DESIGN IMPROVEMENTS OF SIROCCO TYPE FANS BY

MEANS OF COMPUTATIONAL FLUID DYNAMICS AND

STEREOSCOPIC PARTICLE IMAGE VELOCIMETRY
Stefan FRANK1, Manoochehr DARVISH1,
Bastian TIETJEN1 and Adam STUCHLIK1,2
1

University of Applied Sciences HTW Berlin,

Wilhelminenhofstr. 75A, 12459 Berlin, Germany
email: stefan.frank@htw-berlin.de
2

now: I-D-E-A-S, Ulmenstr. 28, 13595 Berlin, Germany

SUMMARY
This paper treats the investigation of the complex internal flow in Sirocco type fans by means of
both numerical and experimental methods. In order to improve the design and efficiency of
these fans, Computational Fluid Dynamics (CFD) with high-quality, high-resolution structured
and unstructured grids are used. Simulations are verified by both integral measurements at a
high-precision chamber test rig according to ISO 5801 and by means of Stereoscopic Particle
Image Velocimetry (Stereoscopic PIV). Comparison between numerical and experimental data
shows a high consistency. Performance curves and efficiencies can be predicted with less than
five per cent deviation in a wide operating range.

INTRODUCTION
Sirocco type fans, sometimes also called squirrel-cage fans, consist of a radial impeller with short
forward curved blades and a scroll housing [1, 2]. They are widely used in industry, especially in
edificial and automotive HVAC applications where relatively high mass flows, size compactness
and low noise levels are required. The main disadvantage is their relatively low efficiency and the
need for empirical designing. The latter results from the fact that the performance curve of Sirocco
fans cannot be derived directly from the Euler turbo-machine equation. Therefore, the nondimensional coefficients of Sirocco fans cannot be found on the Cordier curve [1], but form a small
separate band below it (compare Table 1). The flow in Sirocco fans is mostly unsteady, fully threedimensional, turbulent and shows virtually unavoidable flow separation within the blades and the
scroll, even at the best efficiency point. Experimental optimization of these fans has been
investigated for a long time, e.g. by Roth [2] and many others. The main design parameters such as
blade number, ratio of inlet to outlet diameter etc. can be found in Figure 1a and Table 2
respectively. For instance, in order to optimize the housing, both the scroll width W and the
aperture angle s can be varied. Other parameters including the scroll outlet area have been kept
constant according to Table 2 in this investigation. Figure 1b shows the max. obtainable static
efficiencies for a given set of parameters (Table 2) as a function of scroll aperture angle and flow
_______________________________________________________________________________
FAN 2012

Senlis (France) 18-20 April 2012

FAN 2012
Senlis (France), 18 20 April 2012

coefficient as measured on conventional chamber test rigs. It is evident, that highest efficiencies can
be achieved only for scroll angles between 4,5 and 5,5. As a consequence, the respective nondimensional coefficients and characteristic numbers have optimum values in a very narrow range.
However, if higher flow-rates or pressure coefficients are needed, the optimum design may differ
considerably at the cost of significantly lower efficiencies.
Table 1: Coefficients and non-dimensional numbers for Sirocco type fans

Name

Equation

Flow coefficient

Overall Range

Max. Efficiency

0,3 to 1,0

0,35 0,05

V&
w D 22 n

0,2 to 0,6

0,2 0,03

2 p st
2 D 22 n 2

1,6 to 3,0

2,0 0,1

1,4 to 2,0

2,0 0,1

0,3 to 0,6

0,35 0,05

Flow coefficient w
(based on rotor width w)

w =

Static pressure coefficient

Non-dimensional diameter

= D2 4

Non-dimensional speed

= n

Circumferential
Reynolds number Reu

4 V&
D 32 n

2 p st

2
2
V&

V& 0,5
2
(2 pst )0,75
Reu =

1 105 to 1 106

n D2

The main disadvantage of the experimental optimization is the enormous effort and the need for
idealization of the set-up. Thus, reliable methods for the analytic design and numerical prediction of
fan performance and efficiency - taking into account the actual operating conditions - are needed.
s
60

W
2

Static Efficiency in %

50

40

30

20

5,0

5,5

4,5
4,0

1
D1

10

3,5
3,0

D2

0
0

0,1

0,2

0,3

0,4

Flow Coefficient

Figure 1: a) Geometry parameters of Sirocco type fans

b) Static efficiencies as a function of scroll aperture angle s and design parameters acc. to Table 2 for an outlet
diameter D2 of 160 mm and a Reynolds number Reu of 1 105, after [3]

FAN 2012
Senlis (France), 18 20 April 2012

Table 2: Optimum design parameters for max. efficiency of Sirocco type fans [e.g. 1, 2]

Name

Symbol

Value

40

Ratio of Inlet to Outlet Diameter

D1/D2

0,8

Ratio of Rotor Width to Outlet Diameter

w/D2

0,4

Ratio of Rotor to Scroll Width

w/W

1,1

Blade Inlet Angle

80

Blade Outlet Angle

170

Aperture Angle of Logarithmic Scroll

Blade Number

EXPERIMENTAL SET-UP
The experimental investigations are performed on a special fan test rig with variable speed drive. In
order to determine integral performance values like flow-rate, pressure increase and power the fan
is connected to a highly accurate chamber test rig according to ISO 5801 / DIN 24163
(manufacturer: ILK Dresden). A threefold orifice plate pipe section and a twofold differential
pressure cascade allows for max. uncertainties of less than one per cent of the reading for the flowrate in the entire operating range of 10 to 2.000 m/h. The respective uncertainties for the pressure
increase amount to less than half a per cent of the reading. However, uncertainties of fan efficiency
amount to one to seven per cent depending on the load of the mechanical torque sensor. In addition
to these integral measurements, instantaneous velocity field measurements are performed by
Particle Image Velocimetry (PIV). This is a state-of-the-art optical method of flow measurement
used in research [4, 5].

Stereoscopic PIV

2D-PIV
Scroll

Rotor

CCD
Camera

Image

Tongue

Laser
Lightsheet

Scheimpflug
Condition

CCD 1

CCD 2
Camera 2

Camera 1

Frame CCD1

Reduced
Image

Frame CCD2

Figure 2: Schematic of the set-up for Particle Image Velocimetry (PIV) measurements (top view)
Left: 2D-PIV (2D-2C); Right: Stereoscopic PIV (2D-3C)

FAN 2012
Senlis (France), 18 20 April 2012

Figure 2 shows the schematic set-up for measurements of the two- or three-dimensional velocity
field in the measuring plane (2D plane - 2 velocity components: 2D-2C and 2D-3C, respectively).
Scroll and side walls are manufactured from acrylic glass in optical quality. This allows for the
entrance of the laser light-sheet and the view of the flow field for the CCD cameras (see Figure 3).

Figure 3: Experimental set-up for Stereoscopic Particle Image Velocimetry in two dimensional planes in order to
measure the three dimensional velocity field (i.e. 2D-3C PIV). The two CCD cameras are to be seen in the foreground,
the transparent scroll of the fan with illuminated laser light sheet and the chamber test rig in the background

The PIV measurements are performed using a system by Intelligent Laser Applications ILA, Jlich.
A pressurized atomizer with DEHS provides the seeding of the flow via a cyclone to the inlet of the
auxiliary fan of the chamber test rig. Hence, the particle inlet is about five meters upstream of the
test-section. This allows for a very homogeneous seeding of the flow field with particles of
adequate size and quality. The parameters of the used PIV System are given in Table 3.
Table 3: Parameters of the Particle Image Velocimetry System

Parameter

Value

Resolution of CCD Cameras

2048 Pixel

Max. Energy per Laser Shot

220 mJ

Double Frames per Plane and Operating Point

100

Repetition Rate

5 Hz

In order to evaluate the PIV data, the flow field has to be calibrated for each measuring plane
(Figure 4) using a well-known target with an x,y-grid. The latter is exactly centred via the scroll,
whereas the distance from the wall, i.e. the position of the laser light sheet, is adjusted by special
pins. Since reflections of the laser light can cause severe problems during the evaluation process
some parts of the scroll, namely the tongue, are covered with a Rhodamin containing tape. Thus, the
green laser light is absorbed and orange light is re-emitted. The CCD cameras are equipped with
green interferential filters. As a consequence, high quality pictures can be obtained also in these
potentially difficult areas. The frames are not correlated to the variable rotor position, which results
in a statistical average of the flow field. Each double frame is evaluated by a cross correlation
process followed by several subsequent adaptive cross correlations and global and local filtering

FAN 2012
Senlis (France), 18 20 April 2012

using the VidPIV software [6]. Outliers are identified by means of filtering, subsequently removed
and interpolated. Eventually, all double frames are averaged in order to calculate the velocity field.
The RMS value of velocity for a single interrogation window is about five per cent, thus yielding a
statistical error for its average of about one per cent at a confidence level of 95 %. Hence, the
number of 100 double frames is sufficiently high.

a)

b)
Plane 5

Plane 1

ROTOR

PIV analysis section for

several measuring planes

TONGUE
Figure 4: a) Measuring planes for PIV measurements
b) Single frame of the CCD-camera of the Sirocco fan with scroll, rotor (D2 = 200 mm) and tongue

NUMERICAL FLOW SIMULATION

For the numerical flow calculations commercially available Finite Volume Method software with
Reynolds Averaged Navier-Stokes (RANS) equation is used [7, 8]. Mesh generation is performed
by both structured grids with hexahedral cells [9] and unstructured grids with polyhedral cells [10].
Unstructured grids provide the advantage of higher flexibility than structured grids. For equivalent
cell numbers and comparable cell quality criteria, structured and unstructured grids may yield the
same degree of reliability. The calculations using structured grids are slightly faster than with
unstructured grids. However, this is usually more than compensated by the faster meshing process
for the latter.
Reliable numerical simulations of Sirocco type fans are characterized by a cell number of about
four million (two thirds in the rotor, one third in the stator), the use of state-of-the-art turbulence
models such as the Shear-Stress-Transport (SST) or Realizable k-epsilon model. The adequate
resolution of the boundary layer is crucial for the simulation. As a consequence the nondimensional wall distance y+ has to be controlled down to values about one to two in the rotor,
especially around the blades. In these investigations steady calculations using the "Moving
Reference Frame" approach (also called "Frozen Rotor") are performed exclusively. More details of
the CFD calculations are described elsewhere [9 - 11].
Figure 5a shows a typical performance curve of a Sirocco type fan exhibiting a region of instability
in the throttle range, whereas Figure 5b depicts a comparison of numerical data obtained using
different turbulence models and experimental data. For a wide range of fan operation a quite high
degree of quantitative accordance between experiments and simulations has been obtained. A
fundamental advantage of CFD over any experimental method is that all field properties such as
pressure and velocity are available for the entire flow field. Due to the employment of RANS
equations the stationary macroscopic vortices can be shown here.

FAN 2012
Senlis (France), 18 20 April 2012

b)

Torque in Nm

Static Pressure in Pa

a)

Figure 5: a) Schematic of the performance and power curve

b) Comparison of experimental and numerical data (for polyhedral mesh)

A general assumption in the design of Sirocco fans is that the flow is more or less two-dimensional.
Actually, the flow is three-dimensional almost everywhere in the flow field (also [2]). Figure 6, for
instance, shows simulated streamlines for the Best Efficiency Point of the fan, i.e. at a flow-rate of
450 m/h, rotational speed of n = 1.000 min-1 and a static efficiency of about 60 %. The vortical
flow downstream of the tongue can clearly be identified. Furthermore, single streamlines can be
traced, that lead from the rotor outlet through the gap between inlet nozzle (not shown here) and
shroud back to the rotor inlet. Remarkably, there are hints (e.g. [2]) that when taking design
measures against these backflows, the static efficiency may drop even further.

TONGUE
SHROUD

Vortical Flow

Figure 6: Streamlines from CFD for the Best Efficiency Point (BEP) at n = 1.000 min-1

FAN 2012
Senlis (France), 18 20 April 2012

ROTOR
TONGUE
a)

ROTOR
TONGUE
b)

(B)

(A)

(C)
(D)
ROTOR
TONGUE
c)
Figure 7: 2D flow field of the fan near shroud (plane 5)
in the Best Efficiency Point at a flow-rate of 450 m/h and n = 1.000 min-1
a) Absolute velocity from 2D-PIV measurements
b) Absolute velocity from numerical simulations (CFD)
c) Point wise calculated differences from a) and b)

FAN 2012
Senlis (France), 18 20 April 2012

COMPARISON OF RESULTS
The experimental investigations start with 2D-PIV and just one CCD camera with a resolution of
2048 pixels and a focus length of 50 mm. The camera is oriented perpendicular to the laser light
sheet plane (see Figure 2). Measurements are taken in five planes (as shown in Figure 4a) for
operating points from 150 to 900 m/h. Plane 5 is the most remote measuring plane close to the
shroud, yielding the greatest experimental challenges. For comparison the 2D-PIV and CFD results
at the BEP are shown in Figure 7a and b as colour-plots using the free open source software
ParaView [12]. Concerning the simulations, steady RANS calculations with the SST turbulence
model are performed. A high degree of qualitative accordance between the two methods can be seen
at a glance. However, in order to make a more detailed statement a common evaluation grid with
about six thousand points is generated for the PIV and CFD data. Subsequently, the vectorial
differences of absolute velocity have been calculated point by point. It is worth mentioning, that
using this method the highest possible relative deviation may reach up to 200 % for vectors of equal
value but opposite orientation.

a)

Stereoscopic PIV

Stereoscopic PIV

ROTOR

b)

CFD

TONGUE

CFD

TONGUE
ROTOR

c)
Difference

ROTOR

Difference

TONGUE

Figure 8: Field of 3D absolute velocities near shroud (plane 5)

in the Best Efficiency Point at a flow-rate of 450 m/h and n = 1.000 min-1
a) Stereoscopic PIV measurements of velocity magnitude and vz-component
b) CFD data of velocity magnitude and vz-component
c) Calculated differences between Stereoscopic PIV measurements and CFD data

FAN 2012
Senlis (France), 18 20 April 2012

As it can be seen from Figure 7c, there is a high quantitative consistency between experimental and
simulated values in the core region of the flow. Areas with significantly higher deviations can be
distinguished into two zones: On the one hand, these are boundary areas close to either the scroll
wall (A) and (B) or the rotor (C), where directly reflected laser light is obviously a source of error.
On the other hand, higher deviations can be found downstream of the tongue (D). It is assumed that
these deviations arise mainly from the comparison of "Moving Reference Frame" results for the
CFD and averaged PIV results (concerning the rotor position). A possible remedy is taking PIV
measurements for correlated rotor positions only.
Subsequently Stereoscopic PIV measurements are performed using two CCD cameras (Figures 2
and 3). The cameras can be mounted on a rigid x,y,z-traverse, whereas the optics for the laser lightsheet is mounted on a z-traverse. Thus, cameras and light sheet can be moved simultaneously in the
z-direction (i.e. direction of the scroll width) without re-adjusting the focus, once the correct setting
has been found. The CCD cameras are positioned at an angle of 20 to 25 to the normal of the lightsheet plane. Because of that motorized Scheimpflug tilt devices have to be used in order to keep the
whole measurement plane in focus (see Figure 2). In principle, the ideal angle between each camera
and the normal of the light-sheet plane is 45. However, for this kind of internal flow, the maximum
angle is restricted to about 25. Otherwise, too large portions of the field of interest can not be
evaluated, since they are shadowed either by the scroll, the rotor or the tongue. This applies
especially to the most remote measurement plane 5. The light sheet has a max. thickness of 5 mm,
and again 100 double frames at five Hz are taken for each plane and operating point, respectively.

PLANE 5

Figure 9: Secondary flow (double vortex) in the scroll right above rotor axis as simulated by CFD

Figure 8a shows the z-component of absolute velocity for the Stereoscopic PIV measurements.
Only those areas can be evaluated that are common to both cameras. Unfortunately, some potential
areas of interest, such as the zone directly downstream of the tongue, are excluded from the
stereoscopic evaluation. It can be noticed that the z-component is especially high close to the scroll
(negative z-values) and to the rotor (positive z-values), but relatively low in the core flow. Again,
comparison with the respective simulated values exhibit a high degree of accordance (Figure 8b and
8c). The pronounced vz-component adjacent to the scroll and the rotor indicates a secondary flow in
form of a vortex. Roth [2] has already reported such double vortex systems using hot-wire
anemometry. The respective CFD data for the current investigation is depicted in Figure 9 and
exhibits the same phenomenon. The present results document the suitability of Stereoscopic PIV for
these kinds of internal flows. In addition to the 2D-PIV results even secondary vortex flows have
been detected. It can be concluded, that the quality of the experimental data used for the evaluation
of CFD simulations has been improved significantly.

FAN 2012
Senlis (France), 18 20 April 2012

10

NUMERICAL OPTIMZATION
The possibilities for numerical optimization of Sirocco type fans are shown for a special
application: Based on the experiences made beforehand the Sirocco fan system of a home appliance
shall be optimized using numerical methods. In this case study, ANSYS CFX [14] is applied using
structured ICEM meshes. The geometry is based on the original CAD model (Figure 10a). The
simulations use RANS equations and the SST turbulence model with a highly resolved grid
(hexahedral cells with about five million node points). Starting from a reference configuration,
which is based on previous designs, various new design variants are investigated in detail. Main
variations concern the rotor outlet diameter D2 and the scroll width W (compare Figure 1).

a)

b)
Tongue
Recirculation Zone

Figure 10: Sirocco fan system for a home appliance

a) Structured ICEM mesh of the reference geometry with about five million node points
b) Simulated velocity distributions for the reference geometry (left) and the optimized variant (right)

For evaluation purposes, the reference geometry has been set-up and tested on a chamber test rig,
yielding deviations of less than five per cent between experiment and simulation. The CFD
calculations show that a smaller, yet slightly eccentrically positioned rotor [13] generates a much
more homogeneous velocity distribution in the scroll (Figure 10b, left). Contrary to the optimized
variant (left) the reference geometry (right) exhibits a pronounced vortex and recirculation zone
around and downstream of the tongue. The downstream developing velocity profiles in the scroll
and the subsequent air channel are depicted in Figure 11. Again the velocity distribution is more
homogeneous in the optimized version. For about the same pressure rise, flow-rate and rotational
speed the optimized geometry is equipped with a twelve per cent smaller outlet diameter D2 than the
reference case. This leads to a reduction of shaft power of about 30 % at BEP and thus to a
considerable gain in system efficiency by about 25 percentage points. Table 4 compares the nondimensional coefficients between the reference and the optimized variant. Finally, the coefficients
are closer to the optimum values for max. efficiency as given in Table 1. This means as well, that
there still could be potential for further improvement of the efficiency, given that the required
design changes can be realized in a production home appliance.

FAN 2012
Senlis (France), 18 20 April 2012

11

Figure 11: Velocity distribution in the air channel.

Top: reference geometry. Bottom: Optimized Version
Table 4: Coefficients and non-dimensional numbers for different design variants

Name

Reference

Optimized

Flow coefficient

0,14

0,21

Flow coefficient w (based on rotor width w)

0,13

0,16

Static pressure coefficient

1,1

1,4

Non-dimensional diameter

2,7

2,3

Non-dimensional speed

0,36

0,35

Circumferential Reynolds number Reu

2,1 105

CONCLUSIONS
The performance of Sirocco type fans can be tested on conventional chamber test rigs with low
uncertainties concerning flow-rate, pressure rise and efficiency. However, the conventional
optimization of these systems require a lot of experimental effort and is very time consuming.
Hence, optimization by numerical means such as Computational Fluid Dynamics would be
favourable. In order to do so a reliable data base for the assessment of numerical results has to be
provided. 2D Particle Image Velocimetry measurements can be performed with relatively low
experimental effort, given that an optical access for the laser light-sheet and the CCD camera can be
realized. Problems due to the necessary seeding of the flow and direct reflections of laser-light can
be overcome. Comparison between velocity fields from experiment and simulation are already quite
consistent. However, the third, out-of-plane velocity component remains undetected in principle.
Hence, Stereoscopic Particle Image Velocimetry is applied that provides instantaneous 3D velocity
fields in the measurement plane. Thus, besides the main flow the secondary vortex flow can be
detected and the overall quality of measurement increases significantly. However, this is at the cost
of a higher experimental effort. Furthermore, for this kind of internal flow not the whole velocity
field of interest can be evaluated due to the shadowing of the scroll, the tongue and the rotor.
Nevertheless, data is obtained that could not be achieved otherwise.

FAN 2012
Senlis (France), 18 20 April 2012

12

Numerical calculations are performed using steady RANS equations with SST turbulence model
and highly resolved grids. Both structured and unstructured grids with hexahedral and polyhedral
cells respectively can be applied. For a cell based solver and a comparable number of elements,
both mesh types will yield equally good results at a similar numerical effort. However, mesh
generation is much more flexible and significantly faster with unstructured grids. Comparison
between numerical and experimental data shows deviations of less than five per cent for pressure
rise and efficiency for a wide operating range. Direct and point wise comparison of the velocity
fields show a high degree of consistency between the two methods. Once the numerical method has
been established and proved to be accurate it can be used for the virtual optimization of e.g. Sirocco
fans before building prototypes. Starting from a previous design an optimized design variant could
be developed, that has a considerably smaller size and yields a much higher static efficiency at the
best efficiency point. Therefore, Computational Fluid Dynamics in conjunction with Particle Image
Velocimetry is a useful way to reliably improve Sirocco type fans.

ACKNOWLEDGEMENTS
The investigations have been sponsored by the German Federal Ministry of Education and Research
(BMBF) under contract numbers NUBAT FKZ 17N1507 and NEGOT FKZ 17067X10 (project
coordination by Arbeitsgemeinschaft industrieller Forschung AiF). The laboratory equipment has
partly been sponsored by the European Fund for Regional Development (EFRE IMPACT). The
main author was able to carry out some of the investigations in a research semester in winter
2010/11 granted by his university. All research support is gratefully acknowledged!

BIBLIOGRAPHY
[1] Eck, B. - Fans, Pergamon Press, 1973
[2] Roth, H.W. - Optimization of Sirocco Type Fans (in German), PhD Thesis, TH Karlsruhe, 1980
[3] Teres Duaso, R. - Experimental investigation on the influence of the scroll housing on the performance of Sirocco type fans. Final Project Thesis, Universidad de Zaragoza / HTW Berlin, 2011
[4] Raffel., M.; Willert, C.; Werely, S.; Kompenhans - Particle Image Velocimetry, Springer, 2007
[5] Stuchlik, A.; Frank, S. - Numerical Calculation and Design of Sirocco Type Fans (in German).
Final Report BMBF/AiF-Research Project FKZ 17N1507. HTW Berlin, 2011
[6] VidPIV 4.6 Manual. ILA GmbH, Jlich/Germany
[7] Ferziger, J.H.; Peric, M. - Computational Methods for Fluid Dynamics, Springer, 2002
[8] Star CCM+ 5.04 User Manual. CD Adapco Ltd.
[9] Stuchlik, A.; Frank, S. - Performance Investigations of Sirocco Fans by means of
Computational Fluid Dynamics. Proc. of ISROMAC 13, Hawaii, USA, 2010
[10] Darvish, M. - Numerical Investigations on the Performance Characteristics of Radial Fans
with Forward Curved Blades by means of CFD. Master Thesis, HTW Berlin, 2010
[11] Darvish, M.; Frank, S. - Sirocco Fan Modelling: from Selecting a Proper Turbulence Model to
the Influence of the Cell Shape. fan2012, Paper 073, 2012
[12] ParaView User's Guide (V. 3.10)
[13] Stuchlik, A.; Frank, S. - Eccentrical Rotor Position as a Means of Optimization of Sirocco
Type Fans (in German). HLH, Bd. 62 Nr. 2, 2011
[14] Ansys Inc., "CFX User Guide", Version 12