New API Design Calculations for Sucker-rod Pumping Systems?

FRED D. GRIFFIN*

ABSTRACT The recently published APZ RP 11L: Recommended Practice for Design Calculations for Sucker-rod Pumping Systems takes into account many more of the well variables than had been previously considered in sizing pumping equipment. This paper compares the new API method with the more conventional, simplified method which has been used for so many years. Direct comparisons are made between maximum polished-rod load, minimum polished-rod load, and peak torque on the gear reducer. Both the API and the older method are then compared with measured results taken from 77 wells, covering a wide range of conditions. INTRODUCTION The new API recommended practice was first published in March 1967. It was the final result of a thorough and time-consuming effort that began in 1954. In that year, Sucker Rod Pumping Research, Inc. was created by a group of users and manufacturers of sucker-rod pumping equipment to make a study in depth of the many complex problems associated with artificiallift mechanisms. To perform the necessary work, Midwest Research Institute of Kansas City was retained. Midwest Research Institute, in the beginning, attempted to study sucker-rod pumping systems through use of a mechanical model designed to incorporate all of the important well variables. Although this approach did work, it was not entirely satisfactory. The mechanical model proved difficult t o use and control, and the data obtained could not be easily correlated. It did, however, provide a better understanding of the many complex variables encountered in a sucker-rod pumping system. Midwest then set up an electric analog which was able to produce a large amount of data that could be more easily correlated. Having completed the project, a final report was submitted t o all participating members of Sucker Rod Pumping Research as well as to the API.* D. Schropp (Phillips Petroleum Co., Santa Barbara, Calif.) made an outstanding contribution to the industry by condensing and re-plotting the mass of data contained in Midwest's final report into six families of curves. These were made available to API.
*Lufkin Foundry and Machine Co., Lufkin, Texas ?Presented at the spring meeting of the Rocky Mountain District. API Division of Production, May 1968.
I

The API Committee on Standardization of Production Equipment, at that time headed by Max Halderson, Phillips Petroleum Co., then created a task force to study the final report. Ed Pierce, Sunray DX Oil Co., was chairman of this task group, which included top engineering and production talent from many major users and manufacturers. It was hoped that, after studying Midwest's report, the task group would be in a position to recommend to the Productiqn Equipment Committee that an API recommended practice be adopted. The task force did an outstanding job in taking the mass of data from the report and presenting it in a more usable and convenient form. The curves prepared by Schropp, previously mentioned, form the basis of all calculations in APZ RP 1 I L .

SA portion

of this report is reproduced beginning on p. 232 of this edition of Drilling and Production Practice.

SCOPE OF API RP 111, In any sucker-rod pumping system, polished-rod loads and, consequently, counterbalance and peak torque are complex functions of many well variables. Some of these are : 1. Polished-rod stroke 2. Pumping speed 3. Physical characteristics of the rod string 4. Fluid-column load 5. Polished-rod acceleration pattern 6. Mechanical friction 7. Compressibility 8. Pump submergence 9. Dynamics of the sucker-rod string Neither conventional calculations nor API R P I l L consider items 6. or 7., because the magnitude of these values is almost always unknown. Further, the conventional method usually does not take into consideration pump submergence, although this is not always the case. The API calculation method does consider pump submergence (when it is known) and also one other very important variable not considered in the conventional method, viz., rod dynamics. This is the primary difference between the new API method and the conventional method of calculation. Sucker-rod string dynamics take into account: a, viscous damping; b, unit geometry; c, spring constant of rods and tubing; d . ratio of pumping speed to the natural frequency of the rod string; e, ratio of rod stretch to polished-rod stroke; f, variation of angular velocity of the cranks; g, motor slippage, if electrically driven; and h, system inertia.

NEW API DESIGN CALCULATIONS FOR SUCKER-ROD PUMPING SYSTEMS

221

In the API method some consideration is given to all these dynamic factors, except unit geometry and system inertia. The unit geometry for API RP 11L is assumed to be of the conventional type, i. e., the equalizer bearing is located directly over the gear-reducer crank shaft. Thus, the conventional crankbalanced and air-balanced units both would fall into this category. SIGNIFICANCE OF NON-DIMENSIONAL PARAMETERS Although this paper does not go into the highly technical aspects of API RP I l L , it is important to consider briefly two independent variables that appear throughout this publication. These appear as the abscissa in Fig. 1 through 6 herein (p. 225 through 230). 1. N/NO1, ratio of pumping speed to the natural frequency of the combination rod string. 2 . Fo/Skr, ratio of rod stretch to polished-rod stroke. These variables, or parameters, are the two most important in well-load behavior. The value of Not will usually be of the magnitude of 30 vibrations per minute or greater. Seldom will pumping speed exceed 20 spm. Therefore, this value will vary between 0 and 0.6. Because the pumping speed will never exceed the natural frequency of vibration of the rod string, the higher harmonics (2NOt,3No: 4NO1, etc.) will never be involved. The lower-order harmonics (1/2NO1, 1/ 3 ~ , ' , 1/4N0: etc.) will be in the operating range. It is convenient to think of N/NO1 as "pumping speed", even though it is, in reality, demensionless. Assume, for example, that a tapered rod string with a particular size pump has a natural frequency of vibration of 40 per minute. Then, when we say that N/N,' = 0.25, we are merely saying that the pumping speed is 10 spm and the rod string is operating at a 4th-order frequency, i. e., N/Nol = 0.25 = N/40 or N = 10 spm. Similarly, the dimensionless parameter, F,/Sk,, represents rod stretch. The "actual" rod stretch induced by the fluid load Fo is Folkr, where kr is the spring constant of the rod string. Thus (Fo/kr)/S is the rod stretch expressed as a fraction of the polished-rod stroke. A sucker-rod system having a value of Fo/Skr = 0.4 is one that has 40 percent of the polirhed-rod stroke taken up in rod stretch. When operating at very low speeds, where overtravel is nil, this means that the net stroke at the plunger is 60 percent of the polished-rod stroke. Once again, it is convenient to think of Fo/Skr as rod stretch although it, too, is dimensionless. We should not try to picture the value of Fo/Skr as a measure of the size of the pumping unit. For example, two units shown on Fig. 7 (p. 23 1) (report numbers 303 and 435) each have an Fo/Skr value of 0.269. The unit for report number 303 is a 74-in. stroke unit pumping from 4,300 ft within a 2-in. pump, 718- to 314-in. rods, and requires a 320 API reducer size.

The unit for report number 435 is a 120-in. stroke unit pumping from 9,425 ft with a 1-114-in. pump, 1-in. - 718-in. - 314-in. rods, and requires a 640 API gearreducer size. It is seen that one value of the non-dimensional parameter Fo/Skr can cover a whole series of pumping units, rod strings, and plunger sizes. Thus, thes,e two important non-dimensional parameters, N/No and Fo/SKr, allow us to correlate a whole group of pumping installations without having to consider an infinite number of cases on an individual basis. NOMENCLATURE In comparing the formulas for calculating peak .polished-rod load, minimum polished-rod load, and peak torque for both the conventional and API methods, the following nomenclature will apply. PPRL = peak polished-rod load, pounds MPRL = minimum polished-rod load, pounds PT = peak torque, inch-pounds Wra = weight of rod string in air, pounds Wrf = weight of rod string in fluid, pounds L = pump depth, feet S = polished-rod stroke, inches .. N = pumping speed, strokes per minute = full plunder area, square inches Ap = average area of rod string, square inches A, F = 0.433L(Ap) = total weight of fluid based on full plunger area, pounds = spring constant of the rod string, pounds per kr inch G = modifying factor for conventional units to correct for the deviation from simple harmonic motion-usually has a magnitude of about 1.05 for conventional crank-balanced units and a value of 1 for air-balanced units. GI =same as G , except that it applies to the upstroke on special geometry units only. Usually has a value of about 0.93. G2 =same as G , except that it applies to the downstroke of special geometry units only. Usually has a value of about 1.2. = peak polished-rod load less the weight of the Fl rod string in fluid, pounds F2 = weight of the rod string in fluid less the minimum polished-rod load, pounds T = crank torque without correction factors, inch-pounds T, = torque adjustment constant PEAK POLISHED-ROD LOAD The most widely used conventional method of predicting peak polished-rod load is shown in two forms in Equations ( l ) , (2), and (3) for conventional units. For those who prefer to express fluid load as a function of net plunger area: PPRL = 0.433L (Ap-A,) + Wra + Wra ( ~ ~ 2 / 7 0 , 5 0 (1 0) )

222

FRED D.GRIFFIN

Another approach, which gives identical results, defines fluid load as a function of full plunger area: PPRL = 0.433L ( A p ) + W r f + Wra ( s N ~ / ~ o , ~ o(2) o) PPRL = F o +Wrf+ Wra ( ~ ~ 2 / 7 0 , 5 0 0 ) For units with special geometry: (3)

PEAK TORQUE The conventional method for calculating peak torque for units with conventional geometry is:
PT
= [(PPRL) -

(fifPRL)]x S/2 x G

(7)

For units with special geometry: PT = [(PPR L x ( G I ) - (MPRL ) ] x (G2) .r S/2 The API method for calculating peak torque is: PT = ( 2 T / s ' ~ ~x ) S kr x Ta
(8)

For air-balanced units: PPR L = Fo + Wrf + 0.7 Wra ( ~ ~ 2 / 7 0 , 5 0 0 (3 ) b) By inspection, it is obvious that although the conventional method of predicting peak polished-rod load does consider the acceleration of the rod string, it does not take into account the harmonic effects of avibrating rod string. The API method for predicting peak polished-rod load is as follows: (4) PPR L = Wrf + [(Fl /Skr) (Skr)] The term (F1/Skr)is a non-dimensional parameter taken L plots (F1/Skr) from a curve in APZ RP ~ I which against (N/No) for a series of values of (Fo/Skr). These curves take into account the effect of rod-string harmonics, as well as the normal acceleration effects. The API method does not introduce any modifying factors to take into account units with special geometry.

(9)

The factor ( 2 ~ / ~ 2is k taken ~ ) from a curve in APZ RP IIL which plots ( 2 ~ / ~ 2against k ~ ) (N/No) for various values of (Fo/Skr). Ta is a torque adjustment factor. Both the conventional and API methods of calculating peak torque assume that the peak and minimum polished-rod loads occur at a crank position which results in an optimum mechanical advantage. Thus, the assumption is that the peak and minimum polished-rod loads for conventional units occur at the 75-deg and the 285-deg crank positions. As is commonly known, this is not always the case. When these loads d o come at positions on the crank cycle other than these assumed positions, then errors of considerable magnitude may result no matter which method of calculation is used. The assumption is also made that there is no fluid pound or gas interference. This is not the usual case, but it would be difficult to include these factors in any mathematical formulation. APZ RP 1IL does not include a peak-torque prediction for units with special geometry. Another assumption the API method makes is that the mechanical efficiency of the pumping unit is 100 percent. Some conventional methods of calculating peak torque make this same assumption, although at least one major manufacturer uses a mechanical efficiency of 93 percent. Values tabulated in this paper for the conventional method, however, assume 100 percent mechanical efficiency so they will be on a comparable basis with the API method. It could be argued at length as to just what the mechanical efficiency of a pumping unit - from the polished rod to the slow-speed gear - should be; however, it definitely is not 100 percent. An efficiency of 93 percent seems reasonable when comparing the mechanism of a pumping unit with similar types of machinery. It is recommended that the API consider a mechanical efficiency factor in future revisions of APZ RP I ? L . Still another assumption the API method, as well as the conventional method, makes in predicting peak torque is that the pumping unit is always perfectly

MINIMUM POLISHED-ROD LOAD The conventional formula for minimum polished-rod load for conventional geometry units is:
MPRL = Wrf
-

Wra ( ~ ~ 2 / 7 0 , 5 0 0 )

(5)

For units with special geometry:

For air-balanced units: MPRL

=

Wrf

1.3 Wra ( ~ ~ 2 / 7 0 , 5 0 0 ) (5b)

Here again the deceleration of the rod string is considered, but the dynamic effects are not. The API method for predicting minimum polished-rod load is:

The term (FzISk,) is a non-dimensional parameter taken from a curve in APZ RP I I L which plots (F2/Skr) against (N/No) for a series of values of (F0/Skr). These curves do consider the normal deceleration effects, plus the effects of rod harmonics.

NEW API DESIGN CALCULATIONS FOR SUCKER-ROD PUMPING SYSTEMS

counterbalanced. This is a very naive assumption, to say the least! Of the 77 wells listed in Fig. 7, the average &crease in actual torque on the gear reducer over what it would have been with perfect counterbalance was 19.7 percent. Some units were overloaded well over 50 percent with respect to gear-reducer torque because of an out-of-counterbalance condition. It is regrettably true that many operators do not appreciate fully the real importance of correct counterbalance as a most important influence on the torque imposed on a gear reducer. An out-of-counterbalance condition should always be considered, if we are to be realistic, in determining the size of the pumping-unit, reducer no matter which method of calculation is used. It is, therefore, recommended that a multiplying factor of 1.2 always be applied to the calculated required torque before the final selection of the pumping-unit reducer is made. This recommendation will not necessarily result in a larger .gear reducer than would have been selected previously; but it may do so, depending upon where the calculated torque, as modified, falls. COMPARISON OF PREDICTED AND MEASURED LOADS FOR 77 WELLS In comparing the API method with the conventional method of calculating peak polished-rod load, minimum polished-rod load, and peak torque, it is meaningful to compare both methods with measured results taken from dynamometer studies recorded over a period of several years. In these studies, the peak and minimum loads are taken directly from the dynamometer cards. The counterbalance effect is calculated from the reported position and size of the counterweights. The peak torque on the gear reducer in all cases had been calculated by API Std 11E using accurate torque factors for the polished-rod positions taken at every 15 deg of crank rotation. An effort was made in selecting the dynamometer studies, results of which are recorded in Fig. 7 (p. 231), to cover a wide range of conditions - considering production, depth, pumping speed, and pumping unit size. Dynamometer cards with obvious well abnormalities were not considered. Peak Polished-rod Loads Referring to Fig. 1 and 2 (p. 225 and 226), it is noted that strictly from a standpoint of averages, the API method for predicting peak polished-rod load is very accurate, predicting loads that average only 1.4 1 percent greater than measured. The conventional method predicts peak polished-rod loads that average only 3.43 percent less than measured. Although the "average" deviation for both methods is certainly accurate enough, not too .much can be said for the range of predicted loads. For the API method (excluding the very extreme deviations), 95 percent of

223

the wells fall in a range of from -22 percent to +32 percent. Similarly, the conventional method predicts loads that are measured to be -30 percent to +25 percent. It is disappointing in both methods that the range o f ' deviation of calculated from measured is so great. It should not be so surprising, perhaps, if we stop to consider that so many of the variables that affect well loads are either too complex for mathematical treatment or are variables that depend upon well data that is nearly always unavailable at the time of computation. Minimum Polished-rod Loads In the calculations for minimum polished-rod loads, results show that the API method is definitely superior to the conventional one, see Fig. 3 and 4 (p. 227 and 228). The API method made predictions that averaged only 28 1 lb above measured; whereas, the conventional method predicted minimum loads that averaged 1,299 lb above. Once again, the range of deviation in both methods was wider than would have been desired. Peak Torque Considering the fact that both methods, on an average basis, predict peak loads that deviate from measured only a small amount; and considering the fact that the API method was shown t o be considerably more accurate with respect t o minimum load, then we could expect the API method to be more accurate on a torque comparison. This proves to be the case. Referring to Fig. 5 and 6 (p. 229 and 230), the API method predicts torque values that average 7.26 percent above measured. The conventional method calculates torques that are 18.82 percent below measured results. Thus, the API method is more conservative than the older method of calculation. The wide range of deviation from measured in the torque calculations reflects, of course, the wide range of deviations in both the peak and minimum polished-rod loads discussed previously. DEGREE OF DIFFICULTY OF CALCULATIONS Because the API method takes into account many more variables than the conventional method, it is more difficult and time-consuming to make the necessary calculations. For one not too familiar with API RP I I L , or for one who uses it very infrequently, it will take at least three times as long to make the necessary calculations for designing a pumping-unit-sucker-rod system than does the old conventional way. Although it is more accurate, at least to some degree, the increased difficulty in the calculations will limit its universal use unless a computer program is written which will tabulate all the well loads, reducer torques, rod stresses, etc. within some set limits which have been found to be practical.

224

FRED D. GRIFFIN

I
For example, a computer program could be written around APZ RP 11L which would print out for, say, every 500 ft of depth and in production increments of, say, 100 BID, all conceivable pumping-unit-sucker-rod combinations that would satisfy all the limiting conditions of the program. Some of these conditions might be to limit pumping speed to a minimum of 2 or 3 spm and a maximum of perhaps 20 or 25 spm. Other limitations would have to be a limiting rod stress, a minimum figure for minimum load, etc. It is believed that the computer approach which would provide the user and the manufacturer with printed data in book form is the best course for the API to pursue in obtaining the maximum benefit from API RP 1IL. Some oil companies have done this with their own method of computation and have found that the data books are widely used and widely accepted by field personnel.

CONCLUSIONS AND RECOMMENDATIONS Use of APZ RP 11L makes it possible to: 1. Predict values for peak polished-rod load which are slightly more accurate than the more conventional methods of calculation. 2. Predict values for minimum load which are much more accurate than those determined by conventional methods.

3. Predict gear-reducer torques more accurately, primarily because minimum loads can be more accurately determined. APZ RP 11L makes some broad assumptions in predicting peak torque which can cause a considerable error in unit selection. These are: a. That the maximum and minimum loads occur at the 75-deg and 285-deg crank position, where the pumping-unit torque factors are optimum. Inasmuch as any calculation method for sizing equipment would almost always be made in the absence of a dynamometer card, this assumption seems valid. b. That pumping-unit efficiency is 100 percent. A more realistic value in the range of 90 to 95 percent should be introduced. c. That the pumping unit is perfectly counterbalanced. Years of experience have shown that this is unrealistic. Torque calculations for the gear reducer should include a multiplying factor of 1.2 (in addition to the efficiency factor mentioned in b.) to compensate for out-of-counterbalance condition. TO make API RP IIL more useful, a computer program should be written which will tabulate in book form many sucker-rod pumping systems for a given set of conditions, from which the user can make a selection based on his own judgment.

NEW API DESIGN CALCULATIONS FOR SUCKER-ROD PUMPING SYSTEMS

225

+m
a+soo

PEAK POLISHED ROD LOAD API METHOD

w4-30
*+20

a
a.10
ul

=
L

&-I0

t -20
P

'

f -30-40

. . . . - . . . . . ' . . . . . . . .
0 0

8 o0
0

a .0 . . 8

O o

ow

A .

CUD

I
Y

AW. # V U T l O N ~ l . 4 l %
OCQM~QLL u u n .UCQnmOYLQIQLmUn
k p ~

lslum

-7

+m
MO-

I
2

w-

I
A

w -

CONVENTIONAL METHOD

l60-

+a-

+so

I?+*f'"-oo = o

l,oL

.
0

oO
a :
0 0 0

0 0

ooO+

;-to;
'

" s o o0

; -3dL
-40-

.. . .. . ..
0

0 0

0
U E A s u R E o

8,

8 8

8 0 0

AVB.DNUTION~(-13.43%
O-COIVEIIIQ*Y u n l n WDQM)ITIOUL uome~nv

wnc

-0

MLP*CEOUUIT8

-?I

I
2

w*

I =
fig. 1

2 26

FRED D. GRIFFIN

+m +a

PEAK POLISHED ROD LOAD API METHOD

+to
480-

PEAK POLISHED ROD LOAD CONVENTIONAL METHOD

8+20

0
0
. . 0

e
W

.
00
A

$+lo-

=
b
I

0
-

B % -20 -

o
O

A 0

a 0 ee:.~ Q,
0

.
A

8
0

Oge0
0 0

0 0 O

0 0 *oo 00

h 2-

0
0

~h J

.4

Fig. 2

NEW API DESIGN CALCULATIONS FOR SUCKER-ROD PUMPING SYSTEMS

227

+moo

+--

MINIMUM POLISHED ROD LOAD API METHOD

. .
MINIMUM POLISHED ROD LOAD CONVENTIONAL METHOD

Fig. 3

228
t m

FRED D. GRIFFIN

MINIMUM POLISHED R O D -

+6ooo-

API METHOD

-?I I
tmo@

+am-

N/kd

I
,

MINIMUM POLISHED ROD LOAD CONVENTJONAL METHOD

Fig. 4

NEW API DESIGN CALCULATIONS FOR SUCKER-ROD PUMPING SYSTEMS

229

+to
+a0

0 0 0

PEAK TORQUE API METHOD

+w+40-

+30

0 0 06

r2"
Z+IO
0
n

O O 0

*
0

8
O O 0 OO
0

e-lqro

'
0 0
@

8
0

W4sunLD

8 c-20s >
0 0

08

g-30-

-40-

-80

A ~ ~ A T l W7.28lb W =
DcawlmaKUIIII
C m ~ D u m

-'9,
+70

we

I
.4

W.+#)

PEAK TORQUE CONVENTIONAL ME~HOD

0

w-

:eO
t30

=+,dl

3
8 a
2

= o
0

0,
V

"
YtlWlLD

t -20

00
0

o
O O

aA e
t 9
8

0 0

0 0 0

1-30

400

AVO. OEVIATlON=(-JM0%
*CQM*nmuU*D
CAIRDALUQOU1IL

-?I

wsr

I
.4

3 Fig. 5

230

FRED D.GRIFFIN

PEAKTORQUE API METHOD

+m

~Ftd

.4

a-

PEAK TORQUE CONVENTIONAL METHOD

N/N~

3 Fig. 6

.4

NEW API DESIGN CALCULATIONS FOR SUCKER-ROD PUMPING SYSTEMS

23 1

Fig. 7-Comparison of Predicted and Measured Loads for 77 Wells