You are on page 1of 11

1

CENTRIFUGAL PUMP LOW FLOW AND VANE PASS


RESONANCE VIBRATION: A CASE STUDY

Brandon S. Diegel
Production Operations Engineer
Omaha Public Power District
444 S 16
th
St Mall
Omaha, NE
bsdiegel@oppd.com

Abstract: Many issues are often encountered that can lead to unacceptable levels of vibration on
a horizontal centrifugal pump application. Typically a systematic approach is necessary to
identify all contributing issues and eliminate the root cause of the vibration problem.

This paper seeks to highlight an example of excessive vibration on a newly installed 150 HP
circulating water intake screen wash pump at a 200 MW coal-fired power plant in Omaha,
Nebraska. Investigation revealed the pump vibration was caused by two problems: flow related
vibration due to improper pump size for the specific application and pump structural resonance
excited by impeller vane pass. Several vibration analysis techniques under different pump
operating conditions were used to identify these two problems. The analysis process along with
the steps taken to resolve the vibration problem will be presented within the paper.

Key Words: Amplification factor; damping ratio; half-power method; minimum continuous stable
flow (MCSF); synchronous time averaging; total dynamic head (TDH); vane pass

Introduction: At North Omaha Station Unit 5, a 200 MW coal-fired generating unit, circulating
water is pulled through intake screens located on the Missouri River to provide the cooling source
for the steam condenser. As the intake screens get dirty, the differential pressure increases and
causes a screen wash pump to start. The pump sprays water through nozzles to clean the
screens as they are rotating. Depending on how many screens need to be washed
simultaneously, the flow requirements of the pump can vary significantly. After replacing one of
the screen wash pumps for maintenance reasons, the newly installed pump, which was a
different model and manufacturer, was found to operate with severe vibration in excess of 2.0 in/s
peak overall. After examining the new pump design characteristics, it was found that the new
pump had four vanes compared to six in the old pump. Also, both pumps had a similar design
operating point, but significantly different performance characteristics throughout the required
operating range. The pump was also mounted on a new skid which impacted the structural
natural frequencies. Images of the North Omaha screen houses located on the river and a
rotating screen unit can be seen in Figure 1 and Figure 2.


Figure 1 Figure 2


2

Background Information: The new screen wash pump A consisted of an entirely new pump
skid including the motor, pump, and hollow steel mounting frame. The skid was installed on the
existing steel-plated concrete foundation and was bolted and stitch welded in place. The pump is
a 125 HP single stage centrifugal type rated for 1500 gpm at 250 ft. W.C. total developed head.
The pump is driven by a 150 HP induction motor at 1785 RPM. The pump construction consists
of a four vane semi-open impeller with 8 inch suction and 4 inch discharge piping. The pump is
supported by a single row radial ball bearing and two single row angular contact ball bearings.
The A screen wash pump is connected in parallel with an identical B pump which had been in
service for over a year. Only one pump runs at any given time to meet the necessary operating
requirements. The system piping diagram which shows the pumps and screen wash nozzles can
be seen in Figure 3. The A pump mounting configuration and impeller can be seen in Figure 4
and Figure 5.



Figure 3


Upon startup of the new A pump, plant maintenance personnel observed severe noise and
vibration. The vibration was elevated throughout the entire machine train, but was highest in the
pump horizontal direction exceeding 2.0 in/s peak overall. The vibration was primarily observed
at four times operating speed. Vibration was also observed in excess of .45 in/s peak overall on
the A pump while this pump was shut down and the nearby B pump was running. Due to the
high vibration amplitudes, the pump was considered unsuitable for operation and further
investigation was necessary.



3


Figure 4



Figure 5


Test Data and Observations: Initial vibration data was collected with a CSI 2130 dual channel
analyzer, 100 mV/g magnet-mount accelerometers, and a photo tachometer. Data was collected
across the entire machine train. The highest vibration was found on the pump inboard bearing in
the horizontal direction. Figure 6 shows the spectrum and waveform data on the pump inboard
bearing which indicates an overall peak vibration of 2.42 in/s with a 4x (7173 CPM) peak of 1.97
in/s. The data collected on the motor inboard bearing revealed some 1x vibration with harmonics,
but was also dominated by the 4x frequency component. This data can be seen in Figure 7,
which shows an overall peak vibration of .89 in/s with a 4x (7173 CPM) component of
approximately .45 in/s. Synchronous time averaging revealed that these peaks were exactly 4x
multiples of running speed.
4

Since the pump impeller was known to have four vanes, flow related contributions were initially
suspected. The pump discharge pressure was 137 psi which corresponded to a total dynamic
head of 294 ft. W.C. As seen in the pump curve in Figure 12, this is significantly below the
minimum continuous stable flow range for the pump. It was also noticed that vibration was
elevated on the A pump while only the B pump was running. This vibration was exceeding .45
in/s peak overall and still occurred primarily around 7173 CPM, which was four times the
operating speed, or vane pass frequency, of the identical B pump. This sensitivity to the 7173
CPM excitation from the B pump and the excessive amplitudes observed while running led to
suspicions of resonance on the A pump.



Figure 6



Figure 7
5

To isolate and identify vibration contributions of the motor, it was decided to run the motor
uncoupled while collecting vibration data. As can be seen in Figure 8, the 4x running speed
vibration component was no longer present in the motor data. This data still revealed the
presence of 1x vibration and some running speed harmonics, likely caused by unbalance and
looseness in the motor. Although higher than desired, this was not the primary problem causing
the excessive 4x vibration. The motor was recoupled for further analysis on the pump.



Figure 8

After the motor was recoupled, impact testing was performed on the pump and foundation
structure to test for structural natural frequencies. A rubber hammer was used to impact the
pump skid structure while readings were taken at different locations on the machine train. Figure
9 shows the impact test results measured in the horizontal direction at the pump inboard bearing.
This data revealed a structural natural frequency at 6890 CPM or 114.8 Hz. This would suggest
that the vane pass frequency of the pump, at 7173 CPM, was operating only 4% above this
structural natural frequency.

For additional verification of this amplification due to resonance, it was decided to perform a coast
down test while collecting data filtered at four times operating speed to capture the vane pass
excitation. The 4x filtered coast down data is shown in the Bode and Nyquist plots in Figure 10.
Since the pump was loaded it coasted down very quickly so a limited number of samples were
collected. However, the data clearly shows the 4x filtered peak and phase shift that occurs at
1723 rpm, which corresponds to a 4x filtered frequency of 6892 CPM. This is the same
frequency revealed in the impact test. The Bode plot shows a 4x peak of 2.8 in/s as the vane
pass frequency coasts through this resonance range. This amplification factor (Q factor) and
damping ratio, using the half power method calculated from the Bode plot, were estimated as Q =
13 and = 0.04. These calculations are from Eq. (1) and (2) below. This would suggest that the
resonance was amplifying the original vibration by 13 times.
6


Figure 9



Figure 10
7












The impact and coast down testing confirmed the suspicion of resonance in the A pump
structure and explained the sensitivity to this frequency during the B pump operation. However,
due to the operating point on the pump curve, it was suspected that flow problems were also
contributing to the elevated vane pass vibration. Vibration data was collected while a second
screen was placed into the cleaning operation. This opened additional cleaning nozzles which
increased the flow and decreased the pump discharge pressure from 137 psi to 130 psi. When
this pressure dropped, there was a drastic reduction in audible noise from the pump and the vane
pass vibration dropped from 2.0 in/s to 0.9 in/s. Due to this reduction, it was suspected that the
excessive vibration was caused by flow related problems that were being amplified by the
structural resonance.

Since the B pump had an identical configuration, data was collected on this pump for
comparison. The vibration on this pump was also dominated by the 4x vane pass component,
however, the overall amplitude was much lower at 0.27 in/s peak with one screen in service. This
vane pass vibration was also reduced when a second screen was placed in service and flow
increased in the pump. Impact testing on this pump indicated its natural frequency was lower
than the A pump at 6300 CPM. It was believed that this pump also had low flow vibration
problems, however, since the vane pass frequency was 12% higher than the natural frequency,
the vibration was not being amplified as significantly by resonance.


Corrective Action and Results: The base of the new pump skid was a hollow steel frame with
holes to allow for grout injection during mounting. Since the vane pass frequency of 7173 CPM
was operating above the natural frequency of 6892 CPM, the optimal solution would be to lower
the natural frequency to gain additional separation from the vane pass excitation. It was decided
to fill the pump skid with grout in hope that the additional mass would be greater than the
additional stiffness, which would lower the natural frequency (see Eq. (3)).

8




After the base frame was injected with grout, impact testing was performed again on the pump
inboard bearing to see how the natural frequency had shifted. Figure 11 shows the results of this
test and reveals the new natural frequency of 7734 CPM, or 129 Hz. This shift revealed that the
grout had a greater effect on the system stiffness than the mass, causing the natural frequency to
shift upwards. However, the new natural frequency now had 8% separation from the vane pass
excitation frequency compared to the 4% previously. It was desired to gain at least 15%
separation from the vane pass excitation, however, it was expected the shift would still improve
the flow vibration amplification.



Figure 11


When the pump was started, the dominant vibration frequency was still 4x running speed, but the
overall amplitude was reduced from 2.1 in/s to 0.8 in/s peak. The pump discharge was 140 psi
which corresponded to low flow that was below the minimum continuous stable flow range on the
pump curve.

9

In order to vary the flow characteristics of the pump and determine the vibration at different points
on the curve, additional screens were manually placed into operation. As more screens were
placed in operation, the discharge pressure dropped, and the pump flow was increased. During
this testing, the vibration was reduced from 0.8 in/s with one screen in service to 0.19 in/s with
nozzles for all 5 available screens open. There was also a drastic reduction in audible noise as
the pump passed the minimum continuous stable flow range and entered the more efficient range
of its operating curve. The results of the three flow conditions can be seen in Table I and are
displayed on the pump curve in Figure 12 and the vibration waterfall plot in Figure 13. The
calculation used for TDH can be seen in Eq. (4).

Since the B pump is identical, it also displayed the same vibration improvement with increased
flow. However, this pump did not have the resonance issue excited by the pump vane pass, so
the flow related vibration was not being amplified like the A pump. It was decided to not inject
the B pump base with grout. Since the grout on the A pump increased the natural frequency by
561 CPM, this would have pushed the B natural frequency closer to its vane pass excitation.

Figure 12

Table I
A Screen Wash Pump Vibration
Test #
# Screens in
Service
Discharge
Press. (psi)
TDH (ft.)*
Overall
Vibration
(in/s)
Severity
1 1 140 300 0.81 Warning
2 3 130 283 0.38 Surveillance
3 5 118 264 0.19 Acceptable
*Using 10 psi suction pressure
1 2
3
10






Figure 13


In order to permanently address the pump flow problem, various ways of increasing flow through
the system were explored. The most cost effective solution was determined to be the addition of
a manual bypass valve to the screen wash drain. This bypass valve increased the pump flow by
bypassing water around the screens directly to the drain when only one or two screens were
being washed. The most common screen wash combination is one or two screens at a time.
Since these combinations result in the high back pressure and low flow, the valve will typically be
left open. However, operations desired the flexibility to wash any combination of all five screens if
necessary, so this valve could still be closed when the flow was needed to the screens. With the
bypass valve open, the pump discharge pressure with one screen in service was now reduced to
124 psi. The overall vibration on the A pump was reduced to .24 in/s peak and improved when
additional screens were placed in service. The final vibration in this configuration on pump A
can be seen in Figure 14. The final A pump vibration was still higher than ideal because of the
11

remaining resonance amplification. In order to decrease the vibration further, the mass or
stiffness of the system would need additional modifications to gain greater separation from the
structural natural frequency. Also, as flow is increased near the best efficiency point of the pump,
the vibration reduced further to 0.15 in/s.



Figure 14


Conclusion: Improper pump sizing for the particular application is a common problem that can
lead to vibration issues. In this circumstance, the pump was sized for a best efficiency point that
occurred when all five screens were being washed simultaneously. In reality, operations would
rarely wash more than one or two screens at a time, which resulted in high pump discharge
pressure and low flow. This flow was below the minimum continuous stable flow range for the
new pump, and it could not operate reliably within the entire flow range desired. This problem
was compounded by the structural natural frequency of the newly mounted skid aligning closely
with the pump vane pass excitation.

The resolutions taken on this pump system to address the low flow issue and the vane pass
excited structural resonance will result in a much more reliable pump system. In the worst flow
configuration with one screen in service, the overall vibration has been reduced from 2.1 in/s to
0.2 in/s. This pump system has continued to operate in a satisfactory condition for a year and
half since the modification.

You might also like