CFD Modeling for Motor Fan System

Ya-Chi Chen, Chung-Lung Chen

Rockwell Scientific Company
1049 Camino Dos Rios
Thousand Oaks, CA 91360

Qimin (Jimmy) Dong

Rockwell Automation
6040 Ponders Court
Greenville, SC 29602

Abstrrzci- For fansooled moton, the fan performance performance, based only on the fan blade design, is provided
directly influencesthe thermal performance of the motor and so from the manufacturers.
its lietime. A 10-degree temperature reduction could extend the A commercial CFD software,Fluentm, is used to develop
motor electrical insulation lifetime twofold The purpose of this the CFD model. Three-dimensional flow simulations are
study is to apply Computational Fluid Dynamics (CFD)
modeling techniques to a motor fan system and use the validated performed at various fan speeds for a specific fan model. The
CFD model tu analyze key design parameters. Simulation flow rates and torques are calculated and compared with
results and measured data at various fan speeds showed good experimental measurements and fan law predictions to
agreement between CFD results and the test data In addition, validate the CFD model.
both the calenlated and measured flow rates and torqws foUow The validated CFD model is modified for a paramebic
fan laws very well. The validated CFD model is used to perform study of key design parameters on the fan-bracketaver
a parametric study for the fan system. The effect of axial system. The fan speed is set at IS00 RPM. The parameters
clearance between the fan and the fan cover and between the fan include axial clearance between fan and fan cover, fan cover
and the bracket, the effect of the fan .cover diameter, and the diameter, axial clearance between fan and bracket, and
effect of the number of fan blades and their chord length, on fan
performance are studied. We demonstrated the capability of number of the fan blades and their chord length. All the
applying CFD to complicated geometry to effectively analyze the analyses are carried out by three-dimensional steady-state
design parameters. calculations. The flow rate, torque and fan efficiency are
calculated to examine fan performance. The results and
1. I ~ O D U C T T O N analyses are presented as follows.
Cooling is essential to motor performance. Good thermal 11. NUMERICAL APPROACH
management reduces the risk of overheating and increases
motor lifetime. We have used CFD and thermal modeling The geometry of the fan test system is constructed in the
techniques to motor thermal analysis [I], [2], [3], and shown grid generator (Gambitm). Fig. 1 shows part of the
good correlation between the computed and measured numerical model for the fan-bracket-cover system. A hybrid
temperatures. This therefore proves that thermal modeling mesh is created for the complicated geomeby. Grid quality is
provides useful results to improve motor thermal important to accuracy and stability of the numerical
performance. calculation and is examined to have high quality such as
Numerical modeling reduces the cost of prototype testing small grid skewness. The mesh is exported to the Fluentm
and effectively explores design features. Modeling provides solver and the problem is defined by choosing the appropriate
much information that could be difficult to obtain through flow model and prescribing the boundary conditions.
measurements and M e r can help augment the test
instruments. However, developing a model for an entire
motor can be time consuming. From this viewpoint. we
develop models for-motor components to study and optimize
the design to improve thermal performance [4]. The present
study focuses on shaft-mounted fan system for a totally
enclosed fan-cooled (TEFC) motor. The information for the
integrated fan-blade-cover system performance is generally
lacking from the fan manufacturers test data; i.e., the fan
-
Fig. 1. Fan-bracket-coversystem in CFD model.

82003 IEEE
0-7803-7S17-2/03/$17.00 764

Authorized licensed use limited to: NANKAI UNIVERSITY. Downloaded on January 20,2023 at 15:48:36 UTC from IEEE Xplore. Restrictions apply.
 Fluentm uses a control-volume-based technique to convert In these equations, ,ut is the turbulent viscosity. Gk
the governing equations to algebraic equations that can be represents the generation of turbulent kinetic energy due to
solved numerically [5]. The conservation equations the mean velocity gradients. G, is the generation of
describing laminar flow are [6]
turbulent kinetic energy due to buoyancy. Y, represents the
continuity: contribution of the fluctuation dilatation in compressible
turbulence to the overall dissipation rate. C,, , C,, , and
aP+V.(pU)=O C,,are constants. uk and ue are the turbulent F’randtl
at numbers for k and e , respectively.
Fan motion is simulated using the moving reference frame
Momentum: approach rather than moving the mesh of the fan blades to
reduce computational time. That is done by specifying the
angular velocity of the reference h e and the axis about
=+V. (pU @U)= B Vu which it rotates to the air smrounding the fan blades. No-slip
at
wall boundary condition is imposed on all the solid surfaces
and except the fan blades that have the same rotational velocity as
the surrounding air.
Second-order upwind scheme for momentum equations is
u=-p6+,u(vLr+(vU)’ used to increase accuracy. The default under-relaxation
factors are modified to reduce the change of computed
In the above equations, p is the fluid density, U the variables at each iteration for a stable convergence.
velocity vector, B the body force, D the stIess tensor, p the Convergence of a calculation can be determined by
dynamic viscosity, p the static pressure, and 6 is the monitoring the residuals of computed variables.
Additionally, we monitor fan torque during the iterations. A
Kronecker delta function.
calculation is converged when the residuals are below the
specified convergence criterion and the change of fan torque
An implicit segregated steady-state solver is used. The becomes negligible between iterations.
flow is assumed to be incompressible with cmtaut
properties. The assumption is reasonable because of small
pressure variation for this low Mach number flow and In. RESULTS AND DISCUSSIONS
negligible temperature change. Therefore, the continuity We perform flow simulations for a 13.5-inch fan at various
equation and the momentum equation for a steady-steady fan speeds. The flow rates and the torques are calculated
flow of Newtonian fluid with constant density and viscosity from converged steady-state solutions. A comparison of the
can be simplied as simulation results with test data shows very good agreement
(Fig. 2). The deviation of flow rate is less than 2% and the
vu=o torque is witbin 7%.
B - Vp + ,uV2U= 0 Both the simulation results and test data obey the fan laws.
The fan laws state that the flow rate is linearly proportional to
the fan speed and the torque is related to the square of the fan
The turbulence flow is simulated by the standard k-e speed 17. The curve-fit equations based on the CFD results
turbulence model. The turbulent kinetic energy, k , and its are as follows.
rate of dissipation, E , are obtained from the following
bansport equations: Q = 0.4026N
7 = 4 . 0 4 ~ 1 N2”’
0~

[
V u+L)Vk
’k 1 +G, +G, - P E - Y , =O
Where Q is the volumetric flow rate (ch), 7 is the fan
torque (Ibf-in) and N is the fan speed (RF’M). The fan law
equations can be used to predict the performance for this
and
specific fan at other speeds.
E2
+ C , , G b ) - C , p ~ =0

165

Authorized licensed use limited to: NANKAI UNIVERSITY. Downloaded on January 20,2023 at 15:48:36 UTC from IEEE Xplore. Restrictions apply.
 (I r m , m r ? W 8 " a m s S m r m o
F n Sp.d P P M )
Fig. 3. Flow rate and toque vs. fan cover cl&me.

Fig. 2. Elow rate and torque YS. fan s p e d .

Pressure and velocity are investigated to understand the

flow characteristics and assist in facilitating the
measurements in a fan test rig. The flow path shows a spiral
flow pattern in the test rig. The spiral pattern implies
potential difficulty to get accurate pressure and velocity
measurements.
For fan speed at I800 RPM, the mesh is refmed in the
model to ensure grid independence of the solution. Changes
of the calculated flow rate and torque are about 0.1%. The
solutions are proved to be grid-independent.
The validated CFD model is used to analyze the effects of
design parameters on fan performance. Fan speed is set at
1800 RPM. Fig. 3 shows the effect of the axial clearance
between the fan and the fan cover. S@ulation results show
that both flow rate and torque incrkase as the clearance
increases and approach asymptotes. The percentage change
in the torque is larger than that in the flow rate. Therefore,
the axial clearance between the fan and the fan cover should
be kept to minimumto m i n i i e the torque that relates to fan
input power. However, optimal axial clearance is also
determined under the constraint of installation and
manufactwing tolerance.
Intuitively, flow rate may be expected to increase when the
axial clearance is reduced because the pressure is anticipated
to increase at the upstream regime near the tip of the fan
blade. However, it is opposite from what the simulations
predict. AAer examining the flow path, we can see in Fig. 4
that flow enters the fan assembly toward the center hub
'.
region and moves along the blades. It does not furn radially
and cross the front tip of the fag blades..

Fig. 4. Flow path colored by static prersun (red indicates high and blue
indicates low) at two axial clearance bemm fan and fan aver.

766

Authorized licensed use limited to: NANKAI UNIVERSITY. Downloaded on January 20,2023 at 15:48:36 UTC from IEEE Xplore. Restrictions apply.
 Fig. 5 shows the effect of the fan cover diameter. Flow
rate and fan torque increase as fan cover diameter (or fan
outlet area) increases and the relationship is close to linear.

l r l l . m
(0 32 1. I*
HnbnolBwma

Fig. 7. Flow rate and torque ys. number of fan blades

?I. m $8, I&? 111 1&1 1111 ?PI
F 8 l C a W b )
The chord length of the fan blade is changed to be twice
Fig.5. Row me and torque w. fan mer diameter. and triple of the original. The axial length of the fan cover is
changed accordingly. The flow rate increases as the blade
The axial clearance between the fan and the bracket is chord length increases. A further increase of the blade chord
defmed by the distance &om the fan hub to the end surface of length will not result in further increase of flow rate. It
the bracket The fan cover is modified accordingly to keep shows an optimal chord length (Fig. 8). Once the blade chord
the same clearance between the fan and the fan cover when length exceeds optimal, adverse result of reduced flow rate
the axial clearance between the fan and the bracket is and increased fan torque would occur. In practical
changed. The results show that fan torque increases with the applications, increasing the blade chord length to improve the
clearance between the fan and the bracket, but no clear fan p e r f o m c e is more effective.
relationship is observed for the flow rate (Fig. 6). It suggests
the improvement of fan performance with a reduced
clearance between the fan and the bracket.
l4S

14.0

LO LS I D IS 10 .I 60 I, 10 06 IO
F n Mm ( h )
go
42.0
si0 Fig.8.Raw m e and M g u c w. chord length of fan blade

02 01 all 0.8 I 12 11 ?e For all the parameters, the change in fan efficiency is not
-..a -"had bndul (I") dramatic due to the change in parameters over a reasonable
range. Fan efficiency is the ratio of output air power to fan
Fig. 6 . Row rate and-t w. axial clearance h e n input mechanical power [SI.
fan and bra&&

The effect of the n u m k of fan blades on fan p e r f o m c e Fan efficiency = (Ap Q) / (Tw)
is shown in Fig. 7. The flow rate and torque first increase as
the number of fan blades increases and then approach Where Ap @a) is the difference between the pressure at the
asymptotes. fan outlet and inlet, Q (m3/s) is the volumetric flow rate, T
m-m) is tbe torque requiredto rotate the fan, and w ( d s ) is
the angular fan speed. The pressure, Ap, can be the total
pressure or the static pressure. When the total pressure is
used, the efficiency is called fan total efficiency. When the

167

Authorized licensed use limited to: NANKAI UNIVERSITY. Downloaded on January 20,2023 at 15:48:36 UTC from IEEE Xplore. Restrictions apply.
 static pressure is used, the efficiency. is called fan static REFERENCES
efficiency. C M . Liao, C.L. Chen, and T. Katchcr, ’Thermal Analysis for Design
[I]
of High Performance Motom”, ITherm 1998, Seanle, Wsrhington, pp.
N.CONCLUSIONS 424433.
[2] C.M. Liao, C.L. Cha, and T. Katcher, “Thermal Management of AC
The capability of applying CFD modeling to motor fan loduction Motors Using Computational Fluid Dynamics Modeling’*,
system analysis is demonstrated by developing a CFD model IEhQc‘99, Seattle, Washington, pp. 189-191.
[3] C.M. Liao, C.L. Chen, ‘Tlwmal Analysis and Desip for Power
for the shaft-mounted fan system for a totally enclosed fan- Elechmics of Integrated Motor”, 1999 lotemational Symposium on
cooled (TEFC) motor. The simulation results show good Microeleclmnics, pp. 408414.
agreement with the test data and the fan laws. The validated [4] Y.C. C h a , C.L. C h q Q. Daog, and R Stephason, ‘Thermal
CFD model is used to analyze the effects of design Management for M o l d , ITherm 2002, San Diego, Cdifomia, pp. 545-
551.
parameters on fan performance. The parameters studied were Fluentm 5 User’sGuide, 1998.
[5]
axial clearance between the fan and the fan cover, the fan [6] D.A Anderson, I.C. Taonehill. and RH. Pletcher, Computational Fluid
cover diameter, axial clearance between the fan and the Mechanics and Heat T m f c r , Hemisphere publishing Corporation,
bracket, and the number of fan blades and their chord length. 1984.
The effects of the design parameters on fan performance are [7l M. Tumer, and.C. RoImn, AU You Need to Know about Fans,
Eleefronies Cwling, May, 1996.
presented by calculated flow rates and torques. For all the [8] W.C. Osbome, Fans, Pergamon Pres Ltd., 1966.
parameters, the change in fan efficiency is not dramatic due
to the change in parameters over a reasonable range.

768

Authorized licensed use limited to: NANKAI UNIVERSITY. Downloaded on January 20,2023 at 15:48:36 UTC from IEEE Xplore. Restrictions apply.