introduction to balancing - maintenance · iso 2953 describes balancing machines, and iso 3719 has...
TRANSCRIPT
INTRODUCTION TO BALANCING
WILLIAM C. FOILES
1. INTRODUCTION TO BALANCING
This chapter will introduce the fundamentals of balancing. Balancing
will be defined and explained as to what the process encompasses and
the physical principles behind balancing. This chapter examines aspects
of balancing such as balance weights and auxiliary computations.
The most important rule to understand and remember ifone plans to balance rotating equipment safety rules every-thing. If one is not comfortable with a situation then one shouldevaluate whether to continue or not. To balance one must op-erate and make changes to rotating equipment. This can eas-ily lead to harm to people, the environment, and equipment.Safety comes first.
This material can not attempt to cover the many aspects related to a
safe balancing job. High energy will be present near all balancing activi-
ties. Dangers may include rotating parts, temperature extremes, electri-
cal energy, high pressures, hazardous chemicals, enclosed spaces, explo-
sive gases, poisonous gases (including suffocating gases), heights, falling
material, trips and falls, driving (within the facility or to and from the
facility), magnetic fields, and potentially many other location specific
hazards. Balancing equipment itself may present hazards such as strobe
lights if used and fish tails – a seismic device used to measure shaft abso-
lute vibration; it looks like a fish tail and is used in direct contact with the
shaft; both of these if used for balancing can be dangerous, and a number
of facilities have restricted or banned their use.
One always needs familiarity with the local safety practices prior to
starting any job. Good practice includes reviewing the safety aspects of
any job, balancing included.
2. STANDARDS REGARDING ROTOR BALANCING
What is balancing? ISO DIS 19499 E, currently a draft International
Standard introducing the balancing standards, says the following in re-
gard to balancing.
Date: August 14–17, 2006 UVA Rotordynamics Short Course.
1
2 WILLIAM C. FOILES
Balancing is a procedure by which the mass distribution
of a rotor (or part or module) is checked and, if necessary,
adjusted to ensure that balance tolerances are met.
Rotor unbalance may be caused by many factors, in-
cluding material, manufacture and assembly, wear during
operation, debris or an operational event. It is inevitable
that every rotor, even in series production, has an individ-
ual unbalance distribution.
New rotors are commonly balanced by the manufacturer
in specially designed balancing machines before installa-
tion into their operational environment. Following rework
or repair, rotors can be re-balanced in a balancing machine
or, where appropriate facilities are not available, the rotor
can be balanced in situ, see ISO 20806. In the latter case
the rotor is held in its normal service bearings and support
structure and installed within its operational drive train.
The unbalance on the rotor generates centrifugal forces
when it is rotated in a balancing machine or in-situ. These
forces can be directly measured by force gauges mounted
on the structures supporting the bearings or indirectly by
measuring either the motion of the pedestal or the shaft.
From these measurements the unbalance can be calcu-
lated and balancing achieved by adding, removing or shift-
ing of correction masses on the rotor. Depending on the
particular balancing task the corrections are performed in
one, two or more correction planes.
From this it may not be evident what occurs during rotor balancing.
The first paragraph in the above quote relates more to balancing rigid ro-
tors. Generally, the goal to in situ balancing is to reduce vibration levels
to acceptable levels. When one is balancing it is good practice to agree
upon the goal prior to commencing the balance; otherwise, it can be dif-
ficult to know when one has finished the balance. Balancing a real rotor
in the field can consume a great deal of time, and one will have limita-
tions placed on the time spent.
Take time required into consideration when planning a balance job.
Advice the customer or operator how long the job may take and agree
to have the required time, resources, and number of runs ahead of the
balance. When in doubt use a pessimistic estimate.
2006 UVA ROTORDYNAMICS SHORT COURSE 3
Without too great of a diversion, some coverage of relevant standards
will be given. Many countries’ standard organization have their own stan-
dards regarding balancing, and many of these are adopting the ISO stan-
dards and aligning with these. The aforementioned draft document, ISO
19499, will review the balancing standards if accepted.
ISO 1925 contains the vocabulary used in balancing, and ISO 2041 has
the vocabulary for vibration and somewhat for balancing. This is unusual
for standards to have separate places for the terms, and it makes it diffi-
cult to maintain consistency.
2.1. Balance Quality and ISO 1940. This article does not cover low speed
balancing in significant detail. ISO 1940-1 and ISO 1940-2, parts 1 and
2, cover this subject. The balance quality grade, G, is introduced in ISO
1940-1.
ISO 1940 reasons that the permissible residual imbalance of a rigid ro-
tor (Actually, the term concerning rigid rotor has recently changed, adding
some confusion.) is proportional to the mass of the rotor, i.e. Uper ∼ M .
It then defines a specific residual imbalance, similar to mass eccentricity,
as
eper =Uper
MNext, the standard states that experience has shown that eper varies
inversely proportional to rotor angular velocity, Ω.
Thus, this line of reasoning results in the following.
eperΩ= constant
G , the balance quality grade is defined as eperΩ in units of mm/s. For
a thin disc the balance quality would equate to the peak velocity of the
mass eccentricity. Theoretically, on a soft bearing balance machine, with
the disc running well above the resonance, this would be the peak veloc-
ity of the measured vibration.
Suggested balance grades are in ISO 1940-1 for rigid rotors, or rotors
in a rigid state. Generally, balance grades have a separation factor of 2.5.
This gives typical grades of 6.3, 2.5, 1, etc.
To obtain the permissible imbalance in terms of mass–length or weight–
length, one reverses the equations.
(1) Uper =GM
Ω=
MeperΩ
Ω
For consistent units in the above M is in g, Ω is in rad/s, and Uper is in
g-mm. Use of other units such as ounces, pounds, and rpm is possible.
4 WILLIAM C. FOILES
Below Ui noz is the permissible imbalance in ounce-inches, W is the rotor
weight in pounds, and N is the rotor speed in rpm.
(2)
Ui noz =W lb16 oz
1 lb
1 in
25.4 mm
60 sec-per-min
N rpm(2π rad/rev), or
Ui noz = 6.0153GW
N
2.2. API vs. ISO 1940 Balance Quality Grade. One may notice the simi-
larity in form to Equation 2 of the balance tolerance in API 617, American
Petroleum Institute compressor specification, of 4W /N . This API speci-
fication provides a relatively tight equivalent balance quality grade G of
0.665.
API =4W
N, or
G API =4
6.0154= 0.665
2.3. Other Standards. ISO 2953 describes balancing machines, and ISO
3719 has the symbols for balancing machines and their associated in-
strumentation. Balancing machines have inherent safety concerns; ISO
7475 covers balancing machine enclosures and protective measures for
the measurement station.
ISO 8821 pertains to shaft and fitment key conventions. In the Unites
States this has not been too controversial; mostly, half keys have been
used. However, internationally this has not always been the case. In fact,
because the standard called for full shaft keys and no disc key at one time,
some countries or organizations switched their conventions.
Presently, the ISO 8821 calls for the use of half keys. Since this was not
always the case, one may run into rotors or pieces that were balanced in
a different manner. The standard covers this also. Additionally, practical
methods are examined to balancing with keys.
ISO 10814 deals with the susceptibility and sensitivity of machines to
imbalance. It has some interesting points, but I have seldom seen it used.
It may be too theoretically oriented for most people practicing balancing.
It does contain good pertinent information, which may not make a differ-
ence to the low level balancing practitioner trying to balance a low speed
rigid fan.
Methods and criteria for balancing flexible rotors is covered by ISO
11342. If one desires to balance flexible rotors, one should read this stan-
dard.
2006 UVA ROTORDYNAMICS SHORT COURSE 5
ISO 20806 deals with in-situ balancing of medium and large rotors, cri-
teria and safeguards. This provides guidance on the type of balancing
covered here. ISO 20806 states the following about in-situ balancing.
Balancing is the process by which the mass distribution
of a rotor is checked and, if necessary, adjusted to ensure
that the residual unbalance or the vibrations of the jour-
nals/bearing supports and/or forces at the bearings are
within specified limits. Many rotors are balanced in spe-
cially designed, balancing facilities prior to installation into
their bearings at site. However, if remedial work is carried
out locally or a balancing machine is not available, it is be-
coming increasingly common to balance the rotor in-situ.
In-situ balancing is the process of balancing a rotor in
its own bearings and support structure, rather than in a
balancing machine....
For in-situ balancing, correction masses are added to
the rotor at a limited number of conveniently engineered
and accessible locations along the rotor. By so doing the
magnitude of shaft and or pedestal vibrations and/or Me-
chanical vibration U In-situ balancing of rotors U Guid-
ance, safeguards and reporting unbalance will be reduced
to within acceptable values so that the machine can oper-
ate safely throughout its whole operating envelope. In cer-
tain cases, machines that are very sensitive to unbalance
may not be successfully balanced over the complete oper-
ating envelope. This usually occurs when a machine is op-
erating at a speed close to a lightly damped system mode,
(see ISO 10814), and has load dependent unbalance.
ISO 20806 may be helpful to the balancer. Also, this document can
provide guidance to someone purchasing a balance job; it gives sample
report features in an appendix. I recommend reading and understanding
this document.
API 684 is a tutorial on rotordynamics and balancing. Understand-
ing rotordynamics, forced response, resonances, and mode shapes, will
greatly improve ones skills at balancing from determining the appropri-
ateness of balancing to placing the trial weights.
API 670 covers machinery protection systems. Other ISO standards re-
garding vibration may also pertain to a particular balance; customers of-
ten ask to meet some standard. Some may find some of the ISO criteria
rather generous, and the balancing goal may be much less than one of the
6 WILLIAM C. FOILES
boundaries – in general one should not provide a balance that leaves the
vibration close to an alarm condition, at least as a permanent solution.
3. THE BALANCING PROCESS AND STANDARD ASSUMPTIONS
This section reviews what occurs during a rotor balance in situ and
what assumptions one uses in general to affect a balance.
The common assumptions used in balancing are listed below.
(1) The mass of the balance weight additions or removals are insignif-
icant in comparison to the overall rotor mass. Additionally, this
means that the system dynamic properties do not change as a
function of the balancing process. Balance weights act as 1× or
synchronous forces applied to the rotor. As such, the balance
weight should not deform the structure of the balance plane, ei-
ther.
(2) Linearity, the rotor system responds linearly, specifically to 1×
synchronous force input (balance weights). Given linearity, ro-
tor system response to balance weights attached to the rotor that
generate a force synchronized to rotational speed or frequency
will occur at a frequency equal to 1×. Response to a time delay
in the input force (i.e. angular positioning of the balance weight)
results in a similar delay in the 1× response. This combines with
the typical linearity above to give a complex valued linearity.
(3) Repeatability, the rotor system exhibits repeatable behavior.
(4) Balancing addresses only the 1× response, generally vibration but
sometimes force.
3.1. Balance Weights. Balance weights consist of eccentric addition or
removal of mass from the rotating element. The actual mass involved in
balance weight addition or removal leaves the total rotating mass rela-
tively unchanged.
For example, on a class of large gas turbine that I used to balance regu-
larly with total weight around 100,000 pounds for the rotor I would typi-
cally use from 0.83 pounds (375 g or 13.23) to 1.38 pounds (625 g or 22.05
oz). These weights did not affect the system dynamics, being such a small
fraction of the overall mass of the rotating element.
So, how does such a relatively small amount of mass help to balance
a rotor? One must remember that locally in the rotor force results from
mass eccentricities spinning about the center of rotation. The proper full
units for a balance weight includes both the mass and the eccentricity,
m–e , where m stands for mass and e for the radius or eccentricity of the
mass.
Force is mass times acceleration. For circular motion acceleration is
2006 UVA ROTORDYNAMICS SHORT COURSE 7
a =v 2
e= eω2 v - velocity; ω - angular velocity
For a mass revolving at a circular eccentricity the force is given below.
(3) F = meω2
To use Equation 3 one must use consistent units. A 500 g weight at a 24
inch radius (61 cm) spinning at 3600 rpm generates 9738 pounds of force
(43,345 N). This can be seen by converting 500 g to pounds, 500 g = 1.102
pounds and converting 3600 rpm to rad/sec 3600 rpm = 376.99 rad/sec.
In U.S. customary units pounds represents the force created by the mass
of 500 g at 1 g. 1 g = 386.089 in/sec2. Thus, Equation 3 gives
F = meω2
=1.102
386.089(24)(376.99)2
= 9738 pounds
Calculating the force created by balances weights is good practice. If
there is any question as to the safety of adding the weight, further con-
siderations may be appropriate.
4. LINEARITY
In many ways, linearity is what makes balancing methods work. In
the context of balancing rotor systems linear response to balance weights
implies that the balance weight produces a response at 1× the speed or
frequency of the rotor and only at 1× the frequency of the rotor. This
assumes the use of a linear scale and not one like db’s or other non-linear
measure.
Other measures such as force or strain could be used; although these
have few uses in practice, unless used for a special test. For in situ bal-
ances, vibration values determine the need and quality of the balance;
balance machines may use force measurements. Whatever measure one
uses, it should respond linearly to the force induced by the balance weights;
for example, velocity in mm/s will work for balancing, but V dB’s will not
work in the usual manner, since it is not a linear measure with regard to
the effect balance weights have.
Linearity as used in balancing means two things, generally. First, the
amplitude responds proportionally to the magnitude of the balance weight,
mass-length. Second, the phase of the response changes directly to changes
8 WILLIAM C. FOILES
in the angular location of the balance weights. One may call these ampli-
tude and phase linearity.
The phase linearity corresponds to the response following time delays
to the input force that results from the balance weight angular position-
ing. Both amplitude and phase linearity should depend upon the cir-
cumferential symmetry of the rotor. Many rotors do not have perfectly
symmetric axial cross sections, such as motors and generators, but this
assumption has proved reasonable over the years. Balancing crankshafts
may be another story.
4.1. Linear Response, Influence Coefficients, and the Fundamental Bal-
ance Equation. In this section the following notation will be used.
NOTATION FOR INFLUENCE COEFFICIENTS
H Complex influence coefficient
H Matrix of influence coefficients
h |H |
ω Rotor speed
t time
θ ∠(H) or the lag angle to a unit imbalance at
angle 0
φ Balance weight angle referenced to the bal-
ance plane in lag angles
W Arbitrary balance weight with amplitude, w
and angle φ
R Response due to balance weight
R Multiple response measurements, vector of
measurements
R0 Initial response vector
B Vector of balance weights
Placing a unit balance weight on a balance plane at a particular loca-
tion produces in a linear system a response of amplitude h at the mea-
surement location. We assume that the location on the balance plane
does not matter; the response due to a unit balance weight always pro-
duces an amplitude of h. Of course, this is only the effect due to the bal-
ance weight and does not include other vibration that may be present.
This property comes from linearity and the angular symmetric response
that has been assumed. This response will be sinusoidal of the following
form, h cos(ωt −θ) with the unit balance weight placed at the 0 reference
location for the balance plane.
If one uses an oscilloscope to display the vibration signal and triggers
or starts the display on the shaft trigger signal Figure 1 shows a 1× signal
2006 UVA ROTORDYNAMICS SHORT COURSE 9
with peak-to-peak (pp) amplitude of 3 and a phase lag of 45 degrees, and
a second plot with an amplitude of 4 pp with a phase of 270 degrees.
FIGURE 1. Amplitude of 3 pp, phase of 45 – Amplitude 4
pp, phase 270
timeDiv :
5ms/div
Channel A :
1 V/div
Channel B :
1 V/div
XY
OFF
OffsetB
5
OffsetC
0
OffsetA
0
timeDiv :
5ms/div
Channel A :
1 V/div
Channel B :
1 V/div
XY
OFF
OffsetB
0
OffsetC
0
OffsetA
9
The trigger in Figure 1 starts the waveform displays. To measure phase
one starts on the left side, the time origin of an oscilloscope, and mea-
sures phase from this point. If one knows, as in these instances, that the
waveform is 1× filtered then one can measure the time for one cycle and
the time to the first peak. The phase angle in degrees is given by 360 times
the ratio of the time to the peak divided by the time for one cycle.
One cycle in Figure 1 is 20 ms. The time to the peak of the left display
is 2.5 ms. The amplitude is 3 divisions pp or 3 volts – this assumes that 1
sensor unit equals to 1 volt. Thus, the phase for the left display is
θ1 = 3602.5 ms
20 ms= 45
Similarly, the right display has time lag of 15 ms until it reaches its peak.
This gives the phase lag of the right display as
θ1 = 36015 ms
20 ms= 270
To gain familiarity with the definition of phase lag two more figures are
given. Verify the phase lags in Figure 2 yourself. What are the amplitudes?
Use the lag angle convention, the most common, to measure angles
in the balance plane from the reference 0 angle. With this convention
adding a unit weight at angle φ produces a response of h cos(ωt −θ−φ).
10 WILLIAM C. FOILES
FIGURE 2. Phase of 90 – Phase 0
timeDiv :
5ms/div
Channel A :
1 V/div
Channel B :
1 V/div
XY
OFF
OffsetB
5
OffsetC
0
OffsetA
0
timeDiv :
5ms/div
Channel A :
1 V/div
Channel B :
1 V/div
XY
OFF
OffsetB
0
OffsetC
0
OffsetA
9
The complex valued influence coefficient is generally defined in terms
of amplitude and phase as (h,θ), the response with the lag angle, which
is equivalent to a time delay at the given rotor speed.
At this point something strange occurs in general practice. Either one
views the angles as implicitly negative or accepts the following complex
formulism. Usually, one can overlook this fine point but not always. Basi-
cally, we move to a lag angle world at this point.
The complex formulism used to describe balancing uses Euler nota-
tion for complex exponentials. The complex influence coefficient is given
below, with i 2 =−1. Remember, you are now in the world of lags.
(4)H = (h,θ) = he iθ
= h (cos(θ)+ i sin(θ))
The response to a balance weight W = (w,φ) in this notation would be
(5) R = HW as a multiplication using complex arithmetic
Equation 5 will lead to the basic equation used for balancing from a
single plane to multiple planes and speeds. Write this equation in matrix
from as in Equation 6 for multiple balance planes and/or multiple mea-
surement points. The measurement points may include various sensors,
different speeds, or a variety of operating conditions.
The basis for influence coefficient balancing is given in Equation ??,
rewritten below for reference. The goal of any balance effort is to achieve
an acceptable residual response after balancing, R. Thus, some type of
2006 UVA ROTORDYNAMICS SHORT COURSE 11
minimization is used to find the balances weights B that minimize this
equation. Here bold face represents matrices and vectors.
(6) R = HB+R0
The response R and the influence coefficient matrix, H have the same
number of rows. Again, the response can include sensor locations and
data from various discrete speeds. The column dimension of H matches
the number of balance weights, i.e. the row dimension of H.
4.2. Real and Complex Valued Influence Coefficients. Most mechanical
balancing has used complex influence coefficients where the amplitude
and phase of the vibration response and the balance weights is viewed as
the polar representation of complex number. Open loop magnetic bal-
ancing has preferred real valued influence coefficients. The linear pro-
gramming formulation of min-max mechanical balancing requires the
use of real valued influence coefficients.
Complex arithmetic handles the phase angle information properly, be-
cause the rotor-bearing system is time invariant. The time invariance
means that knowledge of the response to a balance weight at 0 angular
position implies knowledge at all other phase angles; because other an-
gles simply correspond to delays of the input which are the forces created
by the balance weight.
Getting from a complex influence formulation to a real formulation
can be done with the aid of an algebraic isomorphism of the complex
numbers, C, to a subset of 2×2 real matrices, R2×2.
The isomorphism, Φ, is given in Equation (7).
Φ : −→ R2×2
Φ (Z ) =Φ (x + i y) =
(
x −y
y x
)
for z = x + i y ∈C, and x, y ∈R
(7)
The basic properties of this correspondence of the complex numbers
with a subset of 2×2 real matrices follow.
12 WILLIAM C. FOILES
(1) Φ (z1z2) =Φ (z1)Φ (z2) =Φ (z2)Φ (z1).
For z1 = x1 + i y1, and z2 = x2 + i y2
Φ (z1z2) =Φ(
x1x2 − y1 y2 + i (x1 y2 + y1x2))
=
(
x1x2 − y1y2 −(x1 y2 + y1x2)
(x1y2 + y1x2) x1x2 − y1 y2
)
=
(
x1 −y1
y1 x1
)(
x2 −y2
y2 x2
)
=
(
x2 −y2
y2 x2
)(
x1 −y1
y1 x1
)
(2) Φ (z1 + z2) =Φ (z1)+Φ (z2). This is obvious.
(3) Φ (1/z) =(
Φ (z))−1
.
If z = x + i y then
1
z=
x − i y
x2 + y2=
z
|z|2and
Φ
(
1
z
)
=1
x2 + y2
(
x y
−y x
)
=
(
x −y
y x
)−1
=(
Φ (z))−1
(4) Φ (z) =Φ (z) =(
Φ (z))T
, obvious.
(5) |z|2 = det(
Φ (z))
This isomorphism can be extended to apply to complex valued influ-
ence coefficient matrices, H ∈Cn×m , and the image is a 2n×2m matrix in
R2n×2m . For such a complex influence coefficient matrixΦ can be defined
in the following manner.
(8) Φ(H) =
(
ℜ(H) −ℑ(H)
ℑ(H) ℜ(H)
)
The real number valued balance Equation (??) can be completed by
transforming the imbalance vector, B, and the vibration response vector,
R. However, in these cases only one column is required.
B −→
(
ℜ(B)
ℑ(B)
)
and
R −→
(
ℜ(R)
ℑ(R)
)
Whereas the complex version of B ∈Cm×1 the real version is inR
2m×1, and
the real version of R ∈Cn×1 is in R
2n×1. In each case the second column is
not required for the matrix algebra to function properly.
2006 UVA ROTORDYNAMICS SHORT COURSE 13
For fixed rotational speeds and operating conditions a rotor system is
a linear time invariant system, LTI. The real valued influence coefficient
matrix for a LTI rotor system for either mechanical or magnetic open-
loop balancing has the following form.
(9) Φ(H) =
(
Hc −Hs
Hs Hc
)
The c and s subscripts relate to cosines and sines or the corresponding
in-phase and quadrature components of the influence coefficient matrix.
Because LTI rotor systems have this form of influence coefficient matrix,
they lie in the image of the isomorphism, Φ. Thus the real influence coef-
ficient matrix can be transformed to a complex valued one which at time
may be more convenient.
4.3. Generating Influence Coefficients. One applies trial weights to gen-
erate the influence coefficient matrix. Traditionally this has been done
by adding an individual trial weight to balance plane j and measuring
the response due to the balance weight. The same can be done for modal
style balance weights.
The influence coefficient at measurement location i can be calculated
from Equation 6 as follows.
Hi j =Ri −R0
B jusing complex arithmetic
Here Ri indicates measurement at location i from the response vector
of measurements with a single trial weight at balance plane j . R0 is the
reference measurement prior to adding B j , the balance weight in plane
j . One can define modal influence coefficients similarly when the trial
weight has a modal distribution, notably the so-called static and couple
combinations, which are frequently used to balance turbine generators.
More generally Equation 6 can be used to generate influence coeffi-
cients while treating each balance shot as a statistical trial. If one has at
least a linearly independent set of trial weights then the influence coeffi-
cients, H can be thought of as the best fit to Equation 10 that minimizes
the error vector, e, in some fashion.
(10) [R1R2 . . .Rn] = H [B1B2 . . .Bn]+[
R10R2
0 . . . Rn0
]
+e
In Equation 10, responses Ri (a vector of responses at the various mea-
surement points) occurs upon adding balance weights Bi , a vector of the
applied balance weights. Each Ri0 corresponds to the reference run prior
to adding the balance weights; since balance weights may be left in from
14 WILLIAM C. FOILES
run to run, or the balancing may take place on different rotors this nota-
tion is necessary.
If one minimizes the 2-norm of the residuals, e, this produces a re-
gression fit for H. Using the superscript + to denote a pseudo-inverse,
Equation 11 produces the proper estimate for the influence coefficient
matrix.
(11) H =(
[R1R2 . . .Rn]−[
R10R2
0 . . .Rn0
])
[B1B2 . . .Bn]+
Since most of today’s balancing uses computer programs Equation 11
would be easy to implement. If one balances several machines of the
same type or the same machine often, this equation provides a best fit for
the influence coefficients. Some have portrayed the calculation of influ-
ence coefficients as a calibration of the rotor system. However, at best
one can approximate the influence coefficients and not precisely know
them. Thus, using a best fit approach often has advantages.
The formulation in Equation 11 also provides a means to compute the
influence coefficients with various combinations of trial weights, instead
of having to use individual planes. Using individual planes can result in
high vibration; often modal sets of trial weights are used to avoid this.
Superposition. Because of linearity superposition principles can and should
be used. Uses of superposition include the following.
: On an individual balance plane The effect of all the balance weights
on a single balance plane is equivalent to a single balance weight
equal to the vector sum of the individual balance weights on that
plane or conversely achieve the required effect of a balance weight
by using two or more weights in the same balance plane.
: Among two or more balance planes, speeds, or conditions The ba-
sic equation for multiplane balancing results from superposition
or linear response along the rotor and at different speeds or con-
ditions. Conditions could include loadings or peaks of thermal
vectors, as in a turbine or generator.
: Balancing multiple forces in the rotor system
BALANCE WEIGHT CONSOLIDATION
Over the course of years (or perhaps shorter if the balancer does not
take care), an individual balance plane can substantially fill with balance
weights. Some of these weights may effectively cancel the effects of other
weights. When this occurs further balancing can be difficult or impossi-
ble.
2006 UVA ROTORDYNAMICS SHORT COURSE 15
To remedy this one cleans the balance plane by balance weight consol-
idation. Generally, this is easiest with the rotor out in a shop, and some-
times shops may consolidate balance weights. Balance weight consoli-
dation is not the same as stripping the weights in a shop and performing
a new balance. If the balance weights were placed either in situ or in
a high speed balance facility, stripping the weights and replacing them
based on a low speed balance can make life interesting. Balance weight
consolidation simply reduces the number of balance weights while pro-
ducing the same force on the individual balance plane.
To perform balance weight consolidation one establishes a phase ref-
erence system for the balance plane, measures the angular location of
each of the balance weights, and measures the balance weight magni-
tude, both mass and radial distance from the center. Detailed accurate
records must be maintained as with other maintenance work on the ro-
tor. The object is to add a balance weight or balance weights equal to the
vector sum of the original balance weight vectors just measured. It may
not be possible to add a single balance weight to equal the vector sum.
The next section on splitting balance weights describes how to find a
combination of balance weights that equal a desired balance weight.
5. BALANCE WEIGHT SPLITTING
This section is based on an article in Vibrations magazine published by
the Vibration Institute by William C. Foiles in 2005.
5.1. Introduction. Balance weight splitting is the process of installing
more than one balance weight in a balance plane whose sum equals or
approximates the required balance weight. A number of balance pro-
grams and auxiliary programs include this for the user today for a two
weight split.
The necessity of splitting balance weights at locations occurs frequently.
Specific instances requiring weight splitting include occasions:
(1) When the location required to place a balance weight do not exist,
the balance weights need splitting between two or more angular
locations to achieve the required effective balance weight.
(2) When the required balance weight is larger than can be applied at
a single location, or when heavy balance weights are not available
or should not be used in the balance plane.
(3) When the required location has damage or has a weight already
installed.
This section is not meant to be a complete tutorial on the basics of
weight splitting or balancing, nor does this paper cover any safety aspects
16 WILLIAM C. FOILES
necessary to work with rotating equipment when doing a balance job.
The methods shown are for the enjoyment of the readers.
One can not split weights 180 degrees apart to get a solution. Splitting
weights on one side of the desired weights will require a weight removal
on one of the weights; for practical considerations the split angles must
be on either side of the desired weight location and less than 180 degrees
apart.
5.1.1. Two Weight Splits. Sometimes today as in the past the weights are
split graphically. Figure 1 shows the geometry involved in a weight split.
Typically the graphing is done on polar paper.
(1) First one plots the desired balance weight vector, amplitude and
angle.
(2) Next lines at the desired angles to add the split weights are added
to the graph.
(3) Parallel lines to these lines at the desired angle, one-at-a-time, are
constructed through the ends of the desired balance weight vec-
tor to form the required split weight vectors (See Figure 3.). These
parallel lines can be commonly constructed using either a set of
parallels or two triangles.
(4) Usually, although not absolutely required, the parallelogram is com-
pleted by sliding the other weight add vector to the origin as in
Figure 4.
NOMENCLATURE FOR BALANCE WEIGHT SPLIT
~d Desired balance weight addition
α ∠~d
α1 Angle of first balance weight of the split
α2 Angle of second balance weight of the split−→W1 Balance weight vector at angle α1−→W2 Balance weight vector at angle α2
The purpose of the balance weight split is to find−→W1 and
−→W2 such that
−→d =
−→W1 +
−→W2
This two weight add problem has a simple analytical solution using
the. The Law of Sines formula applied to the geometry in Figure 4 gives
the following derivation.
(12)|−→W2|
sin(|α−α1|)=
|−→W1|
sin(|α−α2|)=
|−→d |
sin(180−|α2 −α1|)
2006 UVA ROTORDYNAMICS SHORT COURSE 17
34
5330
315
300285270255
240
225
210
19
51
80
16
515
0
135
120105 90 75
60
4530
15
0
0
5
10
15
0 5 10 15
CW
α1
α 2
~d
FIGURE 3. Balance weight split – Desired balance weight
and angles for the split
The absolute value sign takes care of odd cases. The above equation
leads directly to a solution for the weight splits at the desired angles.
(13) |−→W1| = |
−→d |
sin(|α2 −α|)
sin(180−|α2 −α1|)= |
−→d |
sin(|α2 −α|)
sin(|α2 −α1|)
And
(14) |−→W2| = |
−→d |
sin(|α1 −α|)
sin(180−|α2 −α1|)= |
−→d |
sin(|α1 −α|)
sin(|α2 −α1|)
The was the simple way to get to the formula for weight splitting into
two angles. One can break the equations up into the Cartesian compo-
nents, x and y, and use algebra and trigonometry to arrive at the above
18 WILLIAM C. FOILES
34
5330
315
300285270255
240
225
210
19
51
80
16
515
0
135
120105 90 75
60
4530
15
0
0
5
10
15
0 5 10 15
CW
α1
α 2
~d
−→W1
−→ W2
FIGURE 4. Balance weight split – Geometric Solution
equations or something equivalent. I have not intention of duplicating
the easy solution by the more difficult solution here, but with a little pa-
tience or one of the symbolic computation engines available today one
can get there.
Example for split in two locations. Following the example in Figure 3,
let ~d = (13∡55). This weight is to be split at the two angles α1 = 20
and α2 = 75. Figure 4 shows the graphical solution; below the analytic
solution is followed as above.
2006 UVA ROTORDYNAMICS SHORT COURSE 19
|−→W1| = 13
sin(|75−55|)
sin(|75−20|)= 5.428
|−→W2| = 13
sin(|55−20|)
sin(|75−20|)= 9.103
5.428 (with proper units, which may include the radius) is placed at
20, and 9.103 is placed at 75. One can check that the vectorial sum of
these weights is as desired.
5.1.2. Spitting weights at n angles. For this section on splitting a balance
weight into several locations, the Cartesian coordinate formulation of
the two weight problem will help, the algebra and trigonometry to show
equivalence with the above Law of Sines formulation is left for the curious
reader.
The two equations for the x and y coordinates from Figure 4 result in
Equation 15.
(15)|−→W1|cos(α1)+|
−→W2|cos(α2) = |
−→d |cos(α)
|−→W1|sin(α1)+|
−→W2|sin(α2) = |
−→d |sin(α)
These equations in the two unknowns for the desired split magnitudes,
|−→W1| and |
−→W2| can be solved as usual. One can check that the two weight
split example above solves the Equation 15 above.
The above framework for writing the Cartesian coordinates generalizes
to splitting at more than two angles.
NOMENCLATURE FOR BALANCE WEIGHT SPLIT AT n ANGLES
~d Desired balance weight addition
α ∠~d
α2 Angle of second balance weight of the split−→Wi Balance weight vector at angle αi
Wi |−→Wi | for simplicity, i = 1, . . .n
Li Lower bounds on balance weights, i = 1, . . .n
Ui Upper bounds on balance weights,i = 1, . . .nFor splitting among more than 2 angles, one has the following basic
equations.
20 WILLIAM C. FOILES
(16)
n∑
i=1
|−→Wi |cos(αi ) = |
−→d |cos(α)
n∑
i=1
|−→Wi |sin(αi ) = |
−→d |sin(α)
The equations above, Equation 16, in general have an infinite number
of solutions. Again, the split angles should be on either side of the de-
sired angle, and generally the angles should spread less than 180 degrees.
Without something else, this doesn’t help much.
Adding a goal or optimization objective,such as Equation 17 will lead to
finding solutions. One practical solution is to minimize the total amount
of added weight as below.
(17) minn∑
i=1
Wi
To the above assume all the weights are additions or positive magni-
tudes. The weight removal procedure can be processed by adding weights
at the opposite angles. These equations form a linear programming prob-
lem that can be solved.
On top of the equalities, one may desire to limit the magnitudes of the
weights that can be added. Such extra constraints would give the fol-
lowing equations, Equation 18. Without some constraints the solution,
which minimizes the sum of the weights, only uses the two positions clos-
est to the desired balance weight position.
(18) 0 ≥ Li ≥Wi ≤Ui , i = 1, . . .n
The following linear program incorporates these aspects and will split
a desired balance weight at n desired angles while minimizing the sum
of the split weights, provided a feasible solution exists. Weight bounds,
2006 UVA ROTORDYNAMICS SHORT COURSE 21
both upper and lower, may be applied to the split weights.
minn∑
i=1
Wi such that
n∑
i=1
|−→Wi |cos(αi ) = |
−→d |cos(α)
n∑
i=1
|−→Wi |sin(αi ) = |
−→d |sin(α)
with
0≥ Li ≥Wi ≤Ui , i = 1, . . .n
(19)
The unknowns in the Linear Program 19 are the balance weights, Wi .
A sample routine written for Scilab©
was included in the original Vibra-
tions article.
Examples of multi-angle weight splits. The following example use the
sample Scilab program. The desired balance weight is 450∡100, and
the angles for the split are 75, 90, 105, and 120. Upper bounds were
placed of 250, 100, 100, and 250 were placed respectively on the angular
locations with lower bounds of 0 for each angle. The solution that mini-
mizes the total weight addition is to add 110.35∡75, 100∡90, 100∡105,
and 161.64∡120,
-->bwt=[450 100];
-->angles=[75;90;105;120];
-->lower=[0;0;0;0];
-->upper=[250;100;100;250];
-->split(bwt,angles,lower,upper);
! 110.34707 !
! 100. !
! 100. !
! 161.6394 !
The above example would not have been too difficult to solve manu-
ally, because one always tries to add the most weight closest to the de-
sired angle as possible.
Copyright ©1989-2005. INRIA ENPC. and Scilab is a trademark of INRIA,
www.scilab.org
22 WILLIAM C. FOILES
A more difficult example with both upper and lower bounds follows.
-->ang=[90;120;150;180];
-->lower=[100;100;100;100];
-->upper=[500;250;500;500];
-->split(balwt,ang,lower,upper);
! 145.01279 !
! 250. !
! 372.97723 !
! 100. !
E-mail address: [email protected]