Weak-equilibrium Assumption Research Articles

A PDE-free, neural network-based eddy viscosity model coupled with RANS equations

In fluid dynamics, constitutive models are often used to describe the unresolved turbulence and to close the Reynolds averaged Navier–Stokes (RANS) equations. Traditional PDE-based constitutive models are usually too rigid to calibrate with a large set of high-fidelity data. Moreover, commonly used turbulence models are based on the weak equilibrium assumption, which cannot adequately capture the nonlocal physics of turbulence. In this work, we propose using a vector-cloud neural network (VCNN) to learn the nonlocal constitutive model, which maps a regional mean flow field to the local turbulence quantities without solving the transport PDEs. The network is strictly invariant to coordinate translation, rotation, and uniform motion, as well as ordering of the input points. The VCNN-based nonlocal constitutive model is trained and evaluated on flows over a family of parameterized periodic hills. Numerical results demonstrate its predictive capability on target turbulence quantities of turbulent kinetic energy k and dissipation ɛ. More importantly, we investigate the robustness and stability of the method by coupling the trained model back to RANS solver. The solver shows good convergence with the simulated velocity field comparable to that based on k–ɛ model when starting from a reasonable initial condition. This study, as a proof of concept, highlights the feasibility of using a nonlocal, frame-independent, neural network-based constitutive model to close the RANS equations, paving the way for the further emulation of the Reynolds stress transport models.

Strongly coupled fluid-particle flows in vertical channels. II. Turbulence modeling

In Part I, simulations of strongly coupled fluid-particle flow in a vertical channel were performed with the purpose of understanding, in general, the fundamental physics of wall-bounded multiphase turbulence and, in particular, the roles of the spatially correlated and uncorrelated components of the particle velocity. The exact Reynolds-averaged (RA) equations for high-mass-loading suspensions were presented, and the unclosed terms that are retained in the context of fully developed channel flow were evaluated in an Eulerian–Lagrangian (EL) framework. Here, data from the EL simulations are used to validate a multiphase Reynolds-stress model (RSM) that predicts the wall-normal distribution of the two-phase, one-point turbulence statistics up to second order. It is shown that the anisotropy of the Reynolds stresses both near the wall and far away is a crucial component for predicting the distribution of the RA particle-phase volume fraction. Moreover, the decomposition of the phase-average (PA) particle-phase fluctuating energy into the spatially correlated and uncorrelated components is necessary to account for the boundary conditions at the wall. When these factors are properly accounted for in the RSM, the agreement with the EL turbulence statistics is satisfactory at first order (e.g., PA velocities) but less so at second order (e.g., PA turbulent kinetic energy). Finally, an algebraic stress model for the PA particle-phase pressure tensor and the Reynolds stresses is derived from the RSM using the weak-equilibrium assumption.

Capturing turbulent density flux effects in variable density flow by an explicit algebraic model

The explicit algebraic Reynolds stress model of Grigoriev et al. [“A realizable explicit algebraic Reynolds stress model for compressible turbulent flow with significant mean dilatation,” Phys. Fluids 25, 105112 (2013)] is extended to account for the turbulent density flux in variable density flows. The influence of the mean dilatation and the variation of mean density on the rapid pressure-strain correlation are properly accounted for introducing terms balancing a so-called “baroclinic” production in the Reynolds stress tensor equation. Applying the weak-equilibrium assumption leads to a self-consistent formulation of the model. The model together with a K − ω model is applied to a quasi-one-dimensional plane nozzle flow transcending from subsonic to supersonic regimes. The model remains realizable with constraints put on the model parameters. When density fluxes are taken into account, the model is less likely to become unrealizable. The density variance coupled with a “local mean acceleration” also can influence the model acting to increase anisotropy. The general trends of the behaviour of the anisotropy and production components under the variation of model parameters are assessed. We show how the explicit model can be applied to two- and three-dimensional mean flows without previous knowledge of a tensor basis to obtain the general solution. Approaches are proposed in order to achieve an approximate solution to the consistency equation in cases when analytic solution is missing. In summary, the proposed model has the potential to significantly improve simulations of variable-density flows.

Explicit algebraic Reynolds stress model (EARSM) for compressible shear flows

We develop an explicit algebraic Reynolds stress model (EARSM) for high-speed compressible shear flows and validate the model with direct numerical simulation (DNS) data of homogeneous shear flow and experimental data of high-speed mixing-layers. Starting from a pressure–strain correlation model that incorporates compressibility effects, the weak-equilibrium assumption is invoked to derive the EARSM closure expression. The resulting closure is fully explicit and physically realizable and is a function of mean flow strain rate, rotation rate, turbulent kinetic energy, dissipation rate, and gradient Mach number. Homogeneous shear flow calculations show that the model captures the asymptotic behavior of DNS quite well. Linear EARSM calculations of a plane supersonic mixing-layer are performed, and comparison with experimental data shows good agreement. Salient results are agreement of streamwise velocity similarity profiles, mixing-layer spreading rates, and capturing the Langley curve trend.

Development of an explicit algebraic turbulence model for buoyant flows – Part 1: DNS analysis

An analysis of DNS databases of vertical plane channel flow for forced, mixed and natural convection is proposed. This analysis aims to assess the main features needed to develop an algebraic model for buoyant flows. First, the weak equilibrium assumption, at the root of algebraic models, is investigated. This hypothesis is shown to fail near the velocity maximum and close to the walls but remains valid otherwise, whatever the convection regime. The models for the redistribution term and the pressure scrambling term are then analyzed on the same configurations. A linear form of the Speziale et al. (1991) model is retained for the redistribution term. No model for the pressure scrambling term is fully satisfactory; nevertheless some models are recommended. The buoyant contributions to the pressure term are investigated. Finally, the generalized gradient diffusion hypothesis, which could be used to model the turbulent heat fluxes in order to avoid the coupling with the Reynolds stresses, is shown to be inaccurate.

An explicit algebraic model for the subgrid-scale passive scalar flux

AbstractIn Marstorpet al. (J. Fluid Mech., vol. 639, 2009, pp. 403–432), an explicit algebraic subgrid stress model (EASSM) for large-eddy simulation (LES) was proposed, which was shown to considerably improve LES predictions of rotating and non-rotating turbulent channel flow. In this paper, we extend that work and present a new explicit algebraic subgrid scalar flux model (EASSFM) for LES, based on the modelled transport equation of the subgrid-scale (SGS) scalar flux. The new model is derived using the same kind of methodology that leads to the explicit algebraic scalar flux model of Wikströmet al. (Phys. Fluids, vol. 12, 2000, pp. 688–702). The algebraic form is based on a weak equilibrium assumption and leads to a model that depends on the resolved strain-rate and rotation-rate tensors, the resolved scalar-gradient vector and, importantly, the SGS stress tensor. An accurate prediction of the SGS scalar flux is consequently strongly dependent on an accurate description of the SGS stresses. The new EASSFM is therefore primarily used in connection with the EASSM, since this model can accurately predict SGS stresses. The resulting SGS scalar flux is not necessarily aligned with the resolved scalar gradient, and the inherent dependence on the resolved rotation-rate tensor makes the model suitable for LES of rotating flow applications. The new EASSFM (together with the EASSM) is validated for the case of passive scalar transport in a fully developed turbulent channel flow with and without system rotation. LES results with the new model show good agreement with direct numerical simulation data for both cases. The new model predictions are also compared to those of the dynamic eddy diffusivity model (DEDM) and improvements are observed in the prediction of the resolved and SGS scalar quantities. In the non-rotating case, the model performance is studied at all relevant resolutions, showing that its predictions of the Nusselt number are much less dependent on the grid resolution and are more accurate. In channel flow with wall-normal rotation, where all the SGS stresses and fluxes are non-zero, the new model shows significant improvements over the DEDM predictions of the resolved and SGS quantities.

An alternative assessment of weak-equilibrium conditions in turbulent closure modeling

In formulating polynomial representations of turbulent stress anisotropy tensors, it is necessary to impose equilibrium conditions on the turbulent field in order to reduce the governing transport equations to a set of implicit algebraic equations. Imposing these conditions on the anisotropy tensor rather than the turbulent stress tensor directly results in a set of weak-equilibrium conditions on the evolution of the anisotropy tensor. A second kind of assumption is a requirement that the turbulent transport and viscous diffusion be related to the anisotropy tensor in a manner consistent with both the dynamics and related tensorial properties. In the study reported here, an alternative analysis of the weak-equilibrium hypothesis is performed in order to improve the physical interpretation of this assumption. One important result of the analysis is that previous turbulent weak-equilibrium hypothesis available in the literature impose that the flow has reached a Motion With Constant Relative Principle Anisotropic -Reynolds-Stress History, which means that the eigenvalues of the anisotropic Reynolds stress tensor do not change in time. Since virtually every geometry used to test this kind of hypothesis do not have a change on the cross sectional area and lead to viscometric flows, is natural that at equilibrium conditions such motion occurs. However, the eigenvalues of the Reynolds stress can change in time on turbulent extensional flows due to action of the mean flow in a different process from a turbulent relaxation. In this case, the extended weak-equilibrium assumption has to be related to an objective time derivative which is not purely co-rotational.

Explicit algebraic Reynolds stress and non-linear eddy-viscosity models

This article reviews explicit algebraic Reynolds stress models and other non-linear eddy-viscosity turbulence models utilised in the context of Reynolds-Averaged Navier-Stokes simulations of turbulent flows. Since the 1990s, these modelling classes have become important intermediate classes between the linear eddy viscosity models and full Reynolds stress transport models. The derivation of explicit algebraic Reynolds stress models from the Reynolds stress transport models, and the required simplifications are discussed. The most important simplification is the weak-equilibrium assumption. Properties of the basic weak-equilibrium assumption and its extended form are discussed. Differences between explicit solutions based on complete tensor representation and those based on reduced representation as well as other non-linear eddy-viscosity models are pointed out. The scale-determining models are also discussed briefly.

Observations on the predictions of fully developed rotating pipe flow using differential and explicit algebraic Reynolds stress models

The differences between two differential Reynolds stress models (DRSM) and their corresponding explicit algebraic Reynolds stress models (EARSM) are investigated by studying fully developed axially rotating turbulent pipe flow. The mean flow and the turbulence quantities are strongly influenced by the imposed rotation, and is well captured by the differential models as well as their algebraic truncations. All the tested models give mean velocity profiles that are in good qualitative agreement with the experimental data. It is demonstrated that the predicted turbulence kinetic energy levels vary dramatically depending on the diffusion model used, and that this is closely related to the model for the evolution of the length-scale determining quantity. Furthermore, the effect of the weak equilibrium assumption, underlying the EARSMs, and the approximation imposed for 3D mean flows on the turbulence levels are investigated. In general the predictions obtained with the EARSMs rather closely follow those of the corresponding DRSMs.

An explicit algebraic Reynolds stress and heat flux model for incompressible turbulence: Part II Buoyant flow

This paper presents a derivation of an explicit algebraic model for two-dimensional (2-D) buoyant flows. It is an extension of the work reported in Part I (So et al. [27]). The tensor representation method of Jongen and Gatski [14] is extended to derive an explicit algebraic Reynolds stress model (EASM) for 2-D buoyant flow invoking the Boussinesq approximation. The projection methodology is further extended to treat the heat flux transport equation in the derivation of an explicit algebraic heat flux model (EAHFM) for buoyant flow. Again, the weak equilibrium assumption is invoked for the scaled Reynolds stress and scaled heat flux equation. An explicit algebraic model for buoyant flows is then formed with the EASM and EAHFM. From the derived EAHFM, an expression for the thermal diffusivity tensor in buoyant shear flows is deduced. Furthermore, a turbulent Prandtl number (PrT) for each of the three heat flux directions is determined. These directional PrT are found to be a function of the gradient Richardson number. Alternatively, a scalar PrT can be derived; its value is compared with the directional PrT. The EASM and EAHFM are used to calculate 2-D homogeneous buoyant shear flows and the results are compared with direct numerical simulation data and other model predictions, where good agreement is obtained.

Modelling streamline curvature effects in explicit algebraic Reynolds stress turbulence models

A curvature correction for explicit algebraic Reynolds stress models (EARSMs), based on a formal derivation of the weak-equilibrium assumption in a streamline oriented curvilinear co-ordinate system is presented. The curvature correction is given from the rotation rate of the curvilinear co-ordinate system following the mean flow. Two methods for defining that rotation rate are proposed, one is derived from the strain-rate tensor, and the other is derived from the local mean acceleration vector. Both methods are fully three-dimensional and Galilean invariant and the correction vanishes in cases without curvature or rotation effects. The EARSM proposed by Wallin and Johansson (J. Fluid Mech. 403 (2000) 89) was extended with the proposed curvature corrections and recalibrated in such a way that the original model was retrieved in cases without curvature or rotation effects. Rotating homogeneous turbulent shear flows with vanishing mean vorticity should be close to neutral stability according to linear stability theory, also observed from large eddy simulations. This was used for the recalibration. The importance of the curvature correction and the proposed recalibration is shown for rotating homogeneous shear and rotating channel flows.