You are on page 1of 138

Maria Teodora Pascu

Modern Layout and


Design Strategy for Axial Fans

Moderne Auslegungs- und Entwurfsstrategie für Axialventilatoren

Back cover Front cover


Modern Layout and Design Strategy for Axial Fans

Moderne Auslegungs- und Entwurfsstrategie für


Axialventilatoren

Der Technischen Fakultät der Friedrich–Alexander–Universität Erlangen–Nürnberg

zur Erlangung des Grades

DOKTOR–INGENIEUR

vorgelegt von

Maria Teodora Pascu

Erlangen, 2009
Als Disseration genehmigt von der Technischen Fakultät
der Universität Erlangen-Nürnberg

Tag der Einreichung: 11.12.2008


Tag der Promotion: 24.04.2009

Dekan: Prof. Dr. J. Huber


Berichterstatter: Prof. Dr. Dr. h.c. F. Durst
Prof. Dr. M. Wensing

ii
Acknowledgements

This work received financial support from the Bavarian Science Foundation in the form

of an individual grant, which is gratefully acknowledged.

I would like especially to thank my supervisor, Prof. Dr. Dr. h.c. F. Durst, who from the

first day we met, during the Summer Academy Kuşadasi 2004, captured my attention for

fluid mechanics research, and by supporting my diploma thesis at the Institute of Fluid

Mechanics LSTM Erlangen, brought me in close contact with the topic and stirred my

interest in further academic education. He has enabled me to work in the field of

turbomachines and constantly supported the present work.

I would also like to thank Prof. Dr. M. Wensing for his kind acceptance to review the

present work.

Furthermore, I would like to thank Prof. Dr. A. Delgado, head of the Institute of Fluid

Mechanics LSTM Erlangen, who supported and encouraged the present work, always

taking a genuine interest in the outcome of the investigations.

My deepest acknowledgements go to Dr. J. Jovanović, head of the turbulence research at

LSTM, for his sustained moral support, to whom I dedicate the successful turnout of the

experimental investigations included in the thesis.

I would also like to express my gratitude to Alu Automotive GmbH for the kind support

offered to the present work, and especially to the company manager, Mr. Felix Hellmuth.

I would especially like to thank Dr. Ph. Epple, head of the Turbomachinery Optimization

research group at LSTM Erlangen, for his constant supervision and strong commitment to

the present work.

iii
My warm acknowledgement also goes to Dipl.-Ing. M. Miclea-Bleiziffer for his support

and numerous brainstorming sessions that led to the present layout of the work.

I warmly acknowledge the technical department of the institute for their continual

collaboration, and especially to Mr. F. Kaschak and Mr. C. Bakeberg.

Finally, I would like to thank all my colleagues at LSTM for the wonderful and friendly

working atmosphere, to the workshop and administration, who all contributed to the

achievements in the present work.

Şi cel mai important, doresc să îi mulţumesc familiei mele, pentru sprijinul necondiţionat

si încrederea deplină pe care mi le-au acordat in permanenţă, si cărora le datorez în

întregime tot ceea ce sunt azi.

Erlangen, December 2008 Maria Teodora Pascu

iv
List of Contents

1
5H 6H Introduction and aim of work 11
7H

1.1
8H General introduction
9H 11
10H

1.2
1H Classification of turbomachines
12H 11
13H

1.3
14H Aim of work
15H 11
16H

2
17H 18H Basic equations of fluid mechanics as applied in turbomachines 11
19H

2.1
20H 21H Navier–Stokes equations in rotating systems 11
2H

2.2
23H Energy transfer in turbomachines
24H 11
25H

3
26H 27H Survey of the available design methods for axial impellers 11
28H

3.1
29H Basic features of turbomachinery design
30H 11
31H

3.2
32H Two – dimensional cascade theory
3H 11
34H

3.2.1
35H Aerodynamics forces and governing equations
36H 11
37H

3.3
38H Design methods based on the airfoil theory
39H 11
40H

3.3.1
41H Airfoil families. Mean-line and thickness distribution
42H 11
43H

3.3.2
4H Design parameters
45H 11
46H

3.3.3
47H Cascade losses. Diffusion factor
48H 11
49H

3.4
50H Three-dimensional character of the flow in axial turbomachines
51H 11
52H

4
53H 54H Proposed design strategy for axial fans 11
5H

4.1
56H Mean-line calculation
57H 11
58H

4.2
59H Outlet conditions
60H 11
61H

4.3
62H Meridional flow analysis for axial fans
63H 11
64H

4.4
65H The indirect design problem
6H 11
67H

4.5
68H Parameterization of the total pressure in the span-wise direction for an
69H

axial fan blade 11


70H

4.6
71H Blade shape computation
72H 11
73H

4.7
74H Further design assumptions based on profile analysis
75H 11
76H

4.7.1
7H Static-to-static cascade efficiency
78H 11
79H

4.7.2
80H Total-to-total cascade efficiency
81H 11
82H

4.7.3
83H Profiling the camber line
84H 11
85H

4.8
86H Design Solver (DS)
87H 11
8H

v
4.9
89H DS output
90H 11
91H

5
92H 93H Numerical flow analysis 11
94H

5.1
95H Mathematical model
96H 11
97H

5.2
98H Mesh generation
9H 11
10H

5.3
10H Numerical models and boundary conditions
102H 11
103H

5.4
104H Appropriate performance indicators
105H 11
106H

5.5
107H Optimum span-wise pressure distribution
108H 11
109H

5.6
10H Profile analysis
1H 11
12H

5.6.1
13H Flow domain around the profiles
14H 11
15H

5.6.2
16H Mesh generation
17H 11
18H

5.6.3
19H Numerical results
120H 11
12H

6
12H 123H Experimental validation of the proposed design strategy 11
124H

6.1
125H Investigated impellers
126H 11
127H

6.2
128H Experimental facility
129H 11
130H

6.3
13H Measured parameters
132H 11
13H

6.4
134H Measuring equipment
135H 11
136H

6.5
137H Experimental results
138H 11
139H

6.6
140H Validation of the results
14H 11
142H

7
143H 14H Integrated ideal efficiency for axial fans 11
145H

7.1
146H The Cordier diagram
147H 11
148H

7.2
149H Ideal efficiency for axial fans
150H 11
15H

8
152H 153H Conclusions and outlook 11
154H

vi
Index of Symbols

A [m2] area

M [N m] torque

P [W] power

Q [m3/s] flow rate

R [–] degree of reaction

b [m/s2] acceleration

c [m/s] absolute velocity

l [m] blade chord

n [rpm] rotational speed

r [m] radius

u [m/s] peripheral velocity

w [m/s] relative flow velocity

wm [m/s] meridional component of the relative velocity

wu [m/s] tangential (peripheral) component of the relative velocity

P [Pa] pressure difference

Greek symbols

 [-] diameter coefficient

 [o] gliding angle

 [o] blade angle

 [–] efficiency

 [–] flow coefficient (dimensionless flow)

 [–] pressure ratio between the fan outlet and the inlet static pressures

 [kg/m3] density

vii
 [o] camber turning angle

 [–] temperature ratio between the fan outlet and inlet temperatures

 [–] speed coefficient

 [rad/s] angular velocity

 [–] pressure coefficient (loading)

 [–] circulation

 [–] vorticity

Subscripts

1 blade inlet

2 blade outlet

h hub

t tip, total

s static

d dynamic

viii
Abstract

The present work addresses the application of computational fluid dynamics in the
research and development process of axial fans of the kind used in numerous fields of
engineering and in daily life. In this sense, a modern layout and design strategy for axial
impellers are proposed, as basis for optimization in this engineering field.
Essentially, the strategy is a combined inverse-direct method, based on a design solver
which computes the optimum blade profile according to the flow conditions in the fan,
and does not make use of any predefined profiles.
When applied in a rigorous manner, the proposed design strategy delivers high-
performance design solutions for axial fans, and this is thoroughly confirmed by both
numerical and experimental results.
The design calculation scheme starts with the one – dimensional hypothesis of the mean
streamline, based on which the blade inlet (at all sections) and outlet (at the hub)
conditions are determined. Then, by computing the blade as a succession of several
cascades, the two-dimensional nature of the flow is considered. Finally, the blade profile
is fully resolved by implementing a three-dimensional (meridional) analysis into the
design process. By assuming an arbitrary vortex flow, the optimum pressure distribution
in the span-wise direction is determined and the parameterization of the outlet blade
angle is achieved, as a function of one of the most important constructive characteristics
of an axial fan, i.e. the hub ratio.
The advantages of employing the suggested design strategy as an optimization tool are
first emphasized by fully converged CFD solutions, which show the substantial
improvements in efficiency achieved by the new designs over the reference model, i.e. an
engine cooling fan currently used in the automotive industry. Moreover, the employment
of the non-free vortex assumption at the design stage is proved to be beneficial for the fan
performance, since the design obtained accordingly performs efficiently through a wider
flow range.
Even though modern CFD nowadays achieves excellent flow predictions, phenomena
with impact on the performance are neglected, hence the motivation for the experimental
confirmation of the proposed designs. The performance curves of the non-free vortex

ix
flow design against the reference impeller show an absolute increase in the measured
(total-to-static) efficiency of 10% for the proposed design.
Finally, the present work proposes an analytical computation of the integrated ideal
efficiency for axial fans, a concept which was derived as a response to the incapacity of
the classical Cordier diagram to predict the actual performance of axial impellers
operating in the low-pressure regimes, due to proven inconsistencies for this type of
turbomachine with regard to the definitions of the parameters employed. It is shown that
the proposed design strategy delivers an axial fan whose performance comes very close
to that of the ideal machine.

x
Zusammenfassung
Die vorliegende Arbeit beschäftigt sich mit der Anwendung von numerischen
Strömungssimulationen in der Forschungs- und Entwicklungsarbeit axialer Gebläse, die
im Ingenieursbereich und täglichen Alltag vielfältig zu finden sind. In diesem Sinne wird
für diesen Bereich des Ingenieurwesens eine moderne Auslegungs- und Designstrategie
für axiale Gebläse vorgeschlagen.
Die entwickelte Auslegungsstrategie ist hauptsächlich eine kombinierte inverse–direkte
Methode, die sich auf einen Design-Löser stützt, der in Abhängigkeit der
Strömungsbedingungen im Laufrad, ein optimales Schaufelprofil berechnet, ohne andere
vorgegebene Profile zu nutzen.
Die vorgeschlagene Auslegungsstrategie liefert, wenn sie richtig angewendet wird,
Hochleistungslösungen für das Design axialer Gebläse, die von numerischen und
experimentellen Ergebnissen bestätigt wurden.
Das Auslegungsschema beginnt mit der ein-dimensionalen Hypothese der
Hauptstromlinie, auf der die Bedingungen am Schaufeleintritt (auf allen Querschnitten)
und Schaufelaustritt (an der Nabe) bestimmt werden. Die Zwei-dimensionalität der
Strömung wird mithilfe der Berechnung zwei-dimensionaler Kaskaden berücksichtigt.
Das Schaufelprofil wird im Designverlauf über die Einführung einer drei-dimensionalen
(Merdidionalen) Analyse vollständig berechnet. Nach der Einstellung einer freien
Wirbelströmung werden in der Spannweitenrichtung die optimale Druckverteilung und
die Parametrierung des Austrittwinkels als Funktion der wichtigsten Eigenschaft eines
Axialgebläses - dem Nabe-Gehäuse-Verhältnis - berechnet.
Der Vorteil der vorgeschlagenen Auslegungsstrategie als Optimierungswerkzeug wird
erst durch vollständig konvergierte CFD-Ergebnisse unterstützt. Die neuen Gebläse, zum
Beispiel Lüftungsgebläse der Automobilindustrie, zeigen im Vergleich zum
Referenzmodell erhebliche Verbesserungen des Wirkungsgrads. Außerdem erweist sich
die Anwendung der „Nicht-freien Wirbel“ Annahmen vorteilhaft für das
Leistungsverhalten des Gebläses, da das neue Design über einem weiten Bereich der
Durchfluss-Kennlinie effizienter arbeitet.
Obwohl moderne CFD heutzutage ausgezeichnete Strömungsvorhersagen erreicht und
somit erlauben, aus einer Serie von Auslegungen die beste auszuwählen, wurde die beste

xi
Auslegung am Prüfstand verifiziert. Damit konnte das Auslegungsverfahren
experimentell verifiziert werden.
Nach der Auslegung und Nachrechnung mit CFD, wurde für das beste Laufrad ein
Prototyp gebaut und am Prüfstand vermessen und somit die Auslegung verifiziert.
Ein Vergleich der Kennlinien der mithilfe der „Nicht-freien Wirbel“-Auslegungsmethode
ausgelegten Gebläse mit denen von Referenzgebläsen zeigen in der gemessenen Effizienz
(total -zu- statisch) eine 10 %-ige absolute Erhöhung.
Die vorliegende Arbeit schlägt abschließend eine analytische Berechnung der
integrierten-idealen Effizienz von Axialgebläsen vor. Dieses Konzept wurde als eine
Ergänzung zum klassischen Cordier-Diagramm abgeleitet, da dieses wegen schon
bekannter Widersprüche in den eingesetzten Parametern für diesen Typ von
Turbomaschinen nicht in der Lage ist, die Leistungen von axialen Gebläsen
vorherzusagen. Es wird gezeigt, dass die vorgeschlagene Auslegungsstrategie axiale
Gebläse liefert, deren Wirkungsgrad einer idealen Maschine sehr nahe kommen.

xii
1. Introduction and aim of work
0B

1 Introduction and aim of work


0B

1.1 General introduction


Fluid mechanics is a subject that has developed over centuries to become an independent
field of science, based on laws that are generally accepted, such as conservation of mass,
momentum, and energy, and the basic laws that describe the thermodynamic properties of
the fluids. These laws are best and most condensed written down in terms of tensor
notation:
 Conservation of mass (continuity equation) for Newtonian fluids
   U i 
 0 (1.1)
t xi

 Momentum equation
 U j U j  P  ij
  Ui    gj (1.2)
 t xi  x j xi

 Molecular transport of momentum


 U j U i  2 U k
 ij         ij  (1.3)
 xi x j xk
  3
 Thermal energy equation
 e e  qi U j U j
  Ui  P   ij (1.4)
 t xi  xi x j xi

e  CvT (1.5)

 Molecular transport of heat


T
qi   (1.6)
xi

1
1. Introduction and aim of work
0B

 Equations of state for gas flows


P   RT (1.7)
The above equations describe the ideal gas as a medium, and the different types of flows
are given only by different initial and boundary conditions. These equations are therefore
used nowadays to solve flow problems which occur in different areas of engineering,
science and medicine, and also in nature and daily life, Durst [31]:
15H

 Heat exchanger, cooling and drying technology


 Reaction technology and reactor layout
 Aerodynamics of vehicles and airplanes
 Semiconductor-crystal production, thin-film technology, vapor-phase deposition
processes
 Layout and optimization of pumps, valves and nozzles
 Development of measuring instruments and production of sensors
 Ventilation, heating and air-conditioning techniques, layout and tests, laboratory
vents
 Problem solutions for roof ventilation and flows around buildings
 Production of electronic components, micro-systems analysis engineering
 Layout of stirrer systems, propellers, and turbines
 Sub-domains of biomedicine and medical engineering
 Layout of combustion units
Among the numerous fields, the above equations are used to solve fluid flow problems in
the field of flow machinery in general and turbomachines, in particular. It is the latter
field to which the present thesis attempts to make a contribution.
The flow in turbomachines is turbulent, and there is a crucial difference when modeling
the physical phenomena between laminar and turbulent flow. For the latter, the
appearance of turbulence eddies occurs on a wide range of length scales and typical flow
domains in this case would require computing meshes of 109–1012 grid points, Tu et al.
[103]. With the present-day computing power, the computing requirements for a direct
156H

numerical solution (DNS) of the time-dependent Navier–Stokes equations of the fully

2
1. Introduction and aim of work
0B

turbulent flows at high Reynolds numbers are truly phenomenal. Engineers, however,
require computational procedures that can supply adequate information associated with
the turbulent processes, but wish to avoid the need to predict all the effects caused by
each individual eddy in the flow. This category of CFD users is almost always satisfied
with information about the time-averaged properties of the flow. Therefore, in order to
study the dynamics of turbulence in common engineering applications, Reynolds
turbulence decomposition, for the instantaneous velocity and pressure, and time
averaging of the Navier–Stokes equations are required, yielding in a system of equations
for the mean flow, Jovanović [55]: 157H

U j P   U j 
U i       ui u j    g j (1.8)
xi x j xi  xi 
In Eq.(1.8) the Reynolds equations for turbulent flows are written (for the incompressible
case), including the “Reynolds stresses”  ui u j , which are caused by the turbulent motion.

In this way, new unknowns are introduced, which are in fact correlations of velocity
fluctuations. This yields a system of differential equations which is unclosed, and in order
to close this system, the application of turbulence models is required. Here, the standard
two-equation k-  model should be mentioned, Launder and Spalding [64], which closes
158H

the above system of equations with two additional transport equations: one for the
turbulent kinetic energy k, and the other for the rate of dissipation of the turbulent energy
 . This model is probably the most widely used and validated turbulence model, and its
performance has been assessed against a considerable number of practical flows.
However, despite the many successful applications in handling industrial problems, the
standard k-  model demonstrates only moderate agreement when predicting unconfined
flows, hence numerous turbulence models have been developed recently in order to
obtain accurate solutions of the more complex flow situations, such as the flow in
turbomachines.
Solving such complex flow behaviors is addressed by most computational fluid dynamics
(CFD) commercial codes available today, and due to the proven robustness of the flow
solver for turbomachines, ANSYS CFX was intensively employed in the present work, and
excellent convergence, in terms of both computational time and resources, was obtained
by applying the Shear Stress Turbulence (SST) model, Menter [77]. A detailed 159H

3
1. Introduction and aim of work
0B

explanation of how this turbulence model works and why it is particularly applicable in
the case of turbomachines will be given in Chapter 5 of the thesis.

1.2 Classification of turbomachines


Essentially, the turbomachine is an energy conversion device, converting mechanical
energy to thermal/pressure energy or vice versa. The conversion is done through the
dynamic interaction between a continuously flowing fluid and a rotating machine
component, and both momentum and energy transfer are involved.
In turbomachines, the fluid flows freely between the inlet and outlet of the machine,
without any intermittency. All turbomachines have a freely and continuously rotation part
known as a runner, impeller or rotor, which allows uninterrupted flow through it.
Therefore, the energy transfer between the rotor and the fluid is continuous, as a result of
the rate of change in angular momentum, Aksel [3].
160H

Much has been written on classifying turbomachinery, Wright [112], and without
16H

repeating the entire classification procedure, available in most introductory chapters of


books addressing this field, the author feels that a short summary of the different types
will be helpful for a clear statement of the aim of the present work within the large class
of flow devices to which the terminology refers. In most of the available literature, a
major subdivision can be achieved based on the power criterion, identifying whether
power is added (power absorbing) or extracted from the fluid (power producing). In this
respect, pumps can be assigned to the category of the power-absorbing turbomachines,
and they include liquid pumps, fans, blowers and compressors. Turbines are power-
producing machines and they include windmills, water wheels, modern hydroelectric
turbines, the exhaust side of automotive engine turbochargers, and the power extraction
end of an aviation gas turbine engine. They also operate with various types of fluids,
including gases, liquids, and mixtures of the two.
Another possibility to classify the turbomachines is from the perspective of the fluid
medium handled, either compressible or incompressible, Peng [84]. 162H

The above subdivisions can be summarized as indicated in Figure 1-1:


163H

4
1. Introduction and aim of work
0B

Figure 1-1 Classification of turbomachines, adapted from Aksel [3]


164H

According to the nature of the flow path through the passages of the rotor, the device is
termed an axial-flow turbomachine when the path of the through-flow is wholly or
mainly parallel to the axis of rotation, radial when the path is in a plane perpendicular to
the rotation axis, and mixed-flow turbomachines, when at the rotor outlet both radial and
axial velocity components are present in significant amounts, Dixon [29].
165H

With regard to the first criterion cited, i.e. that of whether power is being extracted from
or added to the working fluid, the present work addresses the category of power-
absorbing turbomachinery, namely fans. Since, as will be shown in the later sections, the
flow in the investigated fans, i.e. air, is characterized by Mach numbers below the
compressibility limit, a further placement in the general classification can be made,
according to the working fluid, i.e. fans operating with incompressible flow.
When considering the nature of the flow path with respect to the axis of rotation, the
present work addresses the axial flow machines, and thus, axial fans.
As a class, axial flow fans include high-capacity, low-head (pressure), single-stage
machines. In small sizes, motor-driven axial fans are sometimes built in two stages owing
to speed limitations but are classed essentially as low head machines, Thwaites [102].
16H

5
1. Introduction and aim of work
0B

1.3 Aim of work


The advent of the gas turbine engine during the Second World War demanded rapid
developments in aerodynamic design and analysis techniques linked to wind tunnel and
model testing, and in response, the field of “Internal Aerodynamics” was born and has
expanded with remarkable speed and complexity over the entire class of turbomachines,
Lewis [68], including axial fans. However, in the area of axial fans, a lack of test and
167H

design data in the early stages of their development is responsible for the attempt of most
designers in this field to make extensive use of airfoil profiles, Stepanoff [100].
168H

The main scope of the design process of axial fans, employing either airfoil theory or
other more direct design methods, is to deliver high-efficiency blades.
The same area is addressed in the present work, but with strong emphasis on developing a
new design strategy for axial fans, which does not make use of any predefined profiles,
such as airfoils, but instead computes high-efficiency profiles, according to the
operational requirements of the investigated impellers. Of course, designing an efficient
profile implies that the required shape has to be aerodynamically efficient and therefore,
the classical design considerations based on the airfoil theory will not be neglected.
However, the method presented is not limited only to such considerations, and even
though the airfoil theory method of axial impeller design is invariably associated with the
free-vortex energy distribution along the radius, variations from this well-known design
assumption are investigated and found more to be appropriate for fan design purposes.
Hence the proposed design method, addressing the field of low-pressure axial fans,
suggests an innovative blend of one-, two-, and three-dimensional flow considerations,
with the aim of delivering the best performing blade profiles, and thus fan models
approaching the ideal flow machine.
The concept of the ideal machine is quantified differently in the literature, and it seems
that there is really no standardized method to determine what the maximum achievable
performance of a specified class of flow machines, might be. Often, most designers refer
to the Cordier diagram in this sense, as presented in Figure 1-2.
169H

Essentially, this diagram delivers, for an optimum pair of rotor dimensions and operating
conditions (quantified by the so-called “diameter number”  , and the “speed number”  ,
respectively), the highest efficiency for the investigated impeller, Eck [33].  and  are
170H

6
1. Introduction and aim of work
0B

derived based on the flow and pressure (head) coefficients, i.e.  and  , respectively. 
and  are essential for the dimensionless analysis in the field of turbomachinery, and
one can interpret them as the equivalent of the Nusselt number (Nu) and Euler number
(Eu) from the Navier–Stokes equations. Based on a given pair  ,   , the flow type of the

investigated rotor is determined on the Cordier diagram: from radial (higher  and
smaller  ), to diagonal and finally, to axial (smaller  and higher  ).

Figure 1-2 Original Cordier diagram, adapted from Cordier [21]


17H

As stated in the above paragraphs, the present work addresses axial flow fans operating in
low-pressure ranges. When identifying the points corresponding to this type of impellers
on the diagram, it seems that they are mostly concentrated in the upper half of the chart,

7
1. Introduction and aim of work
0B

i.e.  opt  0.3 , and when associating a trend-line to cover most of these points, a

“probable” curve of efficiency for low-pressure fans is obtained, as marked in Figure 1-2.
172H

Hence any axial fan, characterized by optimum dimensions and operating conditions,
should deliver its highest efficiency if placed on this curve. However, this concept of
ideal/maximum efficiency, as quantified by the Cordier diagram, is to some extent
inconsistent from the axial impeller point of view. As will be presented in later sections,
axial turbomachines are characterized by integral properties, and all the parameters
influencing the efficiency of the impeller (  and  ) need to be integrated values of the
local ones. Hence the ideal efficiency of axial impellers should represent the result of the
integration of this local efficiency over the entire flow area. This matter is referred to in
the present work and an analytical, integrated expression of the ideal efficiency of axial
fans is proposed.

8
2. Basic equations of fluid mechanics as applied in turbomachines
1B

2 Basic equations of fluid mechanics as applied in


1B

turbomachines

In the previous chapter, a general statement of the essential laws which govern the flow
in turbomachines was made and it was emphasized that full solutions of these equations
are obtained based on the correct treatment of the turbulent Reynolds stresses. However,
before even considering the turbulent aspect of the flow in such machines, a thorough
understanding of these basic equations is required, as they are central for the
Computational Fluid Dynamics (CFD) technique. CFD is fundamentally based on the
governing equations of fluid dynamics, which essentially represent statements of the
conservation laws of physics. The purpose of the present chapter is to introduce the
derivation of these fundamental equations and their employment for the CFD analysis of
turbomachines, where the following physical laws are adopted:
 Mass is conserved for the fluid
 Newton’s second law, the rate of change of momentum equals the sum of all
forces acting on the fluid
 First law of thermodynamics, the rate of change of energy equals the sum of rate
of heat addition to the fluid and the rate of work done on the fluid
In this sense, the derivation of the continuity, momentum, and energy equations in
rotating reference systems will be addressed, as they are applicable to turbomachines, as
in the case of the axial fan depicted in Figure 2-1, where the rotation is about the x3-axis.
173H

9
2. Basic equations of fluid mechanics as applied in turbomachines
1B

Figure 2-1 Axial flow fan within the cartesian coordinate system

2.1 Navier–Stokes equations in rotating systems


Let us consider, for the upcoming derivations, a control volume defined around the fluid
element of known mass m, as shown in Figure 2-2:
174H

Figure 2-2 Control volume around a fluid element

The flow velocity (denoted U in the previous section) will be referred to as c, as it is the
common notation in the theory of turbomachines.
The rate of change in the mass is given by:
dm  
   c dA (2.1)
dt A
The continuity equation states that there is no change in mass with respect to time, hence

10
2. Basic equations of fluid mechanics as applied in turbomachines
1B

dm
0 (2.2)
dt
In the case of the an incompressible fluid, i.e. the fluid density is constant, (2.2) becomes
175H

 
 dA  0
A
c (2.3)

In tensorial form (employing Einstein’s summation), Eq. (2.3) can be written as


176H

ci
 0 , i  1, 2,3 (2.4)
xi
Equation (2.4) expresses the continuity equation for an incompressible, steady flow.
17H

The momentum equation (Newton’s second law) states that the rate of change in the
momentum equals the sum of all forces acting on the fluid element:

dM d    
    c  c dA   F (2.5)
dt dt A A

In (2.5), the sum of all forces acting on the fluid element includes: pressure forces,
178H

friction forces and gravity forces:


d      
 
dt 
 c c dA  Fp  F fr  Fg (2.6)
A

Written in tensorial form, for the cartesian coordinate system, the momentum equation
becomes
c j c j G 1 P 1  ij
 ci    (2.7)
t xi x j  x j  xi
  
gravity pressure friction
force force force

On multiplying by the density, then Eq. (2.7) becomes 179H

 c j c j  G P  ij
  ci      (2.8)
 t xi  x j x j xi

In Eq. (2.8), the following terms can be identified:


180H

G
gj  
x j

where G is the gravitational potential and





11
2. Basic equations of fluid mechanics as applied in turbomachines
1B

where  is the kinematic viscosity and  is the dynamic viscosity.


Equation (2.8) becomes
18H

 c j c j  P  ij
  ci    gj (2.9)
 t xi  x j xi

For the fluid mechanically ideal case of inviscid flow, i.e. ij  0 , the momentum equation

in (2.9) becomes
182H

 c j c j  P
  ci    g j (Euler equations) (2.10)
 t xi  x j

The continuity equation in (2.4) and momentum balance equation, for the viscous flow
183H

case, in (2.9), form a system of four equations with ten unknowns: the flow
184H

velocity c j  c1 , c2 , c3  , the viscous term  ij  11 , 12 , 13 , 22 , 23 , 33  and the pressure P. In

order to close this system, additional equations have to be solved.


According to Durst [31], the viscous term due to the momentum transport  ij is a
185H

c j
function of and is described by the following relation:
xi

 c j ci  2 c
 ij         ij  k (2.11)
 xi x j xk
  3
Considering   const and   const , then

 2 ci  2 ci   ci 
    (2.12)
x j xi xix j x j  xi 
The continuity equation for the incompressible flow states that
ci
0 (2.13)
xi
Equation (2.12) becomes
186H

 2 ci  2 ci
 0 (2.14)
x j xi xix j

On inserting (2.14) into (2.11), then the viscous term in the momentum equation can be
187H 18H

written as

12
2. Basic equations of fluid mechanics as applied in turbomachines
1B

 ij
 2 ci
  2 (2.15)
xi xi
Equation (2.9) becomes
189H

  c j c j  P  2 ci

   ci       gj
  t xi  x j xi2
 (2.16)
 ci  0
 x
 i
Equation (2.16), coupled with the continuity equation (2.13), form the Navier–Stokes
190H 19H

system of equations for incompressible flow, which is a system of four equations with
four unknowns, and hence a closed system. This system, however, in order to describe
accurately the flow in a rotating system, such as the rotor of a turbomachine, needs to be
solved with respect to a rotating (relative) system.
Let us consider a point P and its position with respect to two systems: the absolute
reference system (AS) and the relative reference system (RS), as depicted in Figure 2-3.192H


Its origin is at point OR, it rotates with angular velocity  and its translation from the AS

is given by the vector a .

Figure 2-3 Absolute and relative reference systems at the same time step t. Adapted from Fister [40]
193H


At the same time step t, the position of the point P is given by the vector q with respect

to AS and by r with respect to RS. The rate of change in the position of P with respect to
AS represents the absolute velocity and is given by

 d q
c  OA (2.17)
dt
The rate of change in the position of P with respect to RS is the relative velocity:

13
2. Basic equations of fluid mechanics as applied in turbomachines
1B


 d r
w  OR (2.18)
dt
The rotation of RS with respect to AS creates the circumferential velocity, Fister [40], 194H

which is defined as
  
u r (2.19)
On writing the Navier–Stokes system in (2.16) with respect to RS, then two additional
195H

forces in the momentum balance appear: one due to the centrifugal force, and the other
due to the Coriolis force:
   
 w w    P      
    w    g          r   2   w (2.20)
 t r  r r  
  
centrifugal Coriolis
force force

In tensorial form, (2.20) can be written for all three directions of the cartesian system, i.e.
196H


as indicated in Figure 2-1, whereas for the RS, the angular velocity vector  indicates the
197H

positive x3-direction, Epple [37]: 198H

 w j w j  P  ij
  wi    
  g j   i xi j  i2 x j  2  kij wk i (2.21)
 t xi  x j xi

The system is then completed with the continuity equation:


wi
0 (2.22)
xi
In (2.21) and (2.22), the Navier–Stokes equations with respect to a relative system are
19H 20H

written for the case of the incompressible flow. These are the equations which describe
the rotating frames of reference and, in the case of the turbulent flow in turbomachines,
Reynolds decomposition is applied to this system.
The energy equation for the same relative reference system is obtained by multiplying
(2.21) with w j under the following assumptions: stationary flow, neglecting the molecular
201H

moment transport  ij and considering a constant angular velocity:

D w u2 
2
p
  j  G   0
Dt  2  2 

  w2j p u2 
 wi   G    0 (2.23)
xi  2  2 

14
2. Basic equations of fluid mechanics as applied in turbomachines
1B

On integrating (2.23) according to Figure 2-2, then the following expression for the
20H 203H

energy equation for a rotating frame is obtained:


   w2j p u 2 
   wi    G   wi dAi
 2  
 xi  2 
A 

which finally leads to


1 2 2
W
2
    
c1  c2  u12  u22  w22  w12   (2.24)

Equation (2.24) is the energy equation for a relative system and in the theory of
204H

turbomachines this is referred to as the Euler equation for pumps and turbines (the
difference is made by the sign convention: positive work for turbines, negative for
pumps). This equation relates to the energy transfer in such flow machines and, since it is
essential for the analysis of turbomachines and central to the design process, it will be
treated separately in the following section.

2.2 Energy transfer in turbomachines


1 2 2
The first term in (2.24), 205H

2
 
c1  c2 , represents the energy transfer due to the change in

the absolute kinetic energy of the fluid during its passage between the entrance and exit
1
sections. In a pump or compressor, the discharge kinetic energy from the rotor, c22 , may
2
be considerable. Normally, it is the static head or pressure that is required as useful
energy. Usually, the kinetic energy at the rotor outlet is converted into a static pressure
head by passing the fluid through a diffuser. In a turbine, the change in absolute kinetic
energy represents the power transmitted from the fluid to the rotor due to an impulse
effect. As this absolute kinetic energy change can be used to accomplish a rise in
pressure, it can be called a “virtual pressure rise” or a “pressure rise” which it is possible
to attain. The amount of the pressure rise in the diffuser depends, of course, on the
efficiency of the diffuser. Since this pressure rise comes from the diffuser, which is
1 2 2
external to the rotor,
2
 c1  c2  is sometimes called the “external effect”.
The other two terms in Eq. (2.24) are factors that produce a pressure rise within the rotor
206H

itself, and hence they are called “internal diffusion”. The centrifugal effect, given by the

15
2. Basic equations of fluid mechanics as applied in turbomachines
1B

1 2
term
2
 u1  u22  , is due to the centrifugal forces that are developed as the fluid particles

move outwards towards the rim of the machine, and this effect is produced if the fluid
changes the radius from the entrance to the exit of the impeller. This is not the case with
the axial-flow machines, where the flow particles enter and leave the rotor at the same
radius, and hence u1  u2 .

1 2
The last term,
2
 
w2  w12 , represents the energy transfer due to the change in the relative

kinetic energy of the fluid. If w2  w1 , the passage acts like a nozzle, and if instead

w2  w1 , then it acts like a diffuser.

Figure 2-4 Meridional flow through a turbomachine and flow through an elementary stream tube: a)
meridional flow through a pump or a fan rotor; b) stream tube along the surface of revolution
mapped out by the meridional streamline  0 . Adapted from Lewis [68]
207H

The Euler pump and turbine equation as derived previously is a one-dimensional equation
in the sense that it is applicable to a unit mass of fluid flowing along the line mapped out
by the elementary stream tube illustrated in Figure 2-4b. The circumferential projection
208H

of such infinitely thin stream tubes on to the (x,r) plane leads to the definition of a family
of so-called meridional streamlines illustrated in Figure 2-4a, of which the hub and casing
209H

form the boundary streamlines. It is clear that one Euler equation must be derived for
each meridional streamline during the design phase of a turbomachine and these
equations will lead to a precise specification of the swirl velocity change from cu1 to cu2
required for a specified change in the work of the rotor, W. The Euler equation is thus
central to the design process.

16
3. Survey of the available design methods for axial impellers
2B

3 Survey of the available design methods for axial


2B

impellers

In the field of turbomachinery, multiple design methods and concepts can be identified,
and often important design choices are made based on the designer’s experience. Even
though such design choices appear to be numerous and developed for the particular class
of flow machines under consideration, the design process itself, no matter what its
content, can be laid out in simple steps that need to be followed if the resulting design is
to be a successful one.

3.1 Basic features of turbomachinery design


Much of the established literature on designing flow machines makes valuable references
to the existent design methods, Balje [9], and accordingly three basic steps in the design
210H

analysis are important to designers and analysts of turbomachinery:


 The one-dimensional design or the so-called “mean-line design” method (or
critical path line) usually provides the first step in the design procedure
 Two-dimensional and three-dimensional methods, including blade and vane
definition, supported by methods based on the potential flow analysis
 Advanced viscous 3D calculations on the basis of the Reynolds equations,
including also complex structural calculations
The role of each of these steps needs to be clearly understood for a successful design and
the value of each must not be underrated compared with the others.
The first step includes the preliminary designs, and also some preliminary design
optimization studies, using a variety of one-dimensional design and analysis tools. These
tools attempt to model the basic flow physics at distinct stations through the

17
3. Survey of the available design methods for axial impellers
2B

turbomachine where important events occur or where important thermodynamic and fluid
dynamic book-keeping must be carried out, Japikse and Baines [53]. Usually these 21H

analyses utilize the basic conservation laws (mass, momentum and energy) applied either
along a mean streamline through the machine or along a critical streamline.
Even though the description of this initial step sounds essentially simple, in reality, at this
level, most of the important design decisions must be made, and it is normally the case
that such decisions are entirely based on empirical data and the designer’s experience. It
is difficult, however, to predict, at this level, whether one or the other design choice is
better suited for the application of the machine, and therefore this process involves much
trial and error before a final decision is made.
The design solutions for a basic concept are normally derived for the design point (see
Figure 3-1), and several trial geometries are usually generated.
21H

After the basic one-dimensional configuration has been laid out (i.e. the typical passage
height and mean-line velocity triangles, flow rates, speeds, and power levels are
calculated through the one-dimensional analysis), the designer can proceed to the
specification of the two-dimensional problem, i.e. to specify the actual blade shapes.
Circular arcs, multiple circular arcs, arcs plus straight lines, and polynomials are the most
commonly employed design choices. Traditionally, the first and often most appropriate
tool is the two-dimensional inviscid flow-field analysis. Such calculations enable one to
confront the most essential elements of the flow process, from a physical standpoint, with
reasonably practical numerical calculations.
Equally important at this stage of the design is the choice of important constructive
parameters of the full rotor. In this respect, an excellent summary was presented by
Stepanoff [100], where empirical formulations for critical design parameters, such as the
213H

hub ratio and number of blades, were presented.


Finally, the design process is normally concluded with the performance evaluation of the
rotor, and usually one analyzes the so-called “system characteristic”, which is the map of
the system’s behavior at various operating points. All important parameters reflecting the
performance, i.e. head and efficiency, are observed before one or the other design is
chosen. A typical system characteristic of an axial fan is depicted in Figure 3-1.
214H

18
3. Survey of the available design methods for axial impellers
2B

Figure 3-1 The “design point” (point of maximum efficiency) on a typical system characteristic for
an axial fan

The design point is the point of maximum efficiency on the efficiency curve depicted in
Figure 3-1, and the corresponding flow rate is the design flow, Qdesign . Most designs are
215H

normally evaluated in this manner and the best performing design is usually a blend
between good efficiency rates and pressure (or head) which can be achieved by the
impeller.
Most designs today require optimization to ensure good performance under diverse
operating conditions. Therefore, to carry out an optimization requires repeated analysis,
such as the ones mentioned above, under many different conditions, and the cost of the
two- and three-dimensional analysis flow field calculation is excessive for this approach.
Hence two- and three-dimensional tools are used for detailed refinement of a basic
concept that has been previously optimized with effective mean-line calculations.

3.2 Two – dimensional cascade theory


The operation of any turbomachine is directly dependent upon changes in the working
fluid’s angular momentum as it crosses individual blade rows. A deeper insight into
turbomachinery mechanics may be gained from consideration of the flow changes and
forces exerted within these individual blade rows. The complex three-dimensional flow
can be treated as the superposition of a number of two-dimensional flows. This leads to a
more manageable blade design and profile selection techniques.
For an axial impeller with hub and casing, it is reasonable to assume that the stream
surfaces at the entry to the annulus remain cylindrical as they progress through the

19
3. Survey of the available design methods for axial impellers
2B

machine. Each cylindrical meridional stream surface will then intersect the blade row to
form a circumferential array of blade shapes known as cascade, Lewis [68]. If one such
216H

cylindrical surface were to be unwrapped from the cylindrical coordinate system  x, r 

to a cartesian one  x, y  , then it would have the aspect depicted in Figure 3-2.
217H

unwraping of the
cylindrical surface

from cylindrical to
cartesian coordinate
system through y=rθ

Figure 3-2 Development of a cylindrical blade-to-blade section into an infinite rectilinear cascade in
the cartesian coordinate system. Adapted from Lewis [68] 218H

The full three-dimensional flow could then be modeled by a series of such plane two-
dimensional cascades, one for each of the cylindrical meridional surfaces equally spaced
between hub and casing.
To obtain truly two-dimensional flow, one would require a cascade of infinite extent.
However, cascades must be limited in size, and careful design is needed to ensure that at
least the central regions (where the flow analyses and measurements are made) operate
with approximately two-dimensional flow.
In particular for axial-flow machines of high hub-to-tip ratio, radial velocities are
negligible and, to a close approximation, the flow may be described as two-dimensional,
Dixon [29]. The flow in a cascade is then a reasonable model of the flow in the machine.
219H

For impellers with lower hub-to-tip ratios, the blades will normally have an appreciable
amount of twist along their length, the amount depending upon the chosen design. Even
so, data obtained from two-dimensional cascades can still be of value to a designer
requiring the performance at discrete blade sections of such blade rows.

20
3. Survey of the available design methods for axial impellers
2B

3.2.1 Aerodynamics forces and governing equations


Considering the two-dimensional airfoil cascade depicted in Figure 3-2, the fluid
20H

approaches the cascade from far upstream with a velocity w1 at an angle  1 and leaves far

downstream the cascade with a velocity w2 at  2 . The following analysis assumes that the

fluid is incompressible and the flow is steady. The first assumption is valid since most of
the cascade tests and measurements, even for axial compressors, are made for low Mach
numbers, i.e. below 0.3, when the compressibility effects can be neglected. However, a
correlation between the compressible and incompressible cascades is possible and the
available literature offers detailed techniques on doing so, Csnady [22].
21H

With regard to the steady flow assumption, this is valid for an isolated cascade row, but at
the rotor level, relative motions between successive blade rows appear, causing
unsteadiness.
An explanatory scheme of a portion of an isolated cascade is depicted in Figure 3-3.
2H

Figure 3-3 Forces acting upon an airfoil in a cascade

It is common in turbomachinery analysis to refer to the airfoil aerodynamic properties


with respect to w , which is the velocity of the undisturbed flow, far in front and far

behind the profile. In a cascade, the action of the fluid on the profile can be considered
similar to that taking place on an airfoil in a wind tunnel, provided that the velocity of the
undisturbed flow w is an average of the inlet and outlet relative velocities, Bohl [12]:
23H

21
3. Survey of the available design methods for axial impellers
2B

2
 w  w2u  wm
w  w   1u
2
 and tan    (3.1)

m
2  w1u  w2u
2
In this case, the velocity vectors diagram has the aspect depicted in Figure 3-4.
24H

Figure 3-4 Velocity vectors diagram

Two forces can be recognized as acting upon the profile: one in the axial direction ( Fax )

and the other in the tangential direction ( Ft ). These forces are exerted by unit depth of
blade upon the fluid, exactly equal and opposite to the forces exerted by the fluid upon
the unit depth of blade.
If p1 is the pressure upstream the cascade and p2 is the pressure downstream, then

Fax   p2  p1  bt (3.2)

where b represents the radial length of the blade profile.


When analyzing the velocity diagrams in Figure 3-4, then the following relations can be
25H

developed between the characteristic velocities of the fluid:


w12  wu21  wm2 1
(3.3)
w22  wu22  wm2 2
Because of the incompressibility assumption made in the previous paragraph, the
meridional component of the flow velocity is conserved along the profile, and thus
Q
wm1  wm 2  (3.4)
bt
The energy balance between the inlet and the outlet of the cascade results in

22
3. Survey of the available design methods for axial impellers
2B

p1 w12 p2 w22
   (3.5)
 2  2
With Eqs. (3.3) and (3.4) inserted into Eq. (3.5), the following expression for the axial
26H 27H 28H

force acting upon the profile can be established


 
Fax 
2
 
bt wu21  wu22 
2

bt w12 sin 2  1  w22 sin 2  2  (3.6)

An expression for the tangential force can be obtained by means of the impulse equation:
Ft  m  wu1  wu 2  (3.7)

where m   Q is the mass flow. The volume flow rate is determined from Eq.(3.4): 29H

Q  wmbt
Further manipulation of Eq. (3.7) yields 230H

Ft   wm bt  wu1  wu 2    wm bt  w1 sin  1  w2 sin  2  (3.8)

The resulting force acting on the profile, in the absence of friction (see Figure 3-3), is the 231H

lift, and after replacing expressions (3.6) and (3.8), this force can be written as:
23H 23H

L  Fax2  Ft 2   bt w  wu1  wu 2  (3.9)

Equation (3.9) defines the lift force acting upon a profile in an isolated cascade, in the
234H

ideal case of the frictionless flow. If one considers instead the real case of the flow with
friction, then also a drag force, D, can be included in the force balance.
Considering the unit depth of a cascade blade, the lift force acts perpendicular to the flow
direction, while the drag is parallel to the flow, as indicated in Figure 3-3, and both forces
235H

can be expressed in terms of Fax and Ft , Anderson [5]: 236H

L  Ft sin    Fax cos  


(3.10)
D  Ft cos    Fax sin  

In Eq. (3.10), the angle of the undisturbed flow   is given by (3.1).


237H 238H

The aerodynamic properties of the airfoil are usually presented in terms of dimensionless
coefficients, i.e. the lift and drag coefficients, defined as

23
3. Survey of the available design methods for axial impellers
2B

L
CL 

w2 l
2 (3.11)
D
CD 

w2 l
2
The calculation of the lift force acting on an airfoil can be achieved also by considering
the circulation in a large circuit enclosing the airfoil, and hence, the velocity far in front
the cascade:
L   w (3.12)
Equation (3.12) expresses the Kutta–Joukowski theorem for a single isolated airfoil for
239H

the ideal case of the frictionless flow, when D  0 .


The circulation is the contour integral of velocity around a closed curve and inserting Eq.
(3.9) into Eq. (3.12) yields:
240H 241H

  bt  wu1  wu 2  (3.13)

The Kutta–Joukowski theorem is directly related to the flow at the trailing edge of the
airfoil, Hughes and Brighton [52], and it can be reformulated as follows: for a given
24H

airfoil, at a specified angle of attack, the value of the circulation about the airfoil is such
that the flow leaves the trailing edge smoothly.
Apart from the Kutta–Joukowski theorem, the literature includes numerous methods and
further approximations for the calculation of the lift and drag, for particular families of
airfoils, and an excellent review was given by Thwaites, [102], in his chapter 5.
243H

3.3 Design methods based on the airfoil theory


There are two approaches to blade profile selection which are often referred to as the
direct (analysis) method and the inverse (synthesis) method, Lewis [68]. 24H

The direct method assumes the profile generation through systematic geometrical
techniques, and then series containing geometries generated as such are analyzed by
means of measurements or theoretical investigation, resulting in the determination of the
most efficient profiles and their detailed aerodynamic performance. In engineering
practice, the systematic procedures of the direct method offer a special attraction for
building up experimental and theoretical data for closely related families of cascades.

24
3. Survey of the available design methods for axial impellers
2B

Several direct design methods for generating two-dimensional blade shapes, applicable
for the design of axial turbomachinery cascades, were developed by Korakianitis [60]. 245H

The first method specifies the airfoil shapes with analytical polynomials and it shows that
continuous curvature and continuous slope of curvature are necessary conditions to
minimize the possibility of flow separation and to lead to improved blade designs. The
second method specifies airfoil shapes with parametric fourth-order polynomials, which
result in continuous slope-of-curvature airfoils, with smooth Mach numbers and pressure
distributions. Finally, a third method is presented, in which the airfoil shapes are
specified by using a mixture of analytical polynomials and mapping the airfoil surfaces
from a desirable curvature distribution. This method provides blade surfaces with
desirable performance in very few direct-design iterations.
The inverse method allows the designer to specify the velocity or pressure distribution
along the blade surface and to calculate the profile accordingly.
When prescribing velocity distributions (PVD), two options are available for airfoil or
cascade design by the inverse design method.
The first method permits the designer to specify a prescribed velocity distribution (PVD),
and therefore pressure distribution, on both the upper and lower surfaces, resulting in
automatic synthesis of the entire profile to meet this specification. Although this sounds
attractive, the procedure has its drawbacks. At worst, the designer may choose an
impossible PVD for which there is no corresponding blade profile. At best, the chosen
PVD may lead to an unsuitable profile thickness distribution.
In view of the latter problems, Wilkinson [111] proposed another option for airfoil design
246H

whereby the PVD is limited to the more aerodynamically sensitive upper surface only,
but a profile thickness is also prescribed. In effect, the inverse method then involves
designing the camber line shape required to achieve the desired PVD on the upper
surface. The velocity distribution on the lower surface is simply accepted to adjust freely
to whatever it will. Such design techniques were developed by Ackert [2], Railly [89],
247H 248H

Cheng [19], and Lewis [67].


249H 250H

The inverse design methods, by defining the blade geometry corresponding to a


prescribed pressure distribution, have been the subject of many theoretical and
computational studies and have greatly matured in recent years. They are nowadays

25
3. Survey of the available design methods for axial impellers
2B

reckoned to be a valuable alternative to the iterative use of direct methods, de Vito et. al.
[25]. Although the latter allow an easier control of various geometric parameters to meet
251H

the design requirements, they demand a lot of physical insight by the designer to predict
the changes required to reach a target pressure distribution. Hence inverse methods allow
a more direct interaction with the blade pressure distribution and require much less
computational effort. In this respect, Leonard and Van den Braembussche [66] and 25H

Demeulenaere and Van den Braembussche [26], made remarkable efforts to develop an
253H

efficient inverse design method based on Euler solvers. Their method allows the
definition of the 3D geometry required to obtain a prescribed pressure distribution on the
whole blade surface. Also, by prescribing the blade loading distribution (i.e. the
difference between the blade suction and pressure surfaces), the profile geometry can be
inversely computed, Dang et. al., [23] [105].
254H 25H

3.3.1 Airfoil families. Mean-line and thickness distribution


Since most of the parameters that are critical for the design process of axial fans rely
entirely on the airfoil theory and wind-tunnel measurements of such profiles, the author
feels that a brief explanation of the main characteristics of these airfoils is helpful.
An airfoil can be conceived of as a curved camber line upon which a profile thickness
distribution is symmetrically superimposed, Figure 3-5. 256H

Figure 3-5 Geometrical description of a cascade profile

The traditional approach to the aerodynamical design of axial-flow compressors and fans
is to use various families of airfoils as the basis for blade design, and probably the most
popular ones belong to the NACA family and the British C-series. The development of

26
3. Survey of the available design methods for axial impellers
2B

the NACA profiles started in 1929 in the Langley variable-density wind tunnel. There
exists a huge data base on these profiles and their measured characteristics, and probably
the most comprehensive material on the topic is the summary of Abbot and Von
Doenhoff [1]. 257H

Several series of NACA profiles can be identified (commonly used in the design of axial
compressors and fans are the four and five digit series) and each of these series is
characterized by a different thickness distribution along the mean camber line. All airfoils
in the NACA 4-digit family were designed for the same basic thickness distribution and
the amount of camber (curvature) was systematically varied to produce the family of
related airfoils. By extending the investigation to airfoils having the same thickness
distribution, but moving the positions of the maximum camber far forward on the airfoil,
the 5-digit series was obtained.
The process of combining the mean-line and thickness distribution to obtain the NACA
airfoils is described in Figure 3-6.
258H

Figure 3-6 Sample calculations for the derivation of the NACA 65 series. Adapted from Abbot and
Von Doenhoff [1] 259H

Ordinates of the cambered airfoil are obtained by laying off the thickness distribution
perpendicular to the mean-line. If xU and yU represent the abscissa and ordinate,

respectively, of a typical point of the upper surface of the airfoil and yt is the ordinate of

the symmetrical thickness distribution at the chordwise position x , the upper-surface


coordinates are given by the following relations:
xU  x  yt sin 
(3.14)
yU  y  yt cos 
Analogous relations for the lower surface can be derived.
In Eq. (3.14), yt is the thickness distribution and detailed formulations of the applied
260H

thicknesses to all NACA families are presented in Abbot’s summary of airfoil data.

27
3. Survey of the available design methods for axial impellers
2B

3.3.2 Design parameters


In the process of designing axial impellers, the initial analyses of the flow through the
blade rows assume that the velocity vectors along a blade are constrained to lie along the
mean camber line of the profile. Although this is true exactly on the surface of the profile,
the flow is not generally constrained completely by the blade shape through the entire
flow channel, Wright [112].261H

Since the flow pattern is highly dependent on the geometric characteristics of the channel,
i.e. blade-to-blade distance, it becomes obvious that whatever camber angle value is
prescribed through the design method, it will not coincide with the actual flow angle,
especially in the blade passage. To increase the accuracy of the angle prediction and the
design calculations, one needs to formulate a quantitative model of this flow behavior.
The most important geometric parameters, critical for profile design, are depicted in
Figure 3-5 and the differences between the designed blade angles and the actual flow
26H

angles can be underlined as follows:


 The difference between the prescribed blade angle and the flow angle achieved at
the inlet section is called the incidence, i   1  1 .

 The difference between the blade and flow angles, at the outlet section, is called
the deviation,    2   2 . The deviation reflects the failure to achieve the
expected level of turning of the flow vector from the ideal angle and it is a
function of the geometric and velocity properties of the blade cascade.
There are many empirical correlations available in the literature, which are focused on the
prediction of the flow pattern in the channel by applying various values for the incidence
and deviation angles.
As far as the incidence angle is considered, it frequently refers to the particular inlet angle
for which the leading edge (LE) stagnation point is situated precisely on the end of the
profile camber line, i.e. i  0 , and this is the particular case of the “shock- free” inlet (see
Figure 3-7b).
263H

28
3. Survey of the available design methods for axial impellers
2B

Figure 3-7 Stagnation streamline and the “shock-free” inflow

For greater or smaller inlet angles, the stagnation point will move instead on to the lower
or upper surface respectively, as illustrated in Figure 3-7a and c. The so-called “shock-
264H

free” inlet flow condition ensures the smoothest entry conditions into the cascade and is
thus likely to be close to the “minimum loss” situation, Lewis [68]. In the present work, a
265H

zero incidence assumption is included in the design process.


Including in the design process correlations for the deviation angles accounts for a good
prediction, from the design stage, of the actual flow angles. In this respect, Lieblein [70]
26H

proposed corrections for the NACA 65 series and the British C4 profiles, in terms of
empirical formulations for the incidence and deviation.
Howell [51] developed an analytical method that allowed the modeling of the flow angles
267H

with excellent results: by fixing a “shock-free” entrance of the flow in the cascade, the
vector field can be resolved by expressing the trailing edge (TE) deviation as a simple
function of the channel proportions. Howell’s correlation can be written as follows:
m
 1
(3.15)
 2

where  represents the cascade solidity and is defined as the ratio between the chord
l
length of the profile and the blade spacing (pitch):   .  is the camber turning angle:
t
  1   2
The coefficient m is calculated according to the following expression:

29
3. Survey of the available design methods for axial impellers
2B

  
m  0.41  0.2  1 
 100 
Another empirical correlation for the deviation, derived also for the zero incidence
condition, was proposed by Carter and Hughes, [17]: 268H

1   2
 (3.16)
4 
The outlet flow angle predicted by Carter’s rule is then
2   2   (3.17)
Finally, McKenzie [76] also suggested an empirical formulation of the deviation angle:
269H

1
  1.1  0.31   3
(3.18)

It can be concluded that, with respect to the prediction of the flow angles at critical
stations on the profile, such as the trailing edge, the literature offers numerous methods
(such as the ones presented above) to correct the design angles, so that the desired flow–
blade congruency is achieved. However, most of these methods are empirical and were
determined for a specified class of axial impellers, i.e. axial compressors, and it is the
choice of the designer which of the formulations should be included in the design
process. In the present work, the value of the deviation was assumed to be zero, since it
was found more appropriate for thin profiles; it will be shown in the following section
that, for the fan application of interest for the present work, thin profiles are better suited
rather than airfoils. However, the possibility to assume other values for the deviation is
incorporated in the mathematical routine used for the profile calculation and any of the
presented formulations in Eqs. (3.15), (3.16) and (3.18) can be employed.
270H 271H 27H

3.3.3 Cascade losses. Diffusion factor


The total pressure loss of a cascade depends on many factors. Under normal operating
conditions, the boundary layer on the suction (upper) surface will be much thicker than
that on the pressure (lower) surface of the airfoil, and hence, to a first approximation, the
latter can be neglected. Then, the thickness of the wake (and therefore the total pressure
loss) will be primarily determined by that fraction of the suction surface over which the
velocity difference is negative, since that is where the majority of the boundary layer
growth occurs, Brennen [13] and Lakshminarayana [63]. In other words, the suction
273H 274H

30
3. Survey of the available design methods for axial impellers
2B

surface has a large pressure rise which can cause the viscous boundary layer to separate,
and this is not desirable due to the associated higher losses of the total pressure.
One should visualize the deceleration (diffusion) of the fluid from wmax to w2, where wmax
is the maximum velocity on the suction surface. Hence, a direct correlation between the
total pressure loss and the diffusion (caused by the deceleration of the fluid) on the upper
side is useful. According to Lieblein et al. [72], the amount of diffusion is given by the
275H

diffusion factor, which is defined as


wmax  w2
DF  (3.19)
wmax
Lieblein argued the momentum thickness of the wake, θ, should be correlated with the
diffusion factor. In this respect, several measurements were performed for the NACA65
and British C4 series profiles and it was found out that a value of DF = 0.6 imposes an
upper limit for the allowable diffusion factors, as indicated in Figure 3-8. Above this
276H

value, a dramatic increase in the diffusion in the boundary layer will occur.

Figure 3-8 Wake momentum thickness versus overall diffusion factor DF for NACA 65 and C4
airfoils at minimum loss incidence. Adapted from Lewis [68] 27H

The diffusion factor, as defined in Eq.(3.19), represents an important design parameter


278H

and offers valuable information, from the design stage, on the performance of the
prescribed profile, since a good design should be characterized by limited deceleration in
the flow velocity. Moreover, even though at the design stage a frictionless flow is

31
3. Survey of the available design methods for axial impellers
2B

assumed, by calculating the DF and imposing limitations to its value one can assume that
the real flow with friction will cause fewer losses due to these preventive measurements
taken from the early design stage.

3.4 Three-dimensional character of the flow in axial turbomachines


In the previous section, the most important design considerations in the field of axial
impellers were reviewed and it can be concluded that such methods have the benefit of
the simple two-dimensional flow modeling. However, the designer must not forget that in
truth, the flow in turbomachines is really three-dimensional and any design
considerations based on the 2D cascade theory result in a simplified flow prediction,
which overlooks the 3D character of the flow and its side effects, i.e. secondary flows.
Probably the most notable effort in this area was made by Wu [113], who recognized the
279H

truly three-dimensional nature of the flow in turbomachines and proposed the remarkably
sophisticated computational scheme illustrated in Figure 3-9.
280H

Figure 3-9 S-1 and S-2 stream surface according to Wu. Adapted from Lewis [68]
281H

Wu proposed a general theory of steady three-dimensional flow of a non-viscous fluid in


subsonic and supersonic turbomachines having arbitrary hub and casing shapes and a
finite number of blades. The solution of the three-dimensional direct and inverse problem
was obtained by investigating appropriate combinations of flows on relative stream

32
3. Survey of the available design methods for axial impellers
2B

surfaces whose intersections with a z-plane, either upstream of or somewhere inside the
blade row, form a circular arc or a radial line.
The fully 3D flow was treated by the superposition of a number of 2D flows, but in this
case located on the S-1 and S-2 stream surfaces. S-2 surfaces follow the primary fluid
deflection caused by the blade profile curvature and its associated aerodynamic loading.
Due to the variation of static pressure between the convex surface of the 1st blade and the
concave surface of the 2nd blade, the curvature of each S-2 stream surface will vary, thus
calling for the introduction of several surfaces for adequate modeling, Lewis [68]. 28H

The S-1 surfaces are equivalent to the meridional surfaces of revolution which are
allowed to twist in order to accommodate the fluid movements caused by the variations
of the S-2 surfaces.
The S-1 and S-2 surfaces represent, in fact, a selection of the true stream surfaces passing
through the blade row. By solving equations of motion for the flows on this adaptable
model, successively improved estimates of the S-1 and S-2 surfaces may be obtained,
allowing also the fluid dynamic coupling between them. The iterative approach to
achieve a good estimate of the fully three-dimensional flow was very comprehensively
laid out by Wu [113].
283H

The first major computational scheme based on Wu’s work dealt with axisymmetric
meridional flow located on an averaged S-2 surface, Marsh [74]. 284H

Also based on Wu’s treatment of the flow in turbomachines, Denton [27] developed a
285H

time-marching method, which practically initiated the path for design codes for
compressible three-dimensional flow analysis. Similar efforts were made by Potts [86], 286H

who also developed a time-marching method to study the twisting of the S-1 surfaces
within highly swept turbine cascades.
A second aspect of the three-dimensional character of the flow in turbomachines is the
appearance of the so-called “secondary flows”, which are of great importance in axial
turbomachinery aerodynamics, where boundary layer growth that occurred on the casing
and hub walls of the machine are deflected by the blade rows. Many studies have been
entirely dedicated to the investigation of the boundary layer on the annulus walls of axial
flow machines and it has been concluded that axial impellers operate at nominal
conditions with significant boundary layer separation, Lieblein [71] and Schlichting [94].
287H 28H

33
3. Survey of the available design methods for axial impellers
2B

The separation in the boundary layer is an important issue for the design of axial
turbomachinery blading and major attempts at understanding and solving the complicated
three-dimensional flows which appear due to this separation have been made, Horlock
and Lakshminarayana [49] and Hawthorne and Novak [46]. However, it is generally
289H 290H

agreed that classical boundary layer considerations alone are insufficient for capturing the
full 3D nature of the secondary flows, and full CFD calculations of the whole flow are
generally required, Horlock [50]. 291H

The presence of the 3D secondary flows impacts directly on the performance of the
impeller and therefore methods of controlling the separation phenomenon have to be
included in the design process. One such method is to predict, from the early design
stage, the location where the separation might occur. In this respect, Ramirez Camacho
and Manzanares Filho [90] developed a model for the cascade computation of axial
29H

impellers, able to predict the separation point near the TE and reattaching the flow by
introducing fictitious velocities to achieve the viscous effect of the attached flow.
Other possible methods to control the separation flows are either to introduce a tip
clearance, Gbadebo et al. [41], or through the suction (aspiration) of the boundary layer
293H

on the suction surface of the profile and at the end walls, Gbadebo et al. [42] and 294H

Kerrebrock et al. [59]. Even though both methods yield in immediate increase in the
295H

impeller performance, the first method appears more attractive, especially in the case of
axial fans, since the tip clearance is a design parameter that is fairly easy to control, due
to the general constructive simplicity of the fan casing, which dictates the size of the
clearance. However, an optimum tip gap has to be found and it is normally in the range of
a few percent of the chord length.
Finally, also accounting for the three-dimensional character of the flow, is the meridional
analysis, which is particularly focused on the flow prediction in axial machines. Since the
meridional analysis of axial fans is of major importance for the design strategy proposed
in the present work, a detailed explanation of the concept will be presented in the
following chapter.

34
4. Proposed design strategy for axial fans
3B

4 Proposed design strategy for axial fans


3B

The layout and design strategy proposed in the present work is focused on the blade
design of axial fans that operate under low-pressure regimes, delivering high flow rates,
and thus deals with fans normally employed for cooling purposes.
This method involves a preliminary design stage, based on mean-line performance
calculations, and also a detailed design stage, which includes three-dimensional flow
considerations to generate the initial blade profiles.
Essentially, the design procedure presented below is a design-point method, meaning that
the runner blade is derived for a single point on the system characteristic depicted in
Figure 3-1, i.e. the point of maximum efficiency corresponding to the design flow
296H

rate, Qdesign .

According to the available literature, the design-point technique appears to be an inverse


method. However, the presented design strategy is a combined inverse–direct method
used in an iterative process to generate an actual geometry, where one can specify the
fluid dynamic boundary conditions and the governing equations, and then, effect
solutions for the geometry required to establish these conditions. The approach is very
appealing from the design point of view.
The goal of the design strategy is to deliver high-performance designs for direct industrial
applications. For this, the constructive dimensions and operating regime of an impeller,
currently manufactured in industry and used for engine cooling purposes, are employed
as a reference solution. These details will be referred to later.

4.1 Mean-line calculation


As it was referred in the introduction to section 3.1, the first step in the proposed design
297H

strategy is the one-dimensional flow analysis, i.e. the mean-line calculation. The
advantage of this approach is that Euler’s equation for turbomachinery can be resolved

35
4. Proposed design strategy for axial fans
3B

for each cascade independently. Consequently, the first design consideration is that the
impeller is a succession of such cascades, at different distances from the hub center, and
the mean-line calculations will be performed for each cascade individually.
In the case of axial impellers, the flow particles enter and leave the blade at the same
section, and therefore u1  u2  u . In this case, Euler’s equation states that the total
pressure difference accomplished by a profile assuming inviscid, incompressible flow has
a static and a dynamic component:
 
  
  
pt   w12  w2 2  c2 2  c12 
2    
 (4.1)
 static dynamic 
 pressure pressure 

The first design assumption (confirmed by both CFD and experiments) is that the fluid
has an axial entry in the cascade. This assumption remains valid as long as the entrance in
the impeller is not disturbed, and hence there are no stator vanes in front of the fan blades
which dictate the flow inlet in the rotor.
A simplified scheme of the mean-line calculation is presented in Figure 4-1.
298H

36
4. Proposed design strategy for axial fans
3B

Figure 4-1 Mean-line calculation of the velocity diagram with axial entry

Assuming axial entry of the fluid simplifies the inlet velocity diagram (point 1 in the
above sketch). The tangential component of the absolute flow velocity is zero, i.e. cu1  0 ,
hence there is no pre-whirl induced in the cascade. This implies that
wu1  u (4.2)

where u  2 rn and n represents the rotational speed of the rotor cascade.


With this assumption, Eq.(4.1) becomes
29H

 
pt 
2
w 2
u1 
 wu22  cu22 
2
u 2  wu22   u  wu 2     u  u  wu 2    ucu 2
2


(4.3)

On analyzing the velocity diagrams in Figure 4-1, the following expressions can be
30H

immediately stated for the inlet and outlet blade angles in the cascade:

37
4. Proposed design strategy for axial fans
3B

wm wm
tan  1   (4.4)
wu1 u
wm
tan  2  (4.5)
wu 2
Basically, provided that the flow rate and rotational speed are known, by assuming axial
entry, the inlet blade angle can be calculated.
The meridional component of the relative velocity wm is conserved along the blade
profile, and therefore through out the cascade, and is equal to the ratio of the flow rate to
the flow area. The flow area of the cascade is in fact the flow area corresponding to the
entire impeller, and therefore the area of the ring between the hub and tip sections of the
rotor:


A   rt 2  rh2  (4.6)

The expression of the inlet blade angle becomes


Qd 1
tan  1  (4.7)
  rt 2  rh2  2 rn

For to the derivation of the outlet blade angles, more assumptions are required to solve
the velocity diagram at the cascade exit. This matter will be addressed in the following
section.

4.2 Outlet conditions


In the previous section, the expression for the inlet blade angle was determined according
to the axial entry assumption. This angle is a function of the design point, i.e. flow rate
and rotational speed of the cascade, and, of course, the radius.
Equation (4.1) expresses the total pressure difference achieved by the axial cascade,
301H

according to Euler. It can be observed that it has two components: a static pressure
difference, given by the relative flow velocity gradient, and a dynamic component, given
by the absolute flow velocity gradient.
What normally is of interest for the system characteristic of an axial rotor is the static
pressure component:
  
ps 
2
w 1
2

 w2 2 
2
w 2
u1 
 wu22 
2
u 2
 wu22  (4.8)

38
4. Proposed design strategy for axial fans
3B

If the tangential component of the relative flow velocity at the cascade exit is rewritten
according to Eq. (4.5), then 302H

wm wm
tan  2   wu 2  (4.9)
wu 2 tan  2
Inserting Eq. (4.9) into Eq. (4.8), the following expression of the static pressure
30H 304H

difference is established:

 wm2 
Ps   u 2
  (4.10)
2 tan 2  2 

According to Figure 3-1, the static pressure difference is maximum when the flow rate is
305H

zero and, conversely, the flow rate is maxim at zero pressure difference. If Eq. (4.10) is 306H

written for the highest possible flow rate, i.e. zero pressure difference, then it yields
wm2 2
Qmax 1
u2   4 2 2 2
r n  (4.11)
tan  2
2
A tan 2  2
2

Hence the maximum flow rate is


Qmax  2 rn A tan  2 (4.12)
From Eq. (4.12) it can immediately be observed that the maximum flow rate, at a
307H

specified blade section, is “maximum” when the outlet flow angle


 2 (r )  90 deg ( Qmax   ). Of course, this value is radius dependent, hence it can be
implemented at one blade section only. The question is which section is best suited for
this angle. Applying this angle means, in terms of velocity diagrams, that wu 2  0 and

cu 2  u , i.e. cu 2 is maximum. The tangential component of the absolute velocity cu 2

represents the swirl at the exit of the cascade and the product cu 2 r is a direct measure of
the blade loading. For an axial cascade, the total pressure difference in Eq. (4.3) can be 308H

written in terms of the loading and hence


pt   ucu 2  2 rn  cu 2  Kcu 2 r (4.13)

where K  2  n is a constant and after further equating, it can be concluded that the
swirl velocity downstream of the cascade is inversely proportional to the rotor radius:
K
cu 2  (4.14)
r

39
4. Proposed design strategy for axial fans
3B

Since the swirl velocity cu 2 expresses losses and, according to Eq. (4.13) high swirl
309H

means high blade loading, a good design should limit this component. Moreover, through
deceleration of cu 2 , a recovery in the static pressure can be achieved. Since the hub
section has the highest expected loading and here the profile should not turn more than
axial, this is where a blade angle of 90o is best suited.
For the other blade radii, the outlet blade angles will be determined recurrently, using this
startup value at the hub.

4.3 Meridional flow analysis for axial fans


The previous paragraphs were mainly concerned with the mean-line calculations leading
to the determination of the inlet blade angles, and, by assuming the “maximum flow rate”
condition at the hub section, the derivation of the outlet blade angle at this section was
achieved.
The present section deals with the three-dimensional treatment of the flow in an axial fan
and further design assumptions will be made according to this complex flow treatment.
The starting point of the three-dimensional treatment was made in Figure 2-4, where the
310H

notions of “cascade” and “meridional flow” were introduced. It was shown that fully
three-dimensional flow can be treated, for a more manageable framework, as an
axisymmetric or circumferentially averaged meridional flow, and a series of
superimposed cascade flows to define blade profiles at selected sections from hub to tip.
Previously, the fluid motion through the blade rows was assumed to be two-dimensional
in the sense that radial (span-wise) velocities do not exist. This is not an unreasonable
assumption for axial rotors with a high hub-to-tip ratio. However, for ratios less than 0.8,
the radial velocities through a blade row may become appreciable, the consequent
redistribution of mass flow (with respect to radius) seriously affecting the outlet velocity
profile and, thus, the flow angle distribution, Dixon [29]. Such radial flows are mainly
31H

caused by the imbalance between the strong centrifugal forces exerted on the fluid and
the radial pressures. The flow in an annular passage, in which there is no radial
component of the velocity, with circular streamlines and cylindrical surfaces and which is
axisymmetric, is commonly know as “radial equilibrium flow”. The treatment of the
three-dimensional flow in an axial turbomachine based on the assumption that only radial

40
4. Proposed design strategy for axial fans
3B

flow which may occur is completed within a blade row, the flow outside the row being in
equilibrium, is the radial equilibrium method, Dixon [29]. Since the correct
312H

understanding of this method is central for the proposed design strategy, the author feels
that a brief explanation of the theoretical treatment behind is necessary.

Figure 4-2 Radial equilibrium through: a) a rotor; b) a fluid element (cr = 0)

For a small fluid element of mass dm , of unit depth and at an angle d from the axis,
rotating about the axis, with a tangential velocity cu at a radius r , the radial equilibrium
(centrifugal force is balanced by the pressure force), as indicated in Figure 4-2 b, can be
31H

stated as follows:
1 dp cu2
 (4.15)
 dr r
For incompressible flow, which is normally the case for axial fans, the so – called
stagnation pressure can be defined as
po p cx2 cr2 cu2
    (4.16)
  2 2 2
where cx represents the axial velocity, cr is the radial component of the velocity, and cu
represents the tangential velocity.
Since the radial equilibrium condition, as stated previously, imposes cr  0 , on
differentiating Eq. (4.16) with respect to r one obtains
314H

1 dp0 1 dp dc dc
  cx x  cu u (4.17)
 dr  dr dr dr
dp
Introducing from Eq. (4.15), then
dr
315H

41
4. Proposed design strategy for axial fans
3B

1 dp0 dc c d  rcu 
 cx x  u (4.18)
 dr dr r dr
Equation (4.18) states the equation for the radial equilibrium condition for incompressible
316H

flows. The key element for the present work in this equation was the loading given by the
product  cu r  and the most important design choices were made with respect to this

term.
Equation (4.18) can be applied to two sets of problems:
317H

1) the indirect problem (design method), in which the tangential (swirl) velocity is
specified and the axial velocity is calculated accordingly.
2) the direct problem, in which the swirl angle is given and the axial and tangential
velocities are calculated.
The proposed design strategy is focused on the first method.

4.4 The indirect design problem


The indirect design problem requires specification of the tangential velocity, calculating
the resulting axial velocity. There are several approaches to this problem and they are
generally focused on the product  cu r  in Eq. (4.18).
318H

Probably the most popular design assumption is that the product  cu r  is constant, also

known as the “free-vortex flow” assumption. Essentially, this assumption implies that
 cu r   k (4.19)

When considering an element of the ideal inviscid flow rotating about a fixed axis, as
indicated in Figure 4-2b, then the circulation  (the vortex strength) is involved:
319H


cu  (4.20)
2 r
The vorticity at a point is defined as the ratio between the limiting value of the circulation
  and the elementary area  A , as  A becomes vanishing small:

  lim (4.21)
 A 0  A

Differentiating Eq. (4.20): 320H

d    cu  d cu  r  d r  d   cu rd 

42
4. Proposed design strategy for axial fans
3B

 dc c 
d    u  u  rd dr (4.22)
 dr r 

Equation (4.21) then becomes


321H

d  1 d  cu r 
  (4.23)
dA r dr
d  cu r 
If the vorticity expressed in Eq.(4.23) is zero, then  0 , with the result that
dr
32H

cu r  const (4.24)
Equation (4.24) embodies the absolute condition to be satisfied by a free-vortex flow.
32H

Returning to the expression for the total pressure difference achieved by an axial cascade
in Eq. (4.3), the free-vortex flow assumption can then immediately be translated into a
324H

“constant total pressure” assumption, since


pt   ucu 2  k  cu 2 r  (4.25)

As already mentioned, this is probably the most popular design choice for axial fans,
especially when the two-dimensional airfoil theory is employed in the design process.
However, the use of the three-dimensional analysis with axial fans is not limited only to
this assumption, and the available literature makes references to several other
possibilities, often referred to as “non-free” vortex flow (forced vortex) or “solid body”,
Dixon [29], Lewis [68], Augnier [6], and Carolus [16]. Essentially, these inverse methods
325H 326H 327H 328H

assume some variation of the product  cu r  , deriving the axial velocity according to this

assumption. One of the preferred hypotheses is that of constant swirl velocity,


cu  const .Even though such methods are mentioned, little use has been of them and the
effects of employing a non-free vortex design assumption are not well known.
At this point in the analysis, it becomes of great interest to establish which design choice,
whether free- or non-free vortex flow, delivers, for a specified fan configuration, the best-
performing design. This leads to the motivation for the present work: to determine, for a
given class of fans, the optimum vortex design specification.

43
4. Proposed design strategy for axial fans
3B

4.5 Parameterization of the total pressure in the span-wise direction for an


axial fan blade
Typical design methods for axial fans make intensive use of the free-vortex flow
assumption. Basically, this implies that along the blade, a constant total pressure
difference, according to Euler’s equation, is applied, as indicated in Eq. (4.25).
329H

For the area of interest in the present work, i.e. axial fans used for cooling purposes, the
work of Wallis [108] is relevant, in which an inlet guide vane–rotor–stator installation
30H

was investigated. The system considered was of the free-vortex flow type and several
important parameters, e.g. lift-to-drag ratio, were fixed. This resulted in explicit
expressions for efficiency and total pressure rise as a function of tip speed ratio, hub
ratio, and downstream losses.
Focusing on the same class of fan application, Dugao et al. [30] considered the numerical
31H

design optimization of a rotor–stator configuration for mine ventilation. Employing a


free-vortex design method, a considerable improvement in efficiency was achieved
compared with an existing installation. As an additional advantage, it was found that the
noise emission from the fan installation was reduced.
It thus seems that the free-vortex solution is preferred by most designers in the field.
There is little information available on design techniques which make use of an arbitrary-
vortex assumption, and the differences that such a design consideration might bring about
when compared with the classical zero-vorticity condition.
One of the few references in this respect relates to the work of Sørensen et al. [98] and
32H

[99], which deals with combining an optimization algorithm with the arbitrary vortex
3H

flow, so that a wider range of design alternatives are investigated in an efficient manner.
The fixed design parameters were the tip radius, the number of blades and the tip
clearance. The design variables included the hub radius, the chord distribution, the
camber angle, the total pressure rise and the velocity diagrams, and a series of constraints
were applied to these parameters. The objective function was the efficiency of the rotor,
considered over a design interval of flow rates, not only at the design point, and the
design flow rate was fixed in the centre of the interval. The goal of the optimization was
to maximize the mean value of the aerodynamic fan efficiency.

44
4. Proposed design strategy for axial fans
3B

Since good prospects of employing a non-free vortex flow condition are created by such
results, it seems desirable to undertake the development of a design strategy which, for a
specified fan application, determines whether improvements in the performance can be
achieved by employing the free-vortex flow or whether some variation is better suited.
Hence the motivation of the present analysis was to determine the optimum pressure
distribution along an axial fan blade, by considering both free and non-free vortex flow
assumptions, and parameterizing the solution according to the fan specifications.

Figure 4-3 Sketch of an axial impeller

Considering various sections of the blade (hub, tip and an intermediate section), as
indicated in Figure 4-3, the free-vortex assumption can be written in terms of the total
34H

pressure difference, at different blade sections, as follows:


Pt ,r
1 (4.26)
Pt ,h

The total pressure differences mentioned in Eq. (4.26) are defined as stated in Eq. (4.13).
35H 36H

According to Carolus [16], employing such an assumption introduces, from the design
37H

stage, a correct blade loading (pressure distribution) at the designated section. A proper
blade loading at the tip section (where the radius is larger) demands considerably smaller
values of cu2 (responsible for the swirl) than at the hub section (where the radius is
smaller). This means that close to the hub the absolute flow, characterized by higher cu2,
has to be redirected to the tip section, thus keeping the product cu 2 r constant. The desired
effect of such a consideration is that the flow detachments will occur earlier and hence
the hub section will be more stressed than the tip.

45
4. Proposed design strategy for axial fans
3B

The free-vortex assumption in Eq. (4.26) is directly connected to the swirl component at
38H

the exit of the cascade and it is employed whenever airfoil profiles are used in designing
an axial flow machine. However, there is nothing in the airfoil theory to prevent the
desired head distribution along the blade radius.
By changing the pressure distribution in the span-wise direction, designs providing higher
overall performances may be achieved with a small impact on the dimensions of the
swirl. Hence it is the designer’s task to investigate, for the specified design parameters
and operating conditions, the optimum pressure distribution which balances the
efficiency and the swirl component at the same time.
In the present study, it was assumed that starting from the hub and advancing to the tip,
the Eulerian pressure difference variation can be expressed as a function of the radius:
Pt ,r
 f (r ) (4.27)
Pt ,h

By employing the expression for the total pressure difference in Eq. (4.3) and making use
39H

of the outlet velocity diagram in Figure 4-1, a recurrent relationship between the outlet
340H

blade angles, at different blade sections, can be determined:


 wm   wm 
f (r )2 rh n  2 rh n    2 rn  2 rn  
 tan  2 h   tan  2 r 

wm wm
r  2 n  r 2  f (r )rh2   f (r )rh
tan  2 r tan  2 h
1 2 n 2 r 1
  r  f (r )rh2   f (r ) h (4.28)
tan  2 r rwm r tan  2 h
Equation (4.28) can be further simplified with another assumption made in the previous
341H

section, namely that at the hub, the outlet blade angle is 90o:
1 2 n 2
  r  f (r )rh2  (4.29)
tan  2 r rwm
It can be seen in Eq. (4.29) that the expression for the outlet blade angle can be
342H

parameterized according to the prescribed variation of the total pressure difference in the
span-wise direction. The use of such a formulation can be further extended if the equation
can be expressed in terms of relevant geometric parameters for the design process, such
as the hub ratio.

46
4. Proposed design strategy for axial fans
3B

As already mentioned in the introduction to the present chapter, the major focus of the
present work is the design of axial fans used for cooling purposes (engine cooling fans).
This is an important factor for the parameterization of the pressure variation, since such
devices are normally low-pressure high-flow delivering machines. Hence the pressure
should not be increased to unrealistic values in the range where the impeller will never
operate anyway. For example, for the reference model and its constructional details,
according to the Cordier diagram, a realistic operational value of the pressure difference
achieved by the impeller is below 4000 Pa. Also, rapid variations of the pressure from
one section to the next, which might result from employing exponential or higher order
polynomial laws for the function f  r  , should be avoided.

Another major constraint in the parameterization process of the function f (r ) is the


immediate impact that this expression has on the value of the outlet blade angle. The
angle, at the specified blade radius, delivered by f (r ) should be a realistic value (no
negative values) and should be incorporated in the geometry of the full blade, i.e. to
respect the trend imposed by the neighboring sections and not induce sudden twists in the
blade aspect.
Furthermore, the aim of the present formulation is to derive an expression in which, by
changing one factor, the desired pressure variation along the blade can be achieved.
An iterative routine, in which all of the above constraints were incorporated and solved
simultaneously, gave the following exact expression for the function f (r ) :
x
 rtip 
  r  rh   1
1.35
f (r )  x  (4.30)
 rhub 
The parameter to be changed during the design process is x . It can be immediately seen
that for x  0 the classical assumption of constant pressure (free-vortex flow) is
employed, since f (r )  1 and Eq. (4.26) comes about. For x  1 the pressure variation is
34H

linear with the inverse of the hub ratio, for x  2 parabolic, and so on. The exponential
1.35 in Eq. (4.30) yielded from pressure variation considerations. This exponential was
34H

iterated until a smooth pressure increase from one section to the next was obtained; with
higher degree exponentials the increase was to steep.
Equation (4.30) then becomes:
345H

47
4. Proposed design strategy for axial fans
3B

2 n  2   rtip   
x
1
  h
1.35
  r   x  r  r  1 rh2  (4.31)
tan  2 r rwm    rhub   

Equation (4.31) allows the determination of the outlet blade angle, at the specified radius,
346H

based on the pressure variation from hub to tip. Essentially, this variation has a major
influence on the resulting blade shape and as a consequence, an optimum profile demands
proper prescription of the pressure. Therefore, several test cases of axial fan designs will
be investigated, including the classical constant pressure assumption, in order to
determine, for specified dimensions and operating conditions, the optimum pressure
variation in the span-wise direction, which delivers high efficiency and small losses due
to the swirl component.
Three designs, corresponding to x  0  DesignI  , x  1 DesignII  and x  2  DesignIII 

respectively, are derived and their performance is assessed against the reference model.

4.6 Blade shape computation


The computation of the blade shape is carried out considering the cartesian system
depicted in Figure 4-4. The inlet profile angle is  1 , at an intermediate point along the
347H

profile  m , and the outlet of the profile is described by  2 . Each of the three stations on

the profile is characterized by corresponding coordinates  x, y  .

Figure 4-4 View of the calculated blade shape in (x,y) coordinates

Basically, the blade shape at a specified section can be readily computed by imputing the
y coordinates and angle distribution  (y) , since

48
4. Proposed design strategy for axial fans
3B

y
tan  ( y )  (4.32)
x
To apply effectively any driving action on the fluid, the blade angle is increased from  1

to  2 . The difference between the two,   2   1  , is a measure of the blade curvature

along any given blade section. The increase from  1 to  2 can be estimated by any type
of variation, either linear or higher degree polynomial.
At this point, the presented design strategy requires another assumption, connected to the
blade angle distribution along the calculated profile. It will be assumed that this
distribution is parabolic since, according to Pascu and Epple [82], this assumption is
348H

appropriate for ducted axial fans. Pascu and Epple showed by means of streamline
analysis performed on the designed blades, that very good agreement between the actual
flow angles and the prescribed blade angles is achieved by employing such a design
consideration.
Considering such a distribution  ( y )  Ay 2  By  C :
y  y1     2
y  ym     m (4.33)
y  y2     1
Since both the inlet and outlet conditions are fixed, the blade shape computation is
carried out with only one degree of freedom given by ( m , ym ) . The geometry is iterated
until the axial chord constraints are matched and also corrected for abrupt flow
parameters.

4.7 Further design assumptions based on profile analysis


Considering the flow over a profile, the fluid approaches the profile from upstream with a
velocity w1 at an angle  1 and leaves the profile with a velocity w2 at an angle  2 , as
indicated in Figure 3-3.
349H

49
4. Proposed design strategy for axial fans
3B

Figure 4-5 Forces acting upon the profile for small gliding angles

Figure 4-5 repeats the analysis on the forces acting upon a profile from Figure 3-3, except
350H 351H

that it underlines the presence of  , often referred to as the “gliding” angle of the profile,
which is defined as the drag-to-lift ratio:
D CD
  (4.34)
L CL
In the literature, several optimum values are proposed based on extensive experimental
results, Eckert [34], and it is normally the case that the values of the gliding angle are
352H

very small. The value of  is an important design choice since it denotes the impeller
losses by friction, and thus influences the cascade efficiency and hence the overall
performance of the rotor, as it will be shown in the following paragraphs. The design
value of  will be chosen based on this analysis.

4.7.1 Static-to-static cascade efficiency


Compared with the ideal case of the friction-less flow, when   0 , the drag force
introduced by  in the direction of w (Figure 4-5) calls for a decrease in the pressure
35H

difference.
According to the analysis of the forces acting upon a profile presented in section 3.2.1,354H

the lift force, acting perpendicular to the flow direction, has an axial and a tangential
component. The axial component is given by Eq.(3.2):35H

Fax   p2  p1  bt  Ps bt (4.35)

50
4. Proposed design strategy for axial fans
3B

Considering the case of the flow with friction (the real flow), then the tangential force can
be written as a function of the gliding angle:
Fax
 tan  90      
Ft
Fax 1
 (4.36)
Ft tan      

The definition of the tangential force is given by Eq.(3.8):


356H

Ft   wmbt  wu1  wu 2 

Combining (4.36) and (3.8), the following expression for the axial force acting upon the
357H 358H

profile yields
 wmbt  wu1  wu 2 
Fax  (4.37)
tan      

Substituting Eq. (4.37) into Eq. (4.35), the pressure difference caused by considering the
359H 360H

drag force caused by  in the flow direction, i.e. friction, becomes


Fax  wm  wu1  wu 2 
Ps , fr   (4.38)
bt tan      

The trigonometric expression tan       can be expanded to

tan    tan 
tan        (4.39)
1  tan   tan 
For very small gliding angles, which is normally the case, Bohl [12], the above
361H

expression can be simplified to


  tan  
tan        (4.40)
1   tan  
To facilitate correspondence with the classical cascade evaluation techniques, the
following annotations will be employed, according to Vavra [106]:
362H

wm
 (4.41)
u
wu1  wu 2
 (4.42)
u

51
4. Proposed design strategy for axial fans
3B

The theoretical degree of reaction is given by the ratio of the static pressure difference to
the total pressure gradient, achieved by the cascade, according to Euler’s equation, for the
case of frictionless flow, as defined by (4.8) and (4.3), respectively:
36H 364H

Ps
R (4.43)
Pt
Writing these pressure differences for the case of axial entry, the reaction becomes

2
w 2
u1  wu22  1 wu1  wu 2
R  (4.44)
 ucu 2 2 u

Further more, the expression for the angle   , as given in Eq. (3.1), can also be
365H

reformulated in terms of the dimensionless coefficients defined above:


wm 
tan     (4.45)
wu1  wu 2 R
2
Further equating (4.38), the static pressure difference for the case of flow with friction
36H

can be rewritten as

1 
Ps , fr   u 2 R (4.46)


R
By putting   0 in Eq. (4.46), the static pressure difference, for the ideal, frictionless
367H

flow, can be written as


1
Ps   u 2 (4.47)

R
At this point, a first evaluation of the cascade performance is possible, since expressions
for both the ideal (for the case of frictionless flow) and the real (flow with friction) static
pressure differences are derived. The ratio between the two terms is referred as the static-
to-static cascade efficiency and it has the following form:

Ps , fr 1 
ss   R
(4.48)
Ps R
1 

52
4. Proposed design strategy for axial fans
3B

Equation (4.48) expresses the losses in the static-to-static profile efficiency caused by
368H

considering the drag force acting upon the profile. For the case of ideal frictionless flow,
this efficiency is maximum, i.e. 1, as can be readily observed by putting   0 in the
above formulation.

4.7.2 Total-to-total cascade efficiency


Compared with the ideal flow, in the case of flow with friction, the drag force arises in
the direction of the fluid and, due to this force, work is dissipated, and according to Eck
[32] this can be formulated as
369H

Dw  Ploss wmbt (4.49)


From Eq. (4.49), the expression for the decrease in the total pressure caused by the drag
370H

force can be derived:


Dw
Ploss  (4.50)
wmbt
The definition of the drag force was stated in Eq.(3.11): 371H

 w2 lb
D  CD (4.51)
2
Equation (4.50) then becomes
372H

w w2 l
Ploss  CD  (4.52)
wm 2 t
Returning to the definition of the tangential component of the lift force acting upon the
profile, given by Eq. (3.10), and including the definitions of the lift and drag forces, as
37H

indicated in Eq. (3.11), then the following expression is obtained:


374H

 
Ft  L sin    D cos    CL w2 lb sin    CD w2 lb cos   (4.53)
2 2
However, the tangential force is also given by Eq. (3.8): 375H

Ft   wm bt  wu1  wu 2 

From the identity of the equations, it is found that


 
 wmbt  wu1  wu 2   CL w2 lb sin    CD w2 lb cos  
2 2
Further manipulations of the above equation yield

53
4. Proposed design strategy for axial fans
3B

wmt  wu1  wu 2   w2 lCL  sin     cos    (4.54)

From the velocity triangle depicted in Figure 3-4, it can be written that
376H

wm  w sin   (4.55)
The peripheral relative velocity gradient is
wu  wu1  wu 2
Equation (4.54) then becomes
37H

2wu l 1 
 CL 1    (4.56)
w t tan   

The left-hand term in Eq. (4.56) can be rewritten in terms of the total pressure difference,
378H

as defined in Eq. (4.3): 379H

2Pt l  
 CL 1  
 uw t  tan   

 l  
Pt  uw CL 1   (4.57)
2 t  tan   

Equation (4.57) expresses the total pressure difference achieved by an axial profile
380H

cascade when considering the real case of flow with friction and the drag force acting
accordingly upon the profile. This equation is essential for the design process of the
cascade, since it contains one of the most important design parameters, i.e. the cascade
l
solidity , and it allows the calculation of the optimum solidity, according to the
t
prescribed pressure.
On considering the ideal flow without friction, when the gilding angle is zero, then
 l
Pt  uw CL (4.58)
2 t
In the present work, the cascade solidities were fixed to the values of the reference
impeller (presented below, in Figure 4-7) and with respect to this parameter a design
381H

choice could not be made. However, in the absence of such constraints, Eq. (4.58) can be
382H

used for the dimensioning of an ideal cascade.


At this stage, it is of interest to see how much of the total pressure gradient of the ideal
cascade is reflected in the losses induced by the drag force, expressed in Eq.(4.52):
38H

54
4. Proposed design strategy for axial fans
3B

Ploss w2
 (4.59)
Pt uwm
Equation (4.59) can be written in terms of the reaction in Eq. (4.44) and the
384H 385H

dimensionless flow in Eq. (4.41) as follows:


386H

 R2 
Ploss  Pt      (4.60)
  
At this point, a second performance parameter of the cascade can be defined as the ratio
between the total pressure difference of a real cascade and the total pressure difference of
an ideal cascade, i.e. total-to-total efficiency:

Pt  Ploss  R2 
t  t   1      (4.61)
Pt   

Similarly to Eq. (4.48), the efficiency formulated above is maximum for   0 .


387H

The nature of the fan applications of interest for the present work imposes low-pressure
regimes, and at low-pressure, attributing values to the gliding angle, others than zero, will
only cause additional losses. Hence the present study assumes   0 . However, the
possibility of including in the design process other values of  is incorporated in the
mathematical routine used for the profile computation, and blade shapes according to
such an assumption can be computed.

4.7.3 Profiling the camber line


Another parameter central for the design process is the camber angle,  . Basically, the
resulting shape indicated in Figure 4-4 is the camber line and by attributing some
38H

thickness around it, a profile is obtained. The thickness distribution is again an important
design choice and one has to carefully weigh whether profiling really pays. Most design
methods for axial fans use a variable thickness for the profile (profiling the blade shape),
according to the thickness distribution functions existing in the literature, i.e. NACA and
British-C4 series, Wallis [109]. However, such methods were originally proposed for
389H

axial flow water turbines, where the operating regime requires high pressures. For axial
fans, where the pressures are in the range of thousands of Pa, a variable thickness along
the camber is unnecessary since according to Eckert [34], when measuring two impellers
390H

55
4. Proposed design strategy for axial fans
3B

operating at low-pressure, one with airfoil profiles and the other with constant thickness
(thin profiles), their performances are identical, as shown in Figure 4-6.
391H

Figure 4-6 Performance curves for constant thickness blade and an airfoil, Eckert [34]
392H

Another restriction for profiling results from the narrow pitching induced, i.e. the
distance between two consecutive profiles in a cascade is considerably reduced, and thus,
the cross-section becomes fairly restricted due to profiling, and the high speeds
generated, with the corresponding high friction, often offset the advantages of profiling.
Hence, for the calculated profiles, constant thickness will be applied, thus introducing the
last assumption made at the design stage: due to constant thickness distribution, there will
be no point of maximum thickness, and hence the camber angles, at all points on the
profile, will coincide with the computed blade angles.

4.8 Design Solver (DS)


In the previous sections, all the design assumptions required to compute the blade shape
were made. For summary purposes and also to help with a better understanding of all the
steps of the proposed design strategy, a simplified flow chart of the design steps is
presented in Figure 4-7.
39H

56
4. Proposed design strategy for axial fans
3B

Step 1: Mean-line calculations  Velocity triangles

Axial + Shockless entry  Inlet Maximum flow rate condition + blade


blade angle  1 loading considerations  Outlet blade
angle at the hub  2,h

Step 2: Meridional (3D) flow analysis

- indirect design problem


- arbitrary vortex-flow assumption
- parameterization of the total pressure difference in span-wise direction
 Outlet blade angle at specified section  2  r 

Step 3: Blade angle distribution (parabolic)

Step 4: Profile cascade dimensioning

- initial design value for the gliding angle   0


- constant thickness  camber angles = blade angles
l
- initial values for the cascade solidity correspond to the reference
t
impeller

No. of blades z [–] Characteristic Pitch t [mm] Chord l [mm]


section [mm]
8 147 (hub) 2 r 121
187 t 130
227 z 137
280 (tip) 144

Step 5: Design Solver  Blade Shape

Figure 4-7 Simplified flow chart of the design steps

All of these steps have to be solved for each individual cascade. The DS incorporates all
the design parameters and their derivation into a mathematical routine, which delivers the
blade profile at the specified section, as indicated in Figure 4-8.
394H

57
4. Proposed design strategy for axial fans
3B

START

Cascade Parameters
 air , rhub  147mm, rtip  280mm
n  3000rpm, Qdesign  4m3 / s
z , b, l1  r 

Velocity triangles
u  2 rn
Qdesign
wm 

 rt 2  rh2 
wm
wu1  u  axial entry   tan  1 
wu1

NO YES
Hub
section

2 n  2   rtip   
x
1
 
1.35
  r -  x   r - r  1 rh2   2,h  90 deg
tan  2 r rwm 
h
r
   hub   

wm
tan  2   wu 2
wu 2

( m , ym )

(continued)

58
4. Proposed design strategy for axial fans
3B

 ( y )  Ay 2  By  C
y  y1     2
y  ym     m
y  y2     1

 2 
 
  y  i 
 
 1

x  y tan   y 

xi  xi 1  x

y  x y

 x2  x1    y2  y1 
2 2
l2 

(continued)

59
4. Proposed design strategy for axial fans
3B

( m , ym )

l1  l2

wm
w y 
sin   y 
max w  y   w2
DF 
max w  y 

Ps 
2
 
wu21  wu22 , Pt   ucu 2

Fax  Ps bt , Ft   wmbt  wu1  wu 2 

 0  0

D0

wm 1 wu1  wu 2 Ploss  0
 ,R
u 2 u
ss  1
 R2  Ploss wmbt
Ploss  Pt     , D  t t  1
   w

Ps , fr 1 
R ,   1     R 
2
ss    
Ps R t t   
1 

Figure 4-8 Mathematical structure of the DS

60
4. Proposed design strategy for axial fans
3B

As already stated in the previous section, the proper blade shapes for the fan application
presently investigated are the thin profiles, since it was shown that, at low-pressures,
assuming a variable thickness around the camber line does not bring any improvements
in the performance of the impellers. For such thin profiles, all design assumptions are
made for the case of the frictionless flow   0 and accordingly, the DS completes its
routine on the right branch of the mathematical structure depicted in Figure 4-8. 395H

However, for different impeller applications, such as the axial compressors, which
operate at considerably higher pressures, airfoils are more appropriate according to the
available literature, Aungier [6]. In this case, assuming at the design stage that   0
396H

might deliver profiles with higher aerodynamical performance and an optimum value of
 can be determined by switching the profile calculations to the left branch of the DS.

4.9 DS output
According to the cascade dimensions mentioned in Step 4 in Figure 4-7, the resulting
397H

profiles, in  x, y  coordinates, for the investigated models, i.e. x  0 , x  1 and x  2 , and

also the reference profiles, are presented in Figure 4-9.


398H

61
4. Proposed design strategy for axial fans
3B

inlet

Figure 4-9 Computed camber lines

It can be observed that at the hub section, the camber lines of the three models are
identical, since the inlet angle is always given by the axial entry condition and the outlet
angle is always fixed at 90o. The differences in the computed profiles become obvious at
some distance from the hub, and half-way through the span the characteristic shape of the
camber lines is noticeable, i.e. “S” shape (see sections r227 and r280).
The solidity (the ratio l/t) of the reference varies with the radius, as indicated in Figure
39H

4-10, and it is kept constant for all the models.

62
4. Proposed design strategy for axial fans
3B

Figure 4-10 Variation of the solidity with the blade radius

The computed blade angles, for all three designs, and also the angle distributions of the
reference profiles, are presented in Figure 4-11.
40H

inlet

Figure 4-11 Blade angle distribution

63
4. Proposed design strategy for axial fans
3B

Again, the remark regarding the identical angle distribution at the hub sections applies.
Moreover, it can be observed that the reference cascades are characterized by large angle
gradients, from the inlet to the exit from the cascade, especially from mid-span on, and
the immediate result of such a design assumption can be observed in the relative flow
velocity distributions, as indicated below in Figure 4-12.
401H

Figure 4-12 Relative flow velocity distribution

The reference model is characterized by velocity distributions with huge gradients, very
unlikely to be achieved by the actual flow, according to Carnot’s basic principles for
hydraulic machinery to achieve its maximum efficiency. One of these principles states
that: “the fluid flow should be such that there is no decay, no turbulence, and no sudden
velocity reduction”, Epple [38].
The proposed designs, independent of the corresponding pressure variation, have the
calculated velocities in a fairly uniform range. This is a first indication of the

64
4. Proposed design strategy for axial fans
3B

improvements which can be achieved with the proposed design strategy, right from the
design stage. The same conclusion can also be drawn by analyzing the values of the
diffusion factor, as expressed in Eq. (3.19), shown in Figure 4-13.
402H 403H

Figure 4-13 Variation of the diffusion factor DF with the radius

All of the above results underline the advantage, at least from a design point of view, of
employing the proposed design strategy for the derivation of the blade profiles according
to the operational requirements of the impeller, rather than to use predefined profiles,
which may have been derived for completely different flow conditions, with pressure
and/or velocity values other than those required during normal operation. However, this
statement is made, as already mentioned, strictly from the design point of view, and it
remains to be seen whether the proposed models perform better according to the flow
analysis, and this will be discussed in the following chapter.

65
5. Numerical flow analysis
4B

5 Numerical flow analysis


4B

Previously, the first key element of the design strategy was presented, i.e. a design solver
which computes the optimum blade profile for axial fans, based entirely on the impeller
specifications, and does not make use of predefined profiles, such as those in airfoil
databases.
The next important step in the design process is numerical flow analysis, and the present
work makes intensive use of CFD for the investigation of the proposed designs. There are
many advantages in considering CFD as an integrated part of the design and optimization
process. First, CFD presents the perfect opportunity to study specific terms in the
governing flow equations in a more detailed fashion. Second, CFD complements
experimental and analytical approaches by providing an alternative cost-effective mean
of simulating real fluid flows and substantially reduces lead times and costs in designs
and production compared with an experimental-based approach. With the technological
improvements and competition requiring a higher degree of optimal designs and as new
high-technology applications demand the precise prediction of flow behaviors, CFD is
becoming an integral part of the engineering design and analysis environment and it is
intensively used to predict the performance of new designs before they are manufactured.

66
5. Numerical flow analysis
4B

Figure 5-1 CFD analysis frame work. Adapted from Tu and Liu [103]
40H

An outline of the most important steps involved in full CFD analysis is depicted in Figure
405H

5-1. This is also the sequence that will be followed next.

5.1 Mathematical model

In the previous chapter, a design system for axial fans was presented, which basically
consisted of a quasi-three-dimensional blade design module to generate the initial blading
geometries (detailed design stage), after a mean-line performance calculation was
completed (preliminary design stage). This design system was employed to derive an
optimum solution for a reference impeller (RI), i.e. a baseline model of an axial fan
currently used in the automotive industry for engine cooling purposes. This impeller was
made available by the manufacturer, and for flow investigation purposes, the CAD model
of the blade was also provided.
For the present investigation, four mathematical models of axial fans were derived,
corresponding to x  0 , x  1 and x  2 , and also the reference blade. All four models
are characterized by the same constructive dimension: Dhub  294mm , Dtip  560mm

(corresponding hub ratio 0.5), and all impellers have eight blades. Since the reference
blade was characterized by a thickness of approximately 3mm, when building the CAD
model of the new blades, a constant thickness of 3.2mm was applied symmetrically to the
camber lines presented in Figure 4-9.
406H

67
5. Numerical flow analysis
4B

Exemplary depictions of the mathematical models for the reference impeller


and x  2  DesignIII  are presented bellow:

a) b)

Figure 5-2 CAD models: a) reference impeller; b) x=2(Design III)

The characteristic “S” shape of the new designs can be readily observed, especially
towards the tip section of the rotor, while the reference profile is characterized by circular
profiles. The CAD models of the three new designs were built with Pro/Engineer
Wildfire 3.0 [121].
407H

All four impellers were placed inside a pipe with a tip clearance of approximately 3.5%
from the chord length at the tip section, i.e. 5mm, and hence D pipe  570mm . To bring the

simulation results closer to real-life operating conditions, two extra components were
added to the CAD model: at the inlet of the impeller a rig section long enough that the
flow entering the impeller domain can be considered fully developed (the length of the
rig section was chosen as 3D pipe ), and at the outlet (ambient) another pipe with a length

2 D pipe was included. An exemplary depiction of the CAD models is shown in Figure 5-3.
408H

68
5. Numerical flow analysis
4B

Figure 5-3 CAD model of the full system

5.2 Mesh generation

Following the flow chart depicted in Figure 5-1, the next step in the numerical flow
409H

analysis is the mesh generation of the mathematical model, and probably the most
important matter to be addressed at this level is the mesh-independent results.
Theoretically, the errors in the solution related to the grid must disappear for an
increasing number of grid points. Hence, also from a theoretical point of view, the finer
the mesh, the more accurate are the solutions, and often it is very difficult to determine
the level of fine enough grids, owing to computational time and resources issues. Hence
the motivation of the following investigation was the determination of the optimum grid
structure and size so that mesh-independent results are obtained, with no limitation from
the available computational resources.
All grids were generated using ANSYS ICEM 11.0 [118], applying the Generalized Grid
410H

Interface (GGI) capability of the software. GGI makes it possible to independently


generate a set of meshes for different sections of a problem using any tools or mesh
structure within each component. The individual grids can then be combined in a single
model. For the present models, due to the complexity of the impeller models, tetrahedral

69
5. Numerical flow analysis
4B

elements were chosen. However, tetrahedral cells are not desirable near walls if the
boundary layer needs to be resolved because the first grid point must be very close to the
wall while relatively large grid sizes can be used in the directions parallel to the wall.
These requirements lead to long thin tetrahedra, creating problems in the approximation
of diffusive fluxes, Ferziger and Perić [39]. For this reason, it is preferable that, during
41H

the meshing process, first a layer of prisms or hexahedra near solid boundaries is
generated, starting with a triangular or quadrilateral discretization of the surface, and on
top of this layer, a tetrahedral mesh is generated automatically in the remaining part of
the domain. This, however, depending on the number of the prismatic layers generated,
relates again to the mesh size issue.
Several meshes of the CAD model presented in Figure 5-3 were generated, with and
412H

without the prismatic layers, and the essential parameters for the fan performance, i.e.
torque in the shaft and the static pressure at the inlet of the impeller, were monitored
closely. The results are presented in Figure 5-4.
413H

Figure 5-4 Grid-independent results study (monitored parameters for Qdesign)

For greater transparency, the investigated grid characteristics are summarized in the table
below.

70
5. Numerical flow analysis
4B

Characteristics Number of grid points Prismatic Layers


Mesh 1 approx. 895,000 none
Mesh 2 approx. 1,455,000 none
Mesh 3 approx. 1,700,000 5 layers, initial height 0.09, exponential
growth from the wall
Mesh 4 approx. 2,000,000 10 layers, initial height 0.02, exponential
growth from the wall
Mesh 5 approx. 2,200,000 10 layers, initial height 0.09, linear growth
from the wall
Mesh 6 approx. 2,250,000 none
Mesh 7 approx. 3,000,000 10 layers, initial height 0.09, exponential
growth from the wall
Mesh 8 approx. 3,500,000 none
Mesh 9 approx. 5,500,000 7 layers, initial height 0.09, exponential
growth from the wall
Table 5-1 Mesh characteristics
It can be observed that both the pressure and the torque, for the meshes which do not
include prismatic layers, have strong peaks, and a reasonable trend for their variation as a
function of the grid size can not really be established. This aspect is avoided by
introducing the prismatic layers into the mesh generation process, and the variation of the
monitored parameters becomes fairly stable with increasing number of grid points.
Another parameter essential for the near-wall treatment isY , which basically represents
the dimensionless distance of the first node away from the wall, and for correct
assessment this value should be as small as possible. Considering all these delicate
matters of the near-flow wall treatment, Mesh 9 is found appropriate and hence the grid
characteristics mentioned in Table 5-1 will be employed in the following investigation. A
41H

detail of the mesh around the essential parts of the flow domain, i.e. blade and hub, is
depicted in Figure 5-5 and Figure 5-6.
415H 416H

71
5. Numerical flow analysis
4B

Detail A

Figure 5-5 Final mesh

Detail A

Figure 5-6 Mesh detail indicating the smooth transition from structured to unstructured grid

5.3 Numerical models and boundary conditions

The numerical simulations were carried out with the commercial code ANSYS CFX 11.0.
The simulation type was set initially to be steady. The fluid flow was set to be viscous
(air properties as an ideal gas were considered). The flow was solved with the Navier–
Stokes equations, assuming conservation of mass, momentum and energy.

5.3.1 Flow physics


wm Dtip
Typical values for the Reynolds number, calculated as Re  were around

800,000. One of the main problems in turbulence modelling is the accurate prediction of

72
5. Numerical flow analysis
4B

flow separation from a smooth surface. Standard two-equation turbulence models often
fail to predict the onset and the amount of flow separation under adverse pressure
gradient conditions. To avoid this problem, the model used in the present computations of
the air flow is the Shear Stress Transport model, Menter [77]. The model works by
417H

solving a turbulence/frequency-based model (k-ω) at the wall and k-ε model in the bulk
flow. A blending function ensures a smooth transition between the two models. The
interface between the different frames of reference is taken to be a Frozen Rotor. The
Frozen Rotor model has the advantage of being robust, using less computer resources
than the other frame change models, e.g. the stage interface model, which is not suitable
for applications with tight coupling of components and/or significant wake interaction
effects and may not accurately predict loading. The Frozen Rotor model treats the flow
from one component to the next by changing the frame of reference while maintaining
the relative position of the components. This model must be used for non-axisymmetric
flow domains, such as impeller/volute or classifier/casing.

5.3.2 Boundary conditions


An inlet boundary with specified mass flow was applied to the rig domain, flow direction
normal to the boundary condition and medium turbulence intensity (5%). A rotational
speed of 3000 rpm was applied to the fan domain. The outlet boundary was applied to the
ambient domain by specifying a 25oC temperature (air density 1.05 kg / m3 ). Typical
Mach numbers were around 0.25 (for Mach numbers below 0.3, the fluid can be
considered incompressible, Anderson [5]). Several simulations were carried out for all
418H

models by varying of the flow rate at 2, 3, 4, 5, 6 and 8 m3 / s .

5.3.3 Convergence history


The simulations were completely converged; all runs reached the convergence criterion
(residual type RMS and residual target to default value 10–4). The convergence criterion
value can be increased, but this means also a considerable increase in the computational
time required and not at the cost of significant differences in results. Further, during the
simulations, the most important physical quantities, such as flow rate, pressure and
efficiency, were monitored and it was observed that all these relevant quantities

73
5. Numerical flow analysis
4B

converged properly with the 10–4 residual target. All simulations were solved by parallel
running on 16 CPUs and the typical computation time required for full convergence is in
wall clock seconds of around 5 103 [122]. A detail of the convergence history is depicted
419H

exemplarily in Figure 5-7.


420H

Figure 5-7 Convergence history

5.4 Appropriate performance indicators

A large number of definitions of the efficiency of turbomachines have been given in the
literature, and the correct assessment of turbomachines, especially when it comes to
comparing the performance of more rotors, depends on defining appropriate performance
indicators.
For an axial fan with casing, where the power input to the rotor is in fact the power to the
shaft, the efficiency can be defined as
total (hydrodynamic) energy input to fluid in unit time
Efficiency 
power input to coupling of shaft
In other words, the efficiency of the impeller can be written as the ratio of the hydraulic
power to the shaft power:

74
5. Numerical flow analysis
4B

hydraulic power
Efficiency  (5.1)
shaft power
Ideally, the two parameters should be equal so that the efficiency of the machine is 1.
The shaft power developed by the impeller can be written as the product of the angular
speed and the required torque:
Pshaft   M (5.2)

If the total-to-static pressure difference between the inlet and the outlet of the impeller is
measured, then the hydraulic power is
Phydraulic  Ps Q

When the total pressure difference across the impeller is considered, then the hydraulic
power becomes
Phydraulic  PQ
t

Accordingly, two efficiencies, conveniently named total – to static and total – to – total
efficiencies, can be defined:
Ps Q
t  s 
M (5.3)
PQ
t  t  t
M
Also of high relevance for practical use is the polytropic efficiency (small stage
efficiency), defined as

 poly 
  1 ln  (5.4)
 ln 
p2 T
where the air ratio is   1.4 ,   and   2
p1 T1

5.5 Optimum span-wise pressure distribution

After post – processing the results of the converged numerical simulations, the following
system characteristics for the investigated models were obtained:

75
5. Numerical flow analysis
4B

Figure 5-8 a) Variation of the static pressure difference with the flow rate; b) variation of the torque
with the flow rate

In Figure 5-8a, the system characteristic curves for all four models are plotted. It can be
421H

observed that, while the proposed designs are characterized by similar values of the static
pressure difference between the inlet and the outlet of the impellers, the reference model
has higher pressure values. Since, according to Eq. (5.3), the second parameter, which
42H

essentially influences the efficiency of the impeller, is the torque, in Figure 5-8b the
423H

resulting torque along the rotation axis is plotted. Again, similar values are obtained for
the proposed designs, and much higher values for the reference. There is one point,
however, on both curves, in the behavior of the reference where a sudden decrease in
pressure and torque occurs. Even so, this does not influence the efficiency trends, as
indicated in Figure 5-9.
42H

Figure 5-9 Total-to-static efficiency

76
5. Numerical flow analysis
4B

In Figure 5-9 the total-to-static efficiency, as expressed by Eq. (5.3), is plotted. As


425H 426H

expected, no substantial difference can be observed in the performance of the proposed


designs, at least for the first interval of the investigated flow range. For higher flow rates
however, this is no longer the case, since the differences in the calculated efficiencies
become substantial, i.e. Design III has an absolute increase in efficiency of 16% for 8
m3/s, compared with Design I. The same can be concluded also from the total – to – total
and polytropic efficiency curves, as shown in Figure 5-10, and absolute increases of 13%
427H

and 18%, respectively, for Design III are obtained.

Figure 5-10 a) Total-to-total efficiency; b) polytropic efficiency

At this point, the advantage of considering at the design stage the possibility of a non-free
vortex flow (i.e. assuming that in the span-wise direction the total pressure is not
constant) becomes obvious. Employing such a design assumption impacts directly on the
extension of the flow range under which the fan can effectively operate, since, according
to Figure 5-9 and Figure 5-10, for higher flow rates, Design I – characterized by the free-
428H 429H

vortex flow assumption – shows a rapid decrease in efficiency, whereas Design III,
corresponding to x = 2 non-free vortex flow assumption, performs substantially better.
Since in practice fans often operate far from the design point (and often with low
efficiency), Bolton [14], it therefore seems desirable to choose the design that, for an
430H

interval of flow rates, performs better not only at the design point but also away from this
value.
Hence it can be concluded that for the investigated fan specifications, i.e. with a hub ratio
of 0.5, the best performing design is delivered by the x=2 design assumption, and

77
5. Numerical flow analysis
4B

therefore the optimum span-wise pressure distribution corresponds to a parabolic increase


with the hub ratio, from hub to tip:
2
Pt ,r r 
 2  tip   r  rh   1
1.35
(5.5)
Pt ,h  rhub 
Further more, on comparing the performances of the suggested design and the reference
impeller, significant absolute increases can be observed for Design III: up to 18% in
total-to-static efficiency (which is the relevant performance indicator), up to 4% in the
total-to-total efficiency, and up to 9% in the polytropic efficiency. Hence, as a solution to
the optimization problem of the reference impeller, Design III is suggested.
Note: since the obvious trend in the variation of the x parameter in Eq. (4.30) is to
431H

increase it, a fourth design, corresponding to x  2.5 , was investigated and it was
observed that a further increase in the pressure distribution (more than parabolic with
rtip / rhub ) is not recommended, since decreases in all performance indicators, compared

with x  2 , were obtained.

5.6 Profile analysis

In the previous section, the best performing solution to the optimization problem of a
reference impeller of known dimensions was found, and it was shown that, for the
indicated constructive dimensions (hub ratio 0.5), the optimum variation of the total
pressure difference along the blade can be parameterized according to the equation
2
Pt ,r r 
 2  tip   r  rh   1
1.35
(5.6)
Pt ,h  rhub 
The efficiency curves presented in Figure 5-9 and Figure 5-10 show substantial increases
432H 43H

in the performance of Design III, and represent the integrated values of the efficiency at
the cascade level. They can therefore be referred to as a quantitative analysis.
A qualitative analysis may be the flow aspect around the investigated blades, and this
matter is addressed in Figure 5-11. Even though, from the aspect point of view, the flow
43H

around the proposed designs is very similar, and compared with the reference case
appears much more attached to the blades, it is very difficult to draw conclusions about
the performance of the computed profiles from this cascade perspective. Hence the
motivation of the following section is to perform a thorough flow analysis, directly on the

78
5. Numerical flow analysis
4B

single profiles, for both Design III and the reference case, so that the best performing
profile, from the aerodynamic point of view is determined.

reference

Design I (x=0)

Design II (x=1)

Design III (x=2)

Figure 5-11 Velocity streamlines around the investigated models for Qdesign at r = 227 mm

79
5. Numerical flow analysis
4B

5.6.1 Flow domain around the profiles


In order to carry out the flow analysis over the investigated blade shapes, isolated profiles
were simulated so that the aerodynamic performance of the single profile was captured.
For all characteristic sections (hub, tip, and the two intermediate sections) the profiles of
both blades (Design III and the reference) were investigated, and the flow domain around
the profiles was large enough that the theoretical considerations of the undisturbed flow
far in front of and far behind the profile were satisfied. Before choosing the specific
lengths of the CAD model of the full flow domain around the profiles, several
simulations were carried out so that the dimensions of the domain at which the cascade
influence becomes negligible were chosen. The thickness of the profile was
approximately 10 mm and the flow was investigated on the center-line of the profile,
where no influence from the margins of the side walls was observed. The length of the
chord was kept to the values indicated above in Figure 4-7. A typical CAD of the profiles
435H

and the flow domain around them is depicted in Figure 5-12.436H

Figure 5-12 Model of the flow domain around the considered profiles

Special attention was paid to the setting of the boundary conditions. Excellent
convergence residuals were obtained by setting the following boundary conditions. At the
inlet section, the specified flow velocity wm is

80
5. Numerical flow analysis
4B

Qdesign
wm   22.4m / s (5.7)
  rt 2  rh2 

At the outlet section of the domain, opening boundary to the ambient pressure was
applied; on the lateral surfaces of the domain, symmetry boundary was set; for the top
and bottom surfaces again symmetry was applied since it was observed that, after a
height of the domain of 20 times the chord length from top to bottom, the difference in
results between applying symmetry or opening boundary was negligible. The simulations
were carried out for viscous flow with air as ideal gas.

5.6.2 Mesh generation


In order to calculate a qualitative mesh able to solve accurately the flow around the
profile, and especially the delicate problem of the flow in the near-wall regions, a grid
study was carried out. For this, three different grids were initially generated on the CAD
model depicted in Figure 5-12: a fine tetrahedral mesh (approx. 120,000 cells), a course
437H

hexagonal mesh (approx. 65,000 cells), and a fine hexagonal mesh (approx. 200,000
cells). The hexagonal grids were obtained by dividing the flow domain into blocks, which
were then subdivided into grids with good properties, and O-grid structured topology was
applied for the block corresponding to the profile. The two different hexagonal grids were
calculated by varying the height of the elements along the resulting edges (smaller
heights for the fine mesh).

81
5. Numerical flow analysis
4B

Figure 5-13 Velocity streamlines for the reference profile at r = 280 mm: a) tetrahedral grid; b)
course hexagonal grid; c) fine hexagonal grid

On comparing the flow images, it can be immediately seen that whereas the streamlines
for the tetrahedral grid appear smooth and completely attached to the profile, for the
course hexagonal grid the trajectory deviates slightly and the streamlines are distancing
from the profile, and finally, in the case of the fine hexagonal grid, an obvious flow
detachment from the contour is plotted. As already mentioned in section 5.2, in the near-
438H

wall regions, the structured hexahedra or prismatic layers are preferred, since they resolve
the boundary layer more accurately, and for this reason, the fine hexahedra was used for
further calculations. A detail of the grid generated around the profile is depicted in Figure 439H

5-14.

82
5. Numerical flow analysis
4B

Figure 5-14 Fine hexagonal grid for the reference profile at r=280 mm

5.6.3 Numerical results


A first analysis of the numerical results is the qualitative appreciation of the pressure
contours, as depicted in Figure 5-15. Pressure plots on the blade are helpful in identifying
40H

the different sources of losses in the system, such as the ones caused by sound sources,
i.e. the leading edge of the blade and areas close to tip section, due to the tip clearance,
Carolus et al. [91].
41H

r147 r147

TE
LE

r187 r187

83
5. Numerical flow analysis
4B

r227 r227

r280 r280

Figure 5-15 Pressure contours around the reference (left) and Design III (right)

It can be concluded from Figure 5-15 that at all characteristic sections, the suction on the
42H

reference profile is much more pronounced than for Design III. Additionally, the size of
the suction on the reference profile increases with increase in the radius, indicating an
increase in the boundary layer thickness, causing an associated increase of the diffusion
on the profile. At the same time, the suction on the Design III profile decreases with
increase in the radius and the contour plots indicate fairly uniform distributions on the
upper surfaces, causing the small diffusion predicted at the design stage (see Figure 43H

4-13). Moreover, the “stagnation point” (see Figure 3-7) is much more pronounced on the
4H

reference profiles than on the Design III profiles, indicating that the latter are closer to
“the minimum loss situation”, Lewis [68]. 45H

Such pressure contours can be numerically quantified with the help of the so-called
pressure coefficient, defined as
p  p1
CP  (5.8)
1
 w12
2
where p represents the static pressure along the profile, and p1 and w1 are the static
pressure and velocity, respectively, upstream of the cascade.

84
5. Numerical flow analysis
4B

LE
Pressure (lower)
surface

TE

Suction (upper)
surface

Figure 5-16 CP distributions along the investigated profiles

CP, as defined in Eq. (5.8), is normally plotted against the dimensionless profile length,
46H

defined as the ratio between the position of the camber in the flow direction and the
chord, x / l .
The values of CP are always positive on the pressure (lower) side of the profiles and
negative on the suction side. As expected, the CP distributions agree with the results
depicted by contour plots in Figure 5-15: at all sections, the Design III profiles are
47H

characterized by smaller CP values compared with the reference profiles, indicating


smaller losses through diffusion in the flow around the profiles and, implicitly, higher
performance for the x = 2 shapes. There is one exception, however: at the hub section, CP
for Design III is higher in the region corresponding to the trailing edge, due to the forced
outlet angle, i.e. the maximum flow rate condition and, thus,  2  90 deg .

85
5. Numerical flow analysis
4B

Figure 5-17 Dimensionless velocity distribution

Finally, the performance of the investigated profiles can be analyzed also from the
velocity distribution point of view, as shown in Figure 5-17. To maintain the consistency
48H

of the dimensionless analysis, a velocity coefficient is defined as the ratio of the flow
velocity around the profile to the velocity upstream of the cascade, w / w1 . At all sections,
the velocity coefficients corresponding to Design III indicate smooth variations along the
profile length, with no sudden gradients, compared with the reference shapes.
It can be therefore concluded that, from the profile analysis point of view, the design
strategy delivered aerodynamically superior shapes compared with the reference–
classical circular profiles with a symmetrical thickness distribution.

86
6. Experimental validation of the proposed design strategy
5B

6 Experimental validation of the proposed design


5B

strategy

The most important outcome of the previous chapters is the substantial increase in the
efficiency of the newly designed impellers compared with the reference solution. After a
careful numerical flow analysis, it was concluded in chapter 5 that, for the investigated
49H

hub ratio, Design III, corresponding to the x  2 non-free vortex flow assumption, had
the best performance according to all performance indicators analyzed. Therefore, for the
final comparison with the reference impeller, this design was suggested.
However, the efficiency curves presented were purely numeric, and even though modern
CFD tools achieve excellent flow predictions, some idealization of the flow phenomena
with a significant impact on the performance is nevertheless being made. Hence the
motivation for the following investigation was to capture experimentally the efficiency
increase of the proposed design.

6.1 Investigated impellers

As already mentioned above, two impellers were experimentally investigated, i.e. the
reference impeller, made available by the manufacturer, and a prototype of the proposed
design.
In Figure 6-1a, the reference impeller is depicted, and in Figure 6-1b, a detail of the blade
450H 451H

shape at the tip is shown.

87
6. Experimental validation of the proposed design strategy
5B

Figure 6-1 Detail of the reference impeller: a) the full rotor; b) detail of the blade shape at the tip

The impeller corresponding to Design III (x = 2) was built through the rapid prototyping
method [119]. Eight identical blades were manufactured according to the CAD model of
452H

the blade, shown in Figure 5-2, and captured in the hub. A detail of the prototype blades
453H

is shown in Figure 6-2.


45H

88
6. Experimental validation of the proposed design strategy
5B

Figure 6-2 Prototype blades

Both impellers are characterized by the same hub and tip diameters, i.e. Dhub  294mm

and Dtip  558mm , and of course, the same number of blades, z  8 .

The experimental investigation was focused on the comparison of the system


characteristic and efficiency curves of the two models, and hence the pressure and the
torque (required at the shaft) were measured for each of the investigated flow rate.

6.2 Experimental facility

The experimental analysis was carried out according to the standard norm DIN 24 163
Part 2 for measurements on axial fans [28]: 45H

89
6. Experimental validation of the proposed design strategy
5B

Figure 6-3 Standard norm for measurements on axial impellers, DIN 24 163 Part 2

During the measurements, one of the two wind tunnel facilities available at LSTM
Erlangen was used, characterized by a diameter Dtunnel  2000mm and a total length

Ltunnel  6000mm , as shown in Figure 6-4.


456H

Figure 6-4 Simplified scheme of the experimental setup

The wind tunnel is equipped with an inlet orifice where Venturi nozzles with different
diameters can be fitted, so that the required inlet flow rate is achieved. For the
investigated impellers and their corresponding system characteristics, a nozzle with
Dnozzle  400mm was chosen. The flow was driven by a centrifugal blower, which was
controlled by a frequency converter so that the rotation of the fan could be adjusted. For
flow rate control purposes, the wind tunnel is equipped with a jalousies system, and by
opening/closing the jalousies, the different flow rates on the system characteristic were
achieved.

90
6. Experimental validation of the proposed design strategy
5B

The air stream was guided through a settling chamber consisting of alternate perforated
plates and honeycombs to ensure uniform flow conditions at the outlet of the tunnel,
which was smoothly connected to the impeller inlet.
At the outlet of the wind tunnel, a pipe with a diameter D pipe  560mm and a total length

L pipe  1200mm was installed, and inside this pipe, with a tip gap of approximately 1mm

in radius, the two impellers were mounted.


In Figure 6-5, a detail of the inlet of the impeller is depicted, as it is mounted on the
457H

outlet of the wind tunnel.

Figure 6-5 Inlet of the impeller

In Figure 6-6, the entire measuring facility is presented, with the pipe attached to the
458H

wind tunnel and the impeller discharging into the ambient. The impeller is belt driven by
a motor, which can be observed below the pipe, in the measuring scheme.

91
6. Experimental validation of the proposed design strategy
5B

Figure 6-6 Wind tunnel with pipe and impeller mounted

6.3 Measured parameters

As mentioned previously, the measurements were focused on obtaining the experimental


efficiencies of the two impellers, for comparison purposes. As defined in chapter 5, the
459H

efficiency is a function of achieved pressure difference (static or total), flow rate,


rotational speed and required torque to the shaft:
  f  P, Q, n, M  (6.1)

At the nozzle inlet, both the static pressure at the wall and the ambient temperature were
measured, allowing the calculation of the inlet flow rate according to Bernoulli’s

92
6. Experimental validation of the proposed design strategy
5B

equation. The variation on the investigated flow range was achieved with the jalousies
system.
At a length of approximately 300 mm from the outlet of the tunnel and inlet in the
impellers, a second pressure tap was placed, which measured the pressure difference
between the static pressure at the wall and the ambient, Pinlet . Very close to the impeller
inlet and outlet, two additional thermocouples were placed, so that the temperature
difference achieved was calculated. Additionally, torque and rotational speed
measurements were carried out for each of the investigated flow rates.
A scheme of the measurement points is depicted in Figure 6-7.460H

Figure 6-7 Scheme of the measuring points on the experimental setup

6.4 Measuring equipment

Each pressure tap was connected to an OEI pressure scanner with pressure tubes. The
pressure measurements were made with a Höntzch pressure transducer and the pressure
scanner was used to switch from the first pressure channel (at the nozzle) to the second
(at the impeller inlet).
The torque measurements were done with an HBM T4WA-S3 torque sensor for nominal
values between 5Nm and 1kNm. The output was registered with an HBM SCOUT 55
amplifier.
Both the pressure transducer and torque amplifier were connected to integrating DVs type
DISA 55D31, with voltage output.

93
6. Experimental validation of the proposed design strategy
5B

The impellers were belt driven by an AC motor with a maxim rotation of 3500 rpm and a
nominal power of 18 kW, operating at 3000 rpm, and its control was achieved with a
frequency regulator.
The value of the rotational speed was always maintained at 3000 rpm and for these
readings a Hall-effect sensor, with maximum frequency of 10 MHz and TTL signal
output, was used.
The temperature was measured with K- Type thermocouples.

6.5 Experimental results

After the measurements on both impellers had been performed, the characteristic pressure
and torque values, depicted in Figure 6-8 were obtained. It can be observed that, for the
461H

investigated flow range, the pressure differences between the ambient and the static
pressure at the inlet of the impellers are similar, and in this respect, the behavior of the
two fans does not differ very much.

Figure 6-8 Measured static pressure difference

However, the same cannot be stated when analyzing the torque measured at the shaft,
presented in Figure 6-9.
462H

94
6. Experimental validation of the proposed design strategy
5B

Figure 6-9 Measured torque

It can be readily observed that, throughout the entire flow range, the reference impeller is
characterized by substantially higher torque values than Design III. In fact, starting from
the value of 3m3 / s the absolute increase in the measured torque of the reference is 6%,

whereas for 5m3 / s and 6m3 / s , the difference increases to 11%.


Having completed the pressure difference and torque measurements, the efficiency of the
two impellers can be calculated as the total-to-static efficiency:
Ps Q
total to  static  (6.2)
M
On plotting the calculated efficiency for the two fans as expressed in Eq. (6.2), then the
463H

chart depicted in Figure 6-10 is obtained.


46H

95
6. Experimental validation of the proposed design strategy
5B

Figure 6-10 Total-to-static measured efficiency

Design III is characterized throughout the entire flow range by a considerable increase in
the total-to-static efficiency, and for the interval 4  6 m3 / s this increase is maintained to
11% absolute difference compared with the efficiency curve of the reference. These
differences in the efficiency curves of the two impellers were, of course, expected, since
according to the previous two plots, with no significant difference in the achieved
pressure, Design III is characterized by much smaller torque values for all flow rates and,
according to Eq. (6.2), these are the two most important parameters which influence the
465H

efficiency. Essentially, these results can be generalized to the conclusion that the
proposed design strategy delivered a “torque-optimized” axial fan.
In Figure 6-11, the measured temperature differences for both fans are plotted.
46H

96
6. Experimental validation of the proposed design strategy
5B

Figure 6-11 Measured temperature differences

It can be observed that the cooling capabilities of the impellers are very similar and the
temperature gradients achieved differ by a less than 0.5o for all flow rates investigated.
This result, coupled with the efficiency graphs depicted in Figure 6-10, confirms that the
467H

same cooling effect can be achieved by both impellers, but Design III requires much less
energy, thus resulting in direct cost savings.

6.6 Validation of the results

In chapter 5 of the thesis, based on extensive numerical analysis, the optimum pressure
468H

distribution in the span-wise direction, and hence the optimum vortex-flow design
assumption, were proposed, and it was concluded that Design III was, according to the
numerical prediction, the best performing of the three investigated. These were the
premises for the experimental procedures presented in this chapter, where the reference
impeller and the suggested design were measured according to the standard DIN norm.
The experimental curves for the total-to-static efficiency showed a considerable increase
for Design III throughout the investigated flow range, thus confirming the numerical
advantage presented earlier. However, the numerical investigations of the four models
were carried out for boundary conditions completely different from the actual measuring
stand. Hence the motivation of the following analysis was to compare the numerical

97
6. Experimental validation of the proposed design strategy
5B

prediction of the behavior of the rotors, for the exact conditions of the experimental
setup, with the actual experimental data.
For this, a new CAD model, corresponding to the dimensions in Figure 6-4, was built and
469H

the flow domain of the impellers was adjusted to the same gap, i.e. 1 mm. The mesh
characteristics were kept at the values indicated in section 5.2. Data planes were placed in
470H

the exact locations as the measuring pressure taps and thermocouples.


All of the relevant parameters that influence the performance of the fan were compared
with the measured values, for both the reference impeller and Design III, and excellent
agreement was found, as shown in Figure 6-12 _ Figure 6-14.
471H 472H

Figure 6-12 Numerical and experimental static pressure curves

Figure 6-13 Numerical and experimental torque curves

98
6. Experimental validation of the proposed design strategy
5B

Figure 6-14 Numerical and experimental total-to-static efficiency curves

99
7. Integrated ideal efficiency for axial fans
6B

7 Integrated ideal efficiency for axial fans


6B

7.1 The Cordier diagram


As mentioned in the introduction to this thesis, probably the most often addressed method
in the literature with respect to the concept of maximum efficiency is the Cordier
diagram. Essentially, this diagram is based on the head ( ) and flow (  ) coefficients,
which are defined, according to Strub and Eck [32], as
473H

cm

utip
(7.1)
2Y 2c
  2t  u2
utip utip

where Yt represents the total work achieved by the rotor, and cm is the meridional flow
velocity.

100
7. Integrated ideal efficiency for axial fans
6B

axial rotor radial rotor


Figure 7-1 Different types of rotors as a function of the flow direction. Adapted from Bohl [12]
47H

For the following analysis, first the case of a radial rotor described by an outlet diameter
D2  D , as depicted in Figure 7-1, will be considered. In this case, the representative
475H

flow area is given by the area of a circle with diameter D:


 D2
A (7.2)
4
where utip represents the peripheral velocity at the exit of the rotor; in the case of radial
impellers, this is given by the peripheral velocity calculated at the tip section:
utip   Dn (7.3)

The meridional flow velocity is given by the ratio between the flow rate and the flow
area:
Q 4Q
cm   (7.4)
A  D2
On inserting Eqs. (7.3) and (7.4) into Eq. (7.1), the expression for the flow coefficient
476H 47H 478H

becomes
4Q
 (7.5)
 D3n
2

From Eq. (7.5), a first formulation of the outlet diameter of a radial impeller can be
479H

derived:

101
7. Integrated ideal efficiency for axial fans
6B

D
 4Q  3 (7.6)
2 1 1
 n 3 3 3

The head coefficient in Eq. (7.1) can be written as 480H

2Yt
 (7.7)
 D 2n2
2

From Eq. (7.7), a second expression for the outlet diameter of the impeller can be
481H

derived:
1

D
 2Yt  2 (7.8)
1
 n 2

Of course, Eqs. (7.6) and (7.8) have to be equal, and hence482H 483H

1 1
 4Q  3 
 2Yt  2 (7.9)
2 1 1 1
 n  3 3 3
 n 2

From Eq. (7.9), the following formulation for the rotational speed n can be extracted:
48H

3 1

n
 2Yt  4  2 1 (7.10)
1 3 1
 4Q    2 4 2

Cordier [21] defined the second ratio in Eq. (7.10) as the “speed” number,  :
485H 486H

1
2
 3
(7.11)
 4

Inserting Eq. (7.11) into Eq. (7.10): 487H 48H

3
 2Yt  4 1 2n  Q
n 1
 1
  3
(7.12)
 4Q  2
 2  2Yt  4

Returning to the definitions of the pressure and flow coefficients as expressed in Eq. (7.1) 489H

, and switching the parameter of interest from the diameter D to the rotational speed n,
then the following identity is obtained:
1
4Q

 2Y  2 (7.13)
 D 3 2 1
 D 2

102
7. Integrated ideal efficiency for axial fans
6B

From Eq. (7.13), the following expression for the diameter is derived:
490H

1 1
2Q 2  4 1
D 1 1 1
(7.14)
 2Yt    4 2 2

Similarly to the speed coefficient, Cordier defined a “diameter” number  as


1
 4
 1
(7.15)
 2

Inserting Eq. (7.15) into Eq. (7.14), the following expression for the diameter number can
491H 492H

be established:
2Yt 
4 2
D (7.16)
Q 2
The expressions of  and  in Eqs. (7.11) and (7.16), respectively, are employed in the
493H 49H

Cordier diagram, as shown in Figure 7-2. Cordier found that for the optimum pair of
495H

speed and diameter numbers  opt ,  opt  , any impeller is characterized by its highest

efficiency. On this diagram, the points corresponding to the low-pressure fans were
identified as being located mostly in the upper-half, i.e.  opt  0.3 . By fitting a trend line

to cover all these points, the curve marked in Figure 7-2 was obtained, and this was
496H

termed the “probable” curve of maximum efficiency for axial fans. At this point, it
becomes very interesting to see how close to this curve, indicated by Cordier’s results for
the low-pressure range, the performance of Design III (x = 2) is at the design point. The
corresponding values of  and  for Design III are calculated according to Eqs. (7.11) 497H

and (7.16), respectively, using the characteristic fan dimensions and measured static
498H

pressure difference, as presented in Table 7-1. 49H

103
7. Integrated ideal efficiency for axial fans
6B

Design III characteristics   0.19   0.3   1.71   1.07


Dtip  558mm
Dhub  294mm
n  3000rpm
Qdesign  4m3 / s

Measured head: Ps  1150 Pa


(see Figure 6-8)
50H

Table 7-1 Calculated speed and diameter numbers for Design III

Figure 7-2 The Cordier diagram

104
7. Integrated ideal efficiency for axial fans
6B

On placing Design III on the Cordier diagram, according to the calculated values of the
speed and diameter numbers indicated in Table 7-1, it can be immediately observed that
501H

its efficiency lies exactly on the estimated curve of maximum efficiency, between the
curves for 80 and 85 % efficiency.
However, due to the very similar pressure differences measured for both impellers, i.e.
reference and Design III, as indicated in Figure 6-8, the position of the first impeller in
502H

the Cordier diagram coincides with that of the latter:


 reference  1.73

 reference  1.04
Hence, at least from the Cordier diagram point of view, there is no noticeable difference
in the efficiencies of the two rotors, since they are both characterized by the same values
of the speed and diameter numbers, and they are already located on the curve of
maximum efficiency. Therefore, just from this perspective, there would have been no
need to optimize the reference, and this is definitely contradicted by the results presented
so far. This diagram uses only the rotor characteristic dimensions and operating
parameters, and there is no reference to crucial parameters for the performance of the
impeller, such as the torque. For this reason, for two axial fans with identical dimensions,
which deliver similar pressures at the considered flow rate, no difference in their
performance is obtained as far as the Cordier diagram is concerned. However, by
calculating an appropriate performance indicator, i.e. the total-to-static efficiency in
Figure 6-10, the advantage of one design over the other can be observed, and thus the
503H

need for optimization of the reference.


At this stage of the analysis, some inconsistencies of the Cordier diagram, at least from
the point of view of axial impellers, should be discussed.
First, in the original publication of Cordier, there is no specification of which type of
pressure was being measured, either static or total. However, since it is common for fan
measurements, it can be assumed that it is either the total pressure or the total-to-static
pressure. The total pressure is usually obtained by measuring the total-to-static pressure
and by adding then the dynamic pressure at the exit of the volute. This dynamic
component is calculated as the ratio of the flow rate to the cross section area. In doing so,
the swirl, which is the dominant component of the flow velocity, is neglected. Therefore

105
7. Integrated ideal efficiency for axial fans
6B

the total pressure calculated such will not differ much from the measured total-to-static
pressure. From the above considerations it is clear that any performance indicator should
take into account the actual measured parameters and defined accordingly.
Second, the definition of the flow coefficient for axial impellers employed in this diagram
does not take into account the integral properties of these flow machines. As suggested
by Bohl [12], rather different expression of  should be employed, which takes into
504H

consideration the fact that the characteristic flow area in this case is given by the ring
formed between the hub and the tip sections (see Figure 7-1), and not by the circle with
50H

the tip diameter, as it is in the case of the radial rotors:


4  Dt2  Dh2 
Aaxial  (7.17)
4
The flow coefficient should be then calculated as
4Q
axial  (7.18)
 2  Dt2  Dh2  Dt n

Due to such inconsistencies, the concept of maximum efficiency for axial impellers
cannot be properly quantified by the Cordier diagram. Hence the motivation of the
following analysis is to derive a valid formulation of the ideal efficiency of axial fans,
considering all crucial parameters that influence the performance of such devices.

7.2 Ideal efficiency for axial fans


In the previous section, it was mentioned that the Cordier diagram, as presented in Figure
506H

7-2, links the optimum operating conditions, i.e. flow rate and specific head, with the
optimum diameter and rotational speed, and the intersection reveals the point of
maximum efficiency. Theoretically, any single-stage turbomachine can be plotted on the
Cordier diagram with the help of the dimensionless speed and diameter numbers, as
introduced in Eqs. (7.11) and (7.15):
507H 508H

106
7. Integrated ideal efficiency for axial fans
6B

 P 
2 t  1

 opt 

Dtip
4     4
1
2 Q2 2

1
Q  2
 opt  2  n 3
 3
 Pt  4
4
2  
 
These numbers are formulated based on the flow (  ) and head ( ) coefficients, defined
at the impeller outlet. In the case of the radial rotor, as depicted exemplarily in Figure
509H

7-3, the impeller outlet coincides with the tip section, and thus
cm

utip
(7.19)
2Y 2c
  2t  u2
utip utip

Figure 7-3 Typical velocity diagrams for a radial impeller. Adapted from Epple [37]
510H

However, this is not the case for the axial impellers, where the flow particles enter and
leave the blade at the same radius, and hence the peripheral velocity is constant for a
given blade section:
u1  u2  2 rn (7.20)
In fact, for axial impellers (Figure 7-1b), considering only the peripheral velocity at the
51H

tip is section is an incorrect evaluation, and instead one should calculate the
dimensionless flow and pressure with an integrated value of u over the entire flow area of
the impeller, which corresponds to the ring area between the hub and the tip diameters:

107
7. Integrated ideal efficiency for axial fans
6B

rt
1

uaxial  u 2 rdr (7.21)

 rt  rh2
2
rh

As a result, Eq. (7.19) will be rewritten for the particular case of the axial impeller, at one
512H

blade section:
cm
sec tion  (7.22)
u
2cu 2
 sec tion  (7.23)
u
The expression for the head coefficient can be further developed according to the velocity
diagrams presented in Figure 4-1: 513H

2  u  wu 2   w   w 1 
 sec tion   2 1  u 2   2 1  m  (7.24)
u  u   u tan  2 

Inserting Eq. (7.22) into Eq. (7.24), the following expression is obtained:
514H 51H

 1 
 sec tion  2 1  sec tion  (7.25)
 tan  2 

Let us consider the maximum reaction, at the cascade level, according to Euler’s
equation, defined as the ratio of the static pressure difference that can be achieved to the
total pressure difference:
Ps
Rcascade  (7.26)
Pt
Considering the expressions in Eqs. (4.3) and (4.8), then the reaction expressed in Eq.
516H 517H

(7.26) can be rewritten as


518H

1 u 2  wu22 1 u  wu 2 1  wu 2 
Rcascade    1  (7.27)
2 u  u  wu 2  2 u 2 u 

According to the outlet velocity triangle in Figure 4-1: 519H

wm
wu 2  (7.28)
tan  2
This means that

1 w 1  1 1 
Rcascade  1  m   1  sec tion  (7.29)
2 u tan  2  2  tan  2 

108
7. Integrated ideal efficiency for axial fans
6B

According to Eq. (7.29), the reaction of an axial cascade is always higher than 1/2, and
520H

for very small outlet blade angles, its value approaches infinity. It can be immediately
concluded that the reaction is not really a precise performance indicator, and it cannot be
employed for the determination of a theoretical limit of performance, at least from the
point of view of axial impellers. Instead, different performance indicators have to be
considered, and for this it is very useful to analyze the problem from a practical
perspective.
In the previous chapter, detailed information about the standard measuring techniques for
axial fans was given. It was shown that, in reality, the measured pressure was the
difference between the static pressure at the wall and the ambient pressure, i.e. a static
pressure difference. This statement can be confirmed also from the numerical simulations
point of view, and for exemplification, the velocity streamlines behind the impeller are
depicted in Figure 7-4. It can be readily observed that even at the outlet of the considered
521H

flow domain (approximately 2 m length after the impeller) the swirl is still not decaying,
c22
this swirl being the dominant component in the dynamic pressure  , where
2
c22  c22u  c22m .

Figure 7-4 Swirl development after the impeller

109
7. Integrated ideal efficiency for axial fans
6B

If one wishes to measure the total pressure difference, as it is included in the various
definitions of efficiency, then also the dynamic component (and hence the swirl) would
have to be measured, and this, from the practical point of view, would be very difficult.
In this sense, a reconsideration of the expression for the static pressure difference is
useful, according to the rig measurements conditions, by subtracting the dynamic
component from the total pressure, Epple [36]: 52H


Ps ,rig  Pt  c22 (7.30)
2
The ideal efficiency, following this consideration, can then be expressed as:
Ps ,rig 1 c22
cascade   1 (7.31)
Pt 2 ucu 2
On further equating Eq. (7.31) according to the velocity diagrams, then
523H

1 wm2 1 cu 2
cascade  1   (7.32)
2 ucu 2 2 u
Equation (7.32) can be written in terms of the dimensionless flow and pressure
524H

coefficients, as defined in Eqs. (7.22) and(7.23): 52H 526H

sec
2

cascade  1  tion
 sec tion (7.33)
 sec tion 4
The next step is to express the pressure coefficient in terms of the flow coefficient:
cu 2 u  wu 2  w   w 1   sec tion 
 sec tion  2 2  2 1  u 2   2 1  m   2 1   (7.34)
u u  u   u tan  2   tan  2 
Equation (7.33) then becomes
527H

1 sec 2
1  
cascade  1  tion
 1  sec tion  (7.35)
2  sec tion  2  tan  2 
1  
 tan  2 
Equation (7.35) reveals the theoretical efficiency that can be achieved by an axial cascade
528H

and it is a direct function of the flow coefficient (flow area and flow rate implicitly) and
the value of the outlet blade angle, at the specified cascade radius. To extrapolate from
such a formulation, valid only for a blade section, to the full impeller, an integration
similar to Eq. (7.21), over the entire flow area, is required:
529H

110
7. Integrated ideal efficiency for axial fans
6B

rt
1

 axial  cascade 2 rdr (7.36)

 rt  rh2
2
rh

However, this is associated with the difficulty of the variation of the outlet blade angle
with the radius. For the proposed design strategy, this variation is given by Eq. (4.31): 530H

2 n  2   rtip   
x
1
  h
1.35
  r   x  r  r  1 rh2 
tan  2 r rwm    rhub   

On substituting this into the expression for the cascade efficiency, as defined in Eq.
(7.35), the analytical integration with the radius to the overall efficiency of the impeller
531H

becomes impossible, due to the complicated formulation.


The solution to this problem is to analyze, for known impeller specifications, the
variation of the efficiency with the flow rate, for averaged angle values, between the hub
and the tip sections.
For the prototype considered, an average value between the calculated outlet blade
angles, according to Eq. (4.31), where x = 2, was considered:
532H

radius [mm] 147 187 227 280

2 [o] 90 49.68 36.1 29.63

2, average [o] 51

Table 7-2 Averaged angle for Design III

Equation (7.36) then becomes


53H

 
 
1 
rt
1  1 sec
2
sec tion
axial 

 rt  rh2 2
 r  2 
1  tion
 1 
sec tion  2  tan  2,average
  2 rdr (7.37)
h  
1  
  tan  2,average  
 
The corresponding curve for the integrated efficiency according to Eq. (7.37), for the 534H

investigated flow range, is depicted in Figure 7-5. 53H

111
7. Integrated ideal efficiency for axial fans
6B

Figure 7-5 Design III ideal efficiency

Before commenting on the performance of Design III with respect to the integrated
efficiency curve, first some remarks should be made. For this, it is very useful to identify
on the performance plot three intervals, covering the range of small flow rates, the design
flow rate, and the higher flows range, respectively, as indicated in Figure 7-5.
536H

In the first interval, it can be seen that the integrated ideal efficiency curve starts from
0.5, whereas the measured performance of Design III indicates that for zero flow rate the
efficiency would be zero.
Similarly, in the third interval, the slope of the ideal curve decreases very slowly, while
the measured efficiency drops very suddenly, shortly after the design interval.
Such large differences, which characterize both the first and third intervals, are due to a
series of losses which inevitably appear during normal fan operation, and are not
accounted for in the case of the ideal machine. According to Epple [37], such losses
537H

include:
 shock losses, which appear as soon as the system operates away from the design
point
 friction losses, which account for the energy dissipation due to the contact of the
fluid with solid walls
 mechanical losses, caused by disc friction or bearing losses.

112
7. Integrated ideal efficiency for axial fans
6B

The mechanical losses do not influence the pressure of the system, and hence the
hydraulic power. However, they impact heavily on the power input, i.e. the power to the
shaft. Accordingly, the measured efficiency of the system, as expressed in Eq. (5.1), will
538H

considerably differ from the ideal value.


Because of such losses, differences between the ideal curve and the measured efficiency,
as in the first and the third intervals, cannot be avoided, no matter what the fan system.
Focusing on the middle interval and thus on the design point and the regions around it, it
can observed that, for the design flow rate, i.e. Qdesign  4m3 / s , the efficiency of Design

III is less than 9% (absolute percent) below the ideal value, indicating that the considered
design calculations and assumptions delivered a design solution in close proximity to the
ideal machine, characterized by an infinite number of blades, infinitely small thickness,
and inviscid flow (these are the assumptions under which the Euler equation of
turbomachinery is formulated).
By considering the integral properties of axial impellers, it was shown that the efficiency
expectations suggested by the Cordier diagram for the investigated class of impellers are
not realistic, and cannot be achieved, not even analytically, due to several inconsistencies
with regard to the nature of the flow in such machines. A correct formulation of the
maximum/ideal efficiency concept for axial fans should be the result of the integration,
over the entire flow area, of the performance of each individual cascade, and the present
analysis suggests one analytical method to do so.
It can be concluded, based on the performance plot depicted in Figure 7-5, that the
539H

proposed layout and design strategy delivered an efficient design model for axial fans, as
initial solution to the existing optimization problem, based on the correct theoretical
treatment of the flow in such flow machines, on the one hand, and also incorporates a
series of innovative design assumptions delivering high-efficiency blades, on the other.
Without any further improvements (related to geometry or by employing genetic
algorithms), the outcome is a fan already operating close to its theoretical maximum
achievable performance.

113
8. Conclusions and outlook
7B

8 Conclusions and outlook


7B

In the present work, a layout and design strategy was developed and applied to axial
ventilators/blowers, focused on delivering high-performance design solutions. This
strategy may be employed as an optimization tool in this engineering field. Essentially,
this procedure is a combined inverse–direct design method, which computes the optimum
blade profiles, according the fan specifications and operational requirements, and does
not make use of any predefined profiles, such as airfoils.
Central to the proposed design strategy was the so-called “Design Solver”, in which a
mathematical routine was implemented to solve the flow in three consecutive steps: one-
dimensional (mean-line calculations), two-dimensional (cascade calculation of the blade),
and three-dimensional (meridional analysis).
Most design strategies in this field rely entirely on the extensive data on airfoil profiles
and make use almost exclusively of the free-vortex flow assumption. The present work
was concerned with finding, for given fan dimensions, the optimum vortex-flow
consideration, and the possibility of a non-free vortex flow was also included in the
design process. Moreover, according to the vortex assumption, parameterization of the
pressure distribution in the span-wise direction was possible, which later dictated the
outlet blade angle variation with the radius. All new designs were derived on the basis of
the so-called “reference impeller”, an axial fan with a hub ratio of 0.5, currently used in
the automobile industry for cooling purposes.
Thorough numerical investigation showed that for such fan application (high-flow and
low-pressure), the best performance during operation was achieved by the design
corresponding to the non-free vortex flow assumption, i.e. Design III. Moreover,
compared with the efficiency curves for the reference, substantial increases, of up to

114
8. Conclusions and outlook
7B

18%, were obtained, thus proving the optimizing potential of the suggested layout and
design method.
The advantage of the proposed strategy came about also from aerodynamic analysis
performed on the reference profiles and Design III profiles, and it was shown, by means
of dimensionless pressure and velocity plots, that the latter cause less diffusion in the
flow around them, and thus smaller losses.
Modern Computational Fluid Dynamics tools achieve excellent flow predictions, and due
to time- and cost-reduction considerations, experiments are nowadays being replaced
more often by such numerical methods. However, for validation purposes, the reference
impeller, which was made available by the manufacturer, Alu Automotive GmbH, and a
prototype of Design III were investigated experimentally according to the standard DIN
norm. The relevant parameters for the efficiency were measured, i.e. the static pressure
and the torque required at the shaft to drive the system. It was concluded that although
the two impellers produce similar pressures, the torque required to drive the Design III
impeller was substantially reduced compared with the reference. Of course, this was
immediately observed in the total-to-static efficiency curves, where an absolute increase
of 11% for Design III was obtained. After analyzing the cooling capabilities of the two
impellers, it was observed that similar cooling effects were achieved by both fans, but
with considerable less consumption by Design III, resulting in immediate decreases in
energy and costs.
Another aspect central to this work was the analytical formulation of the integrated ideal
efficiency for axial fans. This concept was derived as a response to the incapacity of the
classical Cordier diagram to predict the actual performance of axial impellers operating in
the low-pressure regimes, due to inconsistencies in the definitions employed. The
suggested analytical formulation can be extended to any type of axial impeller of known
dimensions and design characteristics, thus offering a reliable method to determine the
maximum performance achievable by any fan.
It was shown that, at the design point, the proposed layout and design strategy delivered a
fan operating already in the close proximity to the ideal machine, quantified by the
integrated ideal efficiency curve. This result represents a fully optimized impeller and
underlines the validity of the proposed layout and design method. The method is

115
8. Conclusions and outlook
7B

essentially based on the correct theoretical treatment of the flow in axial fans, and further
improvements in the efficiency of the proposed design can be achieved by employing
genetic algorithms [104], or by improving the geometry of the system, i.e. hub or shroud
540H

[78], thus bringing the impeller closer to the ideal machine. It has to be kept in mind,
541H

however, that these latter methods are more costly and require a good starting solution.
The latter can be given by the present method.
The presented layout and design strategy, although applied to axial fans, has potential to
be employed for other types of axial impellers operating under different pressure regimes,
such as the axial compressors. The proposed design solver handles flow calculations
under both frictionless flow and flow with friction hypotheses. In the present work, the
frictionless flow assumption was employed since it was found appropriate for the case of
low-pressure axial fans. However, for different impeller application, the second
assumption, i.e. design for flow with friction, might be found more effective. In this
respect, a new research venue could determine optimum values for the gliding angle  ,
as function of the different parameters influencing the performance of the impeller, on the
one hand, but also as function of the impeller application, on the other.

116
Bibliography

[1] Abbot, I. H., Von Doenhoff, A. E., “Theory of wing sections. Including a summary
of airfoil data”, Dover Publications, INC., New York, 1959
[2] Ackert, J., “Zum enwurf Dichtstehendes Schaufelgitter”, Schweiz. Bauzeitung, 120,
No.9, 1942
[3] Aksel, Mehmet H., “Notes on Fluid Mechanics”, volume II, version 2.1,
Department of Mechanical Engineering, Middle East Technical University, 2000
[4] Albuquerque, R. B. F., Manzanares-Filho, N., and Oliveira, W., “Conceptual
Optimization of axial-flow hydraulic turbines with non-free vortex design”, Proc. IMech
221, Part A: J Power and Energy, pp. 713-725, 2007
[5] Anderson, John D, “Fundamentals of Aerodynamics”, McGraw-Hill Book
Company, New York, 1984
[6] Aungier, R. H., “Axial-flow compressors. A strategy for aerodynamic design and
analysis”, ASME Press, New York, 2003
[7] Bakir, F., Kouidri, S., Belamri, T., and Rey, R., “Flow study in the impeller–
diffuser interface of a vaned centrifugal fan”, ASME Journal of Fluids Engineering, 127,
pp. 495-502, 2005
[8] Bakir, F., Kouidri, S., Belamri, T., and Rey, R., “On a General Method of Unsteady
Potential Calculation Applied to the Compression Stages of a Turbomachine–Part I:
Theoretical Approach”, ASME Journal of Fluids Engineering, 123, pp. 780-786, 2001
[9] Balje, O., “Turbomachines. A guide to design, selection, and theory”, John Wiley
& Sons, 1981
[10] Beiler, M. G., and Carolus, T. H., “Computation and measurement of the flow in
axial flow fans with skewed blades”, ASME Journal of turbomachinery, 121, pp. 59-66,
1999
[11] Benini, E., and Toffolo, A., “A parametric method for optimal designs of two-
dimensional cascades”, Proc. Inst. Mech. Engrs., Part A, Journal of Power and Energy,
215, pp. 465-473, 2001
[12] Bohl, W., Elmendorf, W., “Stroemungsmaschinen 1“, Vogel Buchverlag, 2004

117
[13] Brennen, Ch. E., “Hydrodynamics of pumps”, Chapter 3 “Two-dimensional
performance analysis”, Concepts NREC, 1994
[14] Bolton, A. N., “Installation effects in fan systems”, Proceedings of the Institution of
Mechanical Engineers, Part A, 204, pp. 201-215, 1990
[15] Bouras, B., Karagiannis, F., Chaviaropoulos, F., and Papailiou, K. D., “Arbitrary
blade section design based on viscous considerations. Blade optimization”, ASME
Journal of Fluids Engineering, vol. 118, pp. 364-369, 1996
[16] Carolus, Th., ”Ventilatoren”, Teubner, 2003
[17] Carter, A. D. S., and Hughes, H. P., “A theoretical investigation into the effect of
profile shape on the performance of airfoils in cascade”, 1946
[18] Chen, B., and Yuan, X., “Advanced aerodynamic optimization system for
turbomachinery”, ASME Journal of turbomachinery, vol. 130, pp. 021005-1–021005-12,
2008
[19] Cheng, K. Y., „Horizontal axis wind turbine and profile aerodynamics“, PhD
Thesis, University of New Castle upon Tyne, 1981
[20] Chung, T. J., “Computational fluid dynamics”, Cambridge University Press, 2002
[21] Cordier,O, “Ähnlichkeitsbedingungen für Strömungsmaschinen”, BWK Bd. 6,
No.10, 1953
[22] Csnady, G. T., “Theory of turbomachines”, McGraw-Hill, New York, 1964
[23] Damle, S., Dang, T. Q., Stringham. J., and Razinsky, E., “Practical use of three-
dimensional inverse method for compressor blade design”, ASME Journal of
turbomachinery, Vol. 121, pp. 321-325, 1999
[24] Dang, T., and Isgro, V., “Euler-based inverse method for turbomachinery blades.
Part 1: 2D cascades” , Am. Inst. Aeronaut. Astronaut. J., 33, pp. 2309-2315, 1995
[25] de Vito, L., Van den Braembussche, R. A., and Deconinck, H., “A novel two-
dimensional viscous design method for turbomachinery blading”, ASME Journal of
Turbomachinery, Vol. 125, pp. 310-316, 2003
[26] Demeulenaere, A., and Van den Braembussche, R., „Three-dimensional inverse
method for turbomachinery design“, ASME Journal of Turbomachinery, 120, pp. 247-
255, 1998

118
[27] Denton, J. D., “An improved time-marching method for turbomachinery flow
calculation”, ASME International Gas Turbine Conference, Wembley, England, Paper
82-GT-239, 1982
[28] DIN – Taschenbuch 254, “Ventilatoren”, Beuth, 1992
[29] Dixon, S. L., “Fluid Mechanics, Thermodynamics of Turbomachinery”, Fourth
Edition, Pergamon Press, 1994
[30] Dugao, Z., Jiang, Z., and Song, J., “Optimization of an axial flow-fan used for
mining local-ventilation”, Comput. Ind. Eng., 31, No.3, pp. 691-696, 1996
[31] Durst, F., “Grundlagen der Strömungsmechanik”, Springer, 2006
[32] Eck, B., “Fans“, 1st English Edition, Pergamon Press, 1973, pp. 234 – 251
[33] Eck, B., “Ventilatoren-Entwurf und Betrieb der Radial-, Axial- und
Querstromventilatoren“, 5th Edition, Springer-Verlag, Berlin 1991
[34] Eckert, B., “Axial- und Radialkompressoren”, Springer, Berlin, 1953
[35] Egorov, I. N., Shmotin, Y. N., and Fedechkin, K. S., “Increasing of axial fan
efficiency basing on optimization strategy”, 6th World Congresses of structural and
multidisciplinary optimization, Rio de Janeiro, 2005
[36] Epple, Ph., Ilic, C., and Durst, F., “Combined Analytical and Numerical Radial Fan
Performance Optimization”, Conference on Modeling Fluid Flow, pp. 1110- 1117, 2006
[37] Epple, Ph, “Modern design and applications for radial fans”, PhD thesis, Institute of
Fluid Mechanics LSTM Erlangen, 2008
[38] Fan, H. Y., “An inverse design method of diffuser blades by genetic algorithms”,
Proc. Instn. Mech. Engrs., Part A, Journal of Power and Energy, 212, pp. 261-268, 1998
[39] Ferziger, J. H., Peric, M., “Computational Methods for Fluid Dynamics”, 3rd
Edition, Springer, 2002
[40] Fister, W., “Fluidenergiemaschinen”, volume 1, Springer-Verlag, 1984
[41] Gbadebo, S. A., Cumpsty, N. A., and Hynes, T. P., “Control of three-dimensional
separations in axial compressors by tailored boundary layer suction”, ASME Journal of
Turbomachinery, Vol. 130, pp. 011004-1-8, 2008
[42] Gbadebo, S. A., Cumpsty, N. A., and Hynes, T. P., “Interaction of tip clearance
flow and three-dimensional separations in axial compressors”, ASME Journal of
Turbomachinery, Vol. 129, pp.99.679-685, 2007

119
[43] Gorla, R., and Khan, A., “Turbomachinery. Design and Theory”, Marcel Dekker,
INC., New York, 2003
[44] Gostelow, J. P., “Cascade aerodynamics”, Pergamon Press Ltd., New York, 1974
[45] Goto, A., Nohmi, M., Sakurai, T., and Sogawa, Y., “Hydrodynamic design system
for pumps based on 3D CAD , CFD and inverse design method”, ASME Journal of
Fluids Engineering, 124, pp. 329-335, 2002
[46] Hawthorne, W. R., and Novak, R. A., “The aerodynamics of turbomachinery”,
Annual Reviews Fluid Mechanics, 1963
[47] He, L., and Sato, K., “Numerical solution of incompressible unsteady flows in
turbomachinery”, ASME Journal of Fluids Engineering, 123, pp. 680-685, 2001
[48] Hellmich, B., and Seume, J. R., “Causes of acoustic resonance in a high-speed axial
compressor”, ASME Journal of Turbomachinery, vol. 130, pp. 031003-1–31003-9, 2008
[49] Horlock, J. H, and Lakshminarayana, B., “Secondary flows: theory, experiment,
and application in turbomachinery aerodynamics”, Annual Reviews Fluid Mechanics,
1973
[50] Horlock, J. H., “Flow through a repeating stage in an axial compressor: a
reconsideration”, Technical Brief, ASME Journal of Turbomachinery, Vol. 126, pp. 677-
679, 2004
[51] Howell, A. R., “Design of axial compressors”, Proc. Instn. Mech. Engrs., 153, 1980
[52] Hughes, W. F., and Brighton, J. A., “Schaum’s Outlines of Fluid Dynamics” Third
Edition, McGraw-Hill Professional, 1999
[53] Japikse, D., and Baines, N. C., “Introduction to turbomachinery”, Concepts ETI,
Inc and Oxford University Press, 1997
[54] Jarret, J. P., Dawes, W. N., and Clarkson, P. J., “An approach to integrated multi-
disciplinary turbomachinery design”, ASME Journal of Turbomachinery, vol. 129, pp.
488-494, 2007
[55] Jovanović, J., “The statistical dynamics of turbulence”, Springer Verlag, 2004
[56] Jiang, J., and Dang, T., “Design method for turbomachines blades with finite
thickness by the circulation method”, ASME Journal of Turbomachinery, 119, pp. 539-
543, 1997

120
[57] Kahane, A., “Investigation of axial-flow fan and compressor rotors designed for
three-dimensional flow”, Technical Note 1652, NACA Langley Memorial Aeronautical
Laboratory, 1948
[58] Kaplan, V., and Lechner, A., “Theorie und Bau von Turbinen – Schnelläufern“,
Oldenburg, Munich, 1931
[59] Kerrebrock, J. L., Epstein, A. H., Merchant, A. A., Guenette, G. R., Onnee, J. F.,
Neumayer, F., Adamczyk, J. J., and Shabbir, A., “Design and test of an aspirated
counter-rotating fan”, ASME Journal of Turbomachinery, Vol. 130, pp. 021004-1-8,
2008
[60] Korakianitis, T., “Hierarchical development of three direct-design methods for two-
dimensional axial turbomachinery cascades”, ASME Journal of turbomachinery, Vol.
115, pp. 314-324, 1993
[61] Kouidri, S, Belamri T., Bakir, F., and Rey, R., “On a general method of unsteady
potential calculation applied to compressor stages of a turbomachine. Part I: Theoretical
approach”, ASME Journal of Fluids Engineering, vol. 123, pp. 780-786, 2001
[62] Kouidri, S, Belamri T., Bakir, F., and Rey, R., “On a general method of unsteady
potential calculation applied to compressor stages of a turbomachine. Part II:
Experimental comparison”, ASME Journal of Fluids Engineering, vol. 123, pp. 787-792,
2000
[63] Lakshminarayana, B., “Fluid dynamics and heat transfer of turbomachinery”, John
– Wiley & Sons, Inc., 1996
[64] Launder, B. E., and Spalding, D. B., “The numerical computation of turbulent
flows”, Computer Methods in Applied Mechanics and Engineering, vol. 3, pp. 269-289,
1974
[65] Lei, V. M., Spakovszky, Z. S., and Greitzer, E. M., „A criterion for axial
compressor hub-corner stall“, ASME Journal of Turbomachinery, vol. 130, pp. 031006-
1–031006-10, 2008
[66] Leonard, O., and Van den Braembussche, R. A., “Design method for subsonic and
transonic cascades with prescribed Mach number distribution”, ASME Journal of
Turbomachinery, Vol. 114, pp. 553-560, 1992

121
[67] Lewis, R. I., “A method for inverse airfoil and cascade design by surface vorticity”,
ASME Paper 82-GT-154, 1982
[68] Lewis, R. I., “Turbomachinery performance analysis”, Arnold, 1996
[69] Lieblein, S., “Analysis of experimental low-speed loss and stall characteristics of
two-dimensional compressor blade cascades”, Lewis Research Center NASA, 1957
[70] Lieblein, S., “Incidence and deviation angle correlations for compressor cascades”,
Lewis Research Center NASA, 1959
[71] Lieblein, S., “Loss and stall analysis of compressors cascades”, Journal of Basic
Engineering, pp. 387-400, 1959
[72] Lieblein, S., Schwenk, F. C., and Broderick, R. L., „Diffusion factor for estimating
losses and limiting blade loadings in axial-flow compressor blades“, Lewis Research
Center NASA, 1953
[73] Madison, R. D., “Fan Engineering”, Buffalo Forge Co, 1949
[74] Marsh, H., “A digital computer for the through flow fluid mechanics of an arbitrary
turbomachine using a matrix method”, A.R.C.R. & M., No. 3509, 1966
[75] Mattingly, J. D., “Elements of gas turbine propulsion”, McGraw – Hill, Inc, 1996
[76] McKenzie, A. B., “The design of axial compressors blading based on tests of a low-
speed compressor”, Proc. Instn. Mech. Engrs., 194, 1980
[77] Menter, F.R., “Zonal two-equation turbulence models for aerodynamic flows”,
AIAA Paper 96-2906, 1993
[78] Neal, D. R., and Foss, J. F., “The application of an aerodynamic shroud for axial
ventilation fans”, ASME Journal of Fluids Engineering, 129, pp. 764-772, 2007
[79] Nerurkar, A. C., Dang, T. Q., Reddy, E. S., and Reddy, D. R., “Design study of
turbomachinery blades by optimization and inverse techniques”, AIAA–96–2555, 1996
[80] Northall, J. D., “The influence of variable gas properties on turbomachinery
computational fluid dynamics”, ASME Journal of Turbomachinery, vol. 128, pp. 632-
638, 2006
[81] Páscoa, J. C., Mendes, A. C., Gato, L. M. C., and Elder, R., “Aerodynamic design
of turbomachinery cascades using an enhanced time-marching finite volume method”,
Computer Modeling in Engineering and Sciences, 6, pp. 537-546, 2004

122
[82] Pascu, M. and Epple, Ph., “Numerical investigation of the validity of axial blade
design”, ASME Proceedings of the International Mechanical Engineering Congress and
Exposition, paper IMECE 2007–42960, 2007
[83] Peng, G., “A practical combined computation method of mean through-flow for 3D
inverse design of hydraulic turbomachinery blades”, ASME Journal of Fluids
Engineering, 127, pp. 1183-1190, 2006
[84] Peng, W. W., “Fundamentals of Turbomachinery”, John Wiley and Sons, 2007
[85] Pierret, S., and Van Den Braembussche, R. A., „Turbomachinery blade design
using a Navier–Stokes solver and artificial neural network“, ASME Journal of
Turbomachinery, Vol. 121, pp. 326-332, 1999
[86] Potts, I., “The importance of S-1 stream surface twist in the analysis of the inviscid
flow through the swept linear turbine cascades”, I. Mech. E., Paper C258/87, 1987
[87] Praisner, T. J., and Clark, J. P., “Predicting transition in turbomachinery. Part I: A
review and new model development”, ASME Journal of Turbomachinery, vol. 129, pp.
1-13, 2007
[88] Praisner, T. J., and Clark, J. P., “Predicting transition in turbomachinery. Part II:
Validation and benchmarking”, ASME Journal of Turbomachinery, vol. 129, pp. 14-22,
2007
[89] Railly, J. W., “Ackert method for the design of tandem cascades”, Engineer, 224
(5827), 405–416, London, 1965
[90] Ramirez Camacho, R. G., and Manzanares Filho, N., “A source wake model for
cascades of axial flow turbomachines”, Journal of the Brazilian Society of Mechanical
Science and Engineering, Vol. 27, pp.288-299, 2005
[91] Reese, H., Kato, C., and Carolus, Th., “Large-eddy simulation of acoustical sources
in a low axial-flow fan encountering highly turbulent inflow”, ASME Journal of Fluids
Engineering, 129, pp. 263 – 272, 2007
[92] Sadeghi, A., and Liu, F., “Computation of mistuning effects on cascade flutter”,
Computational Fluid Dynamics Journal, pp. 640-647, 2001
[93] Schlender, F., and Klingenberg, G., “Ventilatoren im Einsatz: Anwendung in
Geräten und Anlagen” VDI-Verlag, Düsseldorf, 1996

123
[94] Schlichting, H., “Application of boundary layer theory in turbomachinery”, Journal
of Basic Engineering, pp.1-40, 1959
[95] Schoberi, M., “Turbomachinery flow physics and dynamic performance”, Springer,
Berlin, 2005
[96] Scholz, N., “Aerodynamik der Schaufelgitter. Grunlagen, zweidimensionale
Theorie, Anwendungen“, G. BRAUN Verlag, Karlsruhe, 1965
[97] Sieverding, F., Ribi, B., Casey, M., and Meyer, M., “Design of industrial
compressor blade sections for optimal range and performance”, ASME Journal of
Turbomachinery, vol. 126, pp. 323-331, 2004
[98] Sørensen, D. N, Thompson, M. C., and Sørensen, J. N., “Toward improved rotor-
only axial fans–Part I: A numerically efficient aerodynamic model for arbitrary vortex
flow”, ASME Journal of Fluids Engineering, 122, p. 318-323, 2000
[99] Sørensen, D. N, Thompson, M. C., and Sørensen, J. N., “Toward improved rotor-
only axial fans–Part II: Design optimization for maximum efficiency”, ASME Journal of
Fluids Engineering, 122, p. 324-329, 2000
[100] Stepanoff, A. J., “Turboblowers. Theory, Design and Application of Centrifugal
and Axial Compressors and Fans”, John Wiley & Sons, Inc., 1955
[101] Strub, R. A., “Rotoren von Turbomaschinen – Leistung und Schönheit”,
Technische Rundschau Sulzer, Volume 1, 1984
[102] Thwaites, B., “Incompressible aerodynamics. An account of the theory and
observation of the steady flow of incompressible fluid past airfoils, wings and other
bodies”, Dover Publications Inc., New York, 1987
[103] Tu, J., Yeoh, G. H., and Liu, C., “Computational fluid dynamics”, Butterworth–
Heinemann, 2008
[104] Van Den Braembussche, R. A., “Modern turbomachinery component design”,
Conference on Modeling Fluid Flow (CMFF ’06), Budapest, pp. 33-47, 2006
[105] van Rooij, M. P. C., Dang, T. Q., and Larosiliere, L. M., “Improving aerodynamic
matching of axial compressor blading using three-dimensional multistage inverse design
method”, ASME Journal of Turbomachinery, Vol. 129, pp. 108-118, 2007
[106] Vavra, M. H., “Aero-thermodynamics and flow in turbomachines”, John–
Wiley&Sons, Inc., 1960

124
[107] Versteeg, H. G., and Malalaskera, W., “Computational fluid dynamics. The finite
volume method”, Longman Group Ltd., 1995
[108] Wallis, R. A., “Optimization of axial flow fan design”, Mechanical and Chemical
Engineering Transactions, The Institution of Engineers, Australia, MC4, No. 1, p. 31-37,
1968
[109] Wallis, R.A., “Axial Flow Fans”. Newnes London, 1961
[110] Weinig, F., “Die Strömung um die Schaufeln von Turbomaschinen“, 1935
[111] Wilkinson, D. H., “ A numerical solution of the analysis and design problems for
the flow past one or more airfoil or cascades”, A.R.C.R. & M., No. 3545, 1967
[112] Wright, T., “Fluid machinery. Performance, analysis, and design”, CRC Press,
1999
[113] Wu, C. H., “A general theory of three-dimensional in subsonic and supersonic
turbomachines of axial - , radial - , and mixed-flow types”, Lewis Flight Propulsion
Laboratory, Washington, 1952
[114] Wu, C. H., and Brown, C. A., “A method of designing turbomachine blades with
desirable thickness distribution for compressor flow along arbitrary stream filament of
revolution”, Lewis Flight Propulsion Laboratory, Washington, 1951
[115] Yang, H., Nuernberger, D., and Kersken, H. P., “Toward excellence in
turbomachinery computational fluid dynamics: a hybrid structured–unstructured
Reynolds-averaged Navier–Stokes solver”, ASME Journal of Turbomachinery, vol. 128,
pp. 390-402, 2006
[116] Yang, L., Hua, O., and Zhao – Hui, D., “Optimization design and experimental
study of low-pressure axial fan with forward-skewed blades”, International Journal of
Rotating Machinery, volume 2007, Article ID 85275, 2007
[117] Air property calculator: http://users.wpi.edu/~ierardi/FireTools/air_prop.html
0H

[118] ANSYS: http://www.ansys.com/


1H

[119] Hering & Schneider Rapid Prototyping: http://www.hering-schneider.de/


[120] Naca technical reports server: http://ntrs.nasa.gov/search
2H

[121] Pro/Engineer Wildfire: www.ptc.com


3H

[122] Regional Computation Centre Erlangen: http://www.rrze.uni-erlangen.de/


4H

125

You might also like