You are on page 1of 13

2008­01­0735

CFD Simulations of an Automotive HVAC Blower: Operating 
under Stable and Unstable Flow Conditions 
Moulay Bel­Hassan and Asad Sardar
Air International Thermal Systems 

Reza Ghias
ANSYS Inc.

Copyright © 2007 SAE International

ABSTRACT head;   a   compressor;   a   condenser;   a   thermostatic 


expansion   valve   or   orifice   tube   and   associated 
Computational Fluid Dynamics (CFD) is heavily used in  plumbing. The primary function of the HVAC system is to 
automotive  HVAC  industry  in  order   to  reduce   the  time  provide   heated   and   cooled   air   to   the   passenger 
and   cost   in   design,   optimization,   and   development   of  compartment   for   passenger   comfort   and   to   maintain 
different   components   [2].   Correct   prediction   of   the  clear   vision   (both   defogging   and   deicing)   from   the 
aerodynamic   characteristics   of   an   HVAC   blower   is  windshield   and   side   windows.   Air   is   supplied   by   the 
crucial in development of the accurate CFD models for  blower and delivered to the passenger through the duct 
the   whole   HVAC  system.   CFD   models   are   extensively  via the HVAC casing and heat exchangers [1]. 
used   in   the   optimization   of   both   thermal   and   airflow 
characteristics of automotive HVAC [3­5]. In the last decade, the automotive air­conditioning (AC) 
system   market   has   increased   rapidly.   Due   to   this 
In   this   study   we   have   performed   CFD   simulations   for  widespread   use,   large­scale   production   and   cost­
different blower operating conditions in order to assess  effectiveness have been the primary drivers influencing 
the   CFD   results   in   prediction   of   the   aerodynamic  the   automotive   system   design.   Although   these   drivers 
performance   in   an   automotive   HVAC   forward   curved  remain, focus on the improvement of the overall system 
(FC)   centrifugal   blower.   The   realizable   k­ε  turbulence  performance   increases   rapidly.   Therefore   CFD 
model   was   used   on   the   Reynolds   Averaged   Navier­ simulations of the internal flow in AC Units, as well as 
Stokes approach to model complex flow field properly. experimental   investigations   have   received   more 
attention in recent years. Some of the key performance 
Steady   state   analysis  showed   good   correlation   for   the  issues (e.g., evaporator performance, evaporator freeze 
stable   flow   conditions  (high   airflow   and   low  pressure),  control,   total   airflow,   airflow,   and   temperature 
whereas  this approach showed large discrepancies for  distribution)   are   related   to   the   blower.   Better 
unsteady flow conditions (low airflow and high pressure).  understating   of   the   flow   field   characteristics   and 
By a transient simulation and realizable  K­ε  model, the  turbulent structure in such components can lead to the 
CFD   analysis   showed   good   correlation   compared   to  development of AC systems that provides higher thermal 
experimental test results. and   flow   performances   and   significant   noise   level 
reduction.
INTRODUCTION
HVAC   system   resistance   typically   varies   from   low   in 
A   typical   HVAC   system   includes   an   air   conditioning 
panel   full   cold   mode   (maximum   torque   load;   high 
module   comprised   primarily   of   a   blower,   evaporator, 
airflow)   to   high   in   floor   full   hot   (low   torque   load;   low 
heater core, doors, and actuators; an electronic control 
airflow). Characterizing the flow in the maximum torque  transport   equations   allows   the   turbulent   velocity   and 
conditions   is   essential   for   blower   motor   selection,   and  length   scales   to   be   independently   determined.   The 
thereby   determines   the   maximum   power   consumption.  standard  K­ε  model   falls   within   this   class   and   has 
Toward   the   other   end   of   the   spectrum,   high   pressure  become   the   workhorse   of   practical   engineering   flow 
and   low   airflow   operating   conditions   also   need   to   be  calculations   in   the   time   since   it   was   proposed   by 
properly characterized since it determines the maximum  Launder  and Spalding [10]. Robustness,  economy, and 
amount of airflow available for restrictive heater modes  reasonable accuracy for a wide range of turbulent flows 
(further restrictions when adding filters electric heaters).  explain its popularity in industrial flow and heat transfer 
Additionally,   for   enhanced   end   users   comfort,   OEMs  simulations.   It   is   a   semi­empirical   model,   and   the 
require multiple airflow and temperature detent settings.  derivation   of   the   model   equations   relies   on 
In   order   to   satisfy   the   various   blower   operating  phenomenological considerations and empiricism.
conditions   of   the   automotive   HVAC,   complete   blower 
performance   curve   needs   to   be   characterized   (i.e.,  As the strengths and weaknesses of the standard model 
simulated). Table 1 presents the customer specifications  K­ε  have   become   known,   improvements   have   been 
in   terms   of   the   blower   speed   and   the   corresponding  made to the model to improve its performance. Two of 
airflow for different HVAC operating modes (i.e. various  these  variants  are  available   in  FLUENT:  the   RNG  K­ε
flow   system   resistance   characteristics).   HVAC   is  model [11] and the realizable K­ε model [9].
required   to   perform   under   a   wide   range   of   blower 
The standard  K­ε  model [10] is a semi­empirical model 
speeds and airflow rates.
based on model transport equations for the turbulence 
kinetic energy (K) and its dissipation rate (ε). The model 
transport   equation   for  K  is   derived   from   the   exact 
Table 1: Airflow (l/s) quantities for a typical blower setting. Mi  equation, while the model transport equation for   ε  was 
is   the   blower   speed   ranging   from   high   to   low.   OAS   is   for  obtained   using   physical   reasoning   and   bears   little 
outside air.
resemblance   to   its   mathematically   exact   counterpart.  
Air  Inlet  Temp  M5 M4 M3 M2 M1 Low
Delivery Air Door In the derivation of the K­ε model, the assumption is that 
Vent OSA FC 107 93 76 58 38 17 the flow is fully turbulent,  and the effects of molecular 
Bi­level OSA 50 % 124 107 94 71 46 21 viscosity   are   negligible.   The   standard  K­ε  model   is 
Floor OSA FH 78 71 53 38 25 10
therefore valid only for fully turbulent flows.
Defog OSA FH 79 70 56 40 25 9
Defrost OSA FH 71 54 42 32 21 9
The   realizable  K­ε  model   [9] is   a   relatively   recent 
developed   model   and   differs   from   the   standard  K­ε
model in two important ways:
NUMERICAL SIMULATIONS
• The realizable K­ε model contains a new formulation 
FLUENT   V6.3   ­   Realizable  K­ε  model   was   used   as   a 
for the turbulent viscosity.
solver in this study. Flow field was assumed steady state 
incompressible   (except   for   one   case)   and   a   second 
• A new transport equation for the dissipation rate,  ε, 
order  implicit  finite  volume   scheme   was   used   to  solve 
has   been   derived   from   an   exact   equation   for   the 
three   dimensional   Reynolds   Averaged   Navier­Stokes 
transport of the mean­square vorticity fluctuation. 
(RANS) equations for all cases. In the following sections 
some details about turbulence models and numerical set 
up are described.
One   of   the   weaknesses  of  the   standard  K­ε  model  or 
other traditional K­ε models lies within the equation that 
TURBULENCE MODEL
is used to model the dissipation rate (ε). The well­known 
The simplest "complete models'' of turbulence are two­ round­jet anomaly (named based on the finding that the 
equation models in which the solution  of two separate  spreading   rate   in   planar   jets   is   predicted   reasonably 
well,   but   prediction   of   the   spreading   rate   for 
axisymmetric jets is unexpectedly poor) is considered to  This   extra   rotation   effect   has   been   tested   on   single 
be   mainly   due   to   the   modeled   dissipation   equation.   rotating reference  frame systems and  showed superior 
behavior over the standard  K­ε model. However, due to 
The   realizable  K­ε  model   proposed   by   Shih   [7]   was  the nature of this modification, its application to multiple 
intended to address these deficiencies of traditional K­ε reference   frame   systems   should   be   taken   with   some 
models by adopting the following:  caution. 

• A new eddy­viscosity formula involving a variable C  The realizable K­ε model has been extensively validated 
originally proposed by Reynolds [8]. for   a   wide   range   of   flows [9­12],   including   rotating 
• A new model equation for dissipation (ε) based on  homogeneous shear flows, free flows including jets and 
the  dynamic  equation  of  the  mean­square  vorticity  mixing   layers,   channel   and   boundary   layer   flows,   and 
fluctuation separated flows. For all these cases, the performance of 
the model has been found to be substantially better than 
that of the standard  K­ε  model. It is worth to note that 
The   term   "realizable''   means   that   the   model   satisfies  the realizable K­ε model resolves the round­jet anomaly 
certain   mathematical   constraints   on   the   Reynolds  i.e. it predicts the spreading rate for axisymmetric jets as 
stresses, consistent with the physics of turbulent flows.  well as that for planar jets.
Neither the standard K­ε model nor the RNG K­ε model 
is realizable [2]. MULTIPLE ZONES APPROACH

An   immediate   benefit   of   the   realizable  K­ε  model   is  Numerical   simulations   of   multiple   moving   parts   or 
relatively   accurate   prediction   of   the   spreading   rates   in  stationary surfaces which are not surfaces of revolution 
both   planar   and   round   jets.   It   is   also   likely   to   provide  have   encounter   challenges.   For   these   problems,   one 
superior   performance   for   flows   involving   rotation,  must   break   up   the   model   into   multiple   fluid/solid   cell 
boundary   layers   under   strong   adverse   pressure  zones,   with   interface   boundaries   which   separate   the 
gradients, separation, and recirculation. zones.   Zones   with   moving   components   can   then   be 
solved   using   the   moving   reference   frame   equations, 
Both   the   realizable   and   RNG  K­ε  models   have  whereas   stationary   zones   can   be   solved   with   the 
shown substantial improvements over the standard  stationary   frame   equations.   The   manner   in   which   the 
K­ε  model   where   the   flow   features   include   strong  equations   are   treated   at   the   interface   lead   to   two 
streamline   curvature,   vortices,   and   rotation.   Since  approaches which are supported in FLUENT [2]:
the   model   is   still   relatively   new,   it   is   not   clear   in 
exactly   which   instances   the   realizable  K­ε  model  • Multiple Rotating Reference Frames (MRF).
consistently outperforms the RNG model. However,  • Sliding Mesh Model (SMM). 
initial studies have shown that the realizable model 
provides the best performance of all the  K­ε  model  MRF is a steady­state approximation in which individual 
versions   for   several   validations   of   separated   flows  cell   zones   move   at   different   rotational   and/or 
and   flows   with   complex   secondary   flow   features  translational speeds. The flow in each moving cell zone 
[9,12]. is solved using the moving reference frame equations. If 
the zone is stationary, the stationary equations are used. 
  At the interfaces between cell zones, a local reference 
One limitation of the realizable  K­ε  model is production  frame   transformation   is   performed   to   enable   flow 
of   the   non­physical   turbulent   viscosities   in   situations  variables in one zone to be used to calculate fluxes at 
when  the  computational  domain  contains  both rotating  the   boundary   of   the   adjacent   zone.   MRF   is   relatively 
and   stationary   fluid   zones   (e.g.   multiple   reference  simple and more efficient than SMM approach. It should 
frames, rotating sliding meshes). This is due to the fact  be noted that the MRF approach  does not account for 
that   the   realizable  K­ε  model   includes   the   effects   of  the   relative   motion   of   a   moving   zone   with   respect   to 
mean rotation in the definition of the turbulent viscosity.  adjacent zones (which may be moving or stationary); the 
grid remains fixed for the computation. This is analogous 
to freezing  the motion of the moving part in a specific  MESH GENERATION
position and observing the instantaneous flow­field with 
the   rotor   in   that   position.   Hence,   the   MRF   is   often  The   surface   mesh   was   generated   in   Unigraphics 
referred to as the "frozen rotor approach."  (UG).The  surface  mesh  quality  was  improved  using  in 
ICEM­CFD   and   volume   mesh   was   created   in   Tgrid. 
While the MRF approach is clearly an approximation, it  Arbitrary   interface   was   created   between   moving   and 
can   provide   a   reasonable   model   of   the   flow   for   many  stationary   grids.   The   mesh   consisted   of   a   total   of 
applications. For example, the MRF model can be used  3,800,000  tetrahedral cells including  2,650,000  cells in 
for   turbomachinery   applications   in   which   rotor­stator  rotating zone. Figure 1 shows the created mesh that was 
interaction  is  relatively  weak,   and  the  flow   is  relatively  used for this study.
uncomplicated at the interface between the moving and 
stationary zones. In mixing tanks, for example, since the  BOUNDARY CONDITIONS
impeller­baffle   interactions   are   relatively   weak,   large­
The   Steady   state   calculation   using   the  K­ε  realizable 
scale   transient   effects   are   not   present   and   the   MRF 
and multiple rotating reference frame model were used 
model can be used. 
for   high   airflow   regimes   (3023,   3290,   3612   rpm).   For 
Another potential use of the MRF model is to compute a  higher   blower   speeds   and   low   airflow   (3923   rpm,   50 
flow field that can be used as an initial condition for a  cfm), the MRF model failed to deliver good results due 
transient   sliding   mesh   calculation.   This   eliminates   the  to   high   turbulence   and   airflow   instabilities.   The   sliding 
need   for   a   startup   calculation.   The   multiple   reference  Mesh model was employed to simulate this kink of flow. 
frame   model   should   not   be   used,   however,   if   it   is  A   zero   pressure   inlet   was   used   at   the   inlet   boundary. 
necessary  to  actually   simulate  the  transients that   may  The outlet pressure was derived from the experimental 
occur in strong rotor­stator interactions, the sliding mesh  data.   A logarithmic law  of the  wall relative  to  frame  is 
model alone should be used.  used in moving rotor surfaces.

The sliding mesh model approach is, on the other hand, 
inherently unsteady due to the motion of the mesh with 
time. In this approach, motion of stationary and rotating 
components   in   a   rotating   machine   will   give   rise   to 
unsteady  interactions. These   interactions   are generally 
classified as follows:

• Potential   interactions:   flow   unsteadiness   due   to 


pressure waves which propagate both upstream and 
downstream. 
• Wake interactions: flow unsteadiness due to wakes 
from upstream blade rows, convecting downstream. 
• Shock   interactions:   for   transonic/supersonic   flow 
unsteadiness   due   to   shock   waves   striking   the 
downstream blade row.
 
The sliding mesh model accounts for the relative motion 
of   stationary   and   rotating   components   and   make   it 
possible to model the unsteady interaction effects which 
are neglected in MRF approach.
BLOWER TESTING SETUP

The   selected   blower   for   this   study   was   a   single   inlet 


forward   curved   (FC)   centrifugal   fan   having   a   squirrel 
cage   wheel.   Table   2   represents   the   geometric 
characteristics of this blower. 

Table 2: The geometric characteristics of the blower.

Wheel
Fan inlet diameter  127 mm
Fan outlet diameter  160 mm
Fan width  80 mm
Number of blades 43
Inlet blade angle 23 deg
Outlet blade angle 60 deg
Blade chord length 17.6 mm
Pitch to chord ratio 66 %
Scroll
Casing width 85 mm
Cut­off edge distance 13 mm
Expansion angle at θs=210deg 5.8 deg
Diameter of inlet opening 137.5 mm

In order to determine the fan performance, the blower 
and scroll are attached to an FM600 airflow test bench 
which   is   dual   chamber   AMCA   certified   design   (see 
Figures 2 and 3). The test set up and measurements of 
blower aerodynamic performance were made according 
to ANSI/AMCA Standard 210­85 [6].

Figure 1: Shows the mesh in different parts of the blower. The blower inlet configuration was a “free type”; that is 
open   to   atmosphere.   The   fan   total   pressure   was 
measured at the FM600 airflow chamber (averaged over 
It   is   worth   to   mention   that   the   outlet   boundary   was  4   piezometric   static   pressure   tabs).   The   volumetric 
extended in order to reduce the effect of the numerical  airflow was measured in the dual chamber using ASME 
reflections from outlet on the flow field. All steady state  certified   flow   nozzles.   The   test   blower   was   powered 
simulations   were   begun   using   a   first   order   upwind  using the DC Power supply. The rotational speed of the 
scheme   for   pressure,   momentum,   turbulent   equations  blower wheel was measured using a fiber­optic sensor 
and after converging, they were switched to the second  connected   to   a   digital   display   on   the   FM600   airflow 
order upwind. bench.   Signal   pick   was   based   on   reflective   tape 
attached to the rotating wheel.
This  time  step in  unsteady   calculations  was  related to 
the   rotational   speed   of   the   impeller   and   was   small  The   blower   performance   curves   were   obtained   using 
enough   to   get   the   necessary   time   resolution   and   to  fixed voltages (i.e., motor voltages selected were: 12V, 
capture the phenomena due to the blades passage and  13V & 14V). Using the auxiliary FM600 airflow bench fan 
their interactions with the volute casing wall [13]. and  damper­valve combination,   the airflow was set for 
each data point (ranging from 50 cfm to 350 cfm), while 
the   fan   static/total   pressure   and   blower   rpm   were  speeds and outlet pressure.  Last row shows the results 
measured at each data point. Then current draw at each  for unsteady transient simulation.
data   point   was   obtained   through   a   shunt.   This   was 
repeated for each data point until full airflow range was  The same tetrahedral mesh has been used for all cases. 
tested (from 50 to 350 cfm). The range of airflow testing  Also The MRF approach was employed for the first four 
at the various voltages tested allowed enough spread in  configurations,   while   SMM   approach   was   used   for 
blower   rpm   for   matching   different   rpm   and   pressure  configuration 5. Ve Blower Performance Curves: free Inlet, no LPM
06-AIUS-TR-00558
conditions   used   in   CFD   analysis.   The   blower  1400

performance curves based  on pressure versus airflow, 
12 volts

current versus blower speed, and current versus airflow 
1200 13 volts

are shown in Figures 4, 5, and 6 respectively.
14 volts
1000

Pressure (Pa)
800

600

400

200

0
0 50 100 150 200 250 300 350 400
Airflow ( CFM)

    Figure 4: Blower Performance: Pressure vs. airflow.
Ve Blower Performance Curves: free Inlet, no LPM
06-AIUS-TR-00558
30
12 volts

13 volts
25

14 volts

20
Current (Amps)

Figure 2: The blower and scroll attached to an FM600 airflow  15

test bench.
10

12.5 in 29.5 in
Airflow Nozzles 5
location
Blower attaches here
0
2500 2700 2900 3100 3300 3500 3700 3900 4100 4300
Blower Speed ( RPM)

Figure 5: Blower Performance: Current vs. blower speed
Ve Blower Performance Curves: free Inlet, no LPM
06-AIUS-TR-00558
30
12 volts

Airflow direction 25 13 volts


Blower Exit
14 volts

20
4 Static Pressure
Current (Amps)

Taps (piezometric)
15

Figure 3: Shows the locations for measurement of flow 
variables.   10

RESULTS AND ANALYSIS
5

0
Table   3   shows   a   comparison   of   the   volume   rates  0 50 100 150 200 250 300 350 400

between   test   and   numerical   simulation   in   four   blower  Airflow ( CFM)


Figure 6: Blower Performance: Current  vs. airflow air   flow   results   form   test   and   CFD   as   a   function   of 
pressure.   It   shows   clearly   that   results   form   CFD   are 
Simulations   were   carried   out   on   8­CPU,   2.4GHz   Dual 
match with corresponding one in test. 
Core   AMD   Opteron   cluster   with   a   64­bit   operating 
system. It took about 8 hrs to perform 3500 iterations, in  Table 3: Blower air flow rate Comparison.
order to reach a constant mass flow rate at the outlet for 
configurations 1­4 (MRF approach). Configuration 4 was 
used as an initial condition for configuration 5 with SMM  Figure 8 shows the cross sections that have been used 
approach. Using 50 time steps for each wheel rotation, it  in post processing.
took about 25hrs to simulate five rotations of the wheel. 
This shows that SSM approach  is a costly method  for 
Blower Performance: CFD and Test Results Comparison
these kinds of applications and is only recommended for 
cases with unsteadiness.  400
350

Volumteric Airflow (cfm)


300
For configurations 1, 2, and 3, CFD results show very  250

good   correlation   with   test   results.   The   maximum 


Test Results
200
CFD Results

difference of the total air flow from test and CFD is within 
150
100
5%.  50
0
400 600 800 1000 1200
CFD   result   shows   an   overestimate   in   total   airflow  Pressure (Pa)

compared   to   the   test   results   in   configuration   4   (high 


pressure,   high   rpm),   Generally   forward   curved 
Figure 7: Blower performance of the blower based on test and 
centrifugal   fans   show   an   instability   region   (very   high  CFD results.
pressure,   low   airflow)   within   their   blower   performance 
curves  [5,14,15].  This  instability  region  is characterized  Figure   9   shows   static   pressure,   velocity   vector,   total 
by   bi­stable   pressure/airflow   states,   which   lead   to  pressure,   turbulent   kinetic   energy,   and   turbulence 
unsteady   flow   behavior.   The   steady   state   CFD  intensity distributions at the mid section of the wheel for 
simulations are not well suited to simulate these types of  configuration   1.   In   the   vicinity   of   the   cut­off   edge, 
flows (i.e. flow rates that are much lower than at the best  overpressure is not that high compared to the other flow 
efficiency point). By switching to a transient  simulation  regions,   but   the   circumferential   pressure   gradient   is 
(configuration 5) and using the sliding mesh model, the  maximal in the front portion of the casing. Although the 
CFD results were considerably improved (4% difference  results   represent   the   flow­field   at   the   higher  flow   rate, 
between test and CFD results). Figure 7 shows the total  the flow behavior appears much like the case at a low 
Test Pressue Blower Airflow Airflow Test/CFD flow   rate;   the   gradients   along   the   casing   channel   are 
Results vs. increase Speed (cfm) (cfm) difference
CFD (Pa) (rpm) Test CFD (%) steep   enough   to   cause   complete   flow   reversal   in   the 
(1) Steady blade passage while they approach  the cut­off. Except 
state - K­ε
Realizable - 581 3023 350 363 4 in this region,  the static and total pressures  inside the 
second wheel are slightly uniform. This figure also shows that 
order
(2) Steady local regions of negative total pressure are apparent on 
state - K­ε the suction side near the leading edge. This is indicative 
Realizable - 756 3289 275 263 5
second of entry losses.
order
(3) Steady
state - K­ ε Typically a positive total pressure at the interior region of 
Realizable - 936 3612 176 173 2 the   wheel   (blade   inlet)   is   indicative   of   reverse   flow. 
second
order Comparing   Figures   9   and   11   show   that   the   region   of 
(4) Steady
positive total pressure on the blade inlet in low pressure 
state - K­ε
Realizable 1138 3923 50 284 468 flow (configuration  5),  is much  higher  than  in the  high 
- second
order
flow rate (configuration 1). 
(5)Transient
sliding
mesh - K­ε 1138 3923 50 52 4
Realizable -
second
order
In addition to the reverse flow that comes from high to 
low   flow   rate,   the   turbulent   kinetic   energy   (tke)   CFD 
contours (see figures 9 and 11, sections 1 and 2) show 
that:

• tke spreads over a large region circumferentially and 
axially in low flow rate configuration,   whereas it is 
highly concentrated in the lower portion of the blade 
exit region  and  close to the scroll cut­off region  in 
high flow rate configuration.

• On   the   blade   inlet   region,   in   the   low   flow   rate 


configuration,   there   is   significant   amount   of   tke, 
whereas there is negligible tke in the high flow rate 
configuration.
Section 3: Cut­off exit section
Figure 10 and Figure 12 show a larger flow blockage 
Figure 8: Sections of Post­processing
region in the lower half of the wheel in configuration 5 
compared to configuration 1.
Based on the reverse flow and tke results, we can 
conclude that there is a higher level of unsteadiness in 
the low flow rate configuration. This conclusion is 
consistent with previous experimental studies conducted 
on forward curved centrifugal fans [14]. 

The   results   indicate   that   an   increase   in   flow   blockage 


across the blades leads to higher recirculation and flow 
reversal,   which   contributes   to   higher   turbulence   levels 
and   flow   unsteadiness.   This   gives   credence   to   why 
transient CFD model need to be used for such complex 
flows.

Section 1: Circumferential Section

Section 2: Mid section
a

Figure 9: Shows (a) Pressure, (b) Velocity vector, (c) Total 
pressure, (d) Turbulent kinetic energy, and (e) Turbulent 
intensity distributions at section 2 in configuration 1.

c
a

d d

e
e
a

Figure 10: Shows (a) Pressure, (b) Velocity vector, (c) Total 
c
pressure, (d) Turbulent kinetic energy, and (e) Turbulent 
intensity distributions at section 1 in configuration 1.

d
c
e

Figure 11: Shows (a) Pressure, (b) Velocity vector, (c) Total 
pressure, (d) Turbulent kinetic energy, and (e) Turbulent 
intensity distributions at section 2 in configuration 5.

Figure 12: Shows (a) Pressure, (b) Velocity vector, (c) Total 
pressure, (d) Turbulent kinetic energy, and (e) Turbulent 
intensity distributions at section 1 in configuration 5.
3. Eck B., " Fans: Design and Operation of Centrifugal, 
Axial­Flow   and   Cross­Flow   Fans",   Translated   and 
edited   by   Azad   R.S.   and   Scott   D.R.),   Pergamon 
CONCLUTIONS Press, 1973.
4. D. Fischer, "Airflow Simulation Through Automotive 
CFD   analysis   and   blower   performance   measurements  Blowers   Using   Computational   Fluid   Dynamics   ", 
were conducted on a forward curved centrifugal blower  SAE 950438
5. E.   Kwon,   K.   Baek,   N.   Cho,   "Some   Aerodynamic 
(for automotive HVAC applications).
Aspects   of   Centrifugal   Fan   Characteristics   of   an 
Automotive HVAC Blower ", SAE 2001­01­0291
Comparison   of   predicted   total   airflow   rates,   using   a  6. AMCA 210­85, “Laboratory Methods of Testing Fans 
steady   state  K­ε  realizable   model,   showed   a   good  for Fating”
correlation with three test configurations. For the fourth  7. T.­H.   Shih,   W. W.   Liou,   A. Shabbir,   Z. Yang,   and 
J. Zhu., "A New  K­ε  Eddy­Viscosity Model for High 
configuration (high pressure, low airflow), the predicted 
Reynolds   Number   Turbulent   Flows   ­   Model 
total airflow was highly overestimated.  Development   and   Validation",  Computers   Fluids, 
24(3):227­238, 1995.
For   the   fourth   configuration,   the   CFD   results   analysis  8. W. C.   Reynolds,   "Fundamentals   of   turbulence   for 
indicate   that   an   increase   in   flow   blockage   across   the  turbulence modeling and simulation", Lecture Notes 
blades   leads   to   higher   recirculation   and   flow   reversal,  for   Von   Karman   Institute   Agard   Report   No.   755, 
1987.
which   contributes   to  higher   turbulence   levels   and   flow  9. T.­H.   Shih,   W. W.   Liou,   A. Shabbir,   Z. Yang,   and 
unsteadiness. As a result, using a transient approach for  J. Zhu, "A New K­ε    Eddy­Viscosity Model for High 
this configuration was necessary.  Reynolds   Number   Turbulent   Flows   ­   Model 
Development   and   Validation",  Computers   Fluids, 
By switching to a transient simulation, the predicted total  24(3):227­238, 1995.
10. B. E.   Launder   and   D. B.   Spalding.   “Lectures   in  
airflow   was   in   reasonable   agreement   with   measured  Mathematical   Models   of   Turbulence”   ,   Academic 
total airflow. The general flow structure appeared to be  Press, London, England, 1972.
similar   for   all   configurations,   although   with   differences  11. V. Yakhot and S. A. Orszag. “Renormalization Group 
with respect to the location and intensity of the vortices.  Analysis of Turbulence: I. Basic Theory”  Journal of  
Scientific Computing, 1(1):1­51, 1986
12. S.­E.   Kim,   D. Choudhury,   and   B. Patel. 
This   CFD   study   shows   that   numerical   results   for   a 
“Computations   of   Complex   Turbulent   Flows   Using 
forward curved blower are in good agreement with test  the Commercial Code  FLUENT” In  Proceedings of  
results and can be used  for accurate prediction of the  the   ICASE/LaRC/AFOSR   Symposium   on   Modeling  
airflow characteristics of automotive HVAC systems. Complex Turbulent Flows, Hampton, Virginia, 1997.
13. M. Younsi, F. Bakir, S. Kouidri, R. Rey, “Influence of 
Impeller   Geometry   on   the   Unsteady   Flow   in   a 
In   order   to   have   an   accurate   analysis   of   the 
Centrifugal   Fan:   Numerical   and   Experimental 
unsteadiness   in   low   flow   rate   configurations,   it   is  Analysis”, Internal paper of Ecole Superieure d’Arts 
recommended   to   perform   the   transient   simulations   of  et Metiers – France.
the  whole   blower  including   wheel,   inlet   and   scroll  and  14. R.J. Kind, M.G. Tobin, “Flow in a centrifugal fan of 
the squirrel­cage type”, Transactions of the ASME, 
compare the results to corresponding pressure, velocity, 
84/Vol. 112, January 1990.
and   turbulence   results   measured   in   experimental   test.  15. Raj, D.and Swim, W. B., 1981,”Measurements of the 
Specially, the effect of the back­flow through the blade  mean   flow   velocity   and   velocity   fluctuations   at   the 
passages close the cut­off scroll should be evaluated by  exit of an FC centrifugal fan rotor”, ASME Journal of 
Engineering for Power, Vol. 103, pp. 393­399.
experiments. In addition, the position of the stagnation 
point of the cut­off scroll area and the degree of the flow 
blockage both in the inlet and outlet of the blower should 
be confirmed by test results.

REFERENCES

1. “Automotive   Air   Conditioning”,   DELMAR,   Thomson 


Learning, 8th Edition.
2.  Fluent Users Manual

You might also like