Chapter 17

Fan Mechanics

The various parts of a fan must be designed not only for their aerodynamic
function but also for mechanical integrity. Among the mechanical subjects
considered in this chapter are: 1) torque and thrust, 2) centrifugal force, 3)
stresses and strains, 4) bearings, 5) critical speed, 6) balancing, and 7) vibra-
tion isolation. The purpose of these discussions is not to give a complete
design method but, rather, to show some of the important considerations so
that design features will be appreciated when applying and operating fans.

Torque
The shaft torque Ts delivered to a fan by its driving motor can be calcu-
lated from the shaft power Ps and the fan speed N using

K Ps
Ts = , (17.1)
N

where K is 63030 for U.S. customary units of lb-in., hp, and rpm or 1000 2π
for SI units of N-m, kW, and rps. The fan shaft must be able to transmit the
shaft torque without twisting excessively. See Equation 17.28 and the associ-
ated discussions on shaft design.
Because of bearing-torque losses Tb , the impeller torque Ti will be less
than Ts if the fan has bearings:

Ti = Ts − Tb. (17.2)

The blades and their attachments must be able to transmit the impeller
torque without distorting excessively. A rigorous analysis of blade distortions
requires detailed knowledge of the distribution of pressure on the blades.
Some of the distortions can be predicted by using a less rigorous analysis
based on the center-of-pressure concept. Recognizing that the impeller torque
results from the tangential forces of the blades on the air, the equivalent
tangential force per blade Ft can be calculated from the estimated radius r of
the center of pressure and the number of blades n using

‹#1999 Howden Buffalo, Inc.

17-2 FAN ENGINEERING

Ti
Ft = . (17.3)
nr

For some fans, the air load on the blades is insignificant compared with the
mechanical loads due to centrifugal force. For others, especially those with
long cantilevered blades, the air load must be considered when designing the
blades and their restraints.

Axial Thrust
The axial components of the pressure distribution on an impeller lead to
axial thrust. The distribution of axial pressure on the various surfaces of the
impeller will usually be unbalanced. This unbalanced force will tend to move
the impeller toward the inlet in both axial-flow and centrifugal fans. Because
axial thrust must be resisted or counteracted, bearings and bearing supports
(right down to the foundation) must be designed to hold the impeller in place.
The total axial thrust produced by an axial-flow impeller, or net axial
force Fa , can be approximated from

π C p pFT DT
2

Fa = . (17.4)
4

The thrust will be in lb if the fan total pressure pFT is in in. wg, the tip
diameter DT is in ft, and C p is 5.193. Refer to Table 1.4 for C p values for
other units.
For a single-inlet centrifugal impeller, the axial thrust due to unbalanced
pressure forces Fa can be approximated using the inlet diameter Di , the fan
static pressure pFS , and C p as noted above:

π C p K pFS Di
2

Fa = . (17.5)
4

The value of the proportionality constant K can be assumed to be about 1.0

whenever the gage pressure inside the fan housing is positive (blower appli-
cations), when the impeller is completely shrouded (inlet shroud and back-
plate), and when the shaft hole will allow some leakage (no stuffing box). If
the gage pressure in the housing is negative, if the shaft hole is tightly sealed,
or both, the value of K can be assumed to be about 2.0. If the impeller is
completely open (paddle wheel), there will be practically no net thrust. If the
impeller has only one shroud (cone wheel), the value of K will be higher than
for a completely shrouded impeller.
Theoretically, a double-inlet centrifugal rotor should have no axial thrust.
However, some thrust-resisting capacity should be provided to protect against
any unbalanced thrust that might develop because of uneven flow conditions
between the two inlets. This can result from system blockages or from
malfunctioning dampers or variable inlet vanes.

‹#1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-3

Figure l7.1 Axial and Radial Thrust on Impellers

‹#1999 Howden Buffalo, Inc.

17-4 FAN ENGINEERING

The change in momentum resulting from the change in direction of flow

(from axial to radial) in a centrifugal impeller also produces axial thrust. This
momentum thrust Fa ′ tends to move the impeller in the direction opposite the
inlet and can be calculated from the mass flow rate mû and the axial approach
velocity Va ′ using

û a′
mV
Fa ′ = . (17.6)
gc

Fa ′ is negligible compared with the unbalanced pressure thrust Fa at low

flow rates. However, Fa ′ increases with flow rate and usually dominates at
points of rating toward free delivery.
Axial thrust can be reduced or even reversed by building fins on the
backplate exterior. Such fins can be designed to equalize the pressure with
very little flow and only a small power increase. Balancing holes in the
backplate or a balancing chamber, such as often used in high-pressure single-
suction pumps, could also be incorporated but are seldom justified.
The distribution of forces within a fan due to an externally applied vac-
uum or pressure may or may not be unbalanced. Usually, this pressure or
vacuum is considered to act only on the cross-sectional area of the shaft and
then only if just one end extends through the casing to a region of atmospheric
pressure. The resulting thrust will be directed toward or away from the inlet
depending on the direction of the shaft extension and on whether a vacuum or
pressure exists.

Radial Thrust
Unbalanced radial pressures cause radial thrust. Very little radial thrust
can be expected in an axial-flow impeller, but in a centrifugal-fan impeller,
the radial-pressure distribution can become distorted, especially at off-design
points of rating. (Volute-type housings usually cause nearly uniform pressure
distributions over the impeller-discharge area at design.). Any net unbalance
will have to be carried by the bearings and transmitted to the foundation. The
radial thrust from unbalanced pressures is usually negligible compared with
the radial forces produced by rotating mechanical unbalance.

Housing Restraints
The housing, whether floor-mounted or ceiling-suspended, must be
restrained. Although friction and the weight of the unit will sometimes be
enough to keep the fan from moving, more positive restraints should still be
used.
If the impeller is supported by bearings on independent pedestals, the
foundation bolts for the pedestals must transmit to the foundation the axial-
and radial-thrust forces generated in the impeller. Of course, the bearings and
.

‹#1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-5

Figure 17.2 Housing Restraints

the bearing bolts have a similar function. The housing will also be exposed to
thrust forces because of a change in momentum, unbalanced pressure, or both.
For an axial-flow fan housing, the change in momentum, in both the axial
and radial directions, is usually negligible. However, this will not be so if an
appreciable difference exists between the inlet and outlet areas. For a cen-
trifugal-fan housing, changes in momentum always occur in both the axial and
radial directions. The corresponding momentum thrusts may be reinforced or
weakened by any unbalanced pressure forces associated with the housing.
The net axial and radial forces must be transmitted through the foundation
bolts.
If the impeller is supported by bearings mounted on the housing, both the
impeller and housing thrusts must be restrained by the housing's foundation
bolts. The reaction torque at the motor must also be considered in foundation
design.

Centrifugal Force and Centrifugal Moments

Consider a body rotating at constant angular velocity about a fixed axis.
Each element of that body, no matter how small, has an acceleration toward
the center of rotation equal to its linear velocity squared U 2 divided by the
radius r . The restraining force exerted by adjacent elements (called cen-
tripetal force), which equals the product of this acceleration and the mass m ,
must be directed toward the center of rotation. The force exerted by the mass
.

‹#1999 Howden Buffalo, Inc.

17-6 FAN ENGINEERING

on its restraints must equal the restraining force and act in the opposite
direction. This force, known as centrifugal force Fc , can be calculated from

mU 2 mrN 2
Fc = = . (17.7)
gc r CF

For U.S. customary units of lb, lbm ft and rpm, CF is 2934. For SI units of
1 6
N, kg, m, and rps, CF is 1 2π .
2

Equation 17.7 can be used to calculate the centrifugal force related to each
of various elements of a rotating structure by using the mass m of the element
and the radius r of its center of gravity. Similarly, the centrifugal force of the
entire rotating structure will be the same as if the total mass were concentrated
at its center of gravity. The degree to which the structure should be divided
into elements will depend on the purpose of the analysis. For example, the
overall unbalance of a rotor can be expressed in terms of the eccentricity, the
distance between the center of gravity of the whole structure and the center of
rotation. On the other hand, the distribution of centrifugal forces affects
stresses within the structure. To accurately calculate such stresses, these
distributions must be considered rather than the resultant centrifugal forces
acting on the blades, shrouds, etc. A classical strength-of-materials approach,
or perhaps a finite-element analysis, may be required.
If, at any radial location, the mass is not distributed uniformly about the
plane of rotation, centrifugal moments as well as centrifugal force will be
produced. A spectacular example of this is the twisting moment about the
radial axis of an axial-flow fan blade that develops because each radial section
is set at an angle to the plane of rotation. The net centrifugal moment M c can
be determined by dividing the blade into elements in both the radial and
chordwise directions and then calculating the centrifugal force Fci for each
using Equation 17.7. Next, determine the distance xi for each element. This
is measured from the elemental CG to the plane of rotation (passing through
the center of gravity of the entire section at the appropriate radius). Then,
determine the cosine of the angle α i between the radial line through the
elemental CG and the radial axis through the CG of the appropriate section.
Finally, add the products of these factors for all the elements, as shown in

n
M c = ∑ xi Fci cos α i . (17.8)
i =1

Figure 17.3 illustrates how a centrifugal moment is generated by one elemen-

tal mass on an axial-flow blade. Of course, all centrifugal moments must be
resisted by the supporting structure. For some variable-pitch axial-flow fan
blades, the centrifugal moments are large enough to require a counterbalance.
Otherwise, the mechanism for varying the pitch would have to produce very
high torques to turn the blades and, so, would require excessive power.

‹#1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-7

Figure l7.3 Centrifugal Moments

Bearings
Fan bearings must be able to withstand the loads due to the dead weight,
thrust, and unbalance of the rotor assembly. They must also be able to operate
at the intended speed without overheating. Various methods are used to
estimate the temperature rise in sleeve and antifriction bearings, both of which
are used in fans.
When not enough heat is dissipated by natural convection from the pillow
block or other type of bearing housing, some form of forced cooling is neces-
sary. Small fan wheels, called heat slingers, mounted on the shaft between a
hot fan casing and the bearing promote cooling by increasing the circulation
of air over the bearing and by providing extended heat-dissipation surface for
the shaft. Pillow blocks and some bearing liners can be provided with internal
passages through which cooling water or even cooling air can be circulated.
Lubricating oil can be circulated through an external cooler.

‹#1999 Howden Buffalo, Inc.

17-8 FAN ENGINEERING

Most plain journal bearings are furnished with self-aligning and ring-
oiling features. If the bearings are water-cooled, the connections for the water
should be flexible. Even though self-aligning features are built into the
bearings, special precautions should still be taken to line up the original
installation as perfectly as possible. Ring oiling will be adequate if the proper
level is maintained in the reservoir and if peripheral speeds are not excessive.
At high shaft speeds, forced or drop feed may be advisable. Hydrostatic oil
lift is not required if the bearing material is of good quality, but the material
must withstand running dry for a very short period after the shaft has started
from rest. Various babbits and bronzes have different load capabilities. As a
rule of thumb, the diametral clearance between the journal and the bearing
should be about 0.001 inches per inch of shaft diameter plus 0.002 inches.
The axial clearance in thrust bearings usually ranges from 0.008 to 0.012
inches.
The clearance, together with the speed, the load, and the viscosity, deter-
mines the coefficient of friction. With a clearance ratio of 0.001 and a ZN P
of 200, a coefficient of 0.01 can be expected. To calculate ZN P , the vis-
cosity Z should be in centipoises, the speed N should be in rpm, and the
bearing pressure P should be in psi and based on the projected area. De-
creasing the clearance and anything that increases ZN P increases the
coefficient of friction. In starting on a greasy surface, the coefficient may
range from 0.08 to 0.14. If the bearing is perfectly dry, the coefficient may
range from 0.25 to 0.40.
By contrast, antifriction bearings have a coefficient of from 0.001 to
0.002, exclusive of any rubbing seals. Higher values should be expected
during starting. Antifriction bearings must also be selected with loads,
speeds, and heat dissipation in mind.
Every bearing must be protected from excessive vibration, heat, dirt, and
moisture. Any rotor unbalance will produce vibrations at the bearings.
Bearing supports should be isolated from hot fan casings. Seals should
prevent dirt or moisture from entering and lubricant from being lost. How-
ever, bearings in high-velocity air streams are subject to large pressure
differences that can cause loss of lubricant.

Static Stress, Strain, Strength, and Failure

Stress is a mathematical concept useful in analyzing the effects of loads on
structures. There are two kinds: normal stress, that stress normal to the area
on which it acts; and shear stress, that stress along the area. For triaxial stress,
the most general case, stresses are produced in three dimensions with nine
components, as shown in Figure 17.4. Stresses in the third dimension can
often be ignored making possible a simpler analysis. Biaxial stress, as illus-
trated in Figure 17.5, has only four components, two of which are equal. The
maximum normal stress σ 1 and the minimum normal stress σ 2 are called
principal stresses and can be calculated from

‹#1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-9

σ +σ y çæ σ −σ y äã 2

σ 1 ,σ 2 = x ±
å â + τ xy .
x 2
(17.9)
2 2

The maximum shear stress τ max is

çæ σ −σ y äã 2

τ max = ±
å â + τ xy .
x 2
(17.10)
2

The angle φ between the principal axes and the x and y axes can be found
from

2τ x y
2φ = tan −1 . (17.11)
σ x −σ y

Figure 17.4 Three-Dimensional Stress Element

‹#1999 Howden Buffalo, Inc.

17-10 FAN ENGINEERING

When the principal stresses are known, the stress σ φ normal to a plane at an
angle φ to the plane of maximum principal stress can be calculated from

σ1 +σ 2 σ1 −σ 2
σφ = + cos 2φ . (17.12)
2 2

The shear stress τ φ acting along a plane at an angle φ to the plane of maxi-
mum principal stress is

σ1 −σ2
τφ = sin 2φ . (17.13)
2

Figure l7.5 Two-Dimensional Stress Element

‹#1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-11

Strain ε is the linear distortion per unit length due to normal forces.
Shear strain γ is the change in a right angle produced by shear forces.
Equations 17.9 through 17.13 have counterparts for strain. Simply substitute
ε for σ and γ 2 for τ , and use the same subscripts.
For an elastic material, stress is related to strain. In the simplest case,
namely uniaxial tension, stress is proportional to strain, or

σ = Eε , (17.14)

where E is the modulus of elasticity. Similarly,

τ = Gγ , (17.15)

where G is the shear modulus of elasticity. Poisson's ratio ν relates E and

G:

1 6
E = 2G 1 + ν . (17.16)

The relationships between stress and strain are more complicated for
biaxial or triaxial stress. For instance, the principal strains for biaxial stress
are:

σ 1 νσ 2
ε1 = − , (17.17)
E E

σ 2 νσ 1
ε2 = − , and (17.18)
E E

νσ 1 νσ 2
ε3 = − − . (17.19)
E E

These equations can be combined to give the principal stresses in terms of

strains:

σ1 =
1
E ε 1 + νε 2 6
and (17.20)
1− ν 2

σ2 =
1
E ε 2 + νε 16, (17.21)
1−ν 2

which are useful for strain-gage evaluations. Note that, even if the principal
directions are known, both principal strains are needed to determine the
principal stresses.

© 1999 Howden Buffalo, Inc.

17-12 FAN ENGINEERING

Some even simpler analyses can be useful. The average tensile stress s in
a tension member can be computed from the cross-sectional area A at the
relevant plane and the force component F acting along a line perpendicular
to that area by using

F
s= . (17.22)
A

Compressive stresses are also normal stresses but directed opposite to

tensile stresses. Bending stresses in a beam are a particular distribution of
tensile and compressive stresses across the section. Such stresses are of zero
magnitude at the neutral axis. This axis may coincide with the centroidal axis,
depending on the curvature of the beam. The stress in any fiber of a straight
beam can be computed from the bending moment M , the second moment of
the cross section I , and the distance c of the fiber from the neutral axis using

Mc
s= . (17.23)
I

Shear stresses can be produced by any load parallel to the section. Direct
shear V produces an average stress ss over the cross-sectional area A of

V
ss = . (17.24)
A

Torque T produces shear stresses. For a circular shaft with a polar

moment of inertia J , the stress at any radius r is

Tr
ss = . (17.25)
J

Table 17.1 gives A , I , and J for various sections. The second moments
of area I x are about the horizontal axis x − x . The second moment of area
about any parallel axis at a distance y is greater by an amount equal to Ay 2 .
The second moment of any composite section is the total for all the individual
sections about the same axis. The stresses induced in the part may be due to
more than one force.
Table 17.2 gives equations for M and V for various beams.
It was noted above that, in elastic materials, stress is proportional to strain.
Many materials are elastic, at least until the load becomes too great. Figure
17.6 shows what happens to a material like steel during a tension test.
Hooke's law applies up to the proportional limit. The material can be stressed
to the elastic limit without any permanent set. At the yield point, if there is
one, the material stretches without any corresponding increase in load. The
.

© 1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-13

Table 17.1 Properties of Sections

‹#1999 Howden Buffalo, Inc.

17-14 FAN ENGINEERING

Table 17.2 Shear, Moment, and Deflection Formulae

For Beams with Transverse Loads

Adapted from the data of R. J. Roark and W. C. Young: Formulas for Stress and Strain, Fifth
Edition, McGraw-Hill Book Co., Inc. New York, 1975, pp.96-108.

‹#1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-15

Table l7.2 (Cont.) Shear, Moment, and Deflection Formulae

For Beams with Transverse Loads

© 1999 Howden Buffalo, Inc.

17-16 FAN ENGINEERING

Figure l7.6 Stress-Strain Diagram

yield strength corresponds to a small but definite amount of permanent set,

usually 0.2% as shown. Fracture of the test specimen may occur when the
ultimate strength is reached or at a somewhat greater strain. The problem in
applying this information about a material in the design of a structure is that
the actual stresses rarely correspond to those in a simple tension test.
Many failure theories are used to estimate when a part will fail based on
an analysis of the stresses and strains in it compared with the stresses and
strains in the tensile-test specimen at failure. Of course, failure itself must be
defined. For some parts, failure is considered fracture. For others, failure is
inability to perform their function, which might occur if the elastic deforma-
tion exceeds a certain amount. For still others, some permanent set can be
tolerated.
Among the strength theories for static failure are: the maximum-normal-
stress theory, attributed to Rankine; the maximum-shear-stress theory, also
called the Coulomb or Tresca theory, or Guest's law; the Mohr theory; the
maximum-normal-strain theory, often called Saint Venant's theory; the
maximum-strain-energy theory; and the maximum-distortion-energy theory,
commonly called the von Mises-Hencky theory.

‹#1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-17

Each of the static-failure theories requires certain information about the

static strength of the material, and each is best suited to a certain type of
failure. A complete discussion is beyond the scope of this work, but briefly,
for ductile yielding and ductile fracture, the von Mises-Hencky theory is
usually considered the most accurate. Use of the Mohr theory is usually
limited by the accuracy of the available compression and torsion data. And
the more conservative maximum-shear-stress theory is the basis for many
codes. Although the maximum-normal-stress theory is sometimes used to
predict brittle fracture, the Mohr theory is preferred if data are available.
Static (or steady) stresses are produced in a fan by the steady components
of the forces acting on the various parts. The steady components of the fan
pressure produce static stresses in both the rotor and stator parts. The steady
(or constant-speed) rotation of the impeller produces static stresses in the rotor
due to centrifugal force and centrifugal moments. External loads may also
contribute to the static stresses.
Static stresses can be calculated from strain-gage measurements and other
experimental techniques. Analytical methods to determine stresses include
finite-element and strength-of-materials techniques. Although a complete
experimental stress analysis is best, it would be impractical to require such an
analysis for every fan made. Similarly, finite-element analysis will provide
much information, but it may not all be necessary to ensure the integrity of
every fan. Both experimental and finite-element analyses will often reveal
problem areas that might have been missed using less rigorous methods of
analysis. So, their use should be considered, especially for new designs.
Alternatively, larger safety factors must be used with the less rigorous meth-
ods. A simple finite-element model is illustrated in Figure 17.7. More
complicated models are needed to determine stresses and strains at highly
localized areas.
The stress distribution in a centrifugal-fan rotor can be found, at least
approximately, by examining a few equations. The blades can be considered
beams. Bending moments can be determined from the equations in Table
17.2 and the corresponding stresses calculated from Equation 17.23. For
blades that are shrouded both front and back, the beam might be considered
fixed at both ends and uniformly loaded. This suggests that the maximum
stresses occur at the fixed ends, which is usually so. In fact, some fan blades
are designed to operate at speeds that will produce a small permanent set at
the fixed ends. The shrouds can be considered rotating discs with central
holes. Without any external loads, the radial stresses must be zero at the inner
and outer diameters, and the maximum radial stress σ r occurs at the radius
ri ro :

max σ r =
3+ν
8 gc
3 8
ρω 2 ro 2 − ri 2 . (17.26)

‹#1999 Howden Buffalo, Inc.

17-18 FAN ENGINEERING

Figure l7.7 Simple Finite-Element Grid

The maximum tangential stress σ t occurs at the inner radius:

3+ν çæ 1− ν 2äã
max σ t =
å
ρω 2 ro +
â
2
ri . (17.27)
4 gc 3+ν

Of course, the blades and shrouds do interact, so the simple analyses above
will not give precise values.
The stress distribution in an axial-flow fan blade can also be estimated
easily. The blades are tension members relative to centrifugal force and can
be considered cantilever beams (or plates) relative to the axial and tangential
components of the pressure forces. Equations 17.22 and 17.23 can be used.
The centers of gravity for all the sections are usually stacked on a common
radial line to minimize the bending stresses. Sometimes, they are deliberately
offset to compensate for the pressure forces. The supporting hub may be a
complicated structure consisting of a rim and supporting discs. Blade loads
and self-generated loads must be considered.

‹#1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-19

Dynamic Stress, Strain, Strength, and Failure

The discussions in the preceding section dealt with the effects of static
loadings. But, dynamic loadings must also be considered. These are usually
classified as either impact or cyclic.
Fans may be subject to impact loads during quick starts and stops, earth-
quakes, explosions, etc. Stress impact factors based on experience are used in
design. The size of the factor depends on the relationship between the rapid-
ity of the loading and the structure's natural frequencies of vibration. If the
time between zero and maximum load is more than three times the period of
the lowest natural frequency, then the dynamic effects are insignificant and
the factor is 1.0. The factor will definitely be greater than 1.0 if the loading
time is less than one-half the period of the lowest natural frequency of the
structure.
The design of shafts is an example of the use of impact factors. The
design equation is usually expressed in terms of the section modulus J R or
its equivalent πd 3 16 and the maximum allowable shear stress max ss :

πd 3 J
= =
1K M 6 + 1K T6
M
2
T
2

. (17.28)
16 r max ss

The steady-state bending moment M and the steady-state torque T must be

adjusted for any possible shock or impact. The bending moment impact
factor K M ranges between 1.5 and 3.0 and the torque factor KT between 1.0
and 3.0, depending on how suddenly the load is applied. Values of πd 3 16
are listed in Table 17.3. Other parts of the fan should be analyzed similarly.
Failure modes for impact loading resemble those for static loading; that is,
the part may fracture or it may deform too much, either elastically or plasti-
cally. Notches and stress raisers should be avoided. Some parts perform
better if material is strategically removed.
Fans may be subject to cyclic loads due to frequent starts and stops, to
surging or other aerodynamic phenomena, to resonant conditions, etc. Such
loads can produce premature fatigue failures if they are not considered in the
design of the fan. Fatigue cracks typically originate at points of high stress
concentration and propagate until detected or fracture occurs. Low-cycle and
high-cycle fatigue are both possible in fans. Low cycle fatigue is generally
associated with starting and stopping and other major changes in speed. High
cycle fatigue is generally associated with operating at a speed that produces
resonant vibrations.
A material's ability to withstand cyclic loads is indicated by its endurance
limit Sn . Endurance limits are usually determined from tests of standard
specimens. However, actual parts will have different surface finishes, sizes,
or stress distributions and, so, will not perform exactly the same as the test
specimens.

‹#1999 Howden Buffalo, Inc.

17-20 FAN ENGINEERING

Table 17.3 Values of πd 3 16 for Shafts

in in.3

Dia. 1" 2" 3" 4" 5" 6" 7" 8" 9" 10"
0 0.196 1.571 5.301 12.57 24.54 42.41 67.35 100.5 143.1 196.4
1/16 0.236 1.723 5.639 13.16 25.47 43.75 - - - -
1/8 0.280 1.884 5.992 13.78 26.43 45.12 71.02 105.3 149.2 203.8
3/16 0.329 2.055 6.359 14.42 27.41 46.51 - - - -

1/4 0.384 2.236 6.740 15.07 28.41 47.93 74.82 110.3 155.4 211.4
5/16 0.444 2.428 7.136 15.75 29.44 49.39 - - - -
3/8 0.510 2.630 7.548 16.44 30.49 50.87 78.76 115.3 161.8 219.3
7/16 0.583 2,843 7.975 17.16 31.56 52.38 - - - -

1/2 0.663 3.068 8.416 17.89 32.66 53.92 82.83 120.6 168.3 227.3
9/16 0.749 3.304 8.877 18.65 33.79 55.49 - - - -
5/8 0.843 3.551 9.352 19.42 34.94 57.09 87.04 126.0 175.1 235.5
11/16 0.944 3.811 9.845 20.22 36.12 58.72 - - - -

3/4 1.052 4.083 10.35 21.04 37.33 60.38 91.39 131.5 182.0 243.9
13/16 1.169 4.368 10.88 21.88 38.56 62.08 - - - -
7/8 1.294 4.666 11.42 22.75 39.82 63.80 95.89 137.3 189.1 252.5
15/16 1.428 4.977 11.99 23.63 41.10 65.56 - - - -

The design of fans that are subject to impact or cyclic loads requires
consideration of many more factors than can be discussed here. Among the
important fracture-mechanics topics are: mean stress, alternating stress, stress
concentration, toughness, crack initiation, crack propagation, transition
temperature, and residual stresses.
Environmental factors also greatly influence the life of a fan. Creep, creep
rupture, stress-corrosion cracking, corrosion fatigue and any other effects of
temperature or gas composition must be considered.

Vibrations and Critical Speeds

There are two types of vibrations: forced and free. An elastic body will
vibrate freely at one or more of its natural frequencies if its equilibrium is
momentarily disturbed by an external force. The motion will gradually die
down because of damping. If an external force is applied repeatedly, an
elastic body will vibrate at the frequency of the external excitation, whether
this coincides with a natural frequency or not. Resonance occurs when the
excitation frequency coincides with one of the natural frequencies. Some
sources of excitation contain components at several frequencies. Large
amplitude vibrations accompany resonance unless there is considerable
damping in the system, but they can also be caused by large excitation forces
even when the damping is high.

‹#1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-21

The free vibrations of an elastic body can consist of an infinite number of

particle motions. Among these various modes of vibration are certain
principal ones wherein all the particle motions are at the same frequency and
follow a precise amplitude pattern. The most important principal modes of
vibration are those that occur at the lowest frequencies. It is convenient to
classify these modes of vibration according to their numerical order and the
direction of vibration relative to the principal axis of the part or member. The
direction of oscillation may be longitudinal, lateral, or angular. Longitudinal
vibrations are seldom serious in fan parts. Lateral vibrations cause bending.
Angular vibrations lead to torsion. The first bending mode may have a higher
or lower frequency than the first torsional mode, depending on the geometry
and material of the part. The various plates and beams that comprise a fan all
have natural frequencies. The lowest natural frequencies usually occur with
cantilever beams. Unshrouded impeller blades fall into this category.
Shrouding may raise the natural frequency to several times the unshrouded
value.
The lowest natural frequency f n of a uniform cantilever beam can be
calculated from the modulus of elasticity E of the beam material, the second
moment I of the beam cross section about the principal axis, the length l of
the beam, and its mass per unit length w using

.
352 gc EI
fn = . (17.29)
2π wl 4

The lowest natural frequency of an identical beam fixed at both ends

(instead of one) is 6.4 times higher than indicated by Equation 17.29. The
lowest natural frequency of a ring with radius r is

gc EI
f n = 2.68 . (17.30)
wr 4

Equations 17.29 and 17.30 can be used to estimate the lowest natural
frequencies of certain fan parts, such as blades and rings. Calculations are
more complicated for more complicated shapes. Modal analysis testing, as
described in a later section, can be used to find mode shapes and frequencies
for even the most complicated structures. Fan parts can be excited by periodic
aerodynamic forces. And detuning may be necessary to avoid resonance.
A fan, like any rotating elastic structure, will have certain operating
speeds at which objectionable vibrations are likely to occur. These critical
speeds correspond to the various natural frequencies of the system. Since it is
virtually impossible to perfectly balance a fan, there will always be an
excitation force with a frequency corresponding to the operating speed. If one
of the system's natural frequencies coincides with the rotational frequency,
resonance results.

© 1999 Howden Buffalo, Inc.

17-22 FAN ENGINEERING

Figure 17.8 shows the relationships between the nondimensional response

MX me and the frequency ratio f f n for the forced vibration of a system
resulting from rotating unbalance. The total vibrating mass M includes the
rotating mass m , which has an eccentricity e . The system amplitude is X ,
and the phase angle is φ . Note that, as the operating frequency f approaches
the natural frequency f n , the response increases. This increase depends on
the damping factor ζ , where c is the actual damping and cc is the critical
damping. Note also that the phase angle φ changes with frequency ratio.
Figure 17.8 is for a single-degree-of-freedom system. A more typical spec-
trum for a fan that has many degrees of freedom is shown in Figure 17.9.
Critical speeds are often erroneously considered a property of the shaft
only. This is not so; the bearings, supports, foundation, and soil all contribute
to the elastic properties of the system. Special terminology has been sug-
gested to convey this idea. However, we will continue to use the term critical
speed to indicate coincidence between the operating speed and a natural
frequency. The nature of the supporting portion of the system should be
specified whenever a critical speed is calculated. For instance, if details of the
foundation and soil are not known, the critical speed might be calculated for a
fan as if it were on a rigid foundation. If this is done, a considerable margin
between the operating speed and the calculated critical speed should be
provided so that the reduction due to foundation and soil flexibilities will not
place the actual critical speed too near the operating speed.

Figure 17.8 Response to Rotating Unbalance

‹#1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-23

Complex computer programs are available for calculating critical speeds.

These include the effects of the bearings, the supports, the foundation, and the
soil. Also, gyroscopic and impeller flexibility effects can be included when
appropriate. Detailed response calculations are also possible. That is, the
motions at various locations can be calculated for a given excitation force.
Approximate values of critical speed can be predicted from some fairly
simple equations. For a simple system, the maximum static deflection y as
determined from the equations in Table 17.2, can be used in

187.7
N cr = (17.31)
y

to determine the critical speed N cr . The 187.7 factor corresponds to speeds in

rpm and deflections in inches. For SI units, substitute 0.4984 instead.
Equation 17.31 is appropriate for those systems in which the mass can be
considered concentrated at a single point and the static deflection is calculated
at that point. The factors listed in Table 17.2 should be used to convert
deflections for uniform loads to equivalent deflections. If there is more than
one concentrated load, approximate results can be obtained by totaling the
deflections under each load. Do not include any unidirectional deflection,
such as that due to belt pull.

Figure 17.9 Typical Fan Vibration Spectrum

‹#1999 Howden Buffalo, Inc.

17-24 FAN ENGINEERING

Accuracy can be improved in the more complicated cases by using

Rayleigh's method, which involves the equation

m1 y1 + m2 y2 +!+ mn yn
N cr = 187.7 . (17.32)
m1 y1 + m2 y2 +!+ mn yn
2 2 2

When using this method, determine the total deflections under each load.
These deflections result not only from the load at that location but from all
other loads as well. The first approximation based on static loads m1 etc. is
usually within 1% of the correct value. This can be improved by using
2
dynamic load m1 y1 , etc.
Dunkerley's equation can also be used:

1 1 1 1
2
= 2 + 2 +!+ 2 . (17.33)
N cr N1 N2 Nn

In this equation, the actual critical speed N cr is given in terms of the critical
speed N1 , etc. for each mass m1 , etc. in the absence of all others.
The primary critical speed is excited by unbalance. A secondary critical
speed can be observed at one-half the speed of the first. This is caused by
unbalance in conjunction with gravity or, more likely, by a nonuniform
flexibility of the shaft resulting from a keyway or a flat spot on the shaft.
Many other resonances can occur, some of them at speeds below the operating
speed. During startup, the fan must accelerate through these speeds. With
normal damping, a well-balanced fan will easily pass through such criticals,
whereas an unbalanced fan may not.
The natural frequency f n of a simple torsional system can be calculated
from the modulus of rigidity G , the polar moment of inertia J of the shaft
cross section, the length l of the shaft, and the mass moments of inertia I m of
the concentrated masses. When two comparable masses are connected by a
uniform shaft,

1 çæ
gc GJ I m1 + I m2 äã
fn =
2π l åI m1 I m 2 â
. (17.34)

but, when one of the masses is so large compared with the other that the
corresponding end of the shaft can be considered fixed,

1 gc GJ
fn = . (17.35)
2π lI m1

© 1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-25

Torsional critical speeds rarely coincide with operating speeds in fans. When
they do, the easiest way to alter the critical speed is to change the coupling
flexibility. Some coupling designs provide increased damping too.

Balancing
Fan rotors must be balanced; otherwise, vibrations that could damage the
bearings or other parts will be produced. Balancing is accomplished by
redistributing the mass so that the principal inertia axis more nearly coincides
with the axis of rotation. Because perfect balance is never achieved, specifi-
cations usually list the permissible unbalance. This can be stated in terms
such as ounce-inches, which reflect the size and radius of a balancing mass
that would bring the rotor into balance. Also, eccentricity, the distance
between the principal inertia axis and the axis of rotation, can be specified.
Specific unbalance, or unbalance divided by rotor mass, can be specified, too.
However, the most common specifications deal with the vibratory effects of
unbalance. That is, a limit is placed on the displacement, velocity, or accel-
eration that can be measured on a vibrating part.
The amount of unbalance that can be tolerated varies with the speed and
mass of the rotor, the sturdiness of the bearings and the supporting structure,
etc. Table 17.4 lists the vibrations usually associated with certain operating
descriptions for several rotative speeds. Shown are both displacements in
mils and velocities in in./sec.
It is important to distinguish between static and dynamic unbalances. In
Figure 17.10, a rotating impeller is idealized as two axially separated discs on
a shaft between bearings. Centrifugal forces are shown as vectors. Portion A
of the figure depicts a statically unbalanced condition. The vibration will be
.

Figure l7.10 Static and Dynamic Unbalance

‹#1999 Howden Buffalo, Inc.

17-26 FAN ENGINEERING

Table l7.4 Vibration Limits for Fans (Bearing Measurements)

Vibration Severityl Displacement4 - mils peak to peak

Vel2 in./s Quality3 for various fan speeds5 - rpm
6
peak rms rigid flexible6 3600 1800 1200 900 720 600

.025 .018 good good 0.1 0.3 0.4 0.5 0.7 0.8
.040 .028 good good 0.2 0.4 0.6 0.8 1.1 1.3
.062 .044 good good 0.3 0.7 1.0 1.3 1.6 2.0
.100 .071 good good 0.5 1.1 1.6 2.1 2.7 3.2
.16 .11 satisfactory good 0.8 1.6 2.5 3.3 4.1 5.0
.26 .18 satisfactory satisfactory 1.4 2.7 4.1 5.4 6.8 8.1
.40 28 unsatisfactory satisfactory 2.1 4.2 6.3 8.4 10.5 12.6
.62 .44 unsatisfactory unsatisfactory 3.3 6.6 9.9 10.3 16.5 19.8
1.00 .71 unacceptable unsatisfactory 5.3 10.6 15.9 21.2 26.5 31.8
1.56 1.10 unacceptable unacceptable 8.3 16.6 24.8 33.1 41.4 49.7
1
Vibration severity is classified by ranges of rms velocities. Each range is
identified by its highest value in mm/s (not shown).
2
Most portable balancing machines indicate peak (not peak-to-peak) velocity.
3
Balancing qualities listed are ISO judgements. Alarms are often set at 0.3
in./s and shut down at 0.45 in./s.
4
Most portable balancing machines indicate peak-to-peak displacements.
Table values correspond to single-frequency components.
5
Filters should be tuned to the fan speed when measuring vibration severity.
6
Rigid support means that the fundamental natural frequency of the ma-
chine/support system is higher than its main excitation frequency. Flexible
support means that the machine/support system’s fundamental natural fre-
quency is lower than the main excitation frequency.
Adapted from the data of ISO: "Mechanical Vibration of Large Rotating Machines with
Speed Range from 10 to 200 rev/s - Measurement and Evaluation of Vibration Severity in
situ," International Organization for Standardization, ISO-3945-1977, corrected and re-
printed-1978-03-01.

greater at the left-hand bearing than at the right-hand bearing, assuming equal
support stiffness. If the impeller were not rotating, it would tend to assume a
position with the unbalanced mass at the bottom. Static unbalance can be
corrected by adding a single balancing mass at the proper distance on the
opposite side of the impeller, as in portions B and C of the figure. The
method shown in portion B is best, as will soon be evident. Portion C of the
figure depicts a dynamically unbalanced condition. If the impeller is statically
balanced as shown, the vibrations at the two bearings will be of the same
magnitude but in opposite directions. Dynamic unbalance can only be cor-
rected by adding a balancing couple as shown in portion D of the figure.

‹#1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-27

Some balancing situations have not been pictured in Figure 17.10. For
instance, static unbalance can be corrected out of plane without introducing a
dynamic couple: half the balancing mass could be added to each of two planes
on either side of the unbalance. Dynamic unbalance can be corrected out of
plane also. The balancing couple can have any moment arm. Such out-of-
plane corrections will work only if the rotor behaves rigidly. However, if it
behaves flexibly and distorts when it rotates, multiplane balancing may be
necessary. Fortunately, most fans can be balanced with two-plane
corrections.
Balancing will reduce only those vibrations caused by unbalance.
Vibrations caused by looseness between parts, coupling misalignment,
mismatched belts, or other external sources must first be corrected before
balancing is attempted. The effects of a bent shaft can sometimes be balanced
out but only if the bend is slight.
Most fans are balanced before shipping. Machines like that shown in
Figure 17.11 are used extensively. The impeller is mounted on a mandrel and
spun at relatively low speed in very flexible bearings. Every impeller design
can be calibrated for very quick and accurate balancing. However, when the
impeller is placed on its own shaft and bearings, etc., the mounting may differ
from that used for balancing, so a trial run should be made and the balance
touched up if necessary.

Figure l7.11 Shop Balancing of an impeller on a Mandrell

© 1999 Howden Buffalo, Inc.

17-28 FAN ENGINEERING

The impeller can also be balanced on its own shaft in a shop balancing
machine like that shown in Figure 17.12. Here, too, the rotor is balanced at
low speed, and accuracy is possible because of the machine's built-in sensi-
tivity. Touch-up balance may also be needed after erection. This is especially
true for highly stressed wheels and for high-temperature jobs.

Figure 17.12 Shop Balancing of an Impeller and Shaft

‹#1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-29

In the field, fans are usually balanced with the aid of a portable, electronic
balancing machine. Such machines can indicate the vibration amplitudes in
terms of displacement, velocity, or acceleration. They can also indicate
frequency and, with the aid of a strobe light, phase angle, too. Most use a
seismic-type velocity probe and can filter out all frequencies outside the
narrow band of interest. During balancing, the filter is tuned to the rotating
speed. Balancing machines are also useful for diagnosing other problems that
cause vibration. Table 17.5 lists some diagnostic clues. Procedures for using
electronic balancing machines are furnished with the apparatus. Vector
diagrams can be used advantageously.

Table 17.5 Diagnostic Clues

Amplitude Frequency Phase Possible Cause

Steady - 1 × rpm Steady - Unbalance (See text

radial largest single reference for clues to correcting.)
mark
Axial largest 1 or 2 × rpm 1, 2, or 3 Misalignment or bent
reference marks shaft (Check with dial
indicators.)
Unsteady 1 × rpm Unsteady Resonance (Check for
impeller or other
flexibility using modal-
analysis techniques.)
Unsteady 2 × rpm 2 reference marks Looseness (Check
bolts, keys, etc.)
Unsteady rpm ÷ 2 Unsteady Oil whip uncommon
for fans)
Low 60 or 120 Hz 1 or 2 rotating Electrical
marks
Erratic many × rpm Erratic Faulty antifriction
bearings
Erratic 1 or 2 × belt rpm Unsteady Defective belts
(Check by freezing
with strobe.)
Low blade-passing Steady Aerodynamic (Check
cutoff clearance.)

© 1999 Howden Buffalo, Inc.

17-30 FAN ENGINEERING

Balancing is also possible without a machine. The amounts and locations

of the required balancing masses can be determined with a piece of chalk, an
assortment of detachable weights, a good sense of touch, and a little patience.
Any unbalance will displace the shaft, which can usually be observed by
chalking the shaft. This displacement will be fixed with reference to the rotor
and, so, will rotate with reference to a fixed point in space. If a piece of chalk
(or the like) is held at a fixed point so that it just touches the rotating shaft, it
will-leave a mark at the high spot. If the fan is balanced, the mark will extend
all the way around. The greater the unbalance, the shorter the mark will be.
The center of the line would be at the same angular position as the unbal-
anced mass except for lag. The amount by which the mark lags behind the
unbalance will vary, depending on the ratio of operating speed to critical
speed and the amount of damping in the system. For operation well below the
critical speed, the angle of lag will be 0°. For operation closer and closer to
critical, the angle of lag will gradually increase becoming 90° at critical.
Above the critical speed, the angle of lag continues to increase until it be-
comes 180°. For most fans, the angle of lag will be between 15° and 45°.
When a fan is statically out of balance, as shown in Figure 17.10, the
chalk marks at both bearings will be at the same angular position relative to
some point on the rotor. The displacement amplitudes, indicated by the
lengths of the marks, will be equal only if the unbalance occurs midway
between the bearings.
The correct angular and lateral position for a trial mass can be estimated
from an analysis of the chalk marks. If the estimated angular position is
correct but the trial mass is too small, rechalking will produce a longer line in
the same angular position. If the trial mass is too large, the center of the line
will shift 180°. If the trial mass is just right, the line will extend all the way
around the shaft. However, if the angular position is not quite correct, the
center of the line will shift accordingly.
To correct dynamic unbalance, two trial masses should be used. Usually,
they are placed on or near the inlet shrouds for a double-inlet impeller or on
the flange and backplate for a single-inlet impeller. The correct angular
position for each trial mass can be estimated from the chalk mark at the
nearest bearing for Arrangement-3 fans. If the fan is not statically unbal-
anced, these marks will be 180° from each other and about equal in length.
If the chalk marks extend all the way around the shaft but a considerable
vibration on the bearings persists, the trouble can usually be traced to weak
supports or inadequate foundations. Reinforcing the supports, adding mass,
or both may be required.

Vibration Isolation
Vibrations are induced in a fan by unbalanced centrifugal forces and by
aerodynamic forces. Some net force will be transmitted to the supporting
structure. In fact, the supporting structure forms part of the vibrating system,
as discussed in the critical-speed section. Many fans can be installed without
isolators, but the system must be designed accordingly and good balance must
be maintained. Vibration isolation is necessary whenever the vibrations in the

‹#1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-31

supporting structure must be limited because they are annoying or even

destructive.
The amount of force transmitted between members of a vibrating system
depends on the masses, stiffnesses, and damping present. The ratio of trans-
mitted force to impressed force is called transmissibility TR and is related to
the natural frequency f n of the system as well as to the disturbing frequency
f Ignoring damping,

1
TR =
çæ f äã 2
. (17.36)

åf â
n
−1

The natural frequency of a fan mounted on isolators can be determined from

the static deflection y using

1 g
fn = . (17.37)
2π y

Figure 17.13 Transmissibility

‹#1999 Howden Buffalo, Inc.

17-32 FAN ENGINEERING

Equation 17.36 indicates that the transmissibility will be unity when the ratio
of disturbing frequency to natural frequency equals the square root of two.
Transmissibility will be less than unity for higher values of this ratio and
greater than unity for lower values. The effects of damping are shown in
Figure 17.13. If isolators are to be used at all, experience suggests that
transmisability, as calculated from Equation 17.36, should be reduced to 0.20
and preferably to 0.10 or lower.
The materials usually used for vibration mounts are steel springs, rubber-
in-shear, and cork. Cork or a similar material can be used in sheets under
direct loads, since it will compress. Rubber, being virtually incompressible, is
usually bonded to two steel parts and loaded in shear. Steel springs are loaded
directly in compression. Snubbers may be needed to prevent swaying.
Several isolators are usually required for support. These should be installed
according to the distribution of the load, which may be due to thrust forces as
well as to the dead weight of the equipment. If the fan is to be held level, the
deflections must be uniform. This can be accomplished by varying the
durometer of rubber, the sheet area of cork, or the spring constant of a spring.
The deflection should be chosen according to Equations 17.36 and 17.37 to
give the proper natural frequency. Usually, steel springs are required for fan
speeds below about 700 rpm but can be used at any speed. Rubber-in-shear
can be used for speeds above 700 rpm, but steel springs will often be needed
to limit the transmissibility to acceptable values. Cork can be used above
about 1200 rpm with a similar proviso.
It is especially important that a fan and its driving motor be mounted on a
common rigid base. If isolation is required, it should be provided between the
base and the supports so that the base, rather than the isolators, must
withstand any torque or belt pull. Although the additional mass of the base
will limit the system's amplitude of vibration, it will not alter the magnitude of
the forces transmitted. Whenever the vibration mounts must be incorporated
in hangers, a fail-safe design must be used.
Vibratory short circuits must be prevented. Flexible connections must be
used between the fan and any ductwork on either the inlet or the outlet. Fans
used in an air-conditioning cabinet can be isolated from the cabinet, or the
entire cabinet can be isolated from its supports. Steam lines, water lines,
power lines, etc. must be flexibly connected. Inertial masses may be needed
to restrict the amplitude of vibration regardless of whether the fan is mounted
on isolators or not. As a rule of thumb, use a mass of concrete two to three
times the mass of the fan and drive.

Mechanical Testing
Various mechanical tests can be performed on a fan or its parts, either to
provide a basis for design calculations or to verify them. Although
destructive tests are occasionally useful, most fan testing is nondestructive.

© 1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-33

Spin tests may be performed to determine the buckling speed of a blade,

the yielding speed of a shroud, or the bursting speed of an impeller.
Nondestructive spin testing may be used simply to prove that an impeller will
be able to operate at a certain speed. That speed can be the expected
operating speed or some higher speed. Overspeed testing demonstrates that
there is a margin of safety at the normal speed. Overspeeding may also
beneficially redistribute static stresses.
A vacuum test pit is convenient for spin testing. The one illustrated in
Figure 17.14 was made of heavy steel and reinforced concrete and had a
vacuum pump that could provide an absolute pressure of fifty micrometers of
mercury (0.050 mm Hg or 0.00197 in. Hg). The power requirements of the
air turbine, ignoring bearing friction, were thereby reduced to less than
1/15000 of that required to drive the impeller at standard atmospheric
pressure. The heavy construction and underground pit arrangement provided
protection against flying parts.

Figure l7.14 Vacuum Spin-Test Pit

In Figure 17.15, a much larger pit is illustrated. Although not intended for
destructive testing, it, too, was underground and had a very heavy reinforced-
concrete cover to protect against flying parts. Since this pit could not be
evacuated, highly powered drives had to be used for larger impellers at high

© 1999 Howden Buffalo, Inc.

17-34 FAN ENGINEERING

Figure l7.15 Large Test Pit

speeds. (Up to 1500 hp was available at this facility, and wheels up to 193
in.could be tested.) Impellers could be mounted on their own shafts and
bearings and even in their own housings, if desired.

‹#1999 Howden Buffalo, Inc.

 CHAPTER 17 – FAN MECHANICS 17-35

Figure l7.16 Modal-Analysis Test Setup

Strain gages, brittle lacquers, and photoelastic materials can be used to

measure strains so that stresses can be calculated. Brittle lacquers and
photoelastic materials may be useful in defining strain patterns, but because of
the problems associated with rotating impellers, strain gages are used more
extensively for fans. FM transmitters and receivers as well as slip rings have
been used to test fans. Both static and dynamic strains are measured and
recorded. Signals can be recorded during field testing and returned to the
laboratory for processing.
The vibration characteristics of a fan can be found in various ways.
Natural frequencies can be excited by impact, shaking, or explosive shock.
(Vibration at other frequencies can also be forced by shaking.) Then, point
responses can be measured with accelerometers, velocity sensors, or
displacement-measuring devices. From the point responses and excitation
forces, mode shapes can be deduced. Sometimes, if the structure is flexible

© 1999 Howden Buffalo, Inc.

17-36 FAN ENGINEERING

enough, the mode shape can be estimated simply from visual or tactile
observation, but usually, computer solution is required.
Figure 17.16 shows a computer-based, modal-analysis test setup. The test
fan is a heavy-duty centrifugal impeller on its own shaft. The unit is
supported in slings. A geometrical grid is derived from the measured three-
dimensional coordinates of selected points on the fan and recorded in the
computer. During the test, the fan is excited with a hammer blow at one of
the points. The response at each of the points is then measured with an
accelerometer. Since the hammer is equipped with a force gage, the
excitation force can also be measured. The output from the accelerometer
gives the response at all frequencies resulting from the hammer blow. This
time-domain data is transformed by the computer into frequency-domain
information for both the responses and the excitations. (The resulting
frequency spectra are called Fourier transforms, and the computer calculations
fast-Fourier-transform analyses.) Transfer functions are then formed for each
response/excitation pair. Peaks will occur defining the various natural
frequencies. To determine mode shapes and damping ratios, the computer
extracts amplitudes for all pertinent frequencies from the transfer functions for
all the pairs. This information can then be used to animate the geometrical
grid at any frequency. This kind of analysis, which yields mode shapes as
well as frequencies, can be very useful in design and development.
An impact-response test (bump test) will give frequency information but
not mode shapes. Figure 17.17 shows a large industrial airfoil centrifugal
under test. The impeller is mounted on its own shaft but is supported in a
balancing machine. During the test, the fan is bumped, and the response is
measured with an accelerometer. As with the previously described modal
analysis, a frequency spectrum of the response is then formed. No frequency
spectrum can be formed for the excitation because the excitation force is not
measured. The frequency-domain information is then displayed on a CRT or
computer screen. Although limited, this information is still useful in design
and development.
Shaker tests are performed either by mounting the fan on the shaker or by
suspending the fan elastically so that it can be moved by the shaker. The
excitation frequency is controlled. Although the excitation force can be
measured, it is more common to measure the vibration amplitudes of the table.
Point responses can be sensed with an accelerometer and a frequency
spectrum formed as for impact response testing. Or, transfer functions can be
formed as for the modal-analysis testing previously described.
Another use for shaker tests is seismic qualification. In this kind of testing,
the fan must show that it will continue to perform its function after exposure
to a specified shaking pattern. Specifications vary, especially concerning
wave form. But, usually, a frequency sweep is required, and amplitudes are
defined.
Shock tests are specified to demonstrate that a fan will continue to
perform after exposure to a certain shock. Shock-testing machines administer
the shock by dropping a heavy mass through a prescribed arc to strike the
structure on which the test unit is mounted.

© 1999 Howden Buffalo, Inc.

 FAN MECHANICS 17-37

. Shock tests can also be performed by mounting the test fan on a barge and
detonating measured charges in the water at specified distances from the
barge.

Figure l7.17 Impact-Response Testing

‹#1999 Howden Buffalo, Inc.