A3 Vibration Vectors.

qxd 9/10/2002 10:27 AM Page 33

33

Chapter 3

Vibration Vectors

I    ,   vibration signals and the

measurement of phase. Most of the discussion involved the measurement of
vibration signals that are single-frequency sine waves.
In order to make a meaningful phase measurement, a vibration signal must
contain only one (or predominantly one) frequency. However, since typical
machinery vibration signals contain several frequencies, the signal must first be
filtered to a single frequency. Measurement of the amplitude and phase of the fil-
tered signal produces a filtered response, response vector, or vibration vector.
This vector is a powerful tool that provides the foundation for the detection of
many different machine malfunctions and is vital information for balancing.
The vibration vector is the underlying concept for all the Bode, polar, and ampli-
tude-phase-time (APHT) data plots.
We will start with a discussion of the characteristics of unfiltered vibration,
followed by a discussion of the vibration vector. Finally, we will discuss an
important special case of the vibration vector called the slow roll vector.

Unfiltered Vibration
The raw (unfiltered) vibration signal from a transducer is sometimes called
the direct vibration signal; theoretically, there is nothing in the signal path
between the transducer and the instrumentation. In practice, some modifica-
tion of the signal may occur (for example, an output from a monitor that
includes signal processing), but if the circuitry provides no signal processing
besides buffering (something that should always be verified), the output is
assumed to be an exact copy of the original, unfiltered signal, including any dc
offset.
A3 Vibration Vectors.qxd 9/10/2002 10:27 AM Page 34

34 Fundamentals of Vibration

Unfiltered

220˚

Figure 3-1. Unfiltered vibration signal

with its 1X- and 2X-filtered vibration
components. Accurate phase measure- 1X
ment requires that the signal be filtered
to the desired frequency.

T1X

190˚ T2X

2X

Thus, in general, the word unfiltered implies that no modification of the sig-
nal has taken place in the instrumentation, and that it contains all of the fre-
quency components (with amplitude and phase intact) that exist in the incom-
ing transducer signal. Figure 3-1 shows an unfiltered vibration signal and its 1X-
and 2X-filtered frequency components. A number of other frequency compo-
nents are contained in the unfiltered signal.
The amplitude of an unfiltered signal can be accurately measured in peak-
to-peak or peak units (see Chapter 1); however, unless the unfiltered vibration
signal is dominated by a single frequency, it is not possible to measure phase
relationships accurately. Phase measurement requires a signal with a single fre-
quency.
Also, since particular rotor behaviors or malfunctions may be associated
with a specific frequency (for example, rotor unbalance), filtering of the vibra-
tion signal is normally required.

Filtering and the Vibration Vector

Filtering is a signal processing technique that, ideally, rejects all frequencies
that are outside the bandpass region of the filter. The filter used most often on
A3 Vibration Vectors.qxd 9/10/2002 10:27 AM Page 35

Chapter 3 Vibration Vectors 35

machinery vibration signals is the bandpass filter, which removes all signal con-
tent that is above and below the center (bandpass) frequency of the filter. The
center frequency is usually set to either running speed (1X) or a multiple of run-
ning speed if a significant amount of machine vibration occurs at those fre-
quencies (for example, a 5-vane pump impeller would produce five vane-pass
events per revolution, so 5X-filtering might be desirable). Because the rotor
speed changes, some filters automatically adjust the bandpass frequency to
track running speed. Such a filter is called a tracking filter and is commonly used
in rotating machinery applications.
After filtering, the vibration signal is close to a pure sine wave at the band-
pass frequency, and the amplitude and phase of the filtered signal can be meas-
ured using the techniques discussed in the previous chapters.
The amplitude and phase of the filtered signal describe a vibration vector,
which is plotted in the transducer response (UV) plane (Figure 3-2). A vector is
a mathematical object that has both magnitude and direction. The magnitude
of the vibration vector corresponds to the vibration amplitude (in whatever
units are convenient, but usually µm pp or mil pp for shaft relative vibration).
The direction of the vector corresponds to the absolute phase of the filtered
vibration signal.

270˚
Transducer
tn

response
Ro

plane 0˚
+U 220˚

p
90 µm p
90 µm pp

220˚
180˚ phase lag
+V

90˚ Vibration signal

Figure 3-2. The vibration vector. The vector contains the amplitude and absolute
phase information from the filtered vibration signal. It is plotted in the transducer
response plane, where the U axis is aligned with the measurement transducer. The
length of the vector represents the amplitude, and the angular position of the vector
represents the absolute phase of the signal. The angular position is measured from
the U axis in the direction opposite the direction of rotation of the shaft.
A3 Vibration Vectors.qxd 9/10/2002 10:27 AM Page 36

36 Fundamentals of Vibration

The U axis of the plane is aligned with the measurement axis of the trans-
ducer. In the figure, the transducer is mounted 30° from the horizontal. The V
axis is always 90° from the U axis, in the direction opposite of shaft rotation. It
is important to note that the UV axes are independent of any other machine
coordinate system and are associated with each transducer. Each transducer has
its own transducer response plane, and the measurement axis of the transducer
is always aligned with the U axis of its response plane.
The length of the vibration vector is equal to the amplitude of the filtered
vibration. The angle of the response vector relative to the U axis is the absolute
phase lag, measured from the measurement axis in a direction opposite to the
direction of rotation of the machine rotor (Figure 3-3). Thus, depending on the
direction of rotation, the vector can plot in different places. Note that the posi-
tive V axis is always located at 90°, measured opposite the direction of rotation.
To plot a vibration vector, follow these steps:

1) Determine the angular orientation of the transducer relative to

your machine viewpoint. This will define the direction of the U
(measurement) axis.

2) Determine the direction of rotation of the rotor. The positive V

axis will be located 90° from the U axis in a direction opposite to
rotor rotation.

3) Use the absolute phase of the filtered vibration signal to locate

the angular orientation of the vibration vector. Measure the
angle from the U axis, opposite the direction of rotation, toward
the direction of the V axis.

4) The length of the vibration vector is the amplitude of the fil-

tered vibration signal. Typically, displacement units are µm pp
or mil pp, velocity units are mm/s pk or in/s pk, and accelera-
tion units are g pk.

Because the vibration vectors define the response of the machinery to a

variety of factors, it is critical to document this data under a variety of operat-
ing conditions. On critical machinery, where transducers are installed at many
locations, the vibration data from each transducer should be recorded over the
entire operating speed range during startup and shutdown. 1X and 2X vectors
are most commonly measured, but other frequency components should be
A3 Vibration Vectors.qxd 9/10/2002 10:27 AM Page 37

Chapter 3 Vibration Vectors 37

measured if there is a forcing function (such as blade passage) that is at a har-

monic of running speed.
Vibration vectors are also monitored while a machine is running at a con-
stant speed. Changes in operating and load conditions can produce predictable
changes in response vectors, but significant changes outside this envelope could
indicate a change in the machine’s health. Unexpected changes in vibration vec-
tors are important for the early detection of machine internal problems, such as
unbalance, rub, instabilities, and shaft cracks, and external problems, such as
coupling failure, piping strain, and foundation deterioration.
The tip of the vibration vector defines a point in the transducer response
plane. A plot of a set of these points corresponding to different machine condi-
tions provides a powerful visual display of the response of the machine at that
transducer location, whether the machine is starting up, at operating speed, or
coasting down. The plot of a set of startup or shutdown vibration vector points
is equivalent to a polar plot (See Chapter 7), one of the most informative plots
available for diagnosing machinery condition. A set of such points at a steady
operating speed (steady state) produces an APHT plot, and vibration vectors are
monitored during machine operation in acceptance regions in these plots.

Rotn Rotn
270˚ 90˚
+V

50˚
p
20 µm p

0˚ phase lag +U 0˚
+U

50˚
phase lag
20 µ
mp
p

180˚ 180˚
+V

90˚ 270˚

Figure 3-3. V axis orientation versus direction of shaft rotation. The +V axis is located 90° from
the +U axis, against the direction of rotation of the shaft. Phase lag is always measured from U to
V. The same vibration vector will plot differently, depending on the direction of rotation.
A3 Vibration Vectors.qxd 9/10/2002 10:27 AM Page 38

38 Fundamentals of Vibration

Working with Vibration Vectors

When working with vibration vectors, it is important to use a system of
notation that is both convenient and complete. Commonly, vibration vectors are
noted as

r = A ∠Φ (3-1)

where r is the displacement vibration vector, A is the vibration amplitude, ∠

identifies the value that follows as an angle, and Φ (Greek upper case phi) is the
phase angle, usually expressed as positive phase lag. This is a polar representa-
tion of the vector. An equivalent rectangular representation is given by

r = u + jv (3-2)

where j is the square root of −1, and

u = A cos Φ
v = A sin Φ (3-3)

The variables u and v are the rectangular coordinates of the vector in the
transducer response plane. The u coordinate is measured along the U axis (the
transducer measurement axis), and the v coordinate is measured along the V
axis. (As we have already noted, U and V are not the same as the physical XY
coordinate system used to describe the machine.) Vibration vectors are actually
complex numbers, and the j term originates in complex number theory.
Conversion from rectangular form to polar form is performed using these
expressions:

A = u2 + v 2
v  (3-4)
Φ = arctan 2  
u 

where arctan2 represents the arctangent2 function, which takes quadrants into
account.
For example, the vibration vector in Figure 3-2 can be expressed in polar
form as

r = 90 µm pp ∠220˚ (3.5 mil pp ∠220˚)

A3 Vibration Vectors.qxd 9/10/2002 10:27 AM Page 39

Chapter 3 Vibration Vectors 39

and in rectangular form as

r = −69 µm pp − j 58 µm pp (−2.7 mil pp − j 2.3 mil pp)

However, conversion in the opposite direction can lead to difficulty. If a simple

arctangent function is used to calculate the phase lag angle, the result is

v   −58 
Φ = arctan   = arctan   = 40˚
u   −69 

which is incorrect. The situation is shown in Figure 3-4. Adding 180° produces
the correct result: 220°. When using the standard arctangent function, it is a
good idea to sketch the situation, carefully noting the signs of the coordinates,
and verify the result. Engineering and scientific calculators use the arctangent2
function when converting from rectangular to polar coordinates, but they pro-
duce a result between ±180°. When the calculated result is negative, add 360° to
produce a positive phase lag between 0° and 360°.

Rotn Rotn
270˚ 90˚
+V

0˚
+U 220˚ +U 0˚
9
–6 phase lag
90 µm
pp
40˚
8
–5 40˚
90 µ

220˚
mp

phase lag
180˚
p

180˚ 8
–5
+V

9
–6
90˚ 270˚

Figure 3-4. The arctangent function and rectangular to polar conversion. The same vector is plot-
ted for different rotation directions, and the rectangular coordinates are shown. Use of the stan-
dard arctangent function yields a phase lag angle of 40°, which is incorrect; the true phase lag
angle is 220°. When using the arctangent function, it is a good idea to sketch the situation to veri-
fy the calculated angle.
A3 Vibration Vectors.qxd 9/10/2002 10:27 AM Page 40

40 Fundamentals of Vibration

Vibration vectors often need to be added, subtracted, multiplied, and divid-

ed. Addition and subtraction of two vibration vectors can be done graphically, or
the result can be easily calculated using the rectangular form. Multiplication
and division are calculated most easily using the polar form.
Addition is done by adding the u components and v components separately
and then combining the results to define a new vector. The units of measure-
ment of the two vectors must be the same or the result will be meaningless.
Two vibration vectors, r1 and r2, can be added graphically to produce a
resultant vector, r3 (see Figure 3-5). To do this,

1) Plot r1 in the transducer response plane with its tail at the ori-
gin.

2) Plot r2 in with its tail at the origin.

3) Slide r2 over (copy it) so that its tail is at the tip of r1.

4) Draw a new vector from the origin to the tip of the copy of r2.
This vector is the resultant, r3.

Rotn

90˚ 0˚
+V

+U

r2
Figure 3-5. Graphic vector addition. To add r1
and r2, plot the two vectors in the transducer
r3
response plane. Copy r2 so that the tail of r2 is r2 r1
at the tip of r1. The resultant, r3, is the vector
from the origin to the tip of r2.

180˚ 270˚
A3 Vibration Vectors.qxd 9/10/2002 10:27 AM Page 41

Chapter 3 Vibration Vectors 41

The graphical technique is the mathematical equivalent of adding the two

vectors’ u and v components:

r3 = r1 + r2 = (u1 + u2 )+ j (v1 + v2 ) (3-5)

Example
A gas turbine rotates in a Y to X direction at 7450 rpm. Data is taken from a cas-
ing velocity transducer (which provides absolute casing motion) and a shaft rel-
ative displacement transducer. Both transducers are mounted at 45° R. The 1X,
integrated, casing vibration, rc , is found to be 40 µm pp ∠35° (1.6 mil pp ∠35°).
The 1X, shaft relative vibration, rsr , is measured as 30 µm pp ∠120° (1.2 mil pp
∠120°). Find the 1X, shaft absolute vibration vector, rs .

Solution
The shaft absolute vibration vector is found by adding the u and v components
of the casing vibration and shaft relative vibration vectors (Figure 3-6). The
transducer response plane is shown with the U axis aligned with the transduc-
er’s measurement axis. Because rotation is Y to X, the positive V axis is located
90° counterclockwise from the U axis, and the phase angle is measured in that
direction.

rc = 40 ∠35˚ rs = 52 ∠70˚
rsr = 30 ∠120˚

90˚ (33, 23) 90˚

rsr
0˚ 0˚
+U
+V
+U
+V

23

R ot

R ot

rc rs rc
n

(–15, 26)
33

(33–15, 23+26)
rsr
26

5
–1

180˚ 270˚ 180˚ 270˚

60 µm pp full scale 60 µm pp full scale

Figure 3-6. Addition of vibration vectors to find the shaft absolute vibration. Graphically add the
casing absolute vibration vector, rc , to the shaft relative vibration vector, rsr , or add the u and v
components of each vector.
A3 Vibration Vectors.qxd 9/10/2002 10:27 AM Page 42

42 Fundamentals of Vibration

1) Convert both measurements to rectangular form, using Equation

3-3:

rc =(40 µm pp)cos 35˚+ j (40 µm pp)sin 35˚

rc = 33 µm pp + j 23 µm pp

rsr =(30 µm pp)cos 120˚ + j (30 µm pp)sin 120˚

rsr =−15 µm pp + j 26 µm pp

2) Add the components to get the solution in rectangular form

using Equation 3-5:

rs = (33 − 15) µm pp + j (23 + 26) µm pp

rs = 18 µm pp + j 49 µm pp

3) Convert to polar form using Equation 3-4:

2 2
A = (18 µm pp) +(49 µm pp)
 49 µm pp 
Φ = arctan 2  
 18 µm pp 

rs = 52 µm pp ∠70˚ (2.0 mil pp ∠70˚)

Subtraction is done by subtracting the u components and v components

separately and then combining the results to define a new vector. Graphical sub-
traction is performed by adding the negative of one vector to the other. Again,
the units of measurement of the two vectors must be the same or the result will
be meaningless.
Two vibration vectors, r1, and r2, can be subtracted graphically to produce a
resultant vector, r3 (see Figure 3-7). To find r3 = r1 − r2,

1) Plot r1 in the transducer response plane with its tail at the ori-
gin.

2) Find the negative of r2 by changing the phase angle by 180°. This

is also equivalent to multiplying r2 by −1.
A3 Vibration Vectors.qxd 9/10/2002 10:27 AM Page 43

Chapter 3 Vibration Vectors 43

3) Plot −r2 with its tail at the origin.

4) Copy −r2 so that its tail is at the tip of r1.

5) Draw a new vector from the origin to the tip of the copy of −r2.
This vector is the resultant, r3.

The graphical technique is the mathematical equivalent of subtracting the

two vectors’ u and v components:

r3 = r1 − r2 = (u1 − u2 )+ j (v1 − v2 ) (3-6)

Multiplication of two vectors is performed most easily using the polar for-
mat; multiply the amplitudes and add the phase angles:

r3 = (A1 ∠Φ1 )(A2 ∠Φ2 )=(A1 A2 ) ∠(Φ1 + Φ2 ) (3-7)

Division is performed by dividing the amplitudes and subtracting the phase

angles:

A1 ∠Φ1  A1 
r3 = =   ∠(Φ1 − Φ2 ) (3-8)
A2 ∠Φ2  A2 

See the Appendix for examples of multiplication and division of vibration

vectors.

R o tn

90˚ 0˚
+U
+V

Figure 3-7. Graphic vector subtraction. To sub-

tract r2 from r1, plot r1 and r2 in the transduc-
–r2
er response plane. Plot the negative of r2 and
copy it so that the tail of −r2 is at the tip of r1. r3
The resultant, r3, is the vector from the origin r2
to the tip of −r2. r1
–r2

180˚ 270˚
A3 Vibration Vectors.qxd 9/10/2002 10:27 AM Page 44

44 Fundamentals of Vibration

NOTE: For all vector operations, the phase lag angle should be expressed as
a positive number between 0° and 360°. If the calculated phase lag is negative,
add 360°. If the result is greater than 360°, subtract 360°.
Most scientific and engineering calculators can operate directly on complex
numbers and don’t require conversion between polar and rectangular forms.
This vector concept extends to more than just vibration measurement. The
force due to unbalance is a rotating force vector that has a particular angular
position when the Keyphasor event occurs. And Dynamic Stiffness, a very
important concept in machinery behavior, is also expressed as a complex num-
ber, usually in rectangular form. These two vector entities, together with the
vibration vector, are fundamental to understanding the dynamic behavior of
machinery.

The Slow Roll Vector

The slow roll vector is an important application of vector subtraction. The
slow roll vector is a constant, or slowly varying, component of the vibration vec-
tor that represents nondynamic action observed by a transducer. The slow roll
vector will be different for each measurement transducer location. It originates
in mechanical effects, such as a bowed rotor or coupling problem, or in mechan-
ical or electrical runout, and it can distort and obscure the machine’s dynamic
response data (Figure 3-8). Slow roll vector compensation is the technique of sub-
tracting the measured slow roll vector from the transducer vibration vector
(Figure 3-9).
To measure the slow roll vector, we must be able to find an operating condi-
tion where the slow roll is the dominant component of the vibration signal.
Since the 1X dynamic response due to unbalance tends to zero at low speeds,
any 1X vibration measured at these low speeds is considered to be due to
sources other than unbalance. Thus, slow roll vectors are measured in this speed
range, which is called the slow roll speed range. One guideline (and it is only a
guideline) is that the upper limit of the slow roll speed range is about 10% of the
first balance resonance speed of the machine. The slow roll speed range is best
identified using an uncompensated Bode plot (see Chapter 7) of a machine
startup or shutdown.
While noise affects all transducers, slow roll data is usually obtained only
from displacement transducers. Eddy current displacement transducers have a
frequency response that extends to dc (zero speed), while velocity and accelera-
tion transducers do not. Because of this, velocity and acceleration slow roll
response generally cannot be measured, and vibration data from these trans-
ducers is not compensated.
A3 Vibration Vectors.qxd 9/10/2002 10:27 AM Page 45

Chapter 3 Vibration Vectors 45

0˚
1X unbalance
n +U
R ot response

Measured
Figure 3-8. The 1X slow roll vector. The 1X 1X vector
slow roll vector (red) adds to the 1X response l
rol
vector due to unbalance (green). This can w
slo
produce a vibration vector that is significantly 90˚
+V 1X 270˚
different (blue) than the unbalance response
vector. Slow roll vectors can be measured for
any harmonic of running speed.

3000 rpm
180˚

0˚
n +U
ot on
R cti
tra
sub
1X

oll
wr
red

Figure 3-9. Slow roll vector compensation. To Slo l

rol
asu

compensate for slow roll, subtract the slow s low

Me

1 X
roll vector from the vibration vector meas- +V
ured by the same transducer. The resultant is 90˚ 270˚
the slow-roll-compensated, dynamic 1X response is
response vector (green). now unbalance
response

3000 rpm
180˚
A3 Vibration Vectors.qxd 9/10/2002 10:27 AM Page 46

46 Fundamentals of Vibration

Slow Roll Vector Changes

The basic underlying assumption when
1 Cold Runout
using slow roll compensation is that, at running
speed, the shaft probes view the same section of
rotor surface that they view at slow roll.
However, at each axial position on a shaft, the 1 2
shaft surface is likely to produce a different slow
roll vector. Cold
Thermal growth can be one source of slow
roll vector change. During a cold startup, rotors
can grow axially relative to the machine casing;
2 Hot Runout
thus, probes may view a different section of
shaft surface with a different slow roll vector
(Figure 3-10). Because of this, all compensated
1 2 Thermal Growth
transient startup data should be used with cau-
tion. Data from a hot shutdown is much less
likely to have this problem. Hot
Even after the machine reaches thermal
equilibrium at running speed, slow roll data Figure 3-10. Rotor thermal growth and slow
may be invalid. Machines with high operating roll runout measurements. When the rotor is
thermal gradients, such as gas turbines or cold (blue) the probe views probe track 1.
refrigeration compressors, can have different One revolution produces the slow roll wave-
rates of axial rotor and casing growth. In this form shown next to the probe. When the
situation, cold slow roll vectors are question- rotor is hot (red), it grows axially, and the
able, and hot shutdown slow roll vectors are probe now views track 2, which has a differ-
ent slow roll runout. The cold slow roll meas-
likely to be more reliable.
urement is not valid for the hot condition.
In general, it is better to compensate cold
data with cold slow roll vectors and to compen-
sate hot data with hot slow roll vectors.
Even when probes view the same section of
shaft surface, slow roll vectors can change over
a long period of time due to changes in rotor
bow, changes in electrical runout, corrosion of
the viewing surface, or developing machine
problems. Changing slow roll vectors are often a
sign that something significant is happening to
the machine and should be investigated.
A3 Vibration Vectors.qxd 9/10/2002 10:27 AM Page 47

Chapter 3 Vibration Vectors 47

Summary
An unfiltered, or direct, vibration signal is unchanged from the original
transducer vibration signal. It is assumed to contain all of the original frequen-
cy, amplitude, and phase content and the original dc offset, if any.
Filtering removes signal content. Many machinery vibration signals are
bandpass filtered to a multiple of running speed, most often 1X. The filtered sig-
nal is a sine wave with a frequency equal to the bandpass frequency of the filter.
After filtering, the amplitude and absolute phase of the signal can be measured.
A vibration vector is the combination of the amplitude and absolute phase
of a filtered vibration signal. This vector is plotted in the transducer response
plane. Because vibration vectors are complex numbers, they can be added, sub-
tracted, multiplied, and divided.
The slow roll speed range of a machine is the range of speeds where the
dynamic rotor response due to unbalance is insignificant compared to the slow
roll vector; roughly, it is below 10% of the first balance resonance speed of a
machine. The slow roll speed range is best identified using an uncompensated
Bode plot.
Slow roll compensation is the subtraction of the slow roll vector from a
vibration vector at the same measurement location. The resultant vibration vec-
tor will only reflect the dynamic response of the rotor.
A3 Vibration Vectors.qxd 9/10/2002 10:27 AM Page 48