Hydrocarbon Processing - January 2021 - 58

Valves, Pumps and Turbomachinery
Data collection. The pump trip delay was extended from

30 sec to 60 sec to allow more time for vibration data collection. Using the motion amplification camera, the coupling
guard was removed to allow a direct recording of the coupling
movement while the machine was running. The following observations were noted:
1. When the pump started, the vibration level was
4.2 mm pk-pk at a speed of 1,650 rpm. Approximately
13 sec later, the machine reached full speed and the
vibration level was 4.7 mm pk-pk at the I/B side (FIG. 3).
2. At 11:51:48, the machine reached steady state-
the vibration level dropped gradually by 1.5 mm pk-pk
and 1 mm pk-pk at the O/B vertical and horizontal,
respectively, over the steady-state duration of 45 sec.
3. At 11:52:38, the machine tripped at an overall
vibration level of 4.6 mm pk-pk on the I/B side and
at 2.9 mm pk-pk on the O/B side (FIG. 3). During
shutdown, the vibration trend reached 1.5 mm pk-pk
on the pump I/B side, and 0.8 mm pk-pk on the pump
O/B side at low speed (600 rpm).

Data analysis. Investigating the bode plots of the I/B and
O/B sides showed that, during shutdown, the pump O/B plot
dropped following a parabolic shape and then settled around
0.8 mm pk-pk at low speed (500 rpm). Similarly, the pump
I/B plot dropped following a parabolic shape and then settled
around 0.2 mm pk-pk at low speed (500 rpm). This observation indicated the existence of high runout at pump O/B, and

FIG. 3. The trend plot of the pump was extracted from the I/B side.

FIG. 4. Orbital shape during pump startup.

58 JANUARY 2021 | HydrocarbonProcessing.com

a more dominant effect of unbalance on pump I/B. In addition, investigating the orbit plots at pump startup, steady state
and shutdown revealed the following:
* At startup, the orbital shape of pump I/B was round,
and the waveform was sinusoidal, indicating 1X
frequency with a runout of 3 mm pk-pk. Similarly,
at pump O/B, the orbital shape was round, and the
waveform was almost sinusoidal, with slightly clipped
peaks and an overall runout of 2 mm pk-pk (FIG. 4).
During steady state, the orbital shape was slightly
preloaded, and the time waveform remained sinusoidal.
During shutdown, the orbital shape remained preloaded
and the waveform showed rattles, indicating a runout.
* Recording the coupling by using the motion
amplification technology while the machine was at full
speed revealed high vertical motion on the coupling on
the pump side of 11 mm pk-pk (FIG. 5).
The combination of these observations and data showed that
the pump experienced two main forcing frequencies:
1. 1X runout, sinusoidal in nature-most likely
due to a bent shaft or coupling misalignment
2. 1X unbalance, which could have been due to
an imbalance of the rotor and/or coupling.
Root cause 1: Bent shaft (rotor sag). Typically, when an

overhaul is conducted, the pump is dismantled for inspection.
Upon assembly, the impeller is shrink fitted on the shaft.
In this case, a hypothesis was made that, if the fitting procedure was not conducted properly, a residual force would be
exerted on the fitting area. In turn, this force would start to act
during pump shipment to and from the maintenance shop. In
this scenario, the fitting residual force would act downward on
the shaft opposing the upward normal force transmitted from
the truck suspension. This would eventually cause a rotor sag
of 1 mm-2 mm, even if shaft straightening was conducted at
the shop. When the pump was restarted, the vibration trend-
due to shaft sag-would start with high runout on both sides.
When the pump reached full speed, the rotor would start to
rub on both sides, which explained the preload effect observed on both orbit plots at steady state.
In addition, an interesting observation was made on the
vibration trend (FIG. 3). The vibration amplitude started to
drop from 5.2 mm pk-pk to 4.5 mm pk-pk after 21 sec from
startup. The machine runout at cost down (0.2 mm pk-pk) was
also found to be less than the runout during startup (0.8 mm
pk-pk). This result indicated that the runout was not static in
nature like a normal mechanical runout and was affected by
dynamic forces. This observation supported the hypothesis
that the shaft started with a sag, and, when ramped up to full
speed-due to the centrifugal forces-the shaft would begin
to straighten up, leading to less shaft sag. The vibration level
of 4.5 mm pk-pk was still higher than the danger limit of 2.8
mm pk-pk, leading to the emergency shutdown system sending a shutdown signal after the trip delay passed. As a result,
it was expected that, should the pump continue to operate for
a longer time, the vibration would drop below 2.8 mm pk-pk.

Root cause 2: Rotor imbalance. On the latest overhaul,

a visual inspection showed elements removed from the shaft

http://www.HydrocarbonProcessing.com

Hydrocarbon Processing - January 2021

Table of Contents for the Digital Edition of Hydrocarbon Processing - January 2021