Professional Documents
Culture Documents
Summary
The occurrence of electric submersible pump (ESP) failures
caused by spinning diffusers is tolerably frequent in moderateoperating-cost circumstances. However, extremely high wellservicing expenses are associated with many ESP applications.
Examples include deep wells, arctic locations, offshore wells that
must be killed with high-density fluids, and seafloor completions.
In these or other high-cost circumstances, early pump failures of
any mode cannot be tolerated.
Longitudinal compressive force is imposed on the diffuser
stack of ESPs during assembly to prevent diffuser rotation. If this
is done improperly, the diffusers can spin because of torque transferred from the impellers, resulting in early pump failure. This
paper analyzes the mechanics of the spinning-diffuser failure mode
and demonstrates why some pumps with inadequate compressive
force can pass common pump tests but fail in this mode. Equations
are developed to calculate the restraining force as it changes under
varying conditions and the minimum value required to prevent
diffuser spinning. Testing procedures are proposed to emulate the
effects of well conditions conducive to diffuser spinning, thereby
detecting the defective pumps currently being missed. Practical
examples are included that illustrate the utility of the concepts
presented herein.
Introduction
Longitudinal compression is imposed on the diffuser stack of ESPs
during assembly to prevent diffuser rotation. If this is done improperly, the diffusers can spin because of transferred impeller
torque, resulting in an impeded flow condition and early pump
failure. Examination of pumps that have failed in this mode may
disclose evidence of high temperature and diffusers with circumferential wear and/or shortening because of wear caused by relative movement at the contact with adjacent diffusers. This mode of
failure begins at or near the pump top and progresses downward,
as wear further loosens the diffuser stack and the head, developed
by upper stages, deteriorates.
Some pumps that pass API performance specifications during
common testing procedures subsequently fail because of diffuser
spinning. A possible explanation for a portion of these occurrences
is that an otherwise strong pump might meet test standards in spite
of an incipient diffuser-spin condition. It should be pointed out that
a spinning diffuser is audible when horizontal testing is employed
and would not go unnoticed. Most cases of tested pumps failing in
the subject mode are probably the result of service conditions.
The assumption of floating impeller pumps is made throughout
this paper. However, the effects described here are equally applicable to pumps with a fixed impeller design.
Restraining the Diffuser Stack
Impellers transfer torque to diffusers by three means: the mechanical drag of thrust-washer contact, which increases with increasing
impeller thrust; disk friction and other hydraulic drag; and the
dynamic portion of the head generated by the impeller. Torque of
a single impeller is defined in Eq. 1.
62
I = 5,252PI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (1)
At shutoff, it is known that generated head is 12 dynamic and 12
static.1 A reasonable estimate of maximum transferable torque
would be one-half that calculated with Eq. 1, with PI at maximum
value. This assumption is made herein, although experimental data
would be more desirable.
Diffuser rotation is prevented by frictional contact between
stages and with the pump head and base. The contribution of
lateral contact with the housing is neglected here. The static torque
capacity of a diffuser/diffuser contact (or a diffuser/head or diffuser/base contact) is expressed in Eq. 2. The minus sign in this
equation reflects the fact that FN is a compressive force and, therefore, negative.
from well conditions conducive to diffuser spinning could be emulated. Alternatively, this result could be achieved with a device that
would grip the pump near the top, provide only lateral support
elsewhere, and provide either an adjustable pressurized pump suction or variable testing speed. Such extraordinary means might be
justified in high-cost situations, such as arctic locations, offshore
wells that must be killed with high-density fluids, and seafloor
completions. The utility of a test apparatus such as that described
previously is demonstrated in the following examples.
In Example 3, all conditions and equipment are identical to
those of Example 1 except that the well is in an arctic location
where the formation temperature is 63F, as has been reported,2
and a longer compression ring was installed during assembly, resulting in t8.7246104. At this value of t, the pump would
pass both the test-well and horizontal tests of Example 1. A horizontal pump test was performed with the full housing free to
elongate but without application of external force, resulting in
DR26.5105 and FDR2975 lbf. Because 975>964, the diffusers should not have spun during this test. Service conditions
resulted in DR21.3620105 and FDR2205 lbf, assuring
diffuser spinning. This faulty pump would not have been placed in
service if a tensile force had been applied to the housing during
testing that emulated well conditions.
Eq. 6 describes DR2 for a pump with an external tensile force
(FHE) applied to the housing. It was derived from Eq. A-10 by
adding a strain (HE) caused by FHE to the sum of H.
DR2 = AHEH AHEH + ADEDHP + HW + HT
+ HE DP DW DT t. . . . . . . . . . . . . . . (6)
A pump test now can be designed to emulate any set of well
conditions more conducive to diffuser spinning than the preceding
relaxed-housing test. A value of HE is calculated that will
result in HD during testing being equal to what will occur
in service, resulting in the same value of DR2. With this value of
HE, the required external force can be calculated using Eq. 7
(Hooks law).
FHE = HE AH EH. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (7)
The HD term in Example 3 became 8.5648104
when the pump was put in service. It was 7.96244 during the
relaxed-housing test, which is 6.0243105 less. The external force
required to emulate service conditions is calculated to be 5,234 lbf
by substituting this last quantity into Eq. 7. Emulation of the inservice value of DR2 could have been similarly achieved with a
suction pressure (pE) of 830 psi. This value was calculated iteratively with Eqs. 8 and 9. Eqs. 8 and 9 were derived from Eqs. B-6
and B-11, respectively, by adding pE/ to all h terms and eliminating elements not pertinent to pump testing.
HP = di2hP1 + pE 4eHEHdi + eH
Hdi0.5hP1 + pE 2eHEH, . . . . . . . . . . . . . . . . . . (8)
DP = hUn + 1di2 dS2 8ADED 1
1
+ 1 KDB1hUn 1 2 + pE 3. . . . . . . . . . . . . . . (9)
To implement the variable-speed procedure, Eqs. 10 and 11
were derived from Eqs. B-6 and B-11, respectively, by multiplying
all head terms by a speed factor (fS) and eliminating elements not
pertinent to pump testing. The value of fS must be determined
iteratively such that the calculated value of DR2 equals that at
service conditions. The required test speed would be fS0.5T, or
3,631 rpm in this example. This procedure is also applicable to test
well testing and provides a slightly more severe test than the other
alternatives because the threshold value of |FDR2| is proportional to
fS, which would be 1.076964 in this case.
HP = di2hP1 fS 4eHEHdi + eH
HdihP1 fS 4eHEH, . . . . . . . . . . . . . . . . . . . . . . . . . . (10)
DP = hU fSn + 1di2 dS2 8ADED 1
1
Nomenclature
A
AD
AH
At
dD
di
dS
dt
DP
eH
E
ED
EH
fB
fS
F
FDR
FDR2
FHE
FHR
FN
g
h
hi
hP1
hP2
hPn
hS
hST
hU
KDB
KDT
KHT
KK
KS
L
m
n
n
pB
pC
pE
pI
pM
pS
pT
PI
qT
SHx
SHy
Sx
Sy
Sz
t
ti
tT
v
WG
WH1
WM
WP1
WP2
WPn
WS
WU
T
D
DP
DR
DT
DW
H
HP
HR
HT
HW
HE
D
DR1
DR2
H
HR1
t
D
H
C
D
I
i
References
1. Stepanoff, A.J.: Centrifugal and Axial Flow Pumps, second edition,
John Wiley & Sons Inc., New York City (1948) 53.
2. Andrew, J.H. and Augustine, B.G.: Initial Experience with ESPs on
the Alaskan North Slope, paper OTC 7062 presented at the 1992
Offshore Technology Conference, Houston, 47 May.
3. Singer, F.L.: Strength of Materials, Harper & Brothers, New York City
(1951) 17.
Eq. A-9 was derived from Eqs. A-1, A-2, and A-6.
DR1 = AHEHt AHEH + ADED. . . . . . . . . . . . . . . . . . . . . . (A-9)
Eq. A-10 was derived from Eqs. A-8 and A-9 by observing that
DR2=DR1+DR.
DR2 = AHEH AHEH + ADED
HP + HW + HT DP DW DT t.
. . . . . . . . . . . . . . . . . . . . . . . (A-10)
HT = TKHT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-8)
The average pressure-based longitudinal stress imposed on the
diffuser stack equals the product of hU(n+1)/2 times the annular
area between the housing and shaft divided by the effective elastic
area of the diffuser stack. Combining this and the relationship
SE yields Eq. B-9.
H = D. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-3)
DP = 1 pB kDB3 1. . . . . . . . . . . . . . . . . . . . . . . . . . (B-10)
66
Eq. B-11 expresses the total contraction caused by pressurebased forces and was derived by adding Eqs. B-9 and B-10 and
substituting [hU(n1)/2+hP2+. .+hPn+pC/] for pB.
DP = hU n + 1di2 dS2 8AD ED 1
1
hST + 3pI pS DP
DP.
+2hST0.5 DP pI pS 1.5
. . . . . . . . . . . . . . . . . . . . . . . . . (C-7)
Fdt = A h
tT
tT
ti
ti
2
ST t tT
+ pI pS DP dt. . . . (C-4)
Performing the previous integration and substituting the expression for ti shown in Eq. C-3 results in the following.
3mv = AttT
hST + 3pI pS DP
+2hST0.5 DP pI pS 1.5
. . . . . . . . . . . . (C-5)
67