58
ENGINEERING DIPLOMA THESIS by Emeka G. Chijioke Modelling of the Laminar to Turbulent Flow Transition in a Compressor Cascade Modelowanie przejścia laminarno-turbulentnego w przepływie przez wieniec sprężarki Index Number: 209323 Course : Mechanical Engineering (Mechanika i Budowa Maszyn) Thesis Advisor : Dr inż. Sławomir Kubacki Warsaw, January 2012. WARSAW UNIVERSITY OF TECHNOLOGY FACULTY OF POWER AND AERONAUTICAL ENGINEERING DEPARTMENT OF AERODYNAMICS

Dyplom-final Modification

Embed Size (px)

Citation preview

Page 1: Dyplom-final Modification

1

ENGINEERING

DIPLOMA THESIS

by

Emeka G. Chijioke

Modelling of the Laminar to Turbulent Flow

Transition in a Compressor Cascade Modelowanie przejścia laminarno-turbulentnego w przepływie

przez wieniec sprężarki

Index Number: 209323

Course : Mechanical Engineering (Mechanika i Budowa Maszyn)

Thesis Advisor : Dr inż. Sławomir Kubacki

Warsaw, January 2012.

WARSAW UNIVERSITY OF TECHNOLOGY

FACULTY OF POWER AND AERONAUTICAL

ENGINEERING

DEPARTMENT OF AERODYNAMICS

Page 2: Dyplom-final Modification

2

Undersigned, Emeka Chijioke, author of this Thesis.

Aware of criminal liability for making untrue statements I declare that the

following thesis was written personally by myself and that I did not use

any sources but the ones mentioned in the dissertation itself.

I also certify that this version of the work is identical with the attached

electronic version.

--------------------- ------------------

date signature

ACKNOWLEDGEMENTS

I would like to thank Dr. Sławomir Kubacki for his infinite patience, guidance

and support without which this project would never have been completed. Also,

my father and mother and the rest of my family who have always supported me

in everything I have chosen to partake.

Page 3: Dyplom-final Modification

3

ABSTRACT

The purpose of this project is to numerically calculate and understand the flow

behaviour over a NACA 65 Series airfoil using Fluent Solver and to correlate the

data obtained with the provided experimental data from Zaki paper (Zaki T.A,

Wissink J.G, Rodi W and Durbin P.A., Direct simulations of transition in a

compressor cascade: the influence of free-stream turbulence). The flow through

a compressor passage with incoming free-stream grid turbulence is simulated.

At moderate Reynolds number, laminar-to-turbulence transition can take place

on both sides of the airfoil, but proceeds in distinctly different manners (Zaki

and Wissink 2009).

The model for the airfoil 65 series was created, drawn and meshed in Gambit

using geometry data that were sent by Zaki. The model was read into FLUENT

where flow boundary conditions were applied and the discretized Navier-

Stokes equations were solved numerically. The airfoil section pressure

coefficient from the numeric simulation was compared with DNS data from

paper by Zaki and the results agree within 20% for angle of attack 42°.

Page 4: Dyplom-final Modification

4

TABLE OF CONTENTS

Page

ACKNOWLEDGEMENTS………..…………………………………………....... 2

ABSTRACT.............................................................................................................. 3

1. INTRODUCTION.............................................................................................. 6

1.1 Motivation………………………………………………………………….. 6

1.2 Laminar versus turbulent flow.…………………………………………....... 7

1.3 Laminar flow……………………………………………………………….. 7

1.4 Turbulent flow………………………………………………………….…... 8

1.5 Viscosity…………………………………………………………............…. 8

1.6 Boundary layer…………………………………………………………....... 8

1.7 Transitional Modes………………………………………………………..... 9

2. THE GOVERNING EQUATIONS…………………..................................... 11

2.1 Time-averaged Navier-Stokes equations………………………….............. 12

2.2 Transitional model by Walters……………………………………………. 14

2.3 Turbulence model (SST k-ω)……………………………………………... 23

3. OBJECTIVES OF THE THESIS..………………………..………............... 24

4. COMPUTAIONAL DETAILS.……………...……………..…..…................ 24

4.1 Computational mesh……………………………………………………… 24

4.2 Boundary Conditions…………………………………………..………….. 29

Page 5: Dyplom-final Modification

5

4.3 Convergence Criteria…...…………………………………………......…... 32

5. RESULTS AND DISCUSSIONS...………….………………….………….... 34

5.1 Verification of the inlet conditions………………………...……………… 34

5.2 Analysis of the mean velocity and pressure fields……….....….................. 38

5.3 Contour plots of the turbulent to molecular viscosity ratio……………..... 45

5.4 Transition onset …………………………………………….………..….... 49

5.5 Pressure coefficient …………………………………..………………...… 51

5.6 The Skin friction coefficient…...……………...……..……………...……. 55

6. CONCLUSIONS...……………………………....…………………................ 55

7. REFERENCES………………………………………………….….…....…... 56

Page 6: Dyplom-final Modification

6

INTRODUCTION

1.1 Motivation

The ability to accurately predict transitional fluid flow behaviour is important to

the design of engineering systems in a wide variety of applications including

aerospace, automotive, biomedical, heating and cooling, power generation,

marine systems, and chemical processing. Methods for addressing boundary

layer transition in computational fluid dynamics (CFD) simulations can range

from highly empirical approaches based on “engineering insight” for example,

selecting an appropriate transition location and applying a turbulence model

only downstream of this location to direct numerical simulation (DNS).

Almost every flow in nature and in practical engineering applications is

turbulent. After years of research in turbulence, there still does not exist a

precise definition of turbulence. However, some of the characteristics of

turbulent flows can be listed: irregularity, diffusivity, large Reynolds numbers,

three-dimensional vorticity fluctuations, and dissipation (Tennekes and Lumley,

1972). Inspite of all the uncertainties associated with turbulent flows, it has

been encouraging that engineering calculations have been possible with well-

formulated turbulence models.

In 1937, Taylor and von Kármán proposed the following definition of

turbulence:

“Turbulence is an irregular motion which in general makes its appearance in

fluids, gaseous or liquid, when they flow past solid surfaces or even when

neighbouring streams of the same fluid flow past or over one another” (Wilcox,

1994). Turbulence is usually characterized by the presence of a wide range of

length and time scales (Wilcox, 1994).

Page 7: Dyplom-final Modification

7

1.2 Laminar versus turbulent flow

Some flows are smooth and orderly , while others are rather chaotic. The flows

characterized by smooth layers of fluid is called Laminar. Example, the flow of

highly viscous fluid such as oils at low velocities is typically laminar.

The highly disordered fluid flows at high velocities and also characterized

by velocity fluctuations is called Turbulent . The flow of low-velocity fluids such

as air at high velocities is typically turbulent. (Yunus A. Cengel ; John M.

Cimbala).

1.3 Laminar Flow

Laminar Flow is the smooth, uninterrupted flow of air over the contour of the

wings, fuselage, or other parts of an aircraft in flight. Laminar flow is most often

found at the front of a streamlined body and is an important factor in flight. If

the smooth flow of air is interrupted over a wing section, turbulence is created

which results in a loss of lift and a high degree of drag. An airfoil designed for

minimum drag and uninterrupted flow of the boundary layer is called a laminar

airfoil.

The theory in using an airfoil of a particular design was to maintain the

adhesion of the boundary layers of airflow which are present in flight as far aft

of the leading edge as possible. On normal airfoils, the boundary layer would

be interrupted at high speeds and the resultant break would cause a turbulent

flow over the remainder of the foil. This turbulence would be realized as drag

up to the point of maximum speed, at which time the control surfaces and

aircraft flying characteristics would be affected. The formation of the boundary

layer is a process of layers of air formed one next to the other, i.e. the term

laminar is derived from the lamination principle involved.

Page 8: Dyplom-final Modification

8

1.4 Turbulent Flow

Most flows encountered in engineering practice are turbulent. Turbulent

flow is a complex mechanism dominated by fluctuations, characterized by

random and rapid fluctuations of swirling regions of fluid, called eddies,

throughout the flow. In turbulent flow, the swirling eddies transports mass,

momentum, and energy to other regions of flow much more rapidly than

molecular diffusion , greatly enhancing mass, momentum, and heat transfer.

1.5 Viscosity

Viscosity is the measure of internal stickiness of a fluid. Viscosity is caused by

cohesive forces between the molecules in liquid and by molecular collision in

gases . There is no fluid with zero viscosity, so to some degree , all fluid flows

involve viscous effects (Yunus A. Cengel ; John M. Cimbala).

1.6 Boundary layer

As an object moves through a fluid, or as a fluid moves past an object, the

molecules of the fluid near the object are disturbed and move around the

object. Aerodynamic forces are generated between the fluid and the object.

The magnitude of these forces depend on the shape of the object, the speed of

the object, the mass of the fluid passing by the object and on two other

important properties of the fluid; the viscosity, or stickiness, and the

compressibility, or springiness, of the fluid.

Aerodynamic forces depend in a complex way on the viscosity of the fluid.

As the fluid moves past the object, the molecules right next to the surface stick

to the surface. The molecules just above the surface are slowed down in their

collisions with the molecules sticking to the surface. These molecules in turn

slow down the flow just above them. The farther one moves away from the

surface, the fewer the collisions affected by the object surface. This creates a

thin layer of fluid near the surface in which the velocity changes from zero at

the surface to the free stream value away from the surface. This layer is called

the boundary layer because it occurs on the boundary of the fluid.

Page 9: Dyplom-final Modification

9

1.7 Transitional Modes

The important flow characteristic of fluid flow is transition to turbulence.

Transition is the process by which a laminar flow changes to a turbulent flow.

It is known that, typically, the boundary layer flow is laminar over the surface

of the body before it transitions to turbulent flow due to flow instabilities.

Instability of a laminar flow does not immediately lead to turbulence, which is a

severely nonlinear and chaotic stage characterized by macroscopic “mixing” of

fluid particles. Some of the transition modes which lead to turbulence are

natural transition, bypass transition, or separated flow transition. Below are

brief discussion on these different transition modes in summary of what

appears in Mayle (1991).

Natural transition: during this process, initial breakdown of laminar

flow occurs because of amplification of small disturbances, after the initial

breakdown the flow goes through a complex sequence of changes finally

resulting in the chaotic state known as turbulence. Natural transition occurs

when the laminar boundary layer is affected by small disturbances, which grow

into an instability. This instability amplifies within the layer to a point where it

grows and develops into loop vortices with large fluctuations. These highly

fluctuating loop vortices inside the laminar boundary layer develop into

turbulent spots, which then are convected downstream, and eventually, with

time, grow and come together to form a fully developed turbulent boundary

layer.

Bypass transition: this usually occurs at high free-stream turbulence

levels. In this mode of transition, free-stream disturbances influence the

development of turbulent spots that are directly produced within the boundary

layer.

Separated-flow transition: this occurs in the laminar separation bubble.

The flow transitions into turbulent flow over the separated bubble and

reattaches to the surface forming a turbulent shear layer. This usually occurs in

an adverse pressure gradient region that contributes to the separation of the

laminar boundary layer. Separated flow transition is usually found on the

suction surface, near a compressor airfoil’s leading edge, or near the point of

Page 10: Dyplom-final Modification

10

minimum pressure. Turbine blades are likely to have separation along the

suction surface in the trailing edge region. High levels of free-stream

turbulence can cause early transition compared to lower turbulence levels.

In gas turbine engines, the flow is periodically unsteady, so is transition,

and this is called periodic-unsteady transition. In “wake-induced” transition,

the periodic passing of wakes from the upstream blades or obstructions causes

unsteadiness in the flow field and affects transition on the downstream blades.

There also exists something called reverse transition, i.e., transition from

turbulent to laminar flow, which is referred to as “relaminarization.” This is

usually expected to occur at low turbulence levels if the acceleration

parameter, ( )( ) , is greater than 3 x 10-6

. In this equation, U

refers to the velocity in the streamwise direction and x refers to the surface

coordinate in the streamwise direction.

Predicting transition becomes very important for improving the efficiencies of

gas turbine engines. Considering transition will lead to improved designs of

turbomachinery airfoils. A significant amount of research effort has been

devoted to determine the transition regime inside the boundary layer. Since

Direct Numerical Simulation (DNS) and Large Eddy Simulation (LES) are more

computationally expensive using present computing hardware, the Reynolds-

Averaged Navier-Stokes (RANS) equations continue to be better suited for

engineering calculations with the incorporation of appropriate turbulence and

transition models.

Page 11: Dyplom-final Modification

11

2 The Governing Equations

2.1 Time- Average Navier- Stokes Equation (RANS) The Reynolds-averaged Navier–Stokes equations (or RANS equations) are time-averaged equations of motion for fluid flow and are primarily used to describe turbulent flows.

Statistical methods are used to average the fluctuating properties of flow in turbulent case. To obtain the mean values, there are different averaging techniques such as time, spatial and ensemble averaging. For homogenous turbulence with uniform turbulent flow in all directions, spatial averaging is used. For stationary turbulence (i.e. a turbulent flow that, on average, does not vary with time), time averaging is used. Ensemble averaging is the most suitable averaging for flows decaying with time. For the flows that engineers mostly deal with, time averaging is used.

For this thesis, I used the time-averaging technique to couple the models

for both transitional and turbulent cases.

According to Walters (bulletin 80), the Reynolds averaging is also

relevant in the laminar flow region. This means that in principle, the time-

averaging technique can be used basically for the description of the laminar to

turbulent transition.

Time averaging yields an average and a fluctuating part for a certain

variable. These parts could be represented as the part of the instantaneous

parameter, for example, velocity.

( ) ( ) ( ) (1)

where ( ) is expressed as the instantaneous velocity with, ( ); average

and ( ) fluctuating part, see figure 1. below.

Page 12: Dyplom-final Modification

12

Fig.1, Time average velocity.

The starting point for the computation of a steady, incompressible and

viscous flow are the conservation equations for continuity and momentum relating the laws of mechanics to the fluid:

(2)

(3)

Where and are the velocity components, t is time, p is the pressure, is

density rho and is the viscous stress tensor defined by

(4)

Where is the molecular viscosity and is the strain-rate tensor,

(

) (5)

The convective term in Equation (3) is written in conservation form to

simplify the time-averaging process, i.e.,

( )

( ) (6)

𝑈𝑖(𝑥 𝑡)

𝑈𝑖(𝑥)

t

Page 13: Dyplom-final Modification

13

Taking advantage of Equation (2),

tends to zero.

Then combining Equations (3) through (6) the Navier-Stokes equations can be

written in conservation form.

( )

( ) (7)

Time averaging Equations (2) and (7) yields the Reynolds equations of

motion in conservation form as follows,

(8)

(

)

( ) (9)

The time-average conservation of mass, Equation (8) is identical to

instantaneous Equation (2) with the mean velocity replacing the instantaneous

velocity. Subtracting Equation (8) from Equation (2) shows that the fluctuating

velocity, , also has zero divergence. The appearance the correlation

is

the only difference between the time-averaged and instantaneous momentum

equations.

Formula is needed for computing

in order to compute all mean-flow

properties of the turbulent flow under consideration.

Equation (9) can be written using Equation (6) and this yields the

Reynolds average Navier-Stokes equation (RANS)

(

) (10)

The quantity

is known as the Reynolds-stress tensor denoted by .

Thus,

(11)

Page 14: Dyplom-final Modification

14

The Reynolds-stress tensor is a symmetric tensor ( ), and thus has six

independent components that has to be modelled. Therefor to solve Eq. (10), it

is necessary to find enough equations to close the system.

The tensor is modelled using the Boussinesq approximation (Wilcox

book), thus the Reynolds stress tensor is given by

(12)

Where is the turbulent or eddy viscosity, k is the specific kinetic

energy, is the density, the operator is the Kronecker delta.

2.2 Transitional model by Walters

This model described by Walters, is used to predict boundary layer

development and calculate transition onset. The model can be used to

effectively address the transition of the boundary layer from a laminar to a

turbulent regime.

Model Equations. The Governing equations for the new model is

considered to be a three-equation eddy-viscosity type, where three additional

model transport equations are solved for turbulent kinetic energy ( ), the

laminar kinetic energy , and the scale-determining variable ( ), defined here

as ( ⁄ ), where is the isotropic dissipation.

The transport equations in the conservative form provided in the paper

by (Walters and Cokljat) are

( )

(

)

[(

)

] (13)

Page 15: Dyplom-final Modification

15

( )

(

)

[

] (14)

( )

[

(

)

( )

]

[(

)

] (15)

The various terms in the equations represent production, destruction, and transport mechanisms. The inverse turbulent time-scale ( ) is used here for the new model rather than the dissipation rate ( ).

In the equation, the fully turbulent production, destruction, and gradient transport terms (first, third, and fifth terms on the right-hand side of Equation (15)) are comparable to the similar terms in the and equations, (13 and 14). The transition production term (second term on right-hand side) in Eq. (15) is intended to produce a reduction in turbulence length scale during the transition breakdown process. The fourth term on the right-hand side was included in order to decrease the length scale in the outer region of the turbulent boundary layer, which is necessary to ensure correct prediction of the boundary layer wake region.

The production of turbulent and laminar kinetic energy by mean strain is modeled as:

(16)

(17)

where, is the large-scale eddy-viscosity and is the small-scale eddy-

viscosity. The “small-scale” eddy-viscosity concept is defined as

√ (18)

Page 16: Dyplom-final Modification

16

where is the effective small-scale turbulence.

(19)

The kinematic wall effect is included through an effective (wall-limited) turbulence length scale and damping function .

(

) (20)

( ) (21)

The turbulent length scale ( ) is sufficiently large when it is greater than

is the distance from the wall, is a constant and is the turbulent length

scale, defined as

(22)

and in Eq.(19) is the shear-sheltering effect, which refers to the damping of

turbulence dynamics that occurs in thin regions of high vorticity. Its effect is to

prevents transfer of energy from streamwise to wall-normal (and spanwise)

components.

[ (

)

] (23)

The viscous wall effect is formed through the viscous damping function, which is computed in terms of the effective turbulence Reynolds number.

( √

) (24)

(25)

The turbulent viscosity coefficient is defined to satisfy the realizability

constraint following Shih et al.

(

) (26)

Page 17: Dyplom-final Modification

17

A damping function describing the turbulent production due to intermittency is given by

(

) (27)

The production of laminar kinetic energy is assumed to be governed by the large-scale near-wall turbulent fluctuations based on the correlation of pretransitional fluctuation growth with free-stream low-frequency wall-normal turbulent fluctuations. The large-scale turbulence contribution is given by

(28)

where the small-scale contribution is given in equation (19) above. The production term is

(29)

Where is the large- scale turbulent viscosity, and it is modelled as

{ (

)√

( )

} (30)

where

(31)

( ( )

) (32)

Page 18: Dyplom-final Modification

18

[

] (33)

In the new model, shear-sheltering is incorporated through a production damping term, while transition initiation is included through transfer terms in the and equations. The transition process which is the transfer of energy from the laminar kinetic energy to the turbulent kinetic energy assumes the variable to represent the magnitude of fluctuations that display the characteristics of fully turbulent flow, such as strong three dimensionality, multiple length and time-scales, energy cascading, and significant viscous dissipation. The initiation of the transition process in this model is based on local (single-point) flow conditions and adopts an approach for transition initiation based on the concept of shear-sheltering and consideration of relevant time-scales for nonlinear disturbance amplification and dissipation.

The anisotropic (near wall) dissipation terms for and is of the form.

(34)

(35)

The turbulent transport terms (diffusion terms) in the and equations

include an effective diffusivity defined as

√ (36)

To properly reproduce the behaviour of the boundary layer wake region (boundary layer production term), a kinematic damping function is defined

[ (

)

] (37)

Page 19: Dyplom-final Modification

19

Introducing here the Reynolds stress transport equation for incompressible flow provided in (Walters bulletin), Eq. (38).

The area of interest here is on the production term and the pressure strain term.

( )

(

)

[ ( )

] (38)

Uppercase denotes mean quantities, lowercase denotes fluctuating quantities, and the overbar denotes Reynolds averaging.

The production term, ( ), is responsible for the

transfer energy from the mean to the fluctuating flow, and the pressure strain

term, (

), is responsible for the redistribution of energy among the

normal Reynolds stress components and modify the shear stress components. The effect of the pressure strain term is typically modeled as a "return to

isotropy" in which high energy components transfer energy to lower energy components. Moreover, this term is expressed as the sum of a rapid part, which incorporates interactions between turbulent eddies and the mean velocity field, and a slow part that incorporates inter-eddy interactions.

The three main summarized physical mechanisms involved in the RANS-based description of Reynolds stress dynamics for the transitional boundary layer (Walter bulletin book) include:

1) production of one-dimensional streamwise fluctuation energy in the

pretransitional region by entrained freeestream turbulence interacting with the mean strainrate;

2) no generation of three-dimensional (normal and spanwise) fluctuations in the pretransitional region, due to suppression of the pressure strain mechanism found in turbulent flow;

3) transition initiation due to an increase in magnitude of the pressure strain term, which can be viewed as a transfer of energy from the one-dimensional streamwise fluctuations to the three-dimensional fluctuations more indicative of fully turbulent flow.

Page 20: Dyplom-final Modification

20

The transfer of energy from one to another is appropriately interpreted

as energy redistribution (via the pressure-strain mechanism) rather than production (due to interaction with the mean flow) or dissipation (due to viscous mechanisms), Since the total fluctuation energy in the model is

comprised of the sum of and . It is proposed that the suppression of the

pressure strain effect is due to the presence of short molecular diffusion time scales within the pretransitional region, relative to the characteristic time scale of the rapid pressure strain terms. Potential fluctuations associated with energy redistribution are therefore quickly dissipated. shear sheltering here reduces the effective wall-normal length scale of the entrained wall-normal stress component, reducing the diffusion time scale and stabilizing the mechanisms responsible for intercomponent energy transfer.

In the pretransitional region, entrained turbulent fluctuations do not couple strongly with the freestream velocity gradient, it instead act as perturbations on the mean (approximately laminar) velocity profile. Thereafter, giving the turbulent streamwise fluctuations scaled as:

(39)

Where is a fluctuation length scale in the wall-normal direction given by

(40)

Where is the local mean vorticity magnitude. An estimate for the molecular diffusion time scale is constructed as:

(41)

The time scale associated with the rapid pressure strain mechanisms is

proportional to the inverse of the vorticity magnitude:

(42)

The ratio of molecular diffusion to rapid pressure strain time scales is

therefore expressed as:

(43)

Page 21: Dyplom-final Modification

21

Equation (43) as proposed is the relevant local dimensionless quantity that is governing shear sheltering and transition initiation. When the ratio in Eq. (43) is small, pressure strain is suppressed and one-component fluctuations (i.e. laminar kinetic energy) are generated. When the ratio reaches a critical value, the rapid pressure strain term quickly increases in magnitude to generate three- dimensional turbulent fluctuations, and transition begins.

In Equation (13) – (15), The initiation of bypass transition is included through the term , this term is a function of the timescale ratio expressed in Eq.

(43), with transition initiating at a critical value

⁄ (44)

Where is the threshold function which controls the bypass transition

process:

(

) (45)

[(

) ] (46)

The breakdown to turbulence caused by instabilities is considered to be a natural transition production term, given by

(47)

is used for the modelling of the natural laminar to turbulent flow

transition (for low background turbulence intensity Tu<1). is not

responsible for the activation of the transition onset in my case, this is due to

high background of turbulence intensities.

(

) (48)

[( ) ] (49)

Page 22: Dyplom-final Modification

22

( √

) (50)

The sum of the small-scale and large-scale contributions

(51)

which are defined in equations (18) and (30) above, is the turbulent viscosity used in the time-averaged Navier-Stokes equations (9)

The Model Constants

The summary of the model constants for the k-kl-w transition are listed in table 1 below.

Table. 1 Summary of the new model constants.

Page 23: Dyplom-final Modification

23

2.3 Turbulence model (SST K-ω)

Transport equation for the Turbulence SST Model

The two transport equation is given by the following:

[( )

] (52)

[( )

] (53)

and the model is supplemented by the definition of turbulent viscosity:

(54)

The model constants are given in (Menter book) Reference 8.

Page 24: Dyplom-final Modification

24

3 Objectives of the thesis The overall objective of this research is to understand the transitional behavior for the flow over a NACA65 compressor blade cross-section. As predicting the fluid flow behavior from laminar to transition stages is important in many technical applications including turbines, compressors, airfoils (technique which is based on laminar boundary layer) .

To validate the flow model developed by Walters and comparing the numerical results obtained to the DNS results obtained by Zaki (Zaki T.A,

Wissink J.G, Rodi W and Durbin P.A., Direct simulations of transition in a

compressor cascade: the influence of free-stream turbulence). In particular, the computational studies was done for transitional and

turbulent flow on NACA65 airfoil using Fluent software. The turbulent flow use

the K-w STT- two equation model and the transitional flow predictions use the

K-Kl-w three equation model.

4 Computational Details

For computing transition, the approach used in this work involves: 1) obtaining a fully turbulent solution on the whole computational domain, and 2) restarting the fully turbulent solution with the transition model activated.

4.1 Computational Mesh

The mesh generation software, Gambit has been used to generate meshes.

Both structured and unstructured meshes are used throughout this research so

that accurate CFD calculations can be performed over the airfoil sections which

have reasonably complex geometry. Gambit is a modeling software that is

capable of creating meshed geometries that can be read into FLUENT and other

analysis software. Gambit is essential in the process of doing the CFD analysis:

it creates the working environment where the object is simulated. An

important part is creating the mesh surrounding the airfoil, this needs to be

done in all directions on the boundary wall to have a good resolved flow

around the airfoil after calculations.

Page 25: Dyplom-final Modification

25

The mesh and edges must also be grouped in order to set the necessary

boundary conditions effectively. An outline for the Gambit geometry creation is

shown in figure 2.

Fig. 2, Gambit diagram work process

Create Coordinates

Lunch Gambit

Import Coordinates

Define Mesh Boundaries

Create Mesh

Define Boundaries

Export mesh

Page 26: Dyplom-final Modification

26

Boundary layer mesh specification on the airfoil

Table 2, shows the parameters specified to obtain the grid on the airfoil

boundary layer,it was used Gambit for these specifications.

Table 2, airfoil boundary layer mesh specification

Final Mesh

Unstructured grid was applied in some parts and block-structured grid around

the blade and inside the channel see figure 3. Detailed views at the leading and

trailing edge are shown in figure 4.

Fig. 3, Final resultant mesh of the geometry.

Algorithm First row

(a)

Growth

Factor (b/a)

Rows Transition

pattern

Uniform 1.2 24 1.1

Page 27: Dyplom-final Modification

27

Fig. 4, Detail of the grid topology in the leading and trailing edge region.

Page 28: Dyplom-final Modification

28

Refined mesh

Refining the mesh was necessary so as to make sure that the solution is grid

independent. Density of the mesh is higher where the velocity gradients are

higher, that means in the proximities of the leading and trailing edges.

Transitional calculations were done using the refined grid of 350,000 cells.

Detail structure of the refined grid at the leading edge is presented in figure 5.

Structure of the refined grid

Fig. 5, Detail view of refined grid for transitional flow at the leading edge.

Page 29: Dyplom-final Modification

29

4.2 Boundary Conditions

A schematic of the computational domain with the boundary conditions is

shown in Figure 6. The blade geometry is designated V103 as used in the

experiments of Hilgenfeld and Pfitzner 2004. The separation of the top and

bottom computational boundaries is one blade pitch P = 0.59L. The streamwise

extent of the domain is 1.9L. The nean Velocity of the object relative to the

fluid ( ) based on the the axial chord L and the simulation Reynolds number,

Re = 138,500, was calculated given by:

(55)

where : Mean Velocity of the object relative to the fluid (m/s).

L – is the characteristic length, L(m), (axial cord length in this case is used).

is the fluid dynamic viscosity kg/(m·s).

is the fluid density (kg/m³).

At the inflow plane, x/L=−0.4, a mean velocity (Uo cos(α);Uo sin(α); 0) is

prescribed, where the flow angle α = 42o.

Note: The X- and Y-component of flow direction are set because of the

angle of attack: ( ) and ( ) .

Pressure outlet is applied at the flow outlet, this allows the static pressure

to be directly specified and is assumed to be equal to the freestream pressure.

Upper profile and lower profile are set to wall type conditions, and periodic

boundary conditions are applied in the y-direction. Table 4 summarizes the

boundary conditions and the numerical method implemented for all the

simulations both turbulent and transitional flow.

Page 30: Dyplom-final Modification

30

Fig. 6, Cross section through the computational domain at midspan.

A summary of the boundary conditions and the numerical method implemented for all the simulations both turbulent and transitional flow.

Model Steady, two dimensional, incompressible

Viscous model k-w and k-kl-w

Equations Reynolds-average navier-Stokes equations

Density

Viscosity

Turbulent case Transitional case

Pressure Standard Standard

Pressure- velocity coupling Coupled coupled

Momentum First order upwind Second order upwind

Turbulent kinetic energy First order upwind Second order upwind

Pressure Outlet

Velocity Inlet

Materials

Discretization

Page 31: Dyplom-final Modification

31

Laminar kinetic energy First order upwind Second order upwind

Specific dissipation rate First order upwind Second order upwind

Inlet

Velocity components, (method-

magnitude and direction)

The velocity components are established via the inlet

flow angle α =42o, with a velocity magnitude 10m/s

Turbulent case Transitional case

Turbulence intensities

Turbulent length scale Lt=0.008m

kinetic energy

[

]constant

Outlet

Outlet pressure(methods-from

neighbouring cell)

Backflow turbulent intensity

Backflow turbulent length scales Lt=0.008m

Stationary wall

Shear condition No slip

Periodic

Periodic type Translational

Table 3, Numerical method specifications.

Boundary Conditions

The turbulent intensities provided for test case are Tu = {3.25,6.50,10}%

static (gauge) pressure. The value of the specified

static pressure is used only while the flow is subsonic.

if the flow becomes supersonic the specified pressure

will no longer be used.

test case Tu = {3.25,6.50,10}%

Page 32: Dyplom-final Modification

32

4.3 Convergence Criteria

After inputting the mesh and the conditions of the flow, FLUENT iterated until

the solution converged. The scaled residuals show how close the flow is to

converging, the lower the residuals, the closer it is to convergence.

The governing equations were solved sequentially with the pressure-

correction SIMPLE method and momentum interpolation was used for the

pressure-velocity coupling. For the momentum and transport equations, the

scaled residual at each iteration step was obtained as a sum of residuals in all

cells of the computational domain divided by the left hand side of the

discretized equation summed over all cells. For the pressure-correction, the

non-scaled residual was first computed as the sum of the mass imbalance in all

cells. Then, scaling was performed by dividing the non-scaled residual by the

maximum continuity equation residual (normalization factor) reported during

the first five iterations. The multigrid method was used to solve the system of

equations at each iteration step while the Gauss-Seidel method was applied to

eliminate the oscillatory modes of the error at each multigrid level.

Since a numerical approach can only approximate, but never give an exact

solution, a limit for convergence has to be provided. Figure.7, shows the

convergence history for the turbulent flow, and Figure.8, shows the

convergence history for the transitional flow.

The residual criterion was set to 10-6 resulting in 2000 iterations for turbulent

case and 3000 iterations for the transitional case.

Page 33: Dyplom-final Modification

33

Convergence quality

Fig. 7, Convergence history for Turbulent K-omega model.

Fig. 8, Convergence history for transitional flow, K-Kl-w model.

Page 34: Dyplom-final Modification

34

5 Results and Discussions

The meshed geometry was exported from Gambit and was read into the Fluent

solver software. Calculations and observations was made. Computations was

done both on Coarser and Fine Grid, The mesh density is higher where the

velocity gradients are higher, that means in the proximities of the leading and

trailing edges. first the K-omega two-equation (Shear Stress Transport (SST) k-

ω Based Model) was calculated for turbulent prediction and then was applied

three equation turbulence Model K-Kl-omega.

5.1 Verification of the inlet conditions

A test case was done to be sure that the conditions for the incoming flows

are correct, to do this I had to compute for the corresponding turbulent length

scales (Lt) that would correspond to the turbulent intensities (Tu) provided in

the DNS results in Zaki paper.

Figures. 9 and 10 shows the turbulent intensity plots for the turbulent

length scales that were tested. The cases were, Lt = { 0.012, 0.024, 0.006,

0.008}, the results are divided in to two parts :

First part is figure 9, which shows the results for the different

turbulent length scales, Lt = { 0.012, 0.024, 0.006} tested to give the

turbulent intensity (Tu) that would correspond to the turbulent

intensity (Tu) provided in the DNS in Zaki paper. From the results

obtained, I therefore say that the turbulent intensities are increasing

monotonically as the length scales are increased.

Second part is figure 10, which shows the plot of turbulent intensity

for test case Lt= 0.008 compared with the turbulent intensity closer at

the edge of the boundary layers as provided in the DNS in Zaki paper,

and it was the nearest value that matched the DNS result.

Page 35: Dyplom-final Modification

35

From the results, I conclude that the comparable and nearest turbulent

length scale that produced the turbulent intensity provided in the DNS Zaki

paper is the length scale Lt = 0.008. This value was used throughout the rest of

the calculations.

Fig. 9, Calculated turbulent intensities (Tu) compared with the DNS for test

cases Lt = ( 0.012, 0.024, 0.006). Turbulent intensities are increasing

monotonically as the length scales are increased.

0

1

2

3

4

5

6

0 0.2 0.4 0.6 0.8 1 1.2

Turb

ule

nce

inte

nsi

ty, T

u(%

)

Downstream distance, x/L

Turbulence intensity (Tu) for different Turbulent length Scales(Lt)

DNS

TestCase_Lt_0.012

TestCase_Lt_0.024

TestCase_Lt_0.006

Page 36: Dyplom-final Modification

36

Fig. 10, Calculated turbulent intensity (Tu) compared with the DNS for test case

Lt=0.008.The value Lt=0.008 was used for rest of the simulations.

Wall Yplus

In terms of boundary layer theory, y+ is simply a local thickness Reynolds

number. In terms of CFD y+ is a nondimensional distance from the wall to the

first grid point. In a practical sense, we don't resolve the solutions of turbulent

flow by direct numercial simulation. This would require a very fine mesh near

the wall in order to resolve the turbulent eddies in the boundary layer. Also,

turbulence is time varying and random, so CFD models would need to be run as

transient, even if the mean flow is steady state. Modern computers can't

handle this except for the simplest flow, maybe in the future it will be possible

to solve the turbulent Navier Stokes equations by direct numercial simulation

right down to the wall.

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0 0.2 0.4 0.6 0.8 1 1.2

Turb

ule

nce

inte

nsi

ty, T

u(%

) Turbulence intensity (Tu) for different

Turbulent length Scales(Lt)

DNS

TestCase_Lt_0.008

Downstream distance,

Page 37: Dyplom-final Modification

37

Turbulence models deal with the flow in the boundary layer. The boundary

layer is divided into an inner and outer region, and the inner region can be

further subdivided into a laminar (viscous) sublayer and a fully turbulent

region.

Reynolds number, surface roughness, adverse pressure gradient, etc will

change the value of y+ where the velocity profile swicthes from the viscous

sublayer to the log law of the wall. Resolving the viscous sublayer or laminar

boundary layer (say small y+ < 1) is crucial to get a correct shear stress in the

near wall region. The Upper limits on Y+, however depends on the Reynolds

number and becomes larger as Re increases.

The values of y+ are dependent on these two parameters: the resolution

of the mesh and the Reynolds number of the flow, and are defined only in wall-

adjacent cells. The maximal which cannot be exceeded is .

The value of y+ in the wall-adjacent cells dictates how wall shear stress is

calculated.

The equation for y+ is

√ (56)

Where y is the distance from the wall to the cell center, μ is the molecular viscosity, is the density of the air, and is the wall shear stress. Figure 11. indicates range between 0 .005 and 0.07 for the transitional computation. Therefore, I can conclude that the near-wall mesh resolution for this project is very good.

Page 38: Dyplom-final Modification

38

Fig. 11. Wall Y+ range from 0.005 – 0.07 at the lower and upper surface.

5.2 Analysis of the mean velocity and pressure fields

The contour plots of the mean velocity magnitude round the airfoil for the

turbulent and transitional case is shown in figures 12 and 13. The Flow

accelerates or increases significantly on the suction side up to x/C= 0.8 due to

displacement effect of the airfoil and it decelerates starting from x/C= 0.5

towards the trailing edge. It means that the boundary layer becomes prone to

separation on the suction side (adverse pressure gradient).

Flow decelerates gradually along the pressure side up to 80% of the

blade cord. Then slightly accelerates from x/C= 0.6. However there is no big

difference between the turbulent and transitional flow simulation in the bulk of

the flow. Some differences are visible in the separated shear layer close to the

trailing edge.

The velocity is much higher at the upper surface close to the leading

edge of the blade than that at the lower surface. This indicates that there is

lower pressure at the upper surface than the pressure at the lower surface as

shown in the contour plots of the static pressures , figures 19 and 20.

0

0.01

0.02

0.03

0.04

0.05

0.06

0 0.2 0.4 0.6 0.8 1 1.2

Wal

l Yp

lus

Downstream distance x/L

Wall Yplus for transitional case

Upperprofile

Lowerprofile

Page 39: Dyplom-final Modification

39

Fig.12, Mean velocity distribution and contour of velocity magnitude around

airfoil for turbulent case.

Fig. 13, Mean velocity distribution and contour of velocity magnitude around

airfoil for the transitional case.

Page 40: Dyplom-final Modification

40

Figures, 14 and 15 presents detailed views of the velocity vectors at the leading and trailing edge respectively. Looking closer to Fig. 14 at the area around the airfoil, the velocity approaches zero at the wall of the airfoil due to viscous effects.

Fig.14, Velocity vector detail at the leading edge region for turbulent flow.

Attached flow at the wall blade upper and lower surface.

Page 41: Dyplom-final Modification

41

Fig. 15, Velocity vector detail at the trailing edge region for turbulent flow.

The flow on the upper surface decelerates and converges with the flow on the lower surface.

Detail views for the velocity fields for transitional flow are presented below,

figure 16 presents flow details at the trialing edge, at mid-section figure 17 and

at the trailing edge figure 18.

Page 42: Dyplom-final Modification

42

Fig.16, Detail of velocity vectors at the leading edge region for transitional flow. The boundary layer is strongly accelerating close to the leading edge.

Fig.17, Detail of velocity vectors at mid-section of the suction and pressure surface, Transitional case. The velocity approaches zero at the wall of the airfoil due to viscous effects. The suction side prone to mild separation downstream of the blade.

Page 43: Dyplom-final Modification

43

Fig.18, Detail of the velocity vectors at the trailing edge region for transitional

flow. There is low velocity at the trailing edge and at the wake region. The flow on the upper surface decelerates and converges with the flow on the lower surface.

It is seen that the pressure distribution is negative over most of the surface,

except over small regions near the nose and the tail. The pressure distribution

at the upper surface is negative, see figures 19 and 20 as the velocity on the

upper surface is higher than the reference velocity. Whenever there is high

velocity vectors, we have low pressures and vise versa. The phenomenon

complies with the Bernoulli equation.

The static pressure distribution fig. 19 and 20, shows that there is low pressure

shifted towards the leading edge of the suction surface of the blade from x/C

=0.8 (dark blue zone) and it is reaching a high pressure value at the stagnation

region.

Page 44: Dyplom-final Modification

44

At the pressure surface the pressure increases gradually from x/C=0.9 (light

blue zone) towards the trailing edge and finally to the outlet flow plane (light

yellow zone).

From the inlet plane inside the domain, the blue area is approaching to the steady state and the yellow area is approaching to the unsteady state.

The pressure at the trailing edge is related to the airfoil thickness and

shape near the trailing edge. For thick airfoils the pressure here is slightly

positive (the velocity is a bit less than the freestream velocity). For infinitely

thin sections Cp = 0 at the trailing edge. Large positive values of Cp at the trailing

edge imply more severe adverse pressure gradients.

Fig. 19, Contours of static pressure for turbulent flow. The pressure increases

from its minimum value to the value at the trailing edge at the lower surface.

There is no big difference between the turbulent and transitional flow

simulation in the bulk of the flow. Some differences are visible in the separated

shear layer close to the trailing edge.

Page 45: Dyplom-final Modification

45

Fig. 20, Contours of static pressure for transitional flow. Pressure over most of

the upper surface are smaller than those over the bottom surface.

5.3 Contour plots of the turbulent to molecular viscosity ratio

The turbulent viscosity ratio

, is the ratio between the turbulent

viscosity, , and the molecular dynamic viscosity, . The ratio directly says something about how strong the influence of the turbulent viscosity is compared to the molecular viscosity.

Damping of the turbulent viscosity in the laminar zone in the simulations with the transition model by Walters is necessary for the expression Klebanoff modes development. The Klebanoff modes are streamwise-oriented fluctuations which are formed as a result of selective filtering of the freestream fluctuations. If the turbulent viscosity will be kept large in laminar region there will not be a possibility to model the pretransitional boundary layer. The streamwise-oriented structures will be damped by the turbulent fluctuations.

Page 46: Dyplom-final Modification

46

Figures 21 and 22 depicts the contours of turbulent to molecular viscosity ratio for the turbulent and transitional case respectively. Turbulent flow is still laminar, when transitional case at about x/C= 0.8 starts the turbulent flow.

Closer look at the near wall region on suction side (top blade) of the

profile fig (21), the boundary layer is still laminar (dark blue zone). In the

transitional case at about x/C= 0.8, the boundary layer is already turbulent at

this area. In the turbulent simulation the ratio of turbulent to molecular

viscosity is zero (dark blue colour), when the ratio of turbulent to molecular

viscosity for the transitional simulation at x/C=0.8 is 0.96 (thick yellow colour)

Additionally turbulent viscosity is also rather low in the turbulent zone (rear

part of the blade). This is not fully correct. It means that there is delayed onset

to transition, due to long boundary layer laminar flow prediction.

Page 47: Dyplom-final Modification

47

Fig. 21, Contours of Turbulent to molecular viscosity ratio for turbulent flow. Zoomed at x/C=0.8.

Page 48: Dyplom-final Modification

48

Fig. 22, Contours of Turbulent to molecular viscosity ratio for transitional flow. Zoomed at distance x/C= 0.8.

Page 49: Dyplom-final Modification

49

5.4 Transition Onset The ratio of molecular diffusion to rapid pressure strain time scales as described in (Walters bulletin 80 paper) is the important local dimensionless quantity governing shear sheltering and transition initiation. When this ratio is small, pressure strain is suppressed and one-component fluctuations (i.e. laminar kinetic energy) are generated. When the ratio reaches a critical value, the rapid pressure strain term quickly increases in magnitude to generate three-dimensional turbulent fluctuations, and transition begins. See chapter 2.2 above (transitional model by Walters) for full description of this process.

Figure. 23, shows the contour plot of the term as defined in equation (46) above, is a function of the time scale ratio expressed in Eq. (43) with transition initiating at a critical value .

Normal lines are created on the wall surfaces at distances x/L = {0.2, 0.3, and 0.4} respectively to verify the transition onset see figures 23 and magnified view of the contour plot of at the pressure surface at distance x/L = 0.2, the pressure is continuously increasing up to x/C=0.9. This summarizes the fact that transition onset is predicted too late. The too late transition is due to missing model sensitivity to the strong adverse pressure gradient.

In the DNS (Zaki paper), the transition to turbulent started at location x/L 0.2 on the pressure surface at turbulent intensity Tu=3.25% when in computation is at x/L= 0.25, see the skin friction plot figure 29.

According to Mayle the adverse pressure gradient shifts the transition onset

more upstream. This phenomenon is not well reproduced by the Walters

model.

Page 50: Dyplom-final Modification

50

Fig. 23, Contour plot of for transitional flow. Normal distances

created at x/L= 0.2, 0.3, and 0.4. Shows the transition onset at x/C=0.8.

X/L =0.2

Page 51: Dyplom-final Modification

51

5.5 Pressure coefficient This section shows the variations in the pressure coefficient around the airfoil, for turbulent flow, transitional flow and experimental data.

The pressure coefficient is prerequisite for reliable simulation of the flow

over the airfoils/blades. It is a dimensionless parameter which describes the relative pressures throughout a flow field in fluid dynamics defined by the equation

( )

(57)

Where is the static pressure, is the reference pressure and

is the reference dynamic pressure given by

(58)

Where is the reference velocity of the fluid, is the density of air.

The reference pressure is simply a device used to avoid the

numerical round-off errors in the numerical calculations involving pressure.

This can occur when the dynamic pressure change in a fluid, which is what

drives the flow, are small compared to the absolute pressure level.

The reference pressure was computed by subtracting off a suitable

constant reference pressure, to obtain the working pressure that is less prone

to round-off errors.

Generally is plotted against x/C. The blade chord length x/C varies

from 0 at the leading edge to 1.0 at the trailing edge. is plotted "upside-

down" with negative values (suction), higher on the plots presented.

The shape of the pressure distribution is directly related to the airfoil

performance. The features that affects airfoil performance are:

Favourable pressure gradient that leads to laminar flow close the leading

edge of the blade.

Page 52: Dyplom-final Modification

52

Minimum Cp value which determines the maximum speed at upstream of

the blade that can be useful to indicate shock formation and also

determine the extent of pressure recovery.

Adverse pressure gradients that leads to transition and separation. It is

especially important to avoid strong adverse gradients as we approach

the leading edge.

From the computation the flow behaviour along the blade shows an

initial acceleration on the suction surface up to 20% of the chord from the

leading edge. Thereafter a constant deceleration of the flow towards the

trailing edge is noticeable an on the pressure surface the pressure distribution

remains almost constant along the blade for case Tu=3.25% and 6.50% in both

transitional and turbulent case, see figures 24 and 25, below.

The flow agreement are not perfect at the if comparing with the DNS,

NOTE

Figures 24-27, the top curves corresponds to the pressure surface and the

lower curve corresponds to the suction surface. At the suction surfaces some

parts has negative values at the leading edge as the pressure here is lower than

the reference pressure.

Page 53: Dyplom-final Modification

53

Fig. 24, pressure distribution along the blade surface for turbulent

case compared with the transitional case, for case Tu = 6.50%.

Fig. 25, Pressure distribution along the blade surface for the transition

case. Comparison between the increasing turbulent intensities, Tu 3.25%

and 6.50% which was calculated.

-0.4

-0.2

0

0.2

0.4

0.6

0.8

0 0.2 0.4 0.6 0.8 1 1.2

pre

ssu

re c

oef

fici

ent 𝐶𝑝

Distance ×/L

Pressure Coefficient Vs. Dwonstream distance for transitional flow

CFD_Transitionalcase_Tu_6.50%

CFD_Turbulentcase_Tu=6.50%

-0.4

-0.2

0

0.2

0.4

0.6

0.8

0 0.2 0.4 0.6 0.8 1 1.2pre

ssu

re c

oef

fici

ent 𝐶𝑝

Distance ×/L

Pressure Coefficient Vs. Dwonstream distance for transitional flow

CFD_Tu_3.25%

CFD_Tu=6.50%

Page 54: Dyplom-final Modification

54

Fig. 26, pressure distribution along the blade surface for transitional

flow compared with the DNS result by Zaki paper, for the case Tu =

3.25%.

Figure. 27, pressure distribution along the blade surface for transitional flow compared with the DNS for case Tu = 6.50%.

-0.4

-0.2

0

0.2

0.4

0.6

0.8

0 0.2 0.4 0.6 0.8 1 1.2

pre

ssu

re c

oef

fici

ent 𝐶𝑝

Distance ×/L

Pressure Coefficient Vs. Dwonstream distance for transitional flow

DNS_Tu_3.25%

CFD_Tu_3.25%

-0.4

-0.2

0

0.2

0.4

0.6

0.8

0 0.2 0.4 0.6 0.8 1 1.2

pre

ssu

re c

oef

fici

ent 𝐶𝑝

Distance ×/L

Pressure Coefficient Vs. Dwonstream distance for transitional flow

DNS_Tu=6.50%

CFD_Tu=6.50%

Page 55: Dyplom-final Modification

55

5.6 The skin friction coefficient

The skin friction coefficient is defined by

(59)

Where is the local wall shear stress , is the fluid density and is the free-stream velocity (usually taken outside of the boundary layer or at the inlet).

Computations shows that the Walters model is not performing well. Zaki

paper shows that flow with the presence of free-stream perturbations, even at

the lowest intensity , the boundary layer transitions to turbulence.

Transition onset, defined as the point of minimum skin friction, takes place at

x/L=0.2 figure 29, which is upstream of the laminar separation point. The

transitional boundary layer circumvents separation, and transition is complete

at x/L=0.46 for the DNS. There is underprediction for the computational case.

Transition onset is located at x/L=0.24 and transition is completed at x/L= 0.48,

this underprediction compared to the DNS, see the skin friction coefficient

figure 29 for Tu=3.25%.

Figure 28, shows the skin friction for turbulent case along the pressure surface compared with transitional case and DNS, for Tu=6.50%, it shows that the turbulent flow simulation are not able to reproduce the laminar boundary layer. The flow is fully turbulent already at the leading edge.

Comparing the skin friction coefficient Cf figure (28) for the transitional case to that of turbulent case, the turbulent part is predicted but deviated upstream as from the distances x/L=0.04 to x/L=0.9 indicating the transitional flow. There underprediction compared with DNS data.

Figure 29, shows the skin friction along the pressure surface for the turbulent intensities Tu = {3.25, 6.50, 10}%, for transition case (CFD) compared with DNS data. The Transitional location is observed to move upstream monotonically as the turbulent intensity is increased. The earlier transition onset with increasing Tu is physical, this gives that the numerical results are therefore qualitatively correct.

Page 56: Dyplom-final Modification

56

Fig. 28, Skin Friction along the pressure surface for turbulent case compared

with transitional case and DNS data, for the case Tu=3.25%.

Fig. 29, Skin Friction Cf , along the pressure surface for transitional case

compared with the DNS data . For cases, Tu – 3.25%, 6.50% and 10%.

0

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0 0.2 0.4 0.6 0.8 1 1.2

DNS

Transitionalcase

Skin

Fri

ctio

n

Pressure surface skin friction

Downstream distance ×/L

0

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0 0.2 0.4 0.6 0.8 1 1.2

Skin

Fri

ctio

n 𝐶𝑓

Downstream distance ×/L

Pressure surface skin friction coefficient

DNS_Tu=10%

DNS_Tu=6.50%

DNS_Tu=3.25%

CFD_Tu=10%

CFD_Tu=6.50%

CFD_Tu=3.25%

Page 57: Dyplom-final Modification

57

6 Conclusions

Navier–Stokes equation solver was coupled with a transition model and a Reynolds-averaged three equation model, with the Reynolds number 138,500, at flow angle 42o and in velocity 10m/s with various levels of free- stream turbulent intensities, to understand the flow behavior around a NACA65 compressor blade cross-section. The following conclusions are made based on the observations and results analysis.

The transition onset is predicted too late. The too late transition is due to missing model sensitivity to the strong adverse pressure gradient. According to Mayle the adverse pressure gradient phenomenon that shifts the transition onset more upstream is not well reproduced by the Walters model.

The Transitional location is observed to move upstream monotonically as

the turbulent intensity is increased. The earlier transition onset with

increasing Tu is physical, this shows that the numerical results are therefore

qualitatively correct.

Contours of static pressure, pressure coefficient on the entire blade

surface, were compared for both the turbulent and transitional flow.

Because the top surface of the blade is curved, the air travels a greater

distance at a faster rate over the top of the blade than over the bottom of

the blade. This causes a difference in pressure between the top and bottom

of the blade. There is higher pressure on the bottom of the blade than on

the top. At the leading edge the flow is laminar, and at the trailing edge, the

flow is turbulent.

Page 58: Dyplom-final Modification

58

7 REFERENCES

1 Zaki T.A, Wissink J.G, Rodi W and Durbin P.A., Direct simulations of

transition in a compressor cascade: the influence of free-stream

tuebulence.

2 Mayle, R.E., 1991, “ The Role of Laminar-Turbulent Transition in Gas

Turbine Engines,” ASME Journal of Turbomachinery, Vol. 113, pp. 509-

537.

3 Wilcox, D.C., (1994), Turbulence modelling for CFD, DCW Industries, la

Canada, Ca.

4 Walters, D.K and Cokljat D., 2008, “ A Three –Equation Eddy –Viscosity

Model for Reynolds- Averaged Navier-Stokes Simulations of transitional

Flow.

5 Menter, F. R., "Two-Equation Eddy-Viscosity Turbulence Models for

Engineering Applications," AIAA Journal, Vol. 32, No. 8, August 1994, pp.

1598-1605.

6 ECROFTAC Bulletin 80, September 2009, ‘Transition Modelling’.

7 "ANSYS FLUENT 12.1 and FLUENT 6.3.26 ,Flow Modeling Software."

8 Çengel, Yunus A. Fluid mechanics : fundamentals and applications /

Yunus A. Çengel, John M. Cimbala.McGraw-Hill Series in Mechanical

Engineering : McGraw-Hill/Higher Education,cop. 2006.