Voting Thrust Measurements with Other Parameters

Authored by:
Steve Sabin before engaging a machine shutdown sequence. Unlike
other types of machinery measurements, where the question
Corporate Marketing Manager
Bently Nevada Corporation “how much is too much?” can be subjective, the decision of
e-mail: steve.sabin@bently.com how much axial movement is permissible is
straightforward.
For all of these reasons, Bently Nevada recommends
that customers continuously measure axial position and tie
the Danger alarms from such a measurement to automatic

A
lthough the practice of voting thrust position alarms machine shutdown. In some cases, certain processes can be
with other parameters, such as bearing temperature, dangerous or unstable if a machine is shut down suddenly,
is sometimes seen in the field, Bently Nevada does and thrust alarms are more appropriately tied to a sequencer
not recommend it. Proponents of voting between dissimilar that brings the machine down slowly (if possible), rather
measurements generally do so because they believe it helps than an instantaneous trip. However, the initiation of a
guard against false alarms (after all, two independent machine shutdown due to excessive thrust movement can
measurements can’t both be wrong – can they?). However, (and should) be made instantaneously, regardless of how
there are numerous pitfalls to such an approach, which we’ll long it takes to bring the process and machine to a halt.
review below.
Dual Voting
Thrust Position Protection – A Review The nature of a thrust position measurement is that it
Thrust position is an extremely critical measurement observes the gap between the probe tip and the observed
in rotating machinery. When a thrust bearing fails, axial surface, usually the end of a shaft or a thrust collar (Figure 1).
movement of the shaft is no longer constrained and must Because the OK circuitry of a machinery protection system
instead be absorbed by some other part of the machine. uses gap voltage as a measure of the transducer’s integrity,
Sometimes this is a seal, but seals are not designed to a NOT OK condition (gap voltage out of range) in a thrust
support load and they quickly fail under such conditions. position measurement may simply indicate that thrust
More often, the uncontrolled axial movement will simply position has moved so far that the gap is out of the
allow rotating and non-rotating elements to come in contact, transducer’s acceptable range. There is not a problem with
such as blades contacting static elements. The results are the transducer – there is a problem with a machine. This is
catastrophic to the machine and the process, and can be a rarely a problem with radial vibration measurements, as
serious safety risk to personnel. There are documented radial motion is constrained to much smaller distances than
cases of thrust bearing failures leading to fatalities. with axial measurements, and a radial gap will never exceed
the OK limits unless there is a bona-fide transducer
Thrust position can also change very rapidly. Although
problem. Because a thrust position failure can exhibit the
it is not a good design, some machines employ thrust collars
same characteristics as a transducer failure (excessive gap
that are not an integral part of the shaft. If the thrust collar
voltage), a scheme whereby false trips and missed trips can
comes loose, thrust position can change almost
be simultaneously minimized must be used.
instantaneously, allowing operators little time to react and
engage a manual machine shutdown. Finally, because thrust The practice of dual-voting thrust measurements is
position is a non-subjective measurement (alarm limits are widely employed – and is globally recognized as minimum
directly related to axial clearances in the machine and/or acceptable practice for thrust position protection. Two thrust
allowable thrust bearing wear), there is little point in relying probes are installed, each observing the same motion of the
upon interpretation and manual review of thrust alarms shaft or collar. When operating correctly, each probe is

1Q01 | ORBIT | 51
Figure 1.

observing the same motion, and each should display the DANGER relay (or digital output) takes place under the
same movement. By employing AND voting (both probe A conditions shown in Figure 2.
and probe B must exceed their DANGER alarm levels at the Indeed, under condition B where both transducers
same time), the objective of minimizing false trips and have failed (or the axial position has legitimately moved so
missed trips can be realized. With only a single probe, the far that the probes are both outside their linear range), most
ability to distinguish between a transducer failure and machinery operators want to remove the machine from
genuine excessive thrust movement is lost. However, when service anyway, since they are operating without protection
two probes are installed, the likelihood that both transducers or even indication of thrust position. Conversely, if only one
will fail simultaneously is low and activation of the transducer indicates excessive axial movement (gap), and

52 | ORBIT | 1Q01
 actually drop. Therefore, an axial
alarm may occur but not a
temperature alarm. Thus, axial
position and temperature may
not always jointly indicate a
bearing failure.
Such a condition occurred
several years ago on a large
compressor train in a North
American gas processing plant,
and the subsequent failure
resulted in a fatality. The same
plant later destroyed the rebuilt
machine because the operators
ignored the high thrust position
alarms when thrust bearing
temperature once again did not
Figure 2. indicate any abnormalities.
Fortunately, in this second
occurrence, the damage involved
the other does not, it can be reasoned that the transducer is only the machine – not human life. Subsequently,
faulty, since both should always read the same (both are the Bently Nevada thrust monitors were finally
observing the same relative motion). connected to automatic shutdown without any
Proponents of 2-out-of-3 voting use mathematical intervening temperature voting.
probability arguments to show that it provides the optimal 2. Axial rub:
balance of false trips and missed trip protection. However, Sometimes, a severe thrust bearing failure will
it is rarely practical to install three axial transducers in a allow an axial rub to occur. This axial rub
single thrust bearing due to space constraints. Also, 2-out- effectively unloads the thrust bearing by allowing
of-3 voting schemes don’t typically take into account the the rubbing parts to absorb some of the thrust load.
considerations of transducer self-checking capability (NOT In this case, the thrust bearing pads will actually
OK detection). This raises the integrity of a 2-out-of-2 decrease in temperature as the bearing failure
(AND voting) system relative to a system with “dumb” progresses, and will not always go through a high
enough temperature to initiate any temperature
(non- self-checking) inputs. Thus, simple 2-out-of-2 (AND)
alarms.
voting has become the industry norm for thrust position
measurements. 3. Bearing failure:
Failures can occur in a thrust bearing where the
Voting with Other Parameters pads or bearing components deteriorate due to
While the dual-voting convention for thrust electrostatic discharge (see page 23), metallurgical
measurements provides a very high-integrity approach to issues, or other reasons. These do not necessarily
result in a failure mechanism where a pad becomes
both false trips and missed trips, as mentioned previously,
more heavily loaded and its temperature increases.
some still prefer to vote their thrust position alarms with
Axial position is a much more reliable indicator of
bearing temperature or other measurements.
thrust bearing problems.
The following scenarios help to illustrate why this 4. Pad temperature gradients:
is inadvisable:
Depending on where the temperature sensors are
1. Loose or broken thrust collar: embedded in the thrust bearing pads, large
When a thrust collar becomes loose, the shaft is free differences in temperature can occur over the pad's
to move independently of the thrust collar. If geometry. For example, a pad will generally
proximity probes are observing the end of the shaft, experience the highest temperatures at the edge of
they will correctly show the axial movement. the pad where most of the load is carried.
However, the thrust collar will no longer be Temperature variations of 40 or 50 degrees F are not
transferring axial load to the thrust bearing and the uncommon across a pad. Thus, a pad could
bearing temperature may remain the same or experience significant heating during a failure, but

1Q01 | ORBIT | 53
 the location of the
temperature sensing
element may not always
capture the true
temperature at the place
where Babbitt is
breaking down or load
is most concentrated.
This means that Babbitt
could be melting at one
edge of a pad, yet
temperatures could be
much lower – even
normal – at the other
edge of a pad.
Depending on where
the temperature sensors
were located, they may
not generate any Figure 3.
alarms.

5. Temperature Response: shutdown protection is advisable. The customer

Bearing metal has a thermal inertia that generally noted above has since connected all thrust alarms
introduces a delay between the onset of bearing directly to shutdown, without any intervening
failure and a corresponding increase in bearing voting logic using temperature or other parameters.
metal (Babbitt) temperature. Even though the
temperature response may take just a few seconds, Economic Considerations
this is still much longer than the response time for a While the reasons for shutting down as quickly as
position measurement made with proximity probes. possible on excessive thrust movement are firmly rooted in
One of many examples encountered over the years safety concerns, there are also economic considerations.
comes from a recent customer report. While the Other recent customer reports have documented how alarms
axial probes detected excessive movement instantly, from Bently Nevada monitors that are not acted upon can
bearing temperature change was not indicated for
allow excessive thrust movement or radial vibration to
three seconds (after the bearing surface was
occur for many minutes, hours, or even days. During these
completely melted away!), and even then only
events, seal clearances are often enlarged due to excessive
started to climb slowly (Figure 3). Finally, after 30
seconds the machine was shut down manually. movement of the shaft. Even if the bearing is not destroyed,
Fortunately, the damage was limited to a destroyed these enlarged clearances represent a loss of efficiency.
thrust collar and bearing. It could have been much Compounded over months or years, this loss of efficiency
worse. The root cause of the bearing failure was can represent hundreds of thousands, or even millions, of
later determined to be faulty connections to the dollars.
balance piston, and the thrust bearing was
consequently required to absorb an axial load in Summary
excess of its capacity. Instantaneous shutdown
Dual-voting thrust measurements represent a high
would likely have saved the bearing and the thrust
integrity measurement of this important machinery
collar. This failure introduced extended downtime at
parameter. Efforts to vote this measurement with other
a cost of over 250,000 USD per day.
alarms in the machine rarely yield an acceptable balance
As noted above, excessive axial movement leaves
between missed trips and false trips. When making thrust
very little room for interpretation or subjectivity,
measurements, therefore, always employ two probes in a
provided the position alarms have been established
dual-voting configuration, and carefully consider the
correctly in the first place. In a thrust failure, every
second is important and using the fastest implications of voting thrust alarms with other parameters.
measurement available (axial position as measured As the examples in this article indicate, improper voting can
by proximity probes) as the primary means of be both unsafe and expensive.

54 | ORBIT | 1Q01