Professional Documents
Culture Documents
Electric-Submersible-
Pump Failures
Marcus O. Durham, SPE, THEWAY Corp., and Jam.s H. Williams,
SPE, and Dwight J. Goldman, * SPE, Arco Oil & Gas Co.
Summary. Electric submersible Introduction done only on the outer surface. Because so
pumps (ESP's) historically have had ESP's have been used for more than 50 many stages are used and the impellers are
years, typically for moving large volumes not fixed to the shaft, there is considerable
short run lives. Their failures usually distance befween shaft supports.
of fluid where other means of artificial lift
were unexplained and accepted as have not been feasible. Recently, the appli- The pump is coupled to a seal-and-isola-
the norm. Because the equipment cation of these pumps has expanded. It is tion assembly. The pump shaft is coupled
not uncommon to use these pumps to move to the isolation-chamber shaft. Two major
has generally been downhole, find-
500 bbl [SO m 3 ] of fluid from less than functions of this chamber are to support any
ing all the influences that shorten thrust from the pump and to prevent well
2,000 ft [600 m]. The industry has made
equipment life expectancy has been great strides in improving ESP performance. fluid from entering the motor. Because of
difficult. Horizontal surface installa- Run times have improveOsigriificantly over the severe service, this assembly is prone
those existing only 10 years ago. These to failure, with catastrophic results.
tions for water injection have been Various attempts to build an adequate
pumps are now economically competitive
made where machine operating con- with other pumping systems in many envi- sealing assembly have met with less-than-
ditions could be monitored and ronments. 1 phenomenal success. The major problem is
A unique application of ESP technology preventing fluid migration around a moving
equipment performance analyzed.
has been in water-injection systems. Water- shaft while allowing for expansion of the
Unbalanced pump vibration has fluid in the assembly. The first efforts used
injection pumps often must move large
been identified as the cause of many volumes at high pressures. Not surprisingly, a grease-packed housing with the shaft run-
seal problems and ultimately motor ESP's fit this criterion. Although ESP's are ning through the grease. The next major de-
not a panacea for all pumping applications, signs used a manometer-type isolation
and pump failures. In fact, excessive section with a fluid heavier than water to
these examples demonstrate the applicability
vibration appears to be an inherent of ESP technology to an increasing number prevent water migration. In addition, a
mechanical characteristic of ESP's. of applications. Crane-type mechanical seal, usually made
of a ceramic face with a carbon runner, was
To illustrate the severity of this prob- Nevertheless, the significant problems that
still exist with the technology must be ad- placed around the shaft. Later efforts in-
lem, this paper presents data volved multiple manometer-type isolation
dressed before the equipment can achieve
gathered from field installations with adequately long run times for it to be widely sections without the heavy fluid. Current
monitoring equipment, from equip- applied. Current estimates are that ESP's are technology uses an elastomeric bag and one
used in only about 5 % of the lift applications or two manometer-type sections. All these
ment tear-down analysis, and from later designs continue to use the Crane-type
in the petroleum industry. This leaves con-
statistics on ESP failure modes. seal around the shaft. Occasionally, different
siderable room for further application.
materials are used for the seal surfaces, but
History and D.slgn the majority still use ceramic with carbon.
The isolation chamber is coupled to the
Before the problems with ESP's are ad- motor, and its shaft is coupled to the motor
dressed, the construction of the pumping shaft. The motor consists of multiple sec-
system should be analyzed. This analysis tions that are essentially small motors in the
will demonstrate the problems to be solved. same housing and on the same shaft. The
The design of today' s ESP's remains bas- motor-housing diameter is very small com-
ically unchanged from the first pumps. The pared with conventional induction motors.
pump assembly is a slender centrifugal pump Each motor section, with a 5-in. [12.7-cm]
of up to hundreds of stages. The stages con- diameter and IS-in. [46-cm] length, has a
sist of an impeller that "floats" on a shaft. rating of about 10 hp [7.5 kW]. Therefore,
The impeller is mated with a diffuser for 10 to 20 of these sections are often stacked
directing the flow between stages. The im- together to provide the required horsepower.
peller/diffuser assembly is stacked inside a The motor insulating fluid is open to the
small-diameter tubular. isolation chamber fluid. Any problems en-
The impellers are usually made of a ma- countered in the isolation chamber are even-
terial that has been sand cast. The vane tually transmitted directly to the motor and
openings are very small, so machining is will ultimately result in an electrical failure.
'Now with Areo Indonesia, Jakarta. Many motor failures can be attributed,
Copyright 1990 Society of Petroleum Engineers directly or indirectly, to a failure in
the isolation section. The original cause of composition, and temperature, while the ex- "Industry average [ESP]
failure may not be the isolation section, but ternal environment is much cleaner than a
it is eventually involved because it is the cou- wellbore.
run times have
pling and most vulnerable item. A study we conducted of the failure modes Improved from 300
experienced on horizontal centrifugal pumps days a few years ago
Failures in operation since 1982. This represents a
broad spectrum of fluid volumes and horse- to more than 600 days
ESP's historically have short run times.
Even discounting the infantile failures, run power ratings. The predominant recent fail- now. . . . But
ure modes are listed in Table 2. Failure data
times are still relatively short compared with
for six pumps were gathered during an 18-
unfortunately, in some
the life of a project. Industry average run
times have improved from 300 days a few month period; data for two additional pumps environments, system
years ago to more than 600 days now. Some were available over a 30-month period. The life expectancy is only
aggressive users in reasonable environments number of failures in each period is noted.
The remainder of this paper addresses the 30 days."
have seen average run times in excess of 4
years. The best-performing systems have ex- hypothesis that many of the isolation-cham-
tended run lives of 7 to 10 years. But un- ber, motor, and pump failures occur as a re-
fortunately, in some environments, system sult of pump vibration from sources other
life expectancy is only 30 days. than well fluids. If true, then the wrong
What are the differences in these perfor- problem has traditionally been solved. Cor-
mances? Obviously, some concern the en- recting, mitigating, or designing to tolerate
vironment, but it appears that there is a the vibration problem should dramatically
much more subtle problem. improve run times and reduce failures.
Jacobs 2 presented one of the better corre-
lations of ESP failure modes. Although Horizontal Installations
based on the very harsh environments of the The first surface configuration of an ESP
North Sea, the information compares statisti- was installed in May 1981. 3 This was a
cally to other experiences. Table 1 gives a 250-hp [186.5-kW] unit designed t() pump
regrouping of the failure modes of interest 4,400 BID at 2,250 psig [700 m 3 /d at 15.5
to this analysis. MPa]. The pump ran about 5 years, but the
From observing similar failure analyses, thrust chamber had to be replaced periodi-
we can expect that a significant part of the cally and the conventional induction motor
motor-fluid influx problems and many of the experienced several failures. The second
other motor problems result from problems surface installation was made in another field
in the isolation chamber. For example, if in Aug. 1981. Four thrust-chamber failures
fluid exists in both assemblies, the fluid experienced in the next 2 years were attribut-
probably migrated through the protector. ed to normal problems in the development
Similarly, a failure in the isolation-chamber of a new concept.
thrust system would transmit to the motor. The next installation was a replacement
In general, motors have relatively few prob- for a failed 600-hp [447-kW] positive-dis-
lems if the isolation system maintains its in- placement pump. This system had a 6OO-hp
tegrity. [447-kW] motor coupled to an 8,400-B/D
Because of similar statistical comparisons, [1335-m 3 /d] pump at 2,250 psig [15.5
the isolation system has received its share of MPa] pump. Catastrophic failures with this
blame for downhole pump system failures, installation were experienced from the be-
but most of the failures occur in the pumps. ginning. The thrust chamber was replaced
These failures have traditionally been at- 10 times in 3 years; the pump was replaced
tributed to well-fluid conditions like sand or three times and the motor was extensively
to other environmental factors. overhauled. Because of these problems, fu-
Because of the pump's location downhole, ture installations were restricted to 300 hp
gathering information about the operating [224 kW]. This was the first ESP where
performance and conditions of the pump has vibration obviously was a serious and meas-
been very difficult. With the advent of the urable problem.
horizontal pumps on the surface, additional Several other surface installations were
information that can be extrapolated to made in other locations, and other compa-
downhole conditions can be obtained. The nies also began using the technology.
significant feature is that the fluid through Although thrust chambers and motor bear-
the pump is controlled in terms of volume, ings failed after very short run times, the
economics continued to favor these pumps. vironment of the thrust and journal bearings.
One of the later major installations was a The loss of lubrication causes excessive
new waterflood plant where the only pumps bearing wear, which results in uncontrolled
were four surface ESP's operating at 300 shaft movement. Bearing failure from con-
hp [224 kW]. Other pumps have been in- tinued operation causes excessive vibration
stalled, but only the currently operating and motor or pump failure.
pumps are described in Table 3. The pres-
sures and volumes are system averages. Downhole-Pump Failure Analysis
The pumps typically are mounted horizon- When an ESP system fails after a short run
tally on an I-beam with various support con- time, a failure analysis is often performed.
figurations. The motor is a, conventional From participating in many of these tear-
NEMA-type surface motor. The motor shaft down analyses on pumps for all the manu-
is coupled to the thrust-chamber shaft with facturers, from different producing compa-
a metal spline or a flexible coupling. The nies, and in different environments, we have
earth for the Sims Sand Unit (SSU) pumps noted some common failure modes. The ef-
is very compacted from long use as a plant fects of excessive vibration can be observed
facility. The foundation at Northeast Fitts visually on the exterior of the equipment be-
(NEF) was originally somewhat soft, so fore it is opened. When a worn area on the
large concrete supports were poured under housing of the ESP that matches the ribs on
the pumps. The foundation at the Humphrey the armor of the cable is seen, the isolation
"The operating Sand Unit (HSU) was gravel fill typical of chamber usually has failed and the motor has
performance of each new plant locations. experienced a failure.
Observing a problem is only the first step.
pump has been Horizontal.Pump Failure Analysis The next obvious task is to try to determine
monitored since Because horizontal pumps are installed on the cause of the vibration and then to correct
installation. Two major the surface, their failures are much easier the problem. Many possible scenarios can
to investigate than conventional downhole be investigated. If the pump had excessive
failure modes were pumps, where the primary cause of the wear, well fluids containing sand are often
revealed: premature failure may be masked by other conditions. credited with causing the problem when
When a problem occurs, it can be evaluated nothing else is apparent. If the area has no
motor·bearing failure independently before other failures happen. sand or scale and no evidence of other con-
and frequent seal· The conventional air-cooled motors use ditions exists, all too often the failure report
assembly failure." ball bearings on either end of the shaft for would give several possibilities, would state
mechanical support. The bearings ride in a that the cause is unknown, or, worse, would
lubricant-packed race. The motor-bearing indicate that it is just a problem well.
failures on the horizontal pumps have been Experience gathered from surface installa-
caused by a loss of bearing lubricant in a tions and failure analyses of problems with
very short period oftime. This loss allowed these systems indicate that vibration and the
metal-to-metal contact between the race and resulting ESP failures are caused by im-
ball bearing. The friction precipitated a balance from the pump stages and from
tremendous amount of heat and excessive minor deflections of its shaft.
cyclic loading, which led to catastrophic
failure. Vibration is recognized as the cause Vibration
of this failure type. The excessive shaft Vibration is the oscillating motion of a
movement during vibration allows the loss machine from its position of rest. The gener-
of bearing lubrication. al model used for vibrational analysis is a
The thrust chamber uses a Crane-type me- second-order mass(M)-spring(B)-damper(K)
chanical seal to isolate fluids in the pump system (Fig. 1). The Laplace transform
housing from the thrust bearings. The most equation for this system is
common failure mechanism in the thrust
(Ms2 +Bs+K)L(s)=F(s),
chamber is a chip or crack in the ceramic
ring of the seal. The crack permits pumped where L represents displacement and F is
fluid to contaminate the lubricant-filled en- the total force on the system.
~ "- _--I rs
r;;~cu"
6.00
:3 :: [\1\ ~I 'HOWN .RE FOR
'F"lLTERED READINGS TAKEN
}
SSU2 6.5 3.0
ts4,-
4.00 ~---
ON THE MACHINE STRUCTURE;,
- SSU5 >10 1.2
~ ~'r-.,
_O_R--.!'-.!.~~~_~_:_._r--
3.00
~ _~r--
-
""
HSU2 >10 5.0
2.00 - 1'-1\ --- ---
HSU3* 6.0
,
5.0
--~
,- "
1.00
-Pn NEF1 >10 1.0
--
~~ ~ ~
f-I)~
~ --=r-.:
1',11 - 'l~ ~-=-
0.8u I - - ~'h- NEF2 >10 1.0
-- -
~
0.60
~
-
"s; r-~~
~
- ~ ;'d'~
~i)li'f ~
-- --
• After foundation reinforcement.
~
0.40 - -
0.30 1 -
~1: i'\ were conducted on the horizontal pumps and
r\. 1-" "
ts
", 61
0.20 r-
~ ',t ~"
~~ ~"
-
-.- their motors to determine the source and lev-
el of the vibration.
0.10
0.08
0.06
1-
~
'"
-~~ ~ ~I----=-
~~
4~ ~ -1' " -
I"-
~
~
~
1-
~
~
--;r
<'"c:
~
~
First, the motors were run with the pump
disconnected. The vibration amplitUde and
frequency were measured for the motor shaft
with a mechanical analyzer and a connect-
"I
0.04 ~ d'S
/,j.
ed shaft stick. The amplitude of vibration
0.03
1-'q ~I-- I'... was no greater than 0.25 mil [0.06 I!m] for
"I~
0.02
1-1 I' -~ --
~~ any ofthe motors. Because this is well with-
~
in the acceptable vibration tolerances, we
0.01
I I
1--1
I
~j concluded that the motors were not the
source of vibration.
~ -~}
0.008
:S\
-~
Second, the entire pump assembly, with
I 1 -
J
0.006 -- --- ~
~~
0.004 -
.r<'"c: termine the system's natural frequency of
H-I 1 - ~I ~~ vibration. With the analyzer connected to the
0.003 -
0.002 -
0.001
- -~ - - --
I-I f- ~I
II I
I
- - ---- --
1t# -~~
pump, an impulse (blow from a hammer)
was applied to the pump housing while the
pump was not running. The unit vibrated
freely at or near 3,600 rev/min for more
~
0 0 0
0 0 0
N ~ ~
M'
than 5 seconds. This indicates that resonance
was occurring at operating speed.
Third, the analyzer was used to measure
motor and pump-shaft vibration while the
Fig. 2-General machinery vibration severity chart per IRD Mechanalysis. 5 machine was running. The results, measured
with a shaft stick, are given in Table 4. The
If a forcing function is applied to the mass, tudes less than 0.4 mil [0.1 JLm]. Rough op- abnormally high pump-shaft vibration, cou-
it will move, Without adequate damping, the eration begins above 0.8 mil [0.2 I!m], and pled with the first test, indicates that the
system will experience sinusoidal oscilla- impending failure above 3 mils [0.76I!m]. pump shaft was the source of vibration.
tions or vibration. The undamped natural Exciting forces that induce vibration have Several modifications were made to elim-
frequency of vibration is determined by the numerous causes. The more common ones inate resonance from the system. A concrete
ratio of the spring component to the mass are manufacturing defects like bent shafts,
component: foundation was poured beneath each pump.
bad bearings, and unbalanced rotating parts; The steel C-face coupling that connected
Fn = (KIM) 0.5 /h. installation problems like misalignment; and
each motor housing with each pump housing
external effects like electromagnetic fields
Resonance occurs when the frequency re- was removed. A rigid spline-type coupling
and hydraulic forces. When these exciting
sulting from the forcing function is the same forces are imposed on a shaft, the effect is between the pump and motor shafts was re-
as the natural frequency. The displacement, transmitted down the shaft to metal-to-metal placed with a flexible coupling. These modi-
L(s), becomes very large at this frequency. contacts, such as bearings and seal assem- fications changed the spring constant and
If the amplitude is too great, catastrophic blies. The incremental-force amplitudes are damping coefficient of the pump assembly.
failure will occur. Therefore, operating the then added to the natural-frequency ampli- This also totally isolated the electric motor
equipment near its natural frequency should tudes of the machine. As the peak ampli- from the pump. Motor run times immedi-
be avoided. Alternatively, adequate damp- tudes exceed the cyclic fatigue design ratings ately increased from 2 months to more than
ing can be applied or the design must be of any of the machine components, the 1 year. However, this is still very short for
modified to change the mass or the spring machine fails more quickly. motors that often run longer than 20 years.
constant. We have demonstrated that the pump shaft
A guide accepted by the electric motor in- Field Data of ESP's causes pump vibration. Only the
dustry for vibrational evaluation is shown motor life has been improved by isolating
in Fig. 2. Because a conventional motor was The operating performance of each pump
has been monitored since installation. Two the motors from the pumps and by damping
used for the horizontal application, this the resonance. The pump-shaft vibration still
guide is appropriate. Similar results would major failure modes were revealed: prema-
ture motor-bearing failure and frequent seal- exists.
have been achieved with other standards. 6
Fig. 2 correlates the effects of vibrational assembly failure. Table 2 summarizes these
failures for each pump during its period of Correlation of Horizontal and
displacements at various machine operating
speeds. For a 3,600-rev/min machine, such observation. Analyses of the motor bearings Downhole Pumps
as an ESP, this chart indicates that good per- and the seal assembly indicate that vibration Horizontal and downhole centrifugal pumps
formance exists with displacement ampli- caused these failures. Three different tests have very similar failure modes. Although