Juan Uribe

Synthetic Inflow Boundary Conditions for Wall Bounded Flows

The present paper summarizes the research conducted at the University of Manchester over the course of the Desider project on methods of generation of synthetic turbulence. A random procedure, referred to as the Synthetic Eddy Method (SEM), based on the classical view of turbulence as a superposition of coherent structures was developed and compared with other methods of generation of inflow boundary conditions for LES, for different types of wall bounded flows. Synthetic inflow generation methods always show a transition region downstream of the inlet until the turbulence recovers an equilibrium state. The sensitivity of this transition region to variations in the synthetic inflow turbulence energy, length scale and time scale is studied herein with applications to hybrid RANS-LES methods in mind, as RANS results cannot be expected to provide accurate values for these scales to the LES region. A parameterization of the SEM suitable for hybrid RANS-LES simulations is proposed and tested on hybrid RANS-LES simulations of several wall bounded flows, with an emphasis on the flat plat boundary layer.

(v2-/k)-f Turbulence Model and its Application to Forced and Natural Convection

We present the rationale and some validation of a version of Durbin’s elliptic relaxation eddy-viscosity model, which solves a transport equation for the velocity scale ratio v 2 ¯ / k instead of v 2 ¯ . The new model, developed independently at TU Delft and UMIST (in two variants, shows improved robustness, faster convergence and less sensitivity to grid nonuniformities. The two variants differ insignificantly in the formulation of the v 2 ¯ / k and f equations: the UMIST model endeavours to remain close to Durbin’s original one, while TUD variant introduces quasi-linear pressure strain formulation but with some further numerically beneficial simplifications. The model validation in a range of attached, separating and impinging flows with heat transfer, as well as in natural convection in a tall cavity, showed satisfactory predictions in all cases considered.

Development of an Alternative Delayed Detached-Eddy Simulation Formulation Based on Elliptic Relaxation

CD = drag coefficient CDDES = empirical parameter Cf = skin-friction coefficient CL = lift coefficient Cp = pressure coefficient Ce1 = model constant for the dissipation equation Ce2 = model constant for the dissipation equation c = chord length f = elliptic operator fd = delayed detached-eddy simulation blending function h = hill height k = turbulent kinetic energy L = turbulent length scale Re = Reynolds number S = deformation tensor Ub = bulk velocity U∞ = freestream velocity y = distance to the nearest wall y = nondimensional wall distance Δ = large-eddy simulation filter width Δt = time step e = turbulent dissipation κ = von Karman constant ν = molecular viscosity νt = turbulent viscosity Ψ = delayed detached-eddy simulation correction term

Benchmarking of Three Different CFD Codes in Simulating Natural, Forced, and Mixed Convection Flows

In this study, three different CFD codes are assessed in simulating two distinct flow problems with heat transfer. The first case consists of an ascending forced and mixed convection flow, a representative flow in the core of gas-cooled nuclear reactors under “post-trip” conditions, and the second case involves natural convection in an enclosed tall cavity that represents the gas-filled cavities around the nuclear reactor cores. Computations are conducted using the standard k-ω-SST model and the nonlinear k–ε model of Suga. Overall, the results were in satisfactory agreement, and, therefore, the present study represents a successful benchmarking exercise for the nuclear community.

Towards Petascale Computing with Parallel CFD codes

Many world leading high-end computing (HEC) facilities are now offering over 100 Teraflops/s of performance and several initiatives have begun to look forward to Petascale computing5 (1015 flop/s). Los Alamos National Laboratory and Oak Ridge National Laboratory (ORNL) already have Petascale systems, which are leading the current (Nov 2008) TOP500 list [1]. Computing at the Petascale raises a number of significant challenges for parallel computational fluid dynamics codes. Most significantly, further improvements to the performance of individual processors will be limited and therefore Petascale systems are likely to contain 100,000+ processors. Thus a critical aspect for utilising high Terascale and Petascale resources is the scalability of the underlying numerical methods, both with execution time with the number of processors and scaling of time with problem size. In this paper we analyse the performance of several CFD codes for a range of datasets on some of the latest high performance computing architectures. This includes Direct Numerical Simulations (DNS) via the SBLI [2] and SENGA2 [3] codes, and Large Eddy Simulations (LES) using both STREAMS LES [4] and the general purpose open source CFD code Code Saturne [5].

LES and Hybrid RANS/LES of Turbulent Flow in Fuel Rod Bundle Arranged with a Triangular Array

Turbulent flow parallel to fuel rod bundles arranged in a triangular array are computed using LES and hybrid RANS/LES. Inner-channel flow pulsations are captured and the dominant frequency is in agreement with the experiments. The hybrid method shows a lack of accuracy in the near wall region probably due to the formulation of the blending function.

Two-Velocities Hybrid RANS-LES of a Trailing Edge Flow

The flow around a trailing edge is computed with a new hybrid method designed to split the influences of the averaged and instantaneous velocity fields. The model is first tested on channel flows at different Reynolds numbers and coarse meshes giving good predictions of mean velocities and stresses. On the trailing edge flow the predictions of the hybrid model are compared with those using DES-SST on the same coarse mesh. The results of the hybrid model are close to the reference fine LES in terms of mean velocity and turbulent content.

Computation of Flow in a 3D Diffuser Using a Two-Velocity Field Hybrid RANS/LES

The flow inside a three-dimensional diffuser is computed with a two-velocity hybrid RANS/LES model that ensues the separation of dissipative effects of the mean and fluctuating fields to be treated individually; by extracting a running average velocity field from instantaneous quantities. The averaged field is then used to calculate the contribution of the mean shear which is larger than that from the fluctuating one over the near wall region. Results with the hybrid model are presented and compared to those from a DES on the same grid. RANS results obtained with models upon which both these approaches are based are also reported. The RANS results are unable to capture essential characteristics of the flow whereas the hybrid method produces results which generally agree well with the experiment.

Development of a Hybrid RANS/LES Model for Heat Transfer Applications

This work presents a scalar flux model in the framework of a hybrid RANS-LES modelling. The model is tested on a heated channel flow at different Prandtl numbers and on a T-junction. Results show a good agreement with both DNS and experimental data.

Methods used and highlighted results from UMIST

The preferred models of the consortium (SSG, k-w, STT, SCW) have been implemented in a recent open source FV software and tested. Numerical behaviour of two popular versions of the V2F model have been compared and a new formulation proposed, that combines the accuracy of the original version and stability of the degraded “code friendly version”. In the DES framework, the need for reduction of the EVM coefficient has been related to the stress-strain missalignment in rapidly changing time dependent mode and a single transport equation for this coefficient proposed. On the wall function front, the AWF of Craft and Gerasimov entailing an analytical integration of terms over the first cell (body force in particular) proved challenging to implement on unstructured grids, but was achieved. Towards the end of the project, a promising mathematical framework (Robin type conditions) was proposed, based on domain decomposition and mixed Dirichlet and Neumann conditions.