scholarly journals Turbulence Modelling for Secondary Flow Prediction in a Turbine Cascade

1991 ◽  
Author(s):  
J. G. E. Cleak ◽  
D. G. Gregory-Smith
Keyword(s):  
Secondary Flow ◽  
Turbulence Models ◽  
Equation Model ◽  
Turbulent Shear ◽  
Shear Stresses ◽  
Mixing Length ◽  
Turbine Cascade ◽  
Mean Flow ◽  
Complex Flow ◽  
The One

Predictions of secondary flow in an axial turbine cascade have been made using three different turbulence models; mixing length, a one equation model and a k-epsilon/mixing length hybrid model. The results are compared with results from detailed measurements, not only by looking at mean flow velocities and total pressure loss, but also by assessing how well turbulence quantities are predicted. It is found that the turbulence model can have a big influence on the mean flow results, with the mixing length model giving generally the best mean flow. None of the models give good predictions of the turbulent shear stresses in the vortex region, although the k-epsilon model gives quite good turbulent kinetic energy values. The one equation model is the only one to contain a transition criterion. The importance of such a criterion is illustrated, but the present one needs development to give reliable predictions in the complex flow within a blade passage.

Turbulence Modeling for Secondary Flow Prediction in a Turbine Cascade

1992 ◽  
Vol 114 (3) ◽  
pp. 590-598 ◽  
Author(s):  
J. G. E. Cleak ◽  
D. G. Gregory-Smith
Keyword(s):  
Secondary Flow ◽  
Turbulence Modeling ◽  
Turbulence Models ◽  
Equation Model ◽  
Turbulent Shear ◽  
Shear Stresses ◽  
Mixing Length ◽  
Turbine Cascade ◽  
Mean Flow ◽  
Complex Flow

Predictions of secondary flow in an axial turbine cascade have been made using three different turbulence models: mixing length, a one-equation model and a k–ε mixing length hybrid model. The results are compared with results from detailed measurements, not only by looking at mean flow velocities and total pressure loss, but also by assessing how well turbulence quantities are predicted. It is found that the turbulence model can have a big influence on the mean flow results, with the mixing length model giving generally the best mean flow. None of the models give good predictions of the turbulent shear stresses in the vortex region, although the k–ε model gives quite good turbulent kinetic energy values. The one-equation model is the only one to contain a transition criterion. The importance of such a criterion is illustrated, but the present one needs development to give reliable predictions in the complex flow within a blade passage.

Eddy Viscosity Transport Equations and Their Relation to the k-ε Model

1997 ◽  
Vol 119 (4) ◽  
pp. 876-884 ◽  
Author(s):  
F. R. Menter
Keyword(s):  
Boundary Layer ◽  
Shear Stress ◽  
Eddy Viscosity ◽  
Turbulence Models ◽  
Equation Model ◽  
Turbulent Shear ◽  
Boundary Layer Flows ◽  
Turbulent Layer ◽  
Turbulent Shear Stress ◽  
The One

A formalism will be presented which allows transforming two-equation eddy viscosity turbulence models into one-equation models. The transformation is based on Bradshaw’s assumption that the turbulent shear stress is proportional to the turbulent kinetic energy. This assumption is supported by experimental evidence for a large number of boundary layer flows and has led to improved predictions when incorporated into two-equation models of turbulence. Based on it, a new one-equation turbulence model will be derived from the k-ε model. The model will be tested against the one-equation model of Baldwin and Barth, which is also derived from the k-ε model (plus additional assumptions) and against its parent two-equation model. It will be shown that the assumptions involved in the derivation of the Baldwin-Barth model cause significant problems at the edge of a turbulent layer.

Analysis of Turbulent Thermal Convection Between Horizontal Plates

1983 ◽  
Vol 105 (4) ◽  
pp. 789-794 ◽  
Author(s):  
M. Kaviany ◽  
R. Seban
Keyword(s):  
Temperature Distribution ◽  
Kinetic Energy ◽  
Turbulent Kinetic Energy ◽  
Thermal Convection ◽  
Equation Model ◽  
Experimental Results ◽  
Mixing Length ◽  
Mean Temperature ◽  
The Mean ◽  
The One

The one-equation model of turbulence is applied to the turbulent thermal convection between horizontal plates maintained at constant temperatures. A pseudo-three-layer model is used consisting of a conduction sublayer adjacent to the plates, a turbulent region within which the mixing length increases linearly, and a turbulent core within which the mixing length is a constant. It is assumed that the Nusselt number varies with the Rayleigh number to the one-third power. As a result, the steady-state distributions of the turbulent kinetic energy and the mean temperature are obtrained and presented in closed forms. These results include the effects of Prandtl number. The predictions are compared with the available experimental results for different Prandtl and Rayleigh numbers. Also included are the predictions of Kraichnan, which are based on a less exact analysis. The results of the one-equation model are in fair agreement with the experimental results for the distribution of the turbulent kinetic energy and the mean temperature distribution. The predictions of Kraichnan are in better agreement with the experimental results for the mean temperature distribution.

Application of Non-Linear k–ω Model to the Turbulent Flow Inside a Sharp U-Bend

2000 ◽  
Author(s):  
B. Song ◽  
R. S. Amano
Keyword(s):  
Reynolds Number ◽  
Turbulent Flow ◽  
Low Reynolds Number ◽  
Turbulence Models ◽  
Higher Order ◽  
Shear Stresses ◽  
Complex Flow ◽  
Streamline Curvature ◽  
Non Linear ◽  
Low Reynolds

Simulation of the complex flow inside a sharp U-bend needs both refined turbulence models and higher order numerical discretization schemes. In the present study, a nonlinear low-Reynolds number (low-Re) k–ω model including the cubic terms was employed to predict the turbulent flow through a square cross-sectioned U-bend with a sharp curvature, Rc/D = 0.65. In the turbulence model employed for the present study, the cubic terms are incorporated to represent the effect of extra strain-rates such as streamline curvature and three-dimensionality on both turbulence normal and shear stresses. In order to accurately predict such complex flowfields, a higher-order bounded interpolation scheme (Song, et al., 1999) has been used to discretize all the transport equations. The calculated results by using both the non-linear k–ω model and the linear low-Reynolds number k–ε model (Launder and Sharma, 1974) have been compared with experimental data. It is shown that the present model produces satisfactory predictions of the flow development inside the sharp U-bend and well captures the characteristics of the turbulence anisotropy within the duct core region and wall sub-layer.

A Dynamic Time Filtering Technique for Hybrid RANS-LES Simulation of Non-Stationary Turbulent Flow

2019 ◽  
Author(s):  
Tausif Jamal ◽  
D. Keith Walters
Keyword(s):  
Large Eddy Simulation ◽  
Comprehensive Evaluation ◽  
Model Performance ◽  
Turbulence Models ◽  
Mean Flow ◽  
Complex Flow ◽  
Eddy Simulation ◽  
Large Eddy ◽  
Filtering Technique ◽  
Dynamic Time

Abstract Unsteady turbulent wall bounded flows can produce complex flow physics including temporally varying mean pressure gradients, intermittent regions of high turbulence intensity, and interaction of different scales of motion. As a representative example, pulsating channel flow presents significant challenges for newly developed and existing turbulence models in computational fluid dynamics (CFD) simulations. The present study investigates the performance of the Dynamic Hybrid RANS-LES (DHRL) model with a newly proposed dynamic time filtering (DTF) technique, compared against an industry standard Reynolds-Averaged Navier-Stokes (RANS) model, Monotonically Integrated Large Eddy Simulation (MILES), and two conventional Hybrid RANS-LES (HRL) models. Model performance is evaluated based on comparison to previously documented Large Eddy Simulation (LES) results. Simulations are performed for a fully developed flow in a channel with time-periodic driving pressure gradient. Results highlight the relative merits of each model type and indicate that the use of a dynamic time filtering technique improves the accuracy of the DHRL model when compared to a static time filtering technique. A comprehensive evaluation of the results suggests that the DHRL-DTF method provides the most consistently accurate reproduction of the time-dependent mean flow characteristics for all models investigated.

Anisotropic Eddy Viscosity in the Secondary Flow of a Low-Speed Linear Turbine Cascade

2011 ◽  
Author(s):  
G. D. MacIsaac ◽  
S. A. Sjolander
Keyword(s):  
Turbulent Flow ◽  
Secondary Flow ◽  
Eddy Viscosity ◽  
Turbulence Models ◽  
Leading Edge ◽  
Boussinesq Approximation ◽  
Pressure Probe ◽  
Turbine Cascade ◽  
Low Speed ◽  
Sst Turbulence Model

The final losses within a turbulent flow are realized when eddies completely dissipate to internal energy through viscous interactions. The accurate prediction of the turbulence dissipation, and therefore the losses, requires turbulence models which represent, as accurately as possible, the true flow physics. Eddy viscosity turbulence models, commonly used for design level computations, are based on the Boussinesq approximation and inherently assume the eddy viscosity field is isotropic. The current paper compares the computational predictions of the flow downstream of a low-speed linear turbine cascade to the experimentally measured results. Steady-state computational simulations were performed using ANSYS CFX v12.0 and employed the shear stress transport (SST) turbulence model with the γ-Reθ transition model. The experimental data includes measurements of the mean and turbulent flow quantities. Steady pressure measurements were collected using a seven-hole pressure probe and the turbulent flow quantities were measured using a rotatable x-type hotwire probe. Data is presented for two axial locations: 120% and 140% of the axial chord (Cx) downstream of the leading edge. The computed loss distribution and total bladerow losses are compared to the experimental measurements. Differences are noted and a discussion of the flow structures provides insights into the origin of the differences. Contours of the shear eddy viscosity are presented for each axial plane. The secondary flow appears highly anisotropic, demonstrating a fundamental difference between the computed and measured results. This raises questions as to the validity of using two-equation turbulence models, which are based on the Boussinesq approximation, for secondary flow predictions.

Numerical Study of the Turbulent Flow Around a Square Cylinder Near a Wall

2002 ◽  
Author(s):  
Emmanuel Guilmineau ◽  
Patrick Queutey
Keyword(s):  
Reynolds Stress ◽  
Numerical Study ◽  
Turbulence Models ◽  
Equation Model ◽  
Stream Velocity ◽  
Periodic Case ◽  
Square Cylinder ◽  
Mean Velocity ◽  
The One ◽  
Adjacent Wall

Calculations are reported for the flow around a two-dimensional, square cylinder at Re = 22,000 (based on the prism side dimension, D, and the free-stream velocity) placed at various distances from an adjacent wall. The nominal boundary layer thickness is 1.5D. Experiments have indicated that unsteady vortex shedding is suppressed when the wall is relatively close to the cylinder. The turbulent fluctuations are simulated with three turbulence models: the one-equation model of Spalart & Allmaras (1992), the two-equations SST K–ω model (Menter, 1993) and a Reynolds stress Rij–ω closures (Deng & Visonneau, 1999). The paper consists in comparing simulation and experimental results for configurations S/D = 1 (periodic case) and S/D = 0.25 (stationary case). Predicted and measured distributions of the mean velocity, Reynolds stress tensor and surface pressures are compared. Although the agreement is very good in general, observed discrepancies are discussed.

A Numerical Study of the U-Turn Bend in Return Channel Systems for Multistage Centrifugal Compressors

2005 ◽  
Vol 219 (8) ◽  
pp. 749-756 ◽  
Author(s):  
J. M. Oh ◽  
A Engeda ◽  
M. K. Chung
Keyword(s):  
Velocity Profile ◽  
Secondary Flow ◽  
Numerical Study ◽  
Flow Analysis ◽  
Turbulence Models ◽  
Jet Flow ◽  
Complex Flow ◽  
Centrifugal Compressors ◽  
Return Channel ◽  
Channel Systems

A qualitative numerical study of the flow in the U-turn bend of return channel systems for multistage centrifugal compressors is presented. Calculations have been carried out using the flow analysis program FLUENT. The flow in the U-turn bend is highly three-dimensional and complex. The main cause for this is the circumferential variation of the velocity profile at the inlet of the bend. The circumferential variation of the velocity profile is an unavoidable result from the wake/jet flow at the exit of the impeller. In this article, first the effect of the wake/jet flow coming into the U-turn bend is studied. It is shown that the wake/jet flow develops to form the secondary flow in the U-turn bend. The secondary flow, with the high streamline curvature of the flow in the bend, makes the flow inside the bend highly complex. This complex flow is hard to predict with conventional turbulence models that have been developed on the basis of near homogeneity of flows. Comparing the present result with a study that successfully predicted the loss and flow behaviour in the bend, a discussion is presented on the turbulence and the turbulence models. Also, the loss mechanisms in the U-turn bend are discussed in detail.

Prediction of a Laminar Separation Bubble Over a Controlled-Diffusion Compressor Blade

2000 ◽  
Author(s):  
G. V. Hobson ◽  
S. Weber
Keyword(s):  
Turbulence Model ◽  
Separation Bubble ◽  
Turbulence Models ◽  
Leading Edge ◽  
Equation Model ◽  
Implicit Time ◽  
Controlled Diffusion ◽  
Time Marching ◽  
The One ◽  
Edge Separation

The paper describes the comparison of the prediction of the flow through a cascade of controlled-diffusion compressor blades with two Navier-Stokes solvers. Both codes solved the thin-layer N-S equations, however; one code performed implicit time marching whereas the other performed explicit time marching. Flow predictions were accomplished with the implicit code using the algebraic turbulence model of Baldwin and Lomax and the one-equation model of Spalart and Allmaras, while predictions were made with the explicit code using the two-equation model by Wilcox. Predictions were made of the detailed laser-anemometry measurements of the flow field taken previously in a low-speed cascade wind tunnel. Comparisons were also made with the experimentally measured blade surface pressures and flow visualization of the extent of the laminar leading edge separation bubble. The one-equation turbulence model was combined with an intermittency based transition-length model for comparisons with fully turbulent calculations. Both codes predicted the leading-edge separation bubble satisfactorily when using higher order turbulence models.

Mean flow structure in meanders of the Squamish River, British Columbia

1978 ◽  
Vol 15 (11) ◽  
pp. 1833-1849 ◽  
Author(s):  
Edward J. Hickin
Keyword(s):  
Velocity Field ◽  
Secondary Flow ◽  
Flow Structure ◽  
Maximum Velocity ◽  
Shear Stresses ◽  
Channel Migration ◽  
Mean Flow ◽  
Channel Form ◽  
Collapse Model ◽  
Primary Velocity

The primary velocity field and pattern of secondary flow are described for nine consecutive bends of the Squamish River in southwest British Columbia.The velocity field largely can be explained in terms of variation in channel form, advective acceleration responses, and water transfers by secondary flow.The pattern of secondary flow accords with the general model of spiral flow in meanders. Divergences from this ideal pattern can be explained by bend–flow interaction induced by the variable planform geometry of the channel.The strength of secondary circulation increases rapidly as the ratio of the radius of bend curvature to channel width (rm/w) declines from 4.0 to the data minimum of 1.41. There is no discontinuity phenomenon in the flow structure over the measured range of rm/w; the Bagnold separation–collapse model does not apply to the Squamish River.As rm/w declines to values less than 3.0, the maximum velocity filament shifts from the concave to the convex bank zone. The resulting high shear stresses over the point bar and declining shear stresses at the concave bank markedly reduce the channel migration rate.Separation zones developed at the concave bank of tightly curved bends provide the mechanism for completely halting (and indeed reversing) the process of channel migration.

Sign in / Sign up

Export Citation Format

Share Document