Professional Documents
Culture Documents
14 - Cooling Fan Optimization For Heavy Electric Vehicles
14 - Cooling Fan Optimization For Heavy Electric Vehicles
14 - Cooling Fan Optimization For Heavy Electric Vehicles
Electrified Vehicles
A study on performance and noise
Amir Khiabani
Daniel Acebo Alanis
iv
Abstract
Vehicle electrification plays a significant role in the effort to reduce the environmental
impact of the automotive industry. Scania is one of the leading manufacturers of
heavy vehicles which is currently moving towards a sustainable transport system
by manufacturing a new generation of heavy vehicles powered by batteries. One
of the major concerns with these vehicles is related to the noise generated by the
electric axial fans used in the cooling system. This project was conducted with the
purpose of investigating the factors that positively affect both noise and performance
in the electric fans. Based on two different blade design methods and several noise
control techniques, 11 fan models were developed. The fan models created with
design method 1 are equipped with cambered-plate blades, while the models made
with design method 2 consist of airfoil-shaped blades. Moreover, the performance of
these models was analyzed by using theoretical methods and Computational Fluid
Dynamics (CFD). In addition, two empirical approaches were used to estimate the
acoustic energy emitted by the fan models. Furthermore, the developed models
were compared with two commercially available fans. It was found that both design
methods provide similar performance in low pressure differences. On the other hand,
the efficiency and acoustic energy are influenced by the choice of the noise control
methods.
Sammanfattning
Fordonselektrifiering har en väsentlig roll i arbetet med att minska bilindustrins
miljöpåverkan. Scania är en av de ledande tillverkarna av tunga fordon som för när-
varande går mot ett hållbart transportsystem, genom att tillverka en ny generation
tunga fordon drivna med batterier. Ett stort bekymmer med dessa fordon är relat-
erat till det ljud som genereras av de elektriska axialfläktarna som används i kylsys-
temet. Detta projekt genomfördes i syfte till att undersöka de faktorer som positivt
påverkar både buller och prestanda hos de elektriska fläktarna. Baserat på två olika
bladdesignmetoder och flera brusstyrningstekniker, utvecklades 11 fläktmodeller.
Fläktmodellerna som är utformade med konstruktionsmetod 1 är utrustade med
krökformade plattor, medan modellerna som skapades med designmetod 2 består
av vingprofil blad. Dessutom analyserades prestandan för dessa modeller med an-
vändning av teoretiska metoder och strömningsmekaniska beräkningar. Ytterligare
två empiriska tillvägagångssätt användes för att uppskatta den akustiska energin
som släppts ut av fläktmodellerna. Utöver det jämfördes de utvecklade modellerna
med två kommersiellt tillgängliga fläktar. Detta visade att båda konstruktionsme-
toderna resulterar i liknande prestanda vid lågtrycksskillnader, däremot påverkas
verkningsgraden och den akustiska energin av valet av brusstyrningsmetoder.
Keywords: Electric axial fans, Fan design, Blade design, Battery Electric Vehicle,
Fan noise, Fan performance.
v
Acknowledgments
This study was carried out at Scania Group in Södertälje, Sweden. The work was
a collaboration between Vehicle Cooling & Aerodynamics and Electrical Actuators
Design departments.
We would like to take the opportunity to acknowledge everyone who supported
us to accomplish this project. First of all, we would like to extend our deepest
gratitude to Leila Shafiee and Zoltan Kardos who provided guidance and valuable
advice throughout the project. Furthermore, we are grateful to Mattias Chevalier for
taking time out of his schedule to perform all the CFD simulations. Special thanks
to Stina Johansson and Ola Hall for all the support and motivation. We would also
want to thank our examiner, assistant professor Raffaello Mariani, at KTH Royal
Institute of Technology for his help during the project. Last but not least, our deep
and heartfelt gratitude to our families for their relentless and extraordinary patience
and support over the years.
Daniel would also like to acknowledge the monetary support provided by Conacyt
to obtain his master’s degree.
vii
Contents
List of Figures xi
Nomenclature xiv
1 INTRODUCTION 1
1.1 Fan History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 THEORY 4
2.1 Fans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1.1 Axial fan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1.2 Centrifugal fan . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.3 Fan laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.4 Fan performance . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.1 Sound measurement units . . . . . . . . . . . . . . . . . . . . 8
2.2.1.1 Octave bands and weighted sound pressure levels . . 9
2.2.2 Noise in axial fans . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.3 Fan sound law . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.4 Noise control in axial fans . . . . . . . . . . . . . . . . . . . . 12
2.3 Blade Element Theory . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.1 Blade element forces . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.2 Relative flow . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3 METHODS 16
3.1 Design Method 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.1.1 Mean line calculation . . . . . . . . . . . . . . . . . . . . . . . 16
3.1.2 Outlet blade angle . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1.3 Blade shape computation . . . . . . . . . . . . . . . . . . . . . 18
3.1.4 Losses and diffusion factor . . . . . . . . . . . . . . . . . . . . 18
3.2 Design Method 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2.1 Isolated airfoil . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2.2 Cascade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
ix
Contents
4 DESIGN PROCEDURE 30
4.1 Design Method 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.2 Design Method 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5 RESULTS 38
5.1 Design Method 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.1.1 Impeller dimensions . . . . . . . . . . . . . . . . . . . . . . . . 38
5.1.2 Blade characteristics . . . . . . . . . . . . . . . . . . . . . . . 39
5.1.3 Fan dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.2 Design Method 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.2.1 Impeller dimensions . . . . . . . . . . . . . . . . . . . . . . . . 41
5.2.2 Blade characteristics . . . . . . . . . . . . . . . . . . . . . . . 42
5.2.3 Fan dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.3 Fan Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.3.1 Theoretical . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.3.2 CFD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.4 Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.4.1 Approach 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.4.2 Approach 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.4.3 Scania’s fans comparison . . . . . . . . . . . . . . . . . . . . . 50
6 DISCUSSION 52
7 CONCLUSION 55
7.1 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Bibliography 56
x
List of Figures
xi
List of Figures
xii
List of Tables
3.1 Specific sound power levels in eight lowest octave bands for axial fans 29
D.1 CFD results at/near the peak efficiency point at different rotational
speeds. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XIII
D.2 CFD performance results for model 3. . . . . . . . . . . . . . . . . . . XIV
D.3 CFD performance results for model 7. . . . . . . . . . . . . . . . . . . XIV
xiii
Nomenclature
xiv
Nomenclature
xv
1
INTRODUCTION
1
1. INTRODUCTION
electric motors replaced steam engines in order to enhance the quality of human life.
Upon the advent of electric motors, the use of axial-flow fans flourished. Many fan
manufacturing companies, such as Aerex Ltd, started to develop new generations
of axial fans for mine ventilation [2]. A new type of impeller equipped with airfoil-
shaped blades was employed. Furthermore, inlet and outlet guide vanes were used
to improve the efficiency of the fans. Axial-flow fans play a significant role in today’s
growing industries. Automotive industry is one the major business fields in which
this type of fan is highly demanded.
1.2 Background
At the present time, environmental sustainability issues are of huge concern to all
nations of the world. Vehicle electrification is one of the major solutions to con-
trol the environmental damage, including climate change due to greenhouse gases.
Scania, one of the leading heavy vehicle manufacturers, is currently focusing on
innovative technologies with the purpose of electrifying heavy duty vehicles, such
as trucks and buses, as well as infrastructures [4]. Scania’s Battery Electric Ve-
hicle (BEV) is not only compatible with the urban living environment due to low
noise pollution but also considered as an environmentally friendly product with zero
tailpipe emissions. Despite all the efforts made to reduce the noise emitted from the
vehicle by replacing the combustion engine with an electric powertrain, there are
still some components which can generate high-pitched noise. The electric axial fans
used in the cooling system of first-generation BEVs produce loud noise, especially
when running at high rotational speeds. Heavy electrified vehicles are equipped with
batteries which provide the required power to run the electric machine as well as
other components. In order to dissipate the heat from the radiators connected to dif-
ferent cooling circuits, puller electric fans are employed. These fans are responsible
for generating a high capacity of airflow to maintain proper operating temperatures
when the vehicle is travelling at lower speeds. Although this kind of fan works with
appropriate efficiency according to the supplier, its performance still does not fulfill
the requirements from an acoustic point of view, as it is not specially designed for
use in a BEV. Therefore, the main focus of this project is to explore the possibility
of improving the performance of electric fans along with reducing the associated
noise.
1.3 Objectives
The overall objective of the present work is to study the main factors that influence
the performance of a cooling fan with the purpose of optimizing the capacity of
the airflow as well as mitigating the noise generated by the fan. In order to meet
the required cooling capacity for BEV, several fan models are developed based on
two different blade design approaches. This strategy provides the possibility to
design different impellers equipped with cambered-plate and airfoil-sectioned blades.
Moreover, various efficient noise reduction methods are implemented for the purpose
of minimizing the noise emitted by the fan designs.
2
1. INTRODUCTION
The main part of the project, including the blade design, fan dimension, and noise
evaluation process, is conducted by applying suitable design methods in Matlab.
Furthermore, the theoretical results are employed to create fan models using Solid
Edge, a Computer-Aided Design (CAD) software. Then, the performance of the
models are investigated by conducting a Computational Fluid Dynamics (CFD)
analysis in STAR-CCM+. The outcome is compared with the performance parame-
ters of two different electric fans which are already used in Scania’s trucks and buses.
This study will then be used by engineers at Scania to discuss the possibilities to
optimize the performance of future electric axial fans for the use in BEV’s cooling
system.
1.4 Layout
In this section, the overall layout of the project is described and briefly presents the
upcoming chapters.
Chapter 2 covers the theoretical background related to the fundamental principles
of fan performance and noise. Furthermore, it provides information about useful
theories that can be applied in order to design a fan with high efficiency.
In the third chapter, two different design methods which are used to develop fans
with different blade shapes are introduced. Moreover, various noise and performance
estimation methods are presented in chapter 3 with the purpose of analyzing the
fan models.
The design procedures based on the two fan design methods are explained in detail
in chapter 4. This chapter provides an overview of the calculation steps which are
being applied to develop new fan models.
The results from the performance and noise estimations along with the CFD out-
comes are considered in chapter 5. In addition, the characteristics of two commer-
cially available fans are compared with the results related to the developed models.
Finally, chapters 6 and 7 provide a summary of the key outcomes and findings of
the study, and highlight relevant future work.
3
2
THEORY
2.1 Fans
A fan is a device that uses a rotating impeller with the purpose of producing airflow
by creating a pressure difference [5]. Fans can be placed in two main categories
based on flow direction [5], namely:
1. Axial
2. Centrifugal
4
2. THEORY
Based on the shape of the impeller, centrifugal fans are classified as follows:
1. Forward-curved centrifugal fans are equipped with blades which are bent in
the direction of the impeller’s rotation. They provide low noise level since they
operate at low rotational speeds and are suitable for ventilation and cleaning
service applications [7].
2. Radial blade centrifugal fans are characterised by their radial blades which
expand from the hub center without curvature. They are well suited for low-
volume flow rates, high static pressures and temperatures. This type of fan is
5
2. THEORY
the most appropriate centrifugal model for dust extraction applications [7].
3. Backward-curved centrifugal fans are equipped with blades which are curved
in the opposite direction of the impeller’s rotation. This type of fan is more
efficient compared to radial and forward-curved centrifugal fan models. Fur-
thermore, they are capable of operating with a varying static pressure [7].
4. Airfoil blade centrifugal fans are the most efficient of all centrifugal mod-
els. This type of fan provides a high volume of airflow which makes it an
appropriate candidate for industrial, mining, and biochemical applications [7].
Dependent Independent
Law No.
Variables Variables
1a Q1 = Q2 × (D1 /D2 )3 (N1 /N2 )
1b ∆p1 = ∆p2 × (D1 /D2 )2 (N1 /N2 )2 (ρ1 /ρ2 )
1c W1 = W2 × (D1 /D2 )5 (N1 /N2 )3 (ρ1 /ρ2 )
2a Q1 = Q2 × (D1 /D2 )2 (∆p1 /∆p2 )1/2 (ρ2 /ρ1 )1/2
2b N1 = N2 × (D2 /D1 )(∆p1 /∆p2 )1/2 (ρ2 /ρ1 )1/2
2c W1 = W2 × (D1 /D2 )2 (∆p1 /∆p2 )3/2 (ρ2 /ρ1 )1/2
3a N1 = N2 × (D2 /D1 )3 (Q1 /Q2 )
3b ∆p1 = ∆p2 × (D2 /D1 )4 (Q1 /Q2 )2 (ρ1 /ρ2 )
3c W1 = W2 × (D2 /D1 )4 (Q1 /Q2 )3 (ρ1 /ρ2 )
In order to determine the performance of any fan using the relations listed in Table
2.1, it is required to gather all data for another fan of the same series.
The first fan law indicates how pressure difference, volume flow rate, and power are
affected due to the change in fan diameter, rotational speed, or density. The second
fan law shows the influence of modifying pressure difference, diameter, or density on
rotational speed, volume flow rate, and power. In addition, it is possible to predict
the effect of changing density, volume flow rate, or diameter on fan power, rotational
speed, and pressure difference using the third fan law illustrated in Table 2.1.
6
2. THEORY
Normally, performance data consist of information about fan pressure rise ∆p, power
consumption W , volume flow rate Q, and efficiency η, which can be presented in
graphical form, as shown in Figure 2.4.
The initial point on the pressure curve is known as shut-off, and represents the
point of zero volume flow rate. On the other hand, free-delivery point is referred
to the point at which the volume flow rate becomes maximum [10]. The peak of
the pressure curve splits the curve into two regions. The right side of this point is
referred to as the stable section where the operation point can be selected, while the
left part of the peak point represents the stall region of the pressure curve where
the fan is no longer efficient [11].
The performance curve of a given fan depends on the geometry and speed of the
fan, but the operation point depends on the system (installation). Each installation
has a resistance for the required volumetric flow rate [3].
7
2. THEORY
2.2 Noise
2.2.1 Sound measurement units
Noise can be expressed in terms of Sound Power Level (SWL or LW ) and Sound
Pressure Level (SPL or Lp ). Sound power level is logarithmically compared with a
reference power, the picowatt (10−12 watt), while sound pressure level is logarith-
mically compared with a reference pressure of 2·10−5 Pa. The unit to express both
quantities is the decibel scale, dBW for sound power level and dB for sound pressure
8
2. THEORY
level. The sound power and pressure level can be determined as follows:
W
LW = 10 log10 , (2.1)
W0
!
p
Lp = 20 log10 , (2.2)
p0
where W is the sound power of the noise generating equipment, W0 the reference
power, p sound pressure of the noise, and p0 the reference pressure. The main
difference between Equations (2.1) and (2.2) is that the distance is accounted for
when using SPL. Equation 2.3 can be used to relate both measurements,
Qθ
Lp = LW + 10 log10 , (2.3)
4πr2
where r is the distance from the source (m) and Qθ is the directivity factor of the
source in the direction of r. The value of Qθ changes depending on the position of
the fan in the room, as explained in [3].
9
2. THEORY
As summarized in Figure 2.7, blade forces can be steady or unsteady. The former
10
2. THEORY
are those exerted by the fan on a uniform stationary flow which is known as Gutin-
Noise. While the latter, in contrast, occur when the fan operates in a stationary
but non-uniform flow field. Both forces contribute to the overall sound radiation,
but at low speeds, steady forces noise is negligible compared to noise generated
by unsteady forces [13]. A non-uniform flow can be produced by obstructions at
locations upstream of the fan, such as radiator, rods, guide vanes, duct bends, as
well as an asymmetric position of the intake with respect to adjacent floors and
walls. This means that the installation plays an important role on the overall noise
[13], but the evaluation of these effects is beyond the scope of the present work.
Another important aspect to consider is the optimum operational point of the fan.
The optimum performance, both from an operational and acoustic point of view, is
to the right of the stall point in the fan performance curve, typically at the maximum
static efficiency point. At the stall point, the sound suffers a great change in the
spectrum due to a higher level of the broadband component because of the flow
characteristics [14].
11
2. THEORY
Based on Equation (2.5), it can be noted that the sound power level is greatly related
to the velocity and diameter of the fan, two parameters that depend on the desired
performance of the fan.
12
2. THEORY
technique in which the blade trailing edge is in the shape of a sawtooth [15]. The
blade serrations help to diminish the noise generated by the trailing edge in the far
field [18].
Figure 2.8: Fan impeller diagram illustrating an annular blade element of length
dr at radius r [20].
13
2. THEORY
1
dFa = z ρW 2 (Cd cos β + CL sin β) c dr, (2.10)
2
1
dFu = z ρW 2 (CL cos β − Cd sin β) c dr. (2.11)
2
Finally, it can be concluded that the local thrust force dT is identical to the axial
force on each blade section, and the torque applied on each element, dτ , can be
calculated by multiplying the corresponding tangential force by the radius r. Hence:
1
dT = dFa = z ρW 2 (Cd cos β + CL sin β) c dr, (2.12)
2
1
dτ = dFu r = z ρW 2 (CL cos β − Cd sin β) cr dr. (2.13)
2
14
2. THEORY
As it is shown in Figure 2.10, the average tangential velocity felt by the blade section
can be defined as Ωr(1 − b), where b = ω/(2Ω) is the angular induction factor.
Furthermore, the wake rotational speed downstream of the blade is represented by
ω, and Ω is the blade rotating speed. In addition, the axial velocity right before the
impeller can be written as a function of the axial induction factor a and the incoming
flow velocity V∞ , as reflected in the velocity diagram in Figure 2.10. Therefore, the
relative flow angle β can be calculated using axial and average tangential velocities
Ωr(1 − b)
!
β = tan −1
. (2.14)
V∞ (1 − a)
Since the relative flow angle for each blade element is dependent on the rotational
speed of the corresponding element, then the value of β increases from the hub to
tip. Moreover, the relative velocity can be expressed as
V∞ (1 − a)
W = . (2.15)
cos β
Finally, torque and thrust can be expressed in form of induction factors by substi-
tuting Equations (2.14) and (2.15) into Equations (2.12) and (2.13).
15
3
METHODS
Figure 3.1: Mean-line calculation of the velocity diagram with axial entry [21].
16
3. METHODS
Euler’s equation, assuming inviscid and incompressible flow can be written as fol-
lows:
ρh 2 i
∆pt = (w1 − w22 ) + (c22 − c21 ) . (3.1)
2
Since axial entry is assumed, the tangential component of the absolute flow velocity
is zero (cu1 = 0) meaning that wu1 = u. Therefore, Equation (3.1) becomes
ρh 2 i ρh 2 i
∆pt = w1 − w22 + c2u2 = 2
u − wu2 + (u − wu2 )2 = ρu(u − wu2 ) = ρucu2 .
2 2
(3.2)
Given that Qd and N are known, the inlet blade angle γ1 and outlet blade angle
γ2 can be obtained from Figure 3.1. The mean component of the relative velocity
wm is constant throughout the cascade, and is equal to the ratio of Qd and the flow
area. The flow area is the ring area between the radial position of the hub rh and
the tip rt . Therefore, wm can be obtained
Qd
wm = . (3.3)
π(rt − rh2 )
2
17
3. METHODS
Figure 3.2: View of the calculated blade shape in (x,y) coordinates [21].
∆y
tan γ(y) = . (3.8)
∆x
The increase from γ1 to γ2 is to apply a driving action on the fluid, and it can be
calculated by linear, parabolic or higher degree polynomial distribution. According
to Pascu and Epple [22], a parabolic distribution can increase the static efficiency
of the fan. Therefore, the distribution is γ(y) = Ay 2 + By + C:
• y = y1 → γ = γ2
• y = ym → γ = γm
• y = y2 → γ = γ1
By having the inlet and outlet conditions fixed, there is only one degree of freedom
by (γm , ym ), resulting in the complete distribution of profile angles.
18
3. METHODS
wmax − w2
DF = . (3.9)
wmax
According to the research of Lieblein et al. [23], the momentum thickness of the
wake θ was related to the diffusion factor. Several measurements were performed
for two airfoil series, the NACA 65 and British C4 series. It was found out that a
value of DF = 0.6 is the upper limit for the diffusion factor. A higher value than
that, results in a dramatic increase in the diffusion in the boundary layer, as shown
in Figure 3.3.
Figure 3.3: Wake momentum thickness versus overall diffusion factor DF for
NACA 65 and C4 airfoils at minimum loss incidence [21].
c
σ= . (3.10)
s
As expressed in Equation (3.10), the blade solidity is the ratio between the blade
chord length c and the pitch distance s. Blade-to-blade pitch is the distance between
two adjacent blades which is shown in Figure 3.4.
19
3. METHODS
As already stated, the blade solidity factor is of great importance in designing fan
blades due to the aerodynamic effect of the adjacent blades on each other. The
determination of the solidity factor at different radial positions makes it possible to
select an appropriate approach during the fan blade design process.
The isolated airfoil and cascade approaches are valid for low and high solidity rotors,
respectively. It is required to use isolated airfoil technique for the blade solidity range
less than 0.7 and make use of the cascade approach for solidities larger than 1 in
order to achieve a reasonable blade design [25]. Furthermore, the modified isolated
approach can be applied to fulfill the requirement for the mid-range solidities [25].
20
3. METHODS
In order to relate the fan performance parameters with the forces mentioned previ-
ously, a control volume with a width identical to the pitch of a rotor is employed to
surround an isolated airfoil, as illustrated in Figure 3.6. The next step is to derive
the relative mean flow angle βm using
1
tan βm = (tan β1 + tan β2 ), (3.13)
2
where β1 and β2 are the inlet and outlet flow angles, respectively. It is now possible
to derive the net resultant force as a function of the lift and drag forces using the
momentum equation, the velocity triangle shown in Figure 3.6, and Equation (3.13).
Hence, the lift and drag coefficients are
s
CL = 2 (tan β1 − tan β2 ) cos βm − Cd cos βm , (3.14)
c
∆p
!
s
Cd = . (3.15)
c ρW12 /2
Since the drag coefficient of the airfoil is considerably less than the lift coefficient,
the second term in Equation (3.14) is normally neglected [9]. It can be noted that
the pressure difference through the blades and fan performance are strongly related
to the drag and lift coefficients [9].
21
3. METHODS
As the blade-to-blade pitch distance decreases, the solidity factor increases, meaning
that the blade pressure field can be greatly influenced by other blades. Hence, a
mismatch between the true and calculated lift values occurs [27]. Consequently, the
isolated airfoil technique is considered reliable for an axial fan which is characterized
by low solidity and substantial stagger angles γ [28]. As reflected in Figure 3.5,
stagger is the angle between the chord line and the axis of rotation, and is defined
as
γ = βm − α, (3.16)
where α is the angle of attack which depends on lift coefficient of the isolated airfoil.
3.2.2 Cascade
A fan with a high flow coefficient is distinguished by high solidity and low stagger
angles. In this case, a simple design technique called cascade approach is expected
to be utilized [28]. As illustrated in Figure 3.7, the model includes a row of similar
blades located at equal distance s from each other, which demonstrates the actual
impeller in the axial fan. The cascade technique considers the flow deflection across
the blade instead of focusing on the lift force corresponding to each blade [27]. The
major parameters used in the cascade technique procedure are flow deflection angle
ε, camber angle θ, solidity σ, incidence angle i, and stagger angle γ.
The first step in the cascade approach is to select a proper solidity from a cas-
cade performance database corresponding to a desired airfoil series, and thereafter
calculate the camber angle using Equation (3.17) [28].
ε−i
θ= q (3.17)
1 − m s/c
where the flow deflection angle is defined as ε = β1 − β2 and m is the deviation angle
coefficient which is a function of stagger angle.
22
3. METHODS
The initial values of incidence angle and deviation angle coefficient can be allocated
to zero and 0.26, respectively, in order to obtain an appropriate preliminary value for
camber angle θ, according to Wallis [28]. It is now possible to find an approximate
value of stagger angle using the following equation:
θ
γ = (β1 − i) − . (3.18)
2
The final step is to select more precise values of incidence angle and of deviation
angle coefficient from the cascade database and recalculate the camber and stagger
angle.
23
3. METHODS
The series of design curves demonstrated in Figure 3.8 shows how the interference
factor is related to the stagger angle and blade solidity. It can be noted that, as the
solidity decreases, the interference coefficient value moves towards unity, whereas it
approaches zero as the blade solidity increases.
Since the isolated airfoil and cascade data are not valid for intermediate solidities, it
is necessary to apply an interference factor in order to have a reasonable transition
between data corresponding to cascade and isolated airfoil.
Figure 3.8: Interference factor for airfoils with a circular-arc camber line [9].
24
3. METHODS
rotor tips [30]. Tip clearance drag coefficient Cdt is a function of blade span b
and tip clearance CT . The loss due to the tip clearance leakage is determined
as
CT
Cdt = 0.29 CL1.5 . (3.21)
b
4. Annulus loss is referred to the pressure loss due to friction between the fluid
and annular surfaces such as hub and housing walls [9]. In addition, annulus
pressure loss coefficient Cda can be estimated as
0.02s
Cda = , (3.22)
b
where s is the pitch distance.
The pressure loss components explained in this section should be taken into account
in order to predict the fan performance as accurately as possible. Therefore, the
total drag coefficient can be calculated by
Cd = Cdp + Cds + Cdt + Cda . (3.23)
25
3. METHODS
Equation (3.24) is valid for one radial position, only. In order to obtain the total-to-
static pressure difference for the whole impeller, Equation (3.24) has to be integrated
for all radii between hub and tip, taking the area average as below:
rt
!2
1 Z
ρ 2 cm2
∆pt−s,axial = u − 2πr dr . (3.25)
π(rt2 − rh2 ) r 2 sin β2
h
Then, by setting a zero flow rate, i.e. cm2 = 0, in Equation (3.24), the shut-off point
or maximum pressure difference can be calculated by integrating for all radii and
taking the area average as
ρ
∆pt−s,max = u2 , (3.26)
2
Zrt
1 ρ 2
∆pt−s,max,axial = u 2πr dr . (3.27)
π(rt − rh ) r 2
2 2
h
In the same way, the free flow point or maximum flow rate can be calculated by
setting the pressure difference in Equation (3.24) equal to zero and considering that
cm2 = Q/A which leads to
Qt−s,max = uA sin β2 , (3.28)
rt rt
1Z Z
Qt−s,max,axial = uA sin β2 2πr dr = 4π 2 n r2 sin β2 dr . (3.29)
Ar r
h h
Finally, by having both the shut-off point and the free flow point, it is possible to
describe the complete performance curve of the axial fan, using Equations (3.24),
(3.26) and (3.28):
!2
Q
∆pt−s = ∆pt−s,max 1 − . (3.30)
Qt−s,max
The next performance parameter is the efficiency. There are two efficiencies that are
relevant to describe the performance of an axial fan. The total-to-total efficiency
ηt−t of a fan or a pump is defined as the ratio of the total hydraulic power divided
by the shaft power. If no viscous and three-dimensional losses are considered, then
all the mechanical power of the shaft is transferred to the fluid [22]. This can be
expressed as
Whyd,t Q∆pt
ηt−t = = = 1. (3.31)
Wshaf t Q∆pt
Like the pressure difference, the correct efficiency to describe the performance of a
fan is the total-to-static efficiency ηt−s . The total-to-static efficiency is defined as
the ratio of the total-to-static hydraulic power divided by the shaft power
Whyd,t−s Q∆pt−s ∆pt−s
ηt−s = = = . (3.32)
Wshaf t Q∆pt ∆pt
In contrast to the total-to-total efficiency, the efficiency ηt−s will always be less than
one, even if no losses are considered. By rewriting Equation (3.32) with Equation
(3.24), the total-to-static efficiency is
∆pt − (ρ/2)c22 ρ c22
ηt−s = = ηt−t − . (3.33)
∆pt 2 ∆pt
26
3. METHODS
As shown in Equation (3.33), the efficiency ηt−s is less than 1 due to the second
term. This term is independent of the viscosity and it is only a mechanical loss
that has to be accounted for in the performance of fans and pumps [22]. Equation
(3.33) is only valid for one radial section of the blade, and to get the efficiency for
the complete impeller it must be integrated for all radii by taking the area average.
3.3.2 CFD
In order to perform the CFD simulations, 3D models of the 11 designs were created
in Solid Edge 2020. Therefore, the coordinate system of the blade section profiles
obtained with the two design methods was transformed to cylindrical coordinate
system. The CAD models of the 11 fans are shown in Appendix C.
The CFD simulations were performed by Mattias Chevalier from RTGF group in
Scania. In this section, the description of the settings used for the analysis is speci-
fied, as described by Mattias.
The main performance parameters such as volume flow rate, torque, and forces are
measured at three different rotational speeds for each fan model. The simulations
are conducted with STAR-CCM+ v14.06.012, which is developed by Siemens.
The process uses a finite volume approach to discretize the Navier-Stokes equa-
tions which describes the physics of fluids. The flow is assumed to be an ideal gas
considering compressibility effects. In addition, the boundary conditions are set as
stagnation flow inlet and pressure outlet. Furthermore, the airflow is considered to
be turbulent and is modeled using the Reynolds Averaged Navier-Stokes (RANS)
equations. The technique used is the Realizable K-Epsilon All y+ Wall Treatment
turbulence model.
The performance data are measured by installing a shroud around the fan models.
Moreover, each fan assembly is placed between two chambers with a flow resistance
(porous medium) upstream and a blockage downstream, as shown in Figure 3.10.
Figure 3.10: A cut plane view of the fan installation, from left to right: blockage,
fan, fan shroud, and flow resistance.
27
3. METHODS
In order to diminish undesired numerical effects from the boundaries, the computa-
tional domain extends upstream and downstream of the installation. As illustrated
in Figure 3.11, the computational mesh consists of 15 million cells in form of polyhe-
dral inside the domain. In addition, prism layers are used near the walls. Finally, fan
curves are derived by changing the porous medium resistance at a given rotational
speed.
Figure 3.11: A center cut plane view of the computational domain, the flow di-
rection is from right to left.
3.4.1 Approach 1
As stated in Cory [3], the first equation used to approximate the sound power level is
the resulting work done by Beranek, Kamperman, and Alien as shown in Equation
(3.34).
LW = 57.3 + log10 (Qd ) + 20 log10 (∆p) − 10 log10 (η) (3.34)
where
• LW = Overall sound power level of noise transmitted along ducts fitted to
the inlet and outlet of fan operating at/near its peak efficiency (dB ref
10−12 W),
• Qd = Design flow rate (or flow rate at or near peak efficiency) (m3 /s),
• ∆p = Pressure difference at or near peak efficiency (Pa),
• η = Peak efficiency (%).
Then, with Equation (2.3) the sound pressure level can be calculated.
28
3. METHODS
3.4.2 Approach 2
The method presented in Section 3.4.1 results in only one number, namely the overall
power level, but it is also possible to estimate the sound power spectrum of the noise
(LW vs frequency). First, it is necessary to introduce the specific sound power level,
which is defined as
W
H(φ) = , (3.35)
∆p2 Q
where W is the overall sound power in watts. As stated in Crocker [12], Table 3.1
gives specific sound power levels in several octave bands for axial fans. The levels
are given in decibels with reference quantities: W0 = 10−12 W, Q0 = 1 m3 /s, and
∆p0 = 1 kPa.
Table 3.1: Specific sound power levels in eight lowest octave bands for axial fans
[12].
29
4
DESIGN PROCEDURE
30
4. DESIGN PROCEDURE
Next, by solving Equation (4.2) for Diameter Dt , the optimum fan diameter for the
specified performance is found.
√ !0.25
π 2Y
δ= Dt (4.2)
2 Q2d
Then, with Equation (4.3) and Figure 4.2, the hub-to-tip ratio ν can be found:
√
N Qd
Ns = . (4.3)
∆p0.75
u = 2πrn, (4.5)
2πr
s= . (4.6)
z
To estimate the chord length of the blades, the specific speed Ns along with Figure
4.3 is used.
31
4. DESIGN PROCEDURE
Figure 4.3: Specific speed vs. mean head and flow coefficients for axial fans [9].
Where φ is the flow coefficient and ψ the head coefficient. The flow coefficient φ is
defined as
wm
φ= . (4.7)
u
The mean component of the relative velocity wm was introduced in Equation (3.3).
Finally, chord length c is obtained by solving the value from Figure 4.3. The next
step is to calculate the inlet blade angle γ1 and the outlet blade angle γ2 . From
Figure 3.1:
wm
tan γ1 = . (4.8)
u
As mentioned in the previous chapter, Pascu developed Equation (3.7) to estimate
γ2 . In her work, Pascu concluded that assuming a parabolic distribution, i.e. x = 2,
yields better results [21]. This was confirmed by both CFD and experimental results.
Therefore, Equation (3.7) becomes
1 2πn 2
" 2 # !
rt
= r − 2 (r − rh )
1.35
+1 rh2 . (4.9)
tan γ2 rwm rh
Once both angles are calculated, the blade angle distribution γ(y) is resolved in an
iterative process based on the parabolic distribution mentioned earlier. The iteration
is done with a script in Matlab and stops when chord length constraints are matched
and corrected for abrupt changes in blade angle distribution [21].
Another parameter to consider is the camber angle. The shape obtained by the
previous process can be considered as the camber line of the profile. By adding a
thickness distribution, the typical airfoil shapes are obtained. As stated in Pascu
[21], for low pressure axial fans, such as the fan for cooling purposes, this variable
thickness is unnecessary since both profiles, i.e. cambered plates and airfoils, have
identical performances when operating at low pressures. So, for this method, the
blades have constant thickness across the chord direction. Therefore, the camber
angles are the same as the blade angles in this case. In order to test this statement,
design method 2 considers airfoil cross sections.
32
4. DESIGN PROCEDURE
Last, the diffusion factor DF is calculated with Equation (3.9). From the velocity
triangle in Figure 3.1, the velocity distribution w(y) is defined by
wm
w(y) = . (4.10)
sin γ(y)
By trying to keep the diffusion factor below 0.6 along the blade, some of the losses
will be minimized. This can be done by adjusting γ2 to avoid rapid changes from
one section to another [21].
where g is the acceleration due to gravity and H is the head rise. In addition, the
multiplication of head rise and gravitational acceleration can be reformulated as
∆p
gH = . (4.12)
ρair
The next step is to determine the non-dimensional specific diameter ∆s correspond-
ing to ωs using the Cordier diagram illustrated in Figure 4.4. Therefore, the fan tip
diameter Dt can be found as √
∆s Qd
Dt = . (4.13)
(gH)0.25
The specific speed calculated using Equation (4.3) can be used for the purpose of
determining the hub-to-tip ratio ν. As shown in Figure 4.2, the desired hub-tip ratio
can be chosen based on designer’s preference from the specified domain. Hence, the
fan hub diameter Dh can be calculated by considering the hub-to-tip ratio and tip
diameter,
Dh = νDt . (4.14)
The blade solidity calculated using Equation (3.10) is required to fall within the
domain of 0.4−1.1 for the purpose of preventing stall in axial-flow fans. In case
the solidity does not fulfill the requirement mentioned above, another hub-to-tip
ratio should be selected from the range of 0.3−0.7 based on the calculated specific
speed Ns [9]. Before starting the implementation of the blade design approaches,
it is necessary to determine the number of blades z to be used in Equation (4.4).
Since the number of blades has an influence on both the fan efficiency and the
noise pollution generated by the impeller, it is important to keep the number of
33
4. DESIGN PROCEDURE
Figure 4.4: Cordier diagram based on non-dimensional specific speed and diameter
[9].
An essential first step in determining the blade chord length c is to calculate the
flow coefficient φ as below:
Va
φ= , (4.15)
U
where Va and U are axial and tangential velocities, respectively. These velocity
components can be defined as
4Q
Va = , (4.16)
π (Dt2− Dh2 )
U = ωr, (4.17)
where r represents the blade radial position. The axial velocity component is identi-
cal to the absolute inlet velocity V1 and remains constant across the fan blades. This
is due to the fact that the two-dimensional flow is assumed to be incompressible,
meaning that the Mach number is lower than 0.3 [28]. The next step is to estimate
34
4. DESIGN PROCEDURE
the value of φ sc corresponding to the specific speed from Figure 4.3. Thereafter,
the pitch distance s calculated using Equation (4.6) is employed in order to deter-
mine the chord length of the blade at each radial position.
Figure 4.5: Inlet and outlet velocity triangles for axial fans [9].
The local blade solidity needs to be calculated using Equation (3.10) at each radial
position in order to decide which design approach is more suitable for the analysis.
An iterative process shall be used at this stage with the purpose of checking the
accuracy of the model with respect to the hydraulic efficiency ηhyd , stagger angle
γ, and lift coefficient CL . Hydraulic efficiency, also known as stage efficiency, needs
to be initially defined since the outlet flow angles and velocities are dependent of
this parameter. After selecting an appropriate initial value for hydraulic efficiency,
the next step is to make use of the velocity diagram illustrated in Figure 4.5 and
calculate the flow angles and velocities. The mean flow angle βm can be calculated
using Equation (3.13), where the inlet and outlet flow angles are determined as
follows:
U
β1 = tan −1
, (4.18)
Va
U − Vu2
β2 = tan −1
. (4.19)
Va
The tangential component of absolute velocity at the outlet Vu2 highly depends on
hydraulic efficiency and the tip pressure head Ht of the fan, and can be calculated
using
gHt
Vu2 = . (4.20)
ηhyd U
As shown in Figure 4.5, the tangential component of relative velocity at the outlet
Wu2 can be calculated by taking the difference between the tangential velocity and
the tangential component of absolute velocity at the fan outlet. Moreover, the
remaining velocity components such as outlet absolute velocity V2 , inlet relative
35
4. DESIGN PROCEDURE
velocity W1 , and outlet relative velocity W2 can easily be calculated by using the
Pythagorean theorem as follows:
0.5
V2 = Vu2
2
+ Va2 , (4.21)
0.5
W1 = U 2 + Va2 , (4.22)
0.5
W2 = Wu2
2
+ Va2 . (4.23)
Once the blade solidity and the flow angles are determined, it is possible to com-
pute the corresponding lift coefficient using Equation (3.14). The next step is to
determine the Reynolds number at each radial position as
Wm c
Re = , (4.24)
νair
where c is the local chord length and νair is the kinematic viscosity of the air. In
addition, the Reynolds number is a function of mean relative velocity Wm which can
be found using the velocity triangles in Figure 3.6. Hence, Wm becomes
Va
Wm = . (4.25)
cos βm
The next step is to determine the stagger angle based on either angle of attack or
incidence angle, which meets the needs of required lift coefficient. The required
data needed to calculate the stagger angle can be gathered from suitable airfoil or
cascade test data. If the blade solidity is greater than unity, it is necessary to use
the cascade approach based on appropriate test data, such as those shown in Figures
A.1-A.4. On the other hand, the modified isolated approach or isolated technique
can be employed with respect to the solidity condition. In case of low or intermediate
solidities, it is required to use aerodynamic data from one of the airfoils listed in
Table A.1. This table consists of circular-arc and symmetrical airfoils which their
shape and aerodynamic characteristics for various Reynolds numbers are gathered
using XFLR5 software. In order to provide the aerodynamic characteristics, a set of
polars corresponding to each airfoil is generated using Type 1 analysis in XFLR5.
Furthermore high angles of attack are investigated with the purpose of recognizing
pre-stall and stall conditions.
Since an accurate blade design can enhance the fan performance and mitigate the
noise generated by the fan, it is important to select an airfoil which not only fulfills
the requirements for lift but also has a high lift-to-drag ratio CL /Cd [33]. Stall is
an important phenomenon which should be treated when designing a fan impeller.
In order to prevent stall in an axial-flow fan, it is recommended that lift coefficient
shall vary from 0.6 at the tip to 1.3 at the hub when designing a blade using isolated
airfoil approach [33]. In addition, it is necessary to select an airfoil which provides
the required lift without being affected by the stall [33].
After selecting the required aerodynamic data, the corresponding angle of attack
or incidence angle can be employed to calculate the stagger angle using Equation
(3.16) or Equation (3.18). Finally, the hydraulic efficiency can be determined by
1 − Rr − φδr
" #
Rr − φδr
ηhyd =φ + , (4.26)
φ + δr Rr φ + δr (1 − Rr )
36
4. DESIGN PROCEDURE
where Rr is the local reaction for the rotor blade and δr is the drag-to-lift ratio
corresponding to the chosen airfoil which can be computed with
Rr = φ tan βm , (4.27)
Cd
δr = . (4.28)
CL
Moreover, contribution of additional pressure losses to the total drag shall be consid-
ered by using Equation (3.23) for the purpose of predicting the hydraulic efficiency.
The initial value of the hydraulic efficiency should be compared with the value esti-
mated from Equation (4.26) in order to check the accuracy of the model. The design
process needs to be repeated until the desired convergence tolerance is reached. This
procedure runs for all chosen radial positions on the blade in order to select appropri-
ate airfoil profiles which fulfill the requirements for lift, stagger angle and hydraulic
efficiency. The design process steps mentioned above are summarized in the flow
diagram shown in Appendix B.2.
37
5
RESULTS
As mentioned before, the pressure difference required for the Cordier diagram is the
total pressure difference, while the pressure rise specified in Table 5.1 is in form of
total to static. The approximation of the total pressure difference is done with a
script in Matlab and results in a value of 238 Pa. Therefore, the specific head Y
and fan rotational speed n are 194.28 Pa/(kg/m3 ) and 33.33 rps, respectively. Using
Equation (4.1):
√
√ Qd
σ = 2n π = 1.232.
(2Y )0.75
Then, from Figure 4.1, the diameter number is δ = 1.586. Next, by solving Equation
(4.2), the diameter of the fan Dt is obtained as
!−0.25
2δ 2Y
Dt = √ = 368 mm.
π Q2d
After obtaining the diameter, the specific speed Ns is calculated with Equation (4.3)
to find the hub-to-tip ratio ν.
√
N Qd
Ns = = 86895
∆p0.75
38
5. RESULTS
From Figure 4.2 and taking the lowest dashed line, the hub-to-tip ratio is ν = 0.44.
Therefore, the hub diameter is Dh = 0.16 m. Finally, according to Equation (4.4),
the fan must have 5 blades.
s wm 2πr wm 2πr
c=φ = = = 96 mm.
0.61 u (0.61)z 2πrn (0.61)z
After finding the fan dimensions, the shape of the blade section can be found. The
mathematical routine is depicted in Appendix B.1. The blade is separated into 20
equidistant sections from the hub radius rh to the tip radius rt . The blade cross
section shape and blade angle distribution for two different radial positions are shown
in Figures 5.1 and 5.2, respectively.
40
30
20
10
0
0 10 20 30 40 50 60 70 80 90
40
35
30
25
20
15
0 10 20 30 40
39
5. RESULTS
The distribution of γ2 was slightly corrected to decrease the diffusion factor shown
in Figure 5.3. As stated before, this is done to reduce the total pressure loss due to
diffusion in the boundary layer.
0.8
0.6
0.4
0.2
0
80 100 120 140 160 180
* The chord distribution varies linearly from -10 to 10 extra units to the base chord
length (model 1) from root to tip.
40
5. RESULTS
** The chord distribution varies linearly from -20 to 20 extra units to the base chord
length (model 1) from root to tip.
Since the pressure rise used in all calculations is in the form of total pressure, an
iterative process is employed to estimate the total pressure difference. Hence, the
corresponding value of total pressure rise is approximated to 523 Pa. Thereafter,
the rotational velocity is converted to 314 rad/s and used in Equation (4.11) to
determine the non-dimensional specific speed as
√
ω Qd
ωs = = 3.52,
(∆p/ρair )0.75
where the air density is ρair = 1.225 kg/m3 at 15 ◦C. Hence, the non-dimensional
specific diameter ∆s is obtained from Figure 4.4, which is equal to 1.55. It is now
possible to calculate the fan tip diameter using Equation (4.13).
√
∆s Qd
Dt = = 360 mm
(∆p/ρair )0.25
The specific speed Ns is computed with Equation (4.3), and has a value of 83365.
Next, the hub-to-tip ratio ν is obtained from Figure 4.2. Since the upper and lower
limits are defined in Figure 4.2, it is possible to choose different hub-to-tip values
which fall within the corresponding domain. The hub-to-tip value which is selected
for this model is ν = 0.48. Hence, the hub diameter is Dh = νDt = 172 mm, and
the corresponding number of blades becomes five when applying Equation (4.4).
The next step after determining the diameters is to calculate the blade span length
b by taking the difference between tip and hub radii which results in a value of 94
mm.
41
5. RESULTS
Next, the blade is divided into five identical elements, meaning that the blade should
be analyzed at six independent radial positions from hub to tip. The following steps
present how the blade is modeled at the tip, i.e. r = rt . Then, the overall results
will be provided in Table 5.4 at the end of this section.
In order to select an appropriate design method, the local solidity should be calcu-
lated based on other parameters such as local chord length and pitch distance as
shown below.
• Tangential velocity:
U = ωr = 56.52 m/s.
• Flow coefficient:
Va
φ= = 0.25.
U
• Pitch distance:
2πr
s= = 205.9 mm.
z
• Chord length:
s
φ = 0.59 ⇒ c = 87.8 mm.
c
• Solidity:
c
= 0.43.
σ=
s
Given the value of blade solidity, the isolated airfoil technique becomes the most
suitable alternative for designing the blade at the tip. After selecting the best
approach, an iterative process with a convergence tolerance = 0.001 is employed.
As an initial measure, a value of 0.95 is assigned to ηhyd . Then, the required lift
coefficient can be calculated as follows.
• Tangential component of outlet absolute velocity:
gHt
Vu2 = = 8.28 m/s.
ηhyd U
• Lift coefficient:
s
CL = 2 (tan β1 − tan β2 ) cos βm = 0.72.
c
The next step is to compute the tip Reynolds number with respect to the mean
relative velocity.
• Mean relative velocity:
Va
Wm = = 54.26 m/s.
cos βm
• Reynolds number:
Wm c
Re = = 3.24 · 105 ,
νair
where the kinematic viscosity of air is νair = 1.47 · 10−5 m2 /s at 15 ◦C. Since the
isolated airfoil data are only gathered for Reynolds numbers from 5·104 to 5·105
with a step size of 5·104 , an interpolation technique with high accuracy is used to
select the most reasonable aerodynamic data corresponding to each airfoil profile at
different Reynolds numbers. Furthermore, the transition criterion number Ncrit is
set to nine, meaning that the airfoils are analyzed in a wind tunnel under average
disturbance condition [35].
According to the proposed method presented in the previous chapter, NACA 4412
is chosen as the best airfoil profile which not only fulfills the requirements for the
lift but also provides the highest lift-to-drag ratio among all the airfoils in the first
iteration.
1.5 100
80
1
60
0.5
40
0
20
-0.5
0
-1 -20
-10 -5 0 5 10 15 -10 -5 0 5 10 15
The performance of NACA 4412 airfoil shown in Figure 5.4 provides the aerodynamic
information required in order to calculate stagger angle and hydraulic efficiency in
the first iteration. According to the plots illustrated in Figure 5.4, the angle of
attack of the blade tip section becomes α = 2.2° when the lift coefficient is 0.72. In
addition, the profile drag coefficient Cdp has a value of 0.009 since the lift-to-drag
43
5. RESULTS
γ = βm − α = 72.7°.
Moreover, the effect of pressure loss components should be considered with the pur-
pose of estimating the fan performance. Hence, the total drag coefficient becomes:
where tip clearance CT is set to 2 mm. It is notable that the annulus loss associated
with the annular surfaces has the highest influence on the total drag. Finally, the
hydraulic efficiency can be determined as follows.
• Drag-to-lift ratio:
Cd
δr = = 0.092.
CL
• Reaction:
Rr = φ tan βm = 0.93.
• Hydraulic efficiency:
1 − Rr − φδr
" #
Rr − φδr
ηhyd =φ + = 0.72.
φ + δr Rr φ + δr (1 − Rr )
Since the preliminary and calculated values of the hydraulic efficiency do not match
the convergence criteria in the first iteration, it is necessary to use the calculated
value of the hydraulic efficiency as a new input for the next iteration. The iterative
process must continue until all the requirements are satisfied. After 12 iterations,
the hydraulic efficiency and the stagger angle converge, and therefore the process
is accomplished. Since the lift coefficient determined in the last iteration is 0.92,
NACA 4412 is no longer the best performing airfoil. It is notable that NACA 4412
is replaced with NACA 6409 airfoil in order fulfill the requirements for the hydraulic
efficiency and the stagger angle at the blade tip position, i.e. r = rt .
1.5 120
100
1
80
60
0.5
40
20
0
0
-0.5 -20
-10 -5 0 5 10 15 -10 -5 0 5 10 15
44
5. RESULTS
As shown in Figure 5.5, the lift coefficient of 0.92 is achieved at an angle of attack
of α = 2.1, meaning that the corresponding lift-to-drag ratio has a value of 98.7. By
evaluating the aerodynamic data corresponding to NACA 4412 and NACA 6409, it
is possible to conclude that the latter must be employed. The design procedure is
implemented for all selected radial positions from the hub rh to the tip rt with a
step size of 18.8 mm, and the outcomes are presented in Table 5.4.
c s σ βm α γ ηhyd
Radius (mm) (mm) (-) (deg) (deg) (deg) (%) Airfoil
rh 87.8 98.4 0.89 50.7 11.1 39.6 85 S 1223
r2 87.8 119.9 0.73 60.6 7.1 53.5 86 NACA 6409
r3 87.8 141.4 0.62 66.2 5.3 60.9 84 NACA 6409
r4 87.8 162.9 0.54 69.9 3.8 66.1 82 NACA 6409
r5 87.8 184.4 0.48 72.6 2.8 69.8 79 NACA 6409
rt 87.8 205.9 0.43 74.6 2.1 72.5 76 NACA 6409
The efficiency values listed in Table 5.4 must be integrated from hub to tip by taking
the area average in order to provide the total-to-static efficiency of the impeller
ηt−s , which will be presented in Section 5.3.1. As explained in Section 4.2, the same
procedure is applied on each radial position along the blade span, which leads to the
selection of the airfoil profiles shown in Table 5.4. As illustrated in Figure 5.6, S 1223
airfoil profile is used to design the blade at the hub and NACA 6409 is employed
for forming the blade at other radial positions. Furthermore, it can be noticed that
the blade stagger angle increases from 39.6° at the hub to 72.5° at the tip.
45
5. RESULTS
46
5. RESULTS
5.3.2 CFD
In Table 5.7, partial results are shown. The table includes the results for pressure
difference ∆pt−s , volume flow rate Q, torque required at the shaft τshaf t , and total-
to-static efficiency ηt−s . The efficiency is calculated by rewriting Equation (3.32) as
ηt−s = (Q∆pt−s )/(ωτshaf t ).
47
5. RESULTS
The complete table of the performance results along with the pressure distribution
figures are presented in Appendix D.
In order to gain a deeper understanding of the aerodynamic performance, models 3
and 7 are analyzed in detail. The reason to choose these models is to compare two
fans with various blade profiles. Moreover, model 3 has one of the highest efficiencies
among the fan models, while model 7 has a lower efficiency compared to the other
models. Furthermore, these fan models have two of the main noise reduction factors,
such as varying chord along the span and forward-swept blades.
300 60
Performance Fan 3
Performance Fan 7
250 Efficiency Fan 3 50
Efficiency Fan 7
System High
200 40
System Mid
System Low
150 30
100 20
50 10
0 0
0 0.2 0.4 0.6 0.8 1
The performance curves illustrated in Figure 5.8 are derived based on the data
presented in Tables D.2 and D.3. Moreover, the efficiency curves are estimated by
assuming that the torque values corresponding to models 3 and 7 in Table 5.7 are
constant in the operating range.
5.4 Noise
In this section, the sound pressure levels Lp of the 11 models are estimated with
the two approaches previously introduced. Then, the sound pressure levels Lp cor-
responding to the models are compared with the values estimated for two Scania’s
fans.
5.4.1 Approach 1
According to approach 1, the noise power level depends on the volume flow rate and
pressure difference at/near the peak efficiency. With the values of model 1 from
Table 5.7 and Equation (3.34):
48
5. RESULTS
Qθ
Lp = LW + 10 log10 = 73.89 dBA.
4πr2
The sound pressure level results for the fan models are provided in Table 5.8.
5.4.2 Approach 2
From Table 3.1, the specific sound power levels for this fan correspond to the tubeax-
ial fans values. For model 1, the blade pass frequency is calculated to BPF=166,
therefore 7 extra units are added to the frequency band of 125 Hz. Then, by adding
10 log10 Q + 20 log10 ∆p to each specific sound power H, the spectrum is obtained.
Next, the values must be transformed to SPL and weighted for A scale.
To obtain the overall sound pressure level, Equation (3.36) is used with the values
in the last row.
OASP L = 10 log10 10Lpi /10 = 74.77 dBA
X
The results for the other models are listed in Table 5.9.
49
5. RESULTS
Table 5.9: Overall sound pressure level at 2000 rpm using approach 2.
50
5. RESULTS
51
6
DISCUSSION
By evaluating the values of the fan models in Table 5.6, it is possible to notice
the limitations of the theoretical equations. Since the equations to estimate the
performance are only dependent on the outlet blade angle and fan dimensions, some
of them have the same results, even though the number of blades and/or chord
distribution are different. Another aspect to note is the efficiency values. In models
1 to 5 and model 11, the losses are not considered, meaning that the efficiency is
the ideal efficiency, overestimating performance. But even if these equations do
not provide exact results, they provide a good estimation on the magnitude of the
performance. For these reasons, a more complex method is required, such as CFD
analysis.
From Table D.1, it is possible to notice how different factors affect the performance
of the fans. By comparing models 1 and 2, the effect of number of blades can
be observed. Model 1 has 5 blades, while model 2 has 7 blades. At low rotational
speeds, i.e. 1000 and 2000 rpm, it seems that the performance is not greatly affected,
but the torque required is higher for model 2, decreasing the efficiency slightly. But
at a rotational speed of 3000 rpm, even though the performance is almost the same,
the torque required for model 2 is lower, increasing the efficiency. In general, and
according to previous studies, a higher number of blades increased the performance
of the fan, but reduced the efficiency. On the other hand, model 11 has the same
geometry as model 2, but with uneven blade spacing. As stated before, this can
reduce the sound pressure level by 2.96 dBs, but given that the approaches used
to estimate the sound pressure level don’t take into account these types of factors,
it is not possible to observe their influence. At the same time, the performance is
compromised, reducing both the volume flow rate and efficiency.
Another efficient technique to reduce the noise is to design a blade with a larger
tip chord. Both models 3 and 4 have a larger tip chord and a shorter root chord
compared to model 1. This can be observed in the increase in efficiency. While the
performance characteristics are not greatly affected, the torque required for models
3 and 4 is considerably lower than the torque required for model 1, especially at
high rotational speeds.
As mentioned previously, models 5 and 6 are equipped with straight blades and have
larger tip diameters compared to some of the other models. These two models share
an identical number of blades and overall geometry such as same tip diameter and
hub-to-tip ratio, but they differ in the shape of the blades. According to the data
shown in Table D.1, model 5 is able to generate high volume flow rates at higher
pressure rises relative to other fan models, due to its large diameter. In addition,
model 6, which is equipped with airfoil-shaped blades, requires a lower torque, which
52
6. DISCUSSION
in turn enhances the efficiency in comparison with model 5, while they both provide
nearly the same volume flow rate. A possible reason to explain the lower required
torque in model 6 relative to model 5 can be related to less pressure losses generated
by airfoil-shaped blades at higher hub-to-tip ratios.
Fan models 7 and 8 are the modified versions of model 6. These two models are
designed with forward-swept blades with the purpose of reducing the noise. It is
notable that the models have almost the same performance level. Since model 7 is
equipped with forward-curved blades at higher sweep angle relative to model 8, its
performance features are being negatively affected.
In order to analyze the effect of the tip diameter on the performance of a fan,
model 9 and 10 with smaller diameters relative to the other models are examined.
Model 9 has straight and airfoil-shaped blades which can be compared with model
3 since they both have similar performance characteristics. Although fan model 3
is developed based on the first blade design technique and has significantly different
size and dimension, its performance features are identical to model 9. They are not
only the most efficient fans among all the models but also generate almost the same
volume airflow at comparable pressure rise. This is due to various reasons, such as
chord length distribution along the blade span, blade shape, and hub-to-tip ratio.
On the other hand, performance of model 10 is compromised due to the usage of a
high hub-to-tip ratio and forward-curved blades compared to model 9.
Based on all the results in Table 5.7, it is noticeable that there is no clear distinction
between cambered plates and airfoils at low pressure differences, since all of the
models have relative high efficiencies. An experimental test could provide a deeper
insight into the differences.
As illustrated in Figure 5.8, three fictive system curves are used to estimate the
operating point at different system resistances, i.e. high, medium, and low, for fan
models 3 and 7. The results indicate that model 3 provides lower volume flow rate
at relatively lower pressure difference relative to model 7. On the other hand, it
operates at higher efficiencies than model 7, which matches the predicted results
shown in Table D.1.
The noise level estimations are solely done with empirical equations, for this reason
it is not possible to conclude how some factors affect the noise level. On the one
hand, it can be argued that the noise level is reduced by increasing the efficiency of
the fan, this can be observed by comparing model 1 with model 3. On the other
hand, the uneven blade spacing affects directly the efficiency, this is a well-known
design feature to reduce the sound pressure level. Therefore, increasing the efficiency
does not necessarily corresponds to lower noise levels. This was also the conclusion
in the investigation done by Robert C. Mellin [36]. In his work, he also concluded
that increasing the diameter and decreasing the speed will not necessarily reduce
the noise level for a given aerodynamic performance, as shown in the fan sound law
in Equation 2.5.
As stated earlier, the performance values are adjusted in order to have a reasonable
comparison in sound pressure levels between Scania’s and the developed fan models.
As illustrated in Table 5.11, the sound pressure levels of the 11 models are approx-
imately the same as the levels corresponding to the Scania’s fans. Nevertheless, all
the designed models have less number of blades compared to the Scania’s fan 1,
53
6. DISCUSSION
which in turn increases the sound quality. Additionally, the 11 fan models have a
smaller tip diameter relative to the Scania’s fan 1. This is beneficial when there is
a space limitation in the cooling system installation.
54
7
CONCLUSION
Throughout this study, it has been found that several factors influence the perfor-
mance and noise pollution of the electric axial fans. Based on the results discussed
and the studies cited in Section 2.2.4, it can be concluded:
• A larger chord length at the tip relative to the root of the impeller blades
enhances the efficiency and reduces the noise emitted by the fan.
• Irregular spacing between the rotor blades has a negative impact on the fan
performance and efficiency. However, it can significantly reduce the sound
pressure level.
• A forward-swept and curved configuration decreases the efficiency, but also
has a positive effect on mitigating the noise.
• A larger tip diameter and a higher number of blades increment the torque
required which in turn increases the energy consumption of the electric motor
used to run the fan.
• An airfoil-shaped impeller presents similar performance as a cambered-plate
rotor in low pressure differences.
• A small tip clearance leads to a less leakage at the blade tip which in turn
enhances the efficiency and reduces the sound pressure level.
55
Bibliography
56
Bibliography
[15] W. Neise, “Review of fan noise generation mechanisms and control methods,”
in Fan Noise: An International INCE Symposium (A. Guedel, ed.), (Senlis,
France), pp. 45–56, SFA, CETIM, 9 1992.
[16] S. Castegnaro and A. Lazzaretto and M. Masi, “Effectiveness of blade forward
sweep in a small industrial tube-axial fan,” in FAN 2018 International Confer-
ence on Fan Noise, Aerodynamics, Applications and and Systems, (Darmstadt,
Germany), CETIAT, AMCA, 4 2018.
[17] Y. Kumon and M. Ohtsuka, “Development of electric fan propeller featuring
chestnut tiger butterfly wing characteristics,” in Fifth International Symposium
on Aero Aqua Bio-Mechanisms, vol. 3, (Tokyo, Japan), pp. 103–108, J-STAGE,
6 2013.
[18] D. J. Moreau and L. A. Brooks and C. J. Doolan, “On the noise reduction
mechanism of a flat plate serrated trailing edge at low-to-moderate reynolds
number,” in 18th AIAA/CEAS Aeroacoustics Conference (33rd AIAA Aeroa-
coustics Conference), (Colorado Springs, CO, USA), pp. 1–20, AIAA, 6 2012.
[19] G. Ingram, “Wind turbine blade analysis using the blade element momentum
method,” 2011. Durham University.
[20] P.M. Sforza, Theory of Aerospace Propulsion. Woburn, MA, USA: Elsevier
Science, 2016.
[21] M. Pascu, Modern Layout and Design Strategy for Axial Fans. PhD thesis,
Institute of Fluid Mechanics LSTM Erlangen-Nuremberg University, Erlangen,
Germany, 2008.
[22] M. Pascu, M. Miclea, P. Epple, A. Delgado, and F. Durst, “Analytical and nu-
merical investigation of the optimum pressure distribution along a low-pressure
axial fan blade,” Proceedings of the Institution of Mechanical Engineers, Part
C: Journal of Mechanical Engineering Science, vol. 223, no. 3, pp. 643–657,
2009.
[23] S. Lieblein, F. C. Schwenk, and R. L. Broderick, “Diffusion factor for estimating
losses and limiting blade loadings in axial-flow compressor blades,” research
memo, Lewis Research Center NASA, 1953.
[24] D.R. Kirk, “Air-breathing engines: Overview of axial compressors.” https:
//www.slideserve.com/bernad/mae-4261-air-breathing-engines. PPT
Presentation, Online, accessed 2020-05-01.
[25] S. Castegnaro, “Fan blade design methods: Cascade versus isolated airfoil ap-
proach—experimental and numerical comparison,” in ASME Turbo Expo 2016:
Turbomachinery Technical Conference and Exposition (American Society of Me-
chanical Engineers, ed.), vol. 1, ASME, 2016.
[26] S.L. Dixon and C.A. Hall, Fluid Mechanics and Thermodynamics of Turboma-
chinery. Oxford, UK: Butterworth-Heinemann Inc, 2014.
[27] E. Andreadis, “Design of a low speed vane-axial fan,” Master’s thesis, Depart-
ment of Power Engineering and Propulsion, Cranfield University, Cranfield,
UK, 2011.
[28] R.A. Wallis, Axial Flow Fans and Ducts. Malabar, FL, USA: Krieger Publishing
Company, 1993.
[29] A.R. Howell, “Fluid dynamics of axial compressors,” Proceedings of the Insti-
tution of Mechanical Engineers, vol. 153, no. 1, pp. 441–452, 1945.
57
Bibliography
58
A
Design Method 2 Data
I
A. Design Method 2 Data
II
A. Design Method 2 Data
Airfoils
III
B
Flow Diagrams
Start
Cascade Parameters
rhub , rtip ,
n, Qdesign , z, l1
Velocity triangles
u = 2πrn
Qdesign
wm =
π(rt2 − rh2 )
wm
wu1 = u(axial entry) → tan γ1 =
wu1
1 2πn 2
" 2 # !
rt
= r − 2 (r − rh )
1.35
+1 rh2
tan γ2r rwm rh
wm
tan γ2 = → wu2
wu2
(γm , ym )
Continued
IV
B. Flow Diagrams
γ(y) = Ay 2 + By + C
y = y1 → γ = γ2
y = ym → γ = γm
y = y2 → γ = γ1
γ2
γ(y) = γi
γ1
∆x = ∆y tan γ(y)
xi = xi−1 + ∆x
q
l2 = (x2 − x1 )2 + (y2 − y1 )2
Change No
l2 = l1
γm , ym
Yes
Cascade Performance
wm
w(y) =
sin γ(y)
max w(y) − w2
DF =
max w(y)
Stop
V
B. Flow Diagrams
Start
ωs Qdesign , ∆p
Ns
N
∆s Dt Dh ν
b, Va z
Stop Va
rt , ..., rh U φ, β1
Yes
Converged No ηh,i s, c
Vu2 , β2
σ
βm
ηh,f
βm CL , Re
Choose an
appropriate
design approach
γ α, Cd
i, θ
VI
C
CAD Models
VII
C. CAD Models
VIII
C. CAD Models
IX
C. CAD Models
X
C. CAD Models
XI
C. CAD Models
XII
D
CFD Results
Table D.1: CFD results at/near the peak efficiency point at different rotational
speeds.
XIII
D. CFD Results
XIV
D. CFD Results
Figure D.1: Pressure distribution on the suction side of model 1 at 2000 rpm.
Figure D.2: Pressure distribution on the suction side of model 2 at 2000 rpm.
XV
D. CFD Results
Figure D.3: Pressure distribution on the suction side of model 3 at 2000 rpm.
Figure D.4: Pressure distribution on the suction side of model 4 at 2000 rpm.
XVI
D. CFD Results
Figure D.5: Pressure distribution on the suction side of model 5 at 2000 rpm.
Figure D.6: Pressure distribution on the suction side of model 6 at 2000 rpm.
XVII
D. CFD Results
Figure D.7: Pressure distribution on the suction side of model 7 at 2000 rpm.
Figure D.8: Pressure distribution on the suction side of model 8 at 2000 rpm.
XVIII
D. CFD Results
Figure D.9: Pressure distribution on the suction side of model 9 at 3000 rpm.
Figure D.10: Pressure distribution on the suction side of model 10 at 3000 rpm.
XIX
D. CFD Results
Figure D.11: Pressure distribution on the suction side of model 11 at 2000 rpm.
XX