breuer, [6]). In addition, all relevant cases were computed by the conventional LES method aiming at mutual comparison of the predictive capabilities of the afore-mentioned methods with respect to the quality of results and space-time resolution issues.

Investigation of the interaction of a turbulent impinging jet and a heated, rotating disk

The case of a turbulent round jet impinging perpendicularly onto a rotating, heated disc is investigated, in order to understand the mechanisms at the origin of the influence of rotation on the radial wall jet and associated heat transfer. The present study is based on the complementary use of an analysis of the orders of magnitude of the terms of the mean momentum and Reynolds stress transport equations, available experiments, and dedicated Reynolds-averaged Navier–Stokes computations with refined turbulence models. The Reynolds number Rej = 14 500, the orifice-to-plate distance H = 5D, where D is the jet-orifice diameter, and the four rotation rates were chosen to match the experiments of Minagawa and Obi [“Development of turbulent impinging jet on a rotating disk,” Int. J. Heat Fluid Flow 25, 759–766 (2004)] and comparisons are made with the Nusselt number distribution measured by Popiel and Boguslawski [“Local heat transfer from a rotating disk in an impinging round jet,” J. Heat Transfer 108, 357–364...

Introduction of Wall Effects into Explicit Algebraic Stress Models Through Elliptic Blending

In order to account for the non-local blocking effect of the wall, responsible for the two-component limit of turbulence, in explicit algebraic models, the elliptic blending strategy, a simplification of the elliptic relaxation strategy, is used. The introduction of additional terms, dependent on a tensor built on a pseudo-wall-normal vector, yields an extension of the integrity basis used to derive the analytical solution of the algebraic equation. In order to obtain a tractable model, the extended integrity basis must be truncated, even in 2D plane flows, contrary to standard explicit algebraic models. Four different explicit algebraic Reynolds-stress models are presented, derived using different choices for the truncated basis. They all inherit from their underlying Reynolds-stress model, the Elliptic Blending Model, a correct reproduction of the blocking effect of the wall and, consequently, of the two-component limit of turbulence. The models are satisfactorily validated in plane Poiseuille flows and several configurations of Couette–Poiseuille flows.

A Hybrid RANS–LES Model Based on Temporal Filtering

Since large eddy simulation methods (LES) are too CPU-demanding for many complex industrial applications, a multitude of unsteady low-cost strategies have gained prominence over the last decade Frohlich and von Terzi (2008); Sagaut et al (2006). Some of these models can be described as seamless hybrid RANS-LES models since the transition between RANS and LES occurs in a continuous manner so that there is no need to define explicit boundaries between RANS and LES regions. In statistically homogeneous flows, such a model can be seen as an LES with a filter width Δ S continuously going to infinity - a limit that corresponds formally to a RANS approach. However, the majority of flows of practical relevance are inhomogeneous, and in that case such models suffer from an important conceptual weakness due to the fact that the LES yields spatially filtered fields and the RANS yields long-time averaged fields.

Interfacing Stereoscopic PIV measurements to Large Eddy Simulations via Low Order Dynamical Systems

An original method consisting in interfacing dual-time Stereoscopic PIV measurements to a numerical code is proposed to provide inflow data to numerical simulations. It relies on the use of the Proper Orthogonal Decomposition to model the temporal dynamics of the coherent structures of the flow via a low order model and adapt the experimental mesh to the numerical one. LES of a turbulent plane mixing layer is performed to demonstrate the viability of the method.