Donald R.

Smith is a Senior Staff

Engineer at Engineering Dynamics Inc.
(EDI), in San Antonio, Texas. For the past
30 years, he has been active in the field
engineering services, specializing in the
analysis of vibration, pulsation, and noise
problems with rotating and reciprocating
equipment. He has authored and presented
several technical papers. Prior to joining
EDI, he worked at Southwest Research
Institute for 15 years as a Senior Research
Scientist, where he was also involved in troubleshooting and
failure analysis of piping and machinery.
Mr. Smith received his B.S. degree (Physics, 1969) from Trinity
University. He is a member of ASME and the Vibration Institute.
Stephen M. Price is a Senior Staff
Engineer for Engineering Dynamics Inc.,
in San Antonio, Texas. For the past 19
years, he has been actively involved in
solving a variety of problems in systems
that have experienced failures due to
dynamic phenomenon. He has had field and
analytical experience solving problems
with reciprocating and rotating machinery,
along with structural and acoustical
problems. Additionally, Mr. Price has been
involved in development of multiple channel data acquisition and
monitoring hardware and software for critical environments. He
has presented and published several papers in the areas of signal
processing, finite element analysis, fatigue analysis, acoustics, and
reciprocating pumps.
Mr. Price has a B.S. degree (Mechanical Engineering) from
Mississippi State University and an M.S. degree (Mechanical
Engineering) from Purdue University. He is a registered
Professional Engineer in the State of Texas and a member of ASME.
ABSTRACT
Vertical turbine pumps are designed to operate with the pump
shafts in tension (downthrust). Most pump and motor manufactur-
ers caution users to not operate the vertical turbine pumps at
conditions that result in upthrust. Upthrust conditions are usually
encountered for a few seconds each time the pump is started;
however, the pump should not be operated for periods longer than
a few seconds in this condition. Due to lack of instrumentation on
the pumps, it is often difficult to determine if the pump is operating
in an upthrust condition.
Pump impellers are thrust balanced to maintain a downthrust
without excessive loads on the thrust bearings. Changing the pump
operating conditions and the pumped fluid can significantly change
the thrust balance on the pump, resulting in excessive downthrust
that can overload the thrust bearings, or excessive upthrust condi-
tions that can cause undesirable shaft behaviors. Therefore, it is
very important to calculate the thrust loads when changing the
pump operating conditions.
Two case histories are presented that involved two different
abnormal operating conditions that produced an upward thrust on
the pump shaft. In both cases, the upthrust caused the pump shaft
to buckle, resulting in failures of the mechanical seals and destruc-
tion of the shaft. Measured field data are presented for each case.
Diagnostic techniques and instrumentation needed to obtain the
field data required to solve these problems are discussed.
AXIAL THRUST FORCES ON VERTICAL PUMPS
In centrifugal pumps, axial hydraulic thrust forces result from
internal pressures acting on the exposed areas of the rotating
elements. These axial forces are the sum of the unbalanced forces
acting in the axial direction and are affected by many variables
including: impeller type, impeller shroud geometry, and the
location of the impeller relative to the stationary walls (Lobanoff
and Ross, 1992). These pressure forces acting on various
impeller types are illustrated in many textbooks. Theoretically,
the axial forces are balanced on the double-suction impeller;
however, due to many factors, such as nonuniform flows and
nonsymmetrical parts, the axial forces are rarely balanced in
actual service. The pressure forces are not balanced on the
single-suction impeller and thus substantial axial thrust forces
are developed. Thrust bearings are normally required to restrain
these axial forces.
To reduce the axial thrust on a single-suction impeller, the thrust
areas are equalized by drilling holes through the back shroud of the
impeller and maintaining suction pressure in a chamber behind the
impeller. These holes through the impeller shroud are referred to as
balance holes. The arrangement aids in reducing the thrust because
the pressure behind the balance holes will be less than the pressure
above the wear ring, but greater than the suction pressure.
Balancing holes are often considered undesirable due to leakage
back to the impeller suction, which opposes the main flow. This
leakage reduces the efficiency and sometimes can result in recircu-
lation problems.
Vertical turbine pumps were originally developed for pumping
water from wells and have been called deep-well pumps, turbine
well pumps, and borehole pumps. Although these pumps have
been used in other applications, the name vertical turbine pump
has been generally adapted by pump manufacturers (Karassik, et
al., 2001). A typical multistage vertical turbine pump is shown in
Figure 1.
46
UPTHRUST PROBLEMS ON MULTISTAGE VERTICAL TURBINE PUMPS
by
Donald R. Smith
Senior Staff Engineer
and
Stephen Price
Senior Staff Engineer
Engineering Dynamics Inc.
San Antonio, Texas
Figure 1. Typical Multistage Vertical Turbine Pump.
Calculating Net Thrust Forces
In vertical turbine pumps, the axial thrust forces are similar to
those for centrifugal pumps, except that additional forces must also
be considered. Under normal conditions, vertical turbine pumps
have thrust loads acting parallel to the pump shaft that include:

Downward forces due to unbalanced pressure forces.

Downward forces due to the weight of the rotating parts (pump

rotor).

Upward forces due to change in momentum caused by the flow

through the impellers.
At first, it seems that the calculation of the net thrust force is
fairly simple. However, the calculations are complicated by the
particulars of impeller geometry and the nature of the pumping
action. The simplified approach presented here is often modified
by the pump manufacturer based on empirical data for a specific
model of pump.
Downthrust Force
Thrust in the downward direction is due to the unbalanced forces
across the impeller (Figure 2). As shown in the sketch, the
discharge pressure reacts on both upper and lower impeller
shrouds, and the suction pressure reacts on the impeller eye. The
downthrust can be estimated using Equation (1) (Dufour and
Nelson, 1993). The maximum downthrust occurs when the pump
is operated at shutoff conditions due to the higher head.
(1)
where:
F
d
= Downthrust, lb
H = Head, ft
constant = (Net eye area, in
2
)/2.31, lb/ft
SG = Specific gravity
Rotor Wt = Total weight of shaft and impellers, lb
Upthrust Force
An upward force can be produced due to the change in
momentum of the fluid passing through the impeller. The forces
produced can be computed by considering the kinetic energy of the
Figure 2. Downward Thrust Due to Unbalanced Discharge
Pressure Across the Eye of the Impeller.
fluid V
2
/(2g). The upward force increases as the square of the
velocity through the impeller, this explains why the upward forces
can become very high at high flow rates.
For fully radial flow, the upward force can be computed using
Equation (2) (Karassik, et al., 2001). Note, for mixed flow
impellers, the upthrust force is multiplied by some value between
zero and one depending upon the degree to which the flow
becomes radial. Axial flow impellers do not produce upthrust.
(2)
where:
F
u
= Upthrust, lb (N)
V
e
= Velocity in impeller eye, ft/sec (m/sec)
g = 32.2 ft/sec
2
(9.81 m/sec
2
)
c = Unit conversion constant, 2.31 for USCS, 1.02 for SI
A
e
= Net eye area, in
2
(cm
2
)
Computing Axial Forces Using Impeller Thrust Factors
Pump manufacturers generally provide a hydraulic thrust factor
K for each type of impeller. This thrust factor (lb/ft of head) is
based upon operation at the best efficiency point (BEP).
Calculating the thrust using the thrust factor at the BEP will often
result in only an approximate thrust value because a single constant
cannot express the upthrust component that varies with capacity.
Therefore, pump manufacturers typically compute the impeller
thrust using thrust-capacity curves based upon actual measured test
data. For example, one pump manufacturer (Goulds Pumps, 2004)
computes the pump thrust as follows:

The thrust factor K is read from the thrust-capacity curve at

the required capacity and given rpm. This thrust factor combines
the downthrust and upthrust.

The K value is then multiplied by the total pump head times

the specific gravity of the pumped liquid.

If the impeller thrust is excessively high, the impeller can be

hydraulically balanced by reducing the discharge pressure above
the impeller eye by the use of balancing holes and rings (Figure 3).
Hydraulically balancing the impeller significantly reduces the K
PROCEEDINGS OF THE TWENTY-SECOND INTERNATIONAL PUMP USERS SYMPOSIUM 2005 47
Upthrust
Downthrust
Balance Holes
( )( )( ) ( ) F cons t H SG Rotor Wt
d
= + tan
F
F
c
V
g
A
u
e
e
=

1
2
2
Downthrust
Upthrust
value. Note that care should be taken when adding the balance
holes to the impellers. As with the centrifugal pumps, the balance
holes decrease the efficiency by 1 to 5 percent by providing addi-
tional paths for internal recirculation. In addition, the balance holes
can become clogged due to operating with unclean fluids, fluids
with solids, and intermittent service.
Figure 3. Hydraulically Balanced Impeller.
These thrust calculations are usually made by the pump manu-
facturer considering the full range of operating conditions. As
illustrated in the case histories, it is very important to consider
changes in the specific gravity of the fluid and in the operating con-
ditions when computing the downthrust and upthrust forces acting
on vertical turbine pumps.
Symptoms of Upthrust Problems
As stated above, vertical pumps normally operate with the pump
shaft in downthrust (tension). Operating the pumps in upthrust with
the pump shaft in compression can cause many problems (Dufour
and Nelson, 1993).

The compression load can cause the pump shaft to buckle,

resulting in high radial vibration.

Excessive compression loads can cause the shaft to be perma-

nently bowed.

The radial shaft vibration causes wear and damage to the

bearings, especially to carbon bearings (bushings).

The radial shaft vibration also damages the mechanical seals.

Seal failures due to the shaft vibrations are common and they are
often the only indication of excessive upthrust.

Failures of the mechanical seals are also caused by the excessive

upward motion of the shaft, which changes the clearances between
the stationary face and the rotating seal face.

Failures to couplings can occur due to the excessive radial

vibration and the upward forces.

Pump impellers can be damaged due to rubbing on the top of the

bowls.

Driver (motor) thrust bearings often fail if they are not designed
for upthrust loads.

The upward forces can cause the motor rotor to rub against the
stator, resulting in electrical and mechanical damage.
Causes of Excessive Upthrust
The pump manufacturers design the pumps for a certain set of
operating conditions. Case History No. 1 illustrates upthrust
problems that can occur when the fluid properties (specific gravity,
suction, and discharge pressures) are significantly different from
the original design.
Excessive upthrust forces can also occur when a pump is
operated at very high flow rates beyond the BEP. These high flow
rates can occur when a pump is started with low discharge
pressure. To avoid these problems, pump manufacturers
recommend that vertical turbine pumps and vertical can-type
pumps should be started with a partially closed discharge valve that
is fully opened after the discharge pressure is fully developed. Case
History No. 2 discusses pump problems due to upthrust that
occurred when pumps were improperly started at low discharge
pressures with fully-open discharge valves.
CASE 1UPTHRUST PROBLEMS ON
31-STAGE VERTICAL TURBINE PUMPS
Three vertical turbine pumps were installed at a pipeline pump
station (Figure 4). The pumps were referred to as pumps A, B, and
C. The pumps were originally designed for gasoline service
(specific gravity of 0.688) at a suction pressure of 100 psig and a
discharge pressure of approximately 2130 psig. The pump config-
uration is shown in Table 1.
Figure 4. Installation at Pipeline Pump Station.
Table 1. Case 1Pump Configuration.
Prior to being shipped to the plant, the pumps were tested at the
manufacturers shop. This was a hydrostatic test with the pumps
operating on water at a reduced speed of 1795 rpm. During these
UPTHRUST PROBLEMS ON MULTISTAGE VERTICAL TURBINE PUMPS 48
Upthrust
Downthrust
Balance Holes
Item Description
Pump Type Vertical Turbine
Number of Stages 31
Impeller Thrust Balance Configuration 15 Standard, 16 Thrust-Balanced
Driver Induction Motor
Horsepower 600
Running Speed 3570 rpm
shop tests, the pumps operated satisfactorily with shaft vibration
levels of 1.5 mils peak-to-peak, which were considered to be accept-
able and were below the levels specified by the API Standards.
API Standard 610 (1995) provides allowable vibration levels for
vertically suspended pumps operating within the pumps preferred
operating region. For vibration measurements on the pump thrust
bearing housing, the allowable vibration levels are 0.2 in/sec root-
mean-square (rms) (0.3 in/sec zero-to-peak). The pump shaft
allowable vibration can be computed using Equation (3).
(3)
where:
A
u
= Unfiltered displacement determined by fast Fourier
transform (FFT), mils peak-to-peak (not to exceed 4.0 mils
peak-to-peak)
N = Rotational speed, rpm
Using this equation, the allowable shaft vibration levels would be
approximately 2.4 mils peak-to-peak at 1795 rpm and 1.7 mils
peak-to-peak at 3570 rpm.
Operating with Ethane
After the pumps were installed in the field, the plant desired to
operate the pumps in ethane service, in addition to the gasoline.
The specific gravity of ethane is 0.355 and the suction pressure
would be increased to approximately 800 psig. Although the
operating conditions with ethane were significantly different from
the conditions with gasoline, the pump manufacturer reviewed the
pump performance curve and indicated that the pumps could also
operate in ethane service.
When pumps A and B were run with ethane, the vibration levels
at the top of the motor and at the top of the pump thrust stand
(bottom of the motor) were approximately 1 in/sec zero-to-peak,
which exceeded the API 610 (1995) vibration limit. Pump B was
tested with the motor running solo and the vibration levels were
also considered to be high, even in the solo condition. Impact tests
indicated a minor structural natural frequency slightly above the
pump running speed of 60 Hz. The pump manufacturer felt that the
structural natural frequency contributed to the vibration levels.
Next, the connection between the motor and the thrust stand on
pump B was modified in an effort to lower the natural frequency
below the running speed. Impact tests indicated that the natural
frequency was lowered closer to the running speed, which actually
increased the vibration levels. The tests indicated that the vibration
levels were greater in the North-South (N-S) direction, which was
in the direction parallel with the suction and discharge piping. The
vibration levels in the East-West (E-W) direction (perpendicular to
the cutout sections on the thrust stand) were much lower.
Operating Deflection Shape
An operating deflection shape (ODS) was measured on pump A,
which had not been structurally modified. An ODS is the actual
forced vibration response shape of the motor/pump structure at a
particular operating condition. In addition to identifying the forced
vibration operating shape, the animation is useful in identifying
loose joints, weak structural members, and other problem areas.
The animation is created by obtaining response data at multiple
locations in the three orthogonal directions. The data are processed
to produce the animated display.
The ODS model is shown in Figure 5. A selected single frame of
the vibration animation at 1 running speed is shown in Figure 6.
Note that the vibrations are exaggerated in the plot to allow the
ODS to be easily visualized. As shown in the ODS, the maximum
vibration levels were measured near the top of the motor in the N-
S direction. The maximum vibration levels were approximately 6.3
mils peak-to-peak (1.2 in/sec zero-to-peak).
Figure 5. Pump/Motor ODS Model.
Figure 6. Operating Deflection Mode Shape at 1 Running Speed.
The ODS indicated that the motor was rolling or wobbling
on top of the thrust stand. In addition, slipping was observed
between the flanges connecting the discharge head to the barrel.
Subsequent to the ODS testing, several of the flange bolts were
determined to be loose and were tightened; however, tightening the
bolts did not reduce the vibration to acceptable levels.
Pump Shaft Vibration
During the process of obtaining the ODS vibration data on the
motor and thrust stand of pump A, excessive vibrations were
visually observed on the pump coupling. The vibration levels on
the pump coupling hub and the upper pump shaft were estimated
to be in the range of 100 mils peak-to-peak (Figure 7). These
PROCEEDINGS OF THE TWENTY-SECOND INTERNATIONAL PUMP USERS SYMPOSIUM 2005 49
A
N
u
=
10 000 ,
Motor
Thrust Stand
Discharge Head
Barrel
Foundation
Z
Y
X
East-West
North-South
Vertical
Slipping Between
Flanges
Thrust Stand
6.3 mils PP Vibration
Amplitude
Motor Wobbling on
vibration levels were clearly excessive and were considered to be
in the danger level. The observed vibration levels on the motor
coupling hub were significantly lower compared to the pump
coupling hub, which indicated that the coupling was effective in
isolating the vibration between the pump shaft and the motor shaft.
Figure 7. Excessive Vibration on Pump Coupling and Pump Shaft.
Although the observed vibration levels were clearly excessive, it
was decided to obtain a minimum amount of shaft vibration data
using proximity probes. Proximity probes were installed to
measure the pump shaft radial vibration near the mechanical seal
(the probes were installed similarly to those shown in Figure 16 in
Case History No. 2). The probes could not be installed near the
coupling because the estimated vibration levels of 100 mils peak-
to-peak exceeded the range of the proximity probes, which was
approximately 50 mils peak-to-peak.
After the proximity probes were installed, the pump was started,
operated at full speed for only a few seconds, and then shut down.
Figure 8 is a plot of the 1 running speed shaft vibration and phase
(Bode plot) during the coastdown. At operating speed, the pump
shaft vibration near the seal was approximately 36 mils peak-to-
peak. Slow roll data indicated that the shaft runout was
approximately 16 mils peak-to-peak (note that this pump had been
in service for approximately 500 hours at the time of the test).
Figure 8. Bode Plot of Pump Shaft Vibration During Coastdown.
This runout greatly exceeded the allowable runout at the seal.
API 610 (1995) states that the total indicated runout shall not
exceed 0.0005 inch per ft (4 m per 100 mm) of length, or 0.003
in (80 m) over the total shaft length.
The shaft vibration and runout data were reviewed with the
pump manufacturer. Everyone agreed that the vibration levels were
considered to be dangerous and could result in a failure of the
coupling, shaft seal, and/or thrust bearing. The excessive runout
indicated that the shaft was probably bent.
A visual observation of the coupling vibration on pump B, which
was running, indicated that the vibration levels were similar to those
on pump A. Therefore, it was suspected that the shaft on pump B was
also bent. The pump manufacturer recommended that both pumps
should be removed and returned to the factory for further analysis.
Pump C had not been run; therefore, it was decided to measure
the pump shaft runout prior to starting the pump. The shaft was
rotated by hand and the shaft runout at the seal was measured to be
approximately 1 mil peak-to-peak, which was considered to be
acceptable. The pump was subsequently put into operation and the
shaft vibration reached approximately 10 mils peak-to-peak at 1
running speed. Figure 9 is a time-domain plot of the shaft vibration
and pump speed just prior to shutdown and during the coastdown.
As shown, the runout increased to approximately 7.5 mils peak-to-
peak after operation, which indicated that the shaft was bowed or
cocked during this short run.
Figure 9. Time-Domain Plot of Pump Shaft Vibration and Pump
Speed During Coastdown.
The pump manufacturer recommended that the pump shaft be
raised approximately 3/16 inch, which also required adjusting the
seal position. When the pump was restarted, the vibration levels
immediately increased to approximately 18 mils peak-to-peak.
After operating for a few seconds, the mechanical seal began to
smoke and the pump was shut down.
It was determined that the mechanical seal was destroyed and the
pump could not be operated. It was unknown whether the high shaft
vibration contributed to the seal failure (especially since the other
pumps could operate with shaft vibration levels of approximately
31 mils peak-to-peak without immediately destroying the seal). The
plant personnel thought that the seal failure was due to improper
adjustments of the seal set screws after the pump was raised.
Excessive Upthrust Due to Changing
Fluid and Operating Pressures
The problems experienced by the pumps (excessive motor
vibration, excessive shaft vibration, bent shafts, and seal problems)
UPTHRUST PROBLEMS ON MULTISTAGE VERTICAL TURBINE PUMPS 50
Approx. 100 mils p-p
High Vibration Observed
0 500 1000 1500 2000 2500 3000 3500 4000
Speed (RPM)
180
0
-180
D
e
g
r
e
e
s
A
m
p
lit
u
d
e
(
m
ils
P
P
)
10
20
30
40
0
Runout = 16 mils PP
1X Order Track During Coastdown
Pump Shaft Probe Just Above Seal
X Direction
1 DIV
0 5 10 15 20 25 30
Time (secs)
0
2
0
0
0
4
0
0
0
R
P
M
X Probe
Y Probe
U
n
i
t
T
r
i
p
p
e
d
Approx. 10 mils p--p during operation
Runout = 7.5 mils
5 mil / Div
5 mil / Div
were all classical symptoms of excessive upthrust. The pump man-
ufacturer agreed and indicated that the pump thrust had not been
recalculated for operation with ethane. All three pumps were then
returned to the manufacturer for restaging (Figure 10).
Figure 10. Pump Assembly Being Removed.
The 31-stage pump was originally supplied with 15 standard
impellers and 16 hydraulically balanced impellers for operation
with gasoline. For operation with ethane, the pump was reconfig-
ured with 28 standard impellers and only three hydraulically
balanced impellers.
When the pumps were run with ethane with the new impeller
configuration, the vibration levels were significantly reduced. The
maximum vibration levels on the top of motor were less than 1 mil
peak-to-peak. Although no shaft vibration data were obtained,
visual observations indicated that the shaft vibration levels were
also significantly reduced.
It was recommended that proximity probes should be perma-
nently installed on each pump to measure the pump shaft vibration
at the seal. The probes should be monitored on a regular basis to
ensure that the vibration levels are acceptable. Although the motor
vibration levels were reduced with the new impeller configuration,
the field tests indicated that the motor housing vibration was not a
reliable indication of the pump shaft vibration.
The pump manufacturer indicated that the same impeller con-
figuration could not be used for pumping gasoline and ethane. The
original impeller configuration was designed for gasoline and the
new configuration was required for pumping ethane. It was
suggested that one of the pumps could be set up for pumping
gasoline and the other two pumps could be set up for ethane.
A long-term modification would be to consider a more robust
pump design that could operate with a larger range of fluids and
operating conditions. The modifications could include: a larger
diameter shaft, different impellers, and a larger thrust bearing that
could handle higher upthrust and downthrust.
This example illustrates that each pump should be designed to
match the product (specific gravity) and operating conditions
(suction and discharge pressures). The vertical turbine pump
should be set up so that the pump shaft is in tension (downthrust).
For this installation, a different impeller configuration was
required for each product due to the large difference in specific
gravity and operating pressures.
CASE 2UPTHRUST PROBLEM ON
NINE-STAGE VERTICAL CAN-TYPE PUMPS
Three nine-stage vertical pumps were installed at an electric
power station to pump liquid petroleum gas (LPG) with a specific
gravity of 0.53 (Figure 11). The pumps are referred to as pump A,
pump B, and pump C. The pumps draw suction from the bottom of
a 90,000 gallon tank adjacent to the pumps and deliver high
pressure LPG, which is later vaporized downstream to feed two gas
turbines. The design capacity of each pump is approximately 420
gpm at a discharge head of 1600 ft. The pump configuration is
described in Table 2.
Figure 11. Typical Vertical Can-Type Pump.
Table 2. Case 2Pump Configuration.
During normal operation, two pumps operate in parallel with a
third pump in standby. The pump station was configured with a
common recirculation line that is shared by all three pumps. The
pump manufacturer recommended a minimum flow of 111 gpm for
each pump to prevent pump recirculation and to provide sufficient
lubrication to the carbon bushings.
In the design stage, the pump manufacturer performed a
hydraulic analysis of the pump to compute the thrust for a wide
range of operating conditions. Computed thrust values and total
developed head for flow rates from zero to 504 gpm at the
minimum suction pressures are plotted in Figure 12. The down-
thrust values ranged from a maximum of 3585 lb (closed
discharge, zero gpm) to a minimum of 864 lb (maximum suction
pressure and 504 gpm).
The thrust values were extrapolated for high flow rates beyond
the BEP (Figure 12). As shown, at the higher flow rates the thrust
changed from downthrust to upthrust. The pump manufacturer
did not compute the thrust forces for the conditions with zero
PROCEEDINGS OF THE TWENTY-SECOND INTERNATIONAL PUMP USERS SYMPOSIUM 2005 51
Item Description
Pump Type Vertical Can Type
Number of Stages 9
Impeller Thrust Balance Configuration 9 Standard
Driver Induction Motor
Horsepower 150
Running Speed 3550 rpm
Figure 12. Computed Thrust Values and Total Developed Head
Versus Flow Rate.
differential pressure, or for discharge pressure less than the suction
pressure, because these conditions were not normal operating con-
ditions.
Pump Failure History
Pump B
During the initial commissioning, pump B failed after operating
for approximately seven to eight hours. The pump was rebuilt in
the manufacturers shop and returned to service.
A few weeks later, pump B failed for the second time. The pump
mechanical seal was leaking and the pump was taken out of
service. When the pump failed, it had been running together with
pump A. The total running time on each pump was unknown. The
pump was shipped back to the shop for disassembly.
Pump A
Pump A failed approximately one hour after the failure of pump
B. The pump shaft could not be rotated. Pump A was also removed
and taken to the shop for disassembly.
Pump C
Pump C also experienced a failure of the mechanical seal after
operating for a short period. A new mechanical seal was installed
and the shaft runout at the upper seal (below the motor coupling)
was measured with a dial indicator to be approximately 7 mils
peak-to-peak (total indicated reading (TIR) = 7 mils), which was
considered to be excessive. The normal diametrical clearance at the
upper bushing was approximately 8 to 10 mils.
In addition, the shaft was in a slight bind when the shaft was
rotated by hand. This also suggested that the shaft was possibly
bent.
Since the shaft runout (bent shaft) was considered to be
excessive, it was recommended that the pump should not be
started. It was felt that the vibration levels could be excessive and
the pump could experience another catastrophic failure. In
addition, it was felt that it was important to observe the damage on
the internal parts before excessive wear occurred. Therefore, the
pump was removed and also transported to the shop for disassem-
bly.
Mechanical Damage Assessment
Pump A
The pump rotor was seized and the intermediate shaft was bent
several inches (Figure 13). The shaft had to be cut into several
pieces to remove the impellers (Figure 14). The shaft showed
excessive wear at all of the bearing locations (carbon bushings). All
of the carbon bushings were destroyed. The carbon bushings were
broken into several small pieces, which were lodged in the
impellers (Figure 15). In addition, there were also large quantities
of carbon powder in the impellers. All of the wear rings were badly
worn and there was evidence of galling between the wear rings on
the impellers and the wear rings on the stationary parts.
Figure 13. Pump Shaft from Pump A. (Shaft was bent after the shaft
seized. The motor tripped on high amps.)
Figure 14. Damaged Pump Shaft Cut into Pieces to Remove
Impellers.
Pump B
The pump was in similar condition to pump A. Although the
pump shaft did not seize, the shaft was bent and galled. It was
difficult to remove the impellers from the shaft and they had to be
hammered off. The shaft showed similar signs of excessive wear
and heat at all of the carbon bushings. All of the carbon bushings
were similarly destroyed. Again, there were large pieces of the
carbon bushings in the impellers and large quantities of carbon
powder in the impellers.
Pump C
The upper shaft was bowed approximately 3 mils, which was
equal to the API limit. The intermediate shaft was straight with a
maximum bow of approximately 1 to 2 mils. During the disassembly
UPTHRUST PROBLEMS ON MULTISTAGE VERTICAL TURBINE PUMPS 52
0
500
1000
1500
2000
2500
3000
0 100 200 300 400 500 600 700 800 900
Capacity (GPM)
H
e
a
d
(
f
t
)
-2000
-1000
0
1000
2000
3000
4000
T
h
r
u
s
t
F
o
r
c
e
(
L
b
)
Total Developed Head Thrust Force
Upthrust
Downthrust
BEP
Dashed Lines are
Extrapolated Values
Thrust
Head
Net Thrust Calculation
Figure 15. Impeller from Pump A During Disassembly after Pump
Failure. (Note carbon particles and dust from destroyed carbon
bushings.)
of the impellers from the pump shaft, the shaft was still in a bind
and a rub could be detected when the shaft was rotated by hand.
The lower pump shaft was bowed approximately 5 to 6 mils, which
exceeded the API limit. The lower shaft showed wear where the
shaft was in contact with the carbon bushings. The wear pattern
was not completely around the circumference of the shaft, which
indicated that the shaft was bent and was not in even contact with
the bushings.
The carbon bushings showed a light scoring around the circum-
ference. The bushings near the lower impellers (first and second
stages at the bottom end of the shaft) showed almost no wear, while
the bushings at the upper stages showed more wear. The lower
pump shaft was worn where the shaft was in contact with the
bushings.
The bearing damage was thought to be similar to damage to the
gland rings discussed in the manufacturers service manual, which
stated that carbon dust accumulation on an outside gland ring
could be due to: (1) inadequate amount of liquid at the sealing
faces, or (2) liquid film flashing and evaporating between seal
faces and leaving residue that is grinding away the carbon.
The visual inspection indicated that the C pump was apparently
experiencing problems similar to those that led to the severe
damage on pumps A and B. It was felt that the damage on pump C
was limited, because the pump had been in service for only a few
hours.
Task Force Recommendations
A task force was assembled to investigate the pump failures, to
recommend tests to identify the causes for the failures, and to
recommend possible modifications to eliminate the failures. The
inspection had revealed that pumps A and B were almost com-
pletely destroyed and the only items that could be reused were a
few of the flanged column sections. The damage to pump C was
not severe and it was decided to reassemble the pump and use it for
test purposes to obtain data with a pump in service.
The pump failures and the pump operating conditions were
reviewed with the pump manufacturer. At that time, the exact
cause(s) for the failures were not known. Visual inspection of
pumps A and B at the shop suggested that the failures were
possibly due to a lack of liquid flow through the impellers. The
lack of flow would result in a lack of lubrication of the carbon
bushings, which in turn would create excessive heat and the
failures of the carbon bushings. After the carbon bushings failed,
the wear rings on the impellers contacted the stationary parts,
which resulted in galling between the stainless steel parts.
Pump A tripped on excessive motor current after the pump shaft
seized. Pumps B and C tripped on high seal pressure after the
mechanical seals were damaged due to excessive shaft vibration.
Possible Vapor Locking
Although the exact cause(s) for the lack of flow through the
impellers was not known, it was felt that the problem could be due
to vapor locking of the pump impellers. Vapor locking occurs
when an impeller (usually the first stage impeller) ingests a large
quantity of vapor. Usually, small vapor bubbles pass through the
impellers. However, a larger quantity of vapor can cause the
impeller to stop pumping fluid. After the first stage impeller
becomes vapor locked, the second stage impeller can become
vapor locked due to the lack of flow into the impeller. The vapor
locking continues on subsequent impellers until the entire pump is
vapor locked.
The pump service manual stated that it is important to properly
vent the pump barrel and auxiliary seal components, particularly
when pumping fluids like LPG, which are near the vapor pressure.
The manual also stated that the barrel must be continuously vented
back to the vapor side of the suction vessel with liberal size piping.
The pump/piping system was inspected and it was determined
that the pump barrel was not properly vented. Several piping mod-
ifications were recommended to correct potential problems in the
vent and drain systems.
In addition to the piping modifications, several other modifica-
tions were installed in an effort to reduce the possibility of vapor
locking the pumps. These modifications were primarily designed
to prevent cavitation that could produce vapor in the pump.

The pump manufacturer recommended that the first stage

impeller should be replaced with an impeller with a larger diameter
eye to reduce the net positive suction head required (NPSHR).
These impellers were installed on pumps A and B.

The original carbon bushings were replaced with Graphalloy

nickel impregnated carbon bushings.

A larger diameter recirculation valve and piping were installed

to increase the minimum flow rate through each pump from 110
gpm to approximately 250 gpm.

Additional instrumentation was installed on the motor switch

gear to allow the plant distributed control systems (DCS) to
monitor the motor currents. Logic was added to the DCS to trip the
pumps if the motor amps were below the calculated levels for low
flow rates (below the minimum flow rate). This would prevent a
pump from operating with very low flow rates, or no flow condi-
tions.

The pump stainless steel wear rings were replaced with nickel
bronze wear rings to reduce the possibility of galling.
Field Tests
In addition to the possible vapor locking problems, it was felt
that the pump failures could be due to excessive upthrust forces
during certain operating conditions, especially during the startups.
Therefore, field tests were conducted on all three pumps to
determine the minimum flow rates, the effects of parallel pump
operation, and the effects of starting the pumps with the discharge
valve fully and partially opened.
The field tests were designed to obtain additional data to
determine the basic cause(s) of the previous failures. The tests
were designed to measure the pump shaft vibration and deflec-
tions, the suction static pressure and pulsation, the discharge static
PROCEEDINGS OF THE TWENTY-SECOND INTERNATIONAL PUMP USERS SYMPOSIUM 2005 53
pressure and pulsation, the pump flows (total and recirculation),
the temperature of LPG in the pump sump, and the motor amps at
the pump over as wide an operating range as possible.
Instrumentation
During the tests, data were obtained from permanently mounted
transducers (DCS data) and from transducers that were temporar-
ily installed. The following is a list of data that were obtained on
the pumps.

Suction static and dynamic pressureA strain gauge type

pressure transducer was installed in the drain immediately
upstream of the pump to measure the static pressure and dynamic
pressure (pulsation). A second static pressure transducer was
installed in the pump barrel drain line at the pump head. This trans-
ducer functioned similarly to a manometer in an attempt to
measure the static pressure at the pump suction bell (at base of
pump column).

Discharge static and dynamic pressureA pressure transducer

was installed in the vent just downstream of the pump.

Pump shaft vibrationProximity probes were installed to

measure the pump shaft vibration at the mechanical seal and at the
bottom of the motor. The probes were installed radially in pairs (X
and Y, 90 degrees apart) at both locations (Figure 16). These probes
enabled the shaft vibration orbit to be obtained, as well as the shaft
centerline positions. Another proximity probe was mounted in the
vertical direction to measure the shaft thrust direction on the
coupling flange (upthrust or downthrust).
Figure 16. Instrumentation Test Locations.

Vibration of motor and pump housingTriaxial accelerometers

were mounted on the top of the motor and the top of the pump
discharge head to measure structural vibrations in the horizontal (X
and Y) and vertical directions.

Motor currentThe motor currents were measured with clamp-

on amp probes.

Flow ratesTotal flow and recirculation flow rates were

obtained using the output signals from the existing orifice flowme-
ters.

Motor speedThe motor running speed was measured using an

optical tach.

Suction temperature in pump barrelA thermocouple was

installed in the pump barrel drain to measure the temperature in the
pump barrel.
During the field tests, all of the data were simultaneously
recorded using a 32-channel digital tape recorder. For comparison
with the tape recorded data, special files were set up on the DCS to
acquire additional data related to the operation of the LPG pumps.
A spectrum analyzer was used to produce the spectral plots. An
analog-to-digital converter was used to acquire the time-domain
data. Proprietary software was used to control the instrumentation
and acquire the data.
Pump CFirst Start with Discharge Valve Fully-Open
The pump was initially started using the existing operating
procedure with the discharge valve fully-open. Prior to startup, the
pump and piping were vented to remove any vapors and to ensure
that the system was full of liquid. The recycle valve was closed and
the bypass valve around the recycle valve was partially opened.
Before the pump was started, the pump shaft lifted vertically
approximately 20 mils when the suction pressure in the pump can
was increased above 100 psi. The pump manufacturers represen-
tative indicated that this was not uncommon and was due to the
unbalanced force on the pump shaft, which was greater than the
weight of the pump rotor. The lift of 20 mils was the combined
clearance between the coupling hub and the shaft, and the vertical
clearance in the motor thrust bearing. Therefore, when the pump
was started, the rotor was in the upthrust position.
The pump reached full speed in approximately 1.75 seconds.
However, the discharge pressure increased very slowly and
required approximately 30 seconds to increase to 400 psi (Figure
17).
Figure 17. Time-Domain Plot of Pressure, Flow, and Speed During
First Startup of Pump C with Fully-Open Discharge Valve (Vapor
or Flashing in Piping).
The flowmeter, which senses flow to the plant (total flow), did
not indicate any flow until approximately 20 seconds after the
pump started (Figure 17). The flowmeter indicated that the flow
rate increased over a 30 second interval to a maximum of approx-
imately 1200 gpm, which caused the pump to be tripped by the
DCS on excessive flow rate. It is felt that this indicated flow rate
was too high and was incorrect because this flow rate was well
beyond the pump maximum flow rate. The high indicated flow rate
could have been due to vapor or flashing in the piping that could
have caused the orifice flowmeter to read incorrectly.
UPTHRUST PROBLEMS ON MULTISTAGE VERTICAL TURBINE PUMPS 54
Axial
Probe
Rigid
Coupling
Radial
Probes
Optical
Tach
0 5 10 15 20 25 30 35 40 45 50
Time (secs)
0
2
0
0
4
0
0
P
r
e
s
s
u
r
e
(
P
S
I
G
)
F
l
o
w
(
G
P
M
)
0
1
0
0
0
5
0
0
S
p
e
e
d
(
R
P
M
)
4
0
0
0
0
D
ischarge
Suction
Total Flow
Recycle Flow
M
o
t
o
r
S
t
a
r
t
T
r
i
p
The pump shaft operated in the upthrust condition for approxi-
mately 25 seconds until the discharge pressure was increased to
approximately 200 psi. This upthrust condition is not desirable and
can result in bending of the pump shafts and excessive radial
vibration.
Pump CSecond Start with Discharge Valve Fully-Open
Prior to restarting the pump, the shaft runout was measured. The
runout was increased from 7.5 mils to approximately 12 mils. This
runout was measured with the pump barrel pressurized, which meant
that the shaft was again in the upthrust position. Initially, it was
thought that the permanent shaft bending was increased from 7.5
mils to 12 mils; however, it was later determined that the increased
bending was a temporary condition due to the upthrust on the shaft.
The pump was started using the same procedure with the
discharge valve fully-open. Again, the discharge pressure increased
very slowly and required approximately 25 seconds to reach 500 psi
(Figure 18). During this start, the maximum indicated flow rate was
approximately 850 gpm, which could have been the actual flow rate.
The motor current data shown in Figure 19 did not remain constant
and the motor amps reduced slightly during this period of high
indicated flow rate, which suggests that these flow rates were real.
Figure 18. Time-Domain Plot of Pressure, Flow, and Speed During
Second Startup of Pump C with Fully-Open Discharge Valve (No
Vapor in Piping).
The shaft was in the upthrust position for approximately 15 to 20
seconds until the discharge pressure increased above 200 psi. As
shown in Figure 19, the shaft vibration decreased as the pump
discharge increased above 200 psi, because the pump downthrust
increased with the higher discharge pressure. As discussed, the
pump is designed to operate in downthrust with the shaft in tension.
Pump CStartup with Partially-Closed Discharge Valve
Based upon the results of the previous tests, it was recom-
mended that the pump should be started with a partially-closed
discharge valve to cause the pump to develop discharge pressure in
a shorter time period. The discharge valve was initially closed and
then opened approximately two to three turns, and then the pump
was started.
The pump reached full speed in approximately 1.5 seconds (Figure
20) and the discharge pressure immediately increased to approxi-
mately 350 psi in approximately 1 second. The pump shaft position
changed from upthrust to downthrust in approximately 0.5 seconds
(Figure 20) compared to approximately 15 to 25 seconds when the
pump was started with the discharge valve in the fully-open position.
Figure 19. Time-Domain Plot of Shaft Radial Vibration, Shaft Axial
Position, Flow, and Motor Amps During Second Startup of Pump
C with Fully-Open Discharge Valve.
Figure 20. Time-Domain Plot of Shaft Radial Vibration, Shaft Axial
Position, Pressure, and Running Speed During Startup of Pump C
with Partially-Closed Discharge Valve.
As shown in Figure 20, the pump shaft radial (X, Y) vibrations
were reduced when the pump shaft changed to the downthrust
position. The radial vibration levels remained fairly constant over a
large flow range after the pump switched to the downthrust condition.
Based upon this test, it was recommended that the starting
procedure should be modified to start the pumps with the discharge
valve almost closed. After a few seconds of operation, the
discharge valve was slowly opened. This starting procedure
generally agreed with the procedure recommended by the pump
manufacturer in the pump manual.
After this startup, the pump continued to operate satisfactorily with
fairly constant vibration levels, even when the flow rates were varied.
Pump AStartup with Pump C Running
Pump A was essentially a new pump with all new parts, except
for a few column spacers. The pump had a new first stage impeller
and Graphalloy

bushings.
PROCEEDINGS OF THE TWENTY-SECOND INTERNATIONAL PUMP USERS SYMPOSIUM 2005 55
0 5 10 15 20 25 30 35 40 45 50
Time (secs)
0
2
0
0
4
0
0
P
r
e
s
s
u
r
e
(
P
S
I
G
)
F
l
o
w
(
G
P
M
)
0
1
0
0
0
5
0
0
S
p
e
e
d
(
R
P
M
)
4
0
0
0
0
Suction
Discharge
Total Flow
Recycle Flow
1 DIV
1.5 1.75 2 2.25 2.5 2.75 3 3.25 3.5 3.75 4
Time (secs)
0
5
0
0
P
r
e
s
s
u
r
e
(
P
S
I
G
)
Suction Pressure
Pump C Discharge Pressure
Axial Position Shifted Out of Range of Probe
Note: Downthrust Occurred in Approx 0.5 sec
After Startup
Vibration Was Higher Than A or B Pumps
Speed Profile
0 RPM
4000 RPM
Note: "Noisy" Tach Signal
Because Pump C Has a Bent Shaft
Upper Shaft X
Upper Shaft Y
10 mil / Div
10 mil / Div
10 mil / Div
Shaft Axial
1 DIV
0 5 10 15 20 25 30 35 40 45 50
Time (secs)
Shaft Radial Vibration - 20 mils / Div
Shaft Axial Position - 20 mils / Div
Pump Discharge Head Vibration - 5 mils / Div
Total Flow - 1000 GPM / Div
Recycle Flow - 1000 GPM / Div
Motor Current - 500 Amp / Div
Prior to startup, the shaft was rotated by hand and the runout on
the pump shaft at the mechanical seal was measured to be approx-
imately 4 mils with no suction pressure. The runout was increased
to approximately 12 mils when the shaft was in upthrust with 180
psi suction pressure. This increase in runout was due to the shaft
being in compression in the upthrust condition.
Pump A was started with pump C in service. Since another pump
was in service, the pump was started with the discharge valve fully-
open. As shown in Figure 21, the pump shaft shifted to the
downthrust position in approximately 0.5 seconds after the startup.
The discharge pressure began to increase immediately as the pump
running speed ramped to full speed. After the pump shifted to the
downthrust position, the vibration levels were immediately
reduced from approximately 9 mils peak-to-peak to the original
runout value of 4 mils peak-to-peak.
Figure 21. Time-Domain Plot of Shaft Radial Vibration, Shaft Axial
Position, Pressure, and Running Speed During Startup of Pump A
with Pump C Operating.
Pump C was tripped approximately nine seconds after pump A
was started. After pump C tripped, the pump C shaft axial position
switched to the upthrust position when the discharge pressure was
reduced to the suction pressure (Figure 22).
Figure 22. Time-Domain Plot of Flow, Shaft Axial Position,
Pressure, and Motor Amps of Pumps A and C During Startup of
Pump A with Pump C Operating and Shutdown of Pump C.
When pump A was shut down, the vibration levels again
increased to approximately 11 mils peak-to-peak during the
coastdown when the pump returned to the upthrust condition. This
shaft vibration was almost identical to the runout measured before
the startup when the shaft was in the upthrust condition, which
indicated that the shaft was not bent during the startup or during
operation.
Test Results
The following conclusions are based upon the measured data,
inspections of the failed pumps, and various technical references.

The original startup procedure with the fully-open discharge

valve was undesirable because the pump shaft operated with high
vibration levels in the upthrust position for approximately 15 to 25
seconds, which caused the pump shaft to be permanently bent.

The startup procedure with the partially-closed discharge valve

was much better. The pump operated in the upthrust position for
only 0.5 seconds and the shaft was not permanently bent.

Startup with the discharge valve fully-open was satisfactory

when another pump was in service, because the pump was started
against a high discharge pressure.

There was no evidence of major cavitation, which suggests that

large volumes of vapor were not being produced in the pumps due
to cavitation at the impellers.

The proximity probes showed that the pump shaft was deflected
radially at certain operating conditions, which caused the pump
shaft to be forced against the bushings. It is thought that the force
against the bushings would be increased with a bent shaft. The
increased force with a bent shaft would result in higher tempera-
tures at the bushings. This increased temperature could then cause
the LPG to vaporize. Since the pumps did not originally have a
continuous vent system, the vapor levels could increase in the
pump and could eventually result in a vapor lock condition in the
impellers. This vapor lock condition could then result in the lack of
cooling and the eventual failure of the pumps.

During the field tests, the vent systems were installed and there
was no indication of a large increase in temperature in the pump
can, which suggested that vapor was not collecting in the pump can.

During a shutdown, the pump shaft vibration increased to

excessive levels because the pump shaft was in the upthrust
condition for approximately 25 seconds as the pump is coasting
down. Pump shutdowns should be minimized to reduce seal wear.

Operation in the upthrust condition caused excessive shaft

vibration and was thought to be the major cause of the previous
seal failures.

No additional failures occurred after the startup procedures were

modified.

The field tests indicated that the vibration levels on the motor
and the pump discharge housing were low, even when the shaft
vibration levels were excessive; therefore, the motor and pump
housing vibration should not be used to evaluate vertical pumps.
However, the pump shaft vibrations measured with proximity
probes were indicative of possible problems and could be used to
evaluate the condition of the pumps.
SUMMARY
As discussed, upthrust forces can result in excessive vibration,
bent pump shafts, and damage to the seals and bearings. In some
instances, high-level upthrust forces can cause catastrophic failures
of the pump.
The following is a list of recommendations to prevent and avoid
excessive upthrust forces.

The pump manufacturer should be requested to perform a

hydraulic analysis if the operating conditions (suction pressure,
UPTHRUST PROBLEMS ON MULTISTAGE VERTICAL TURBINE PUMPS 56
1 DIV
4.5 4.75 5 5.25 5.5 5.75 6 6.25 6.5 6.75 7
Time (secs)
0
5
0
0
P
r
e
s
s
u
r
e
(
P
S
I
G
)
Suction Pressure Pump A Discharge Pressure
Discharge Line Pressure
Axial Position Shifted Out of Range of Probe
Note: Downthrust Occurred in less than 0.5 sec
After Startup
Vibration Was Reduced to Approx 4 mils PP
When Downthrust Was Achieved
Speed Profile
0 RPM
4000 RPM
Shaft - X Probe 10 mil / Div
Shaft - Y Probe 10 mil / Div
Shaft Axial Probe - 20 mil / Div
1 DIV
0 5 10 15 20 25 30
Time (secs)
0
5
0
0
P
r
e
s
s
u
r
e
(
P
S
I
G
)
Suction Pressure
Pump A Disch Pressure Pump C Dsch Pres
C--->
A
-
-
>
Pump A Shaft Axial Position
20 mil / Div
20 mil / Div
C Shaft Axial Position
200 Amps RMS / Div Pump A Motor Current
Pump C Motor Current 200 Amps RMS / Div
P
u
m
p
A
S
t
a
r
t
P
u
m
p
C
T
r
i
p
discharge pressure, and flow rates) and/or fluid properties (specific
gravity) are significantly changed from the original design values.

When operated alone, the pump should be started with a

partially-closed discharge valve. After a few seconds, the valve can
be fully opened because the discharge pressure will be fully
developed.

For operation in parallel with another pump that is running, the

pump can be started with a fully-opened discharge valve because
the pump will be started against a high discharge pressure.

The field tests indicated that upthrust conditions occur for a

short time period during the startup and shutdown of the pumps;
therefore, the number of startups and shutdowns should be
minimized, if possible. In situations where coastdown times are
lengthy, a braking system should be considered.

Shaft upthrust problems were not easily detected by measuring

vibrations on the motor and pump thrust stand with accelerometers
or velocity transducers. Therefore, proximity probes should be
installed to monitor the pump shaft vibration. Another proximity
probe could also be installed axially to measure upthrust.
REFERENCES
API 610, 1995, Centrifugal Pumps for Petroleum, Heavy Duty
Chemical, and Gas Industry Services, Eighth Edition,
American Petroleum Institute, Washington, D.C.
Dufour, J. W. and Nelson, W. E., 1993, Centrifugal Pump
Sourcebook, New York, New York: McGraw-Hill, Inc.
Goulds Pumps, 2004, Pump Fundamentals, Technews,
www.gouldspumps.com, Retrieved April 14, 2004.
Karassik, I., Messina, J., Cooper, P., and Heald, C., 2001, Pump
Handbook, Third Edition, New York, New York: McGraw Hill.
Lobanoff, V. S. and Ross, R. R., 1992, Centrifugal Pumps: Design
& Application, 2nd Edition, Houston, Texas: Gulf Publishing
Company.
ACKNOWLEDGEMENTS
The authors wish to thank the staff of Engineering Dynamics
Inc., especially Charles Hill and Mark Broom for their assistance
in preparing this paper.
PROCEEDINGS OF THE TWENTY-SECOND INTERNATIONAL PUMP USERS SYMPOSIUM 2005 57