SPE

SPE 15367

Real~ime Treating Pressure Analysis Aided by h!ew Correlation

by D.L. Lord and J .M. McGowen, Hallibution Services
SPE Members

Copyright 1966, society of Petroleum Engineers

This psper was prepared for presentation at the 61st Arnual Technical Conference and Exhibitionof the society c! Petroleum Engineers held in New
Orleans, LA October 5-8, 1966.

This paper waa selected for presentation by an SPE Program Committee following review o’ information contained in an abatract submittedby the
author(s). Contents of the paper, as presented, have not been reviewed by the Society of Petroleum Engineers and are aubjact to CO,t ?ctionby the
author(s).The material, as presentad, #sea not necessarily reflect any positionof the Society of Petroleum Engineers, its officers, or mef ‘Jers. Papers
preeentsd at SPE mwtings are subj~t to publication reviaw by Editorial Committees of *he Society of Petroleum Engineers. Perm:seionto copy is
restricted to an abstract of not more than 300 words. illustrations may not be copied. The abstract should contain conspicuousacknowledgmentof
where and by whom the paper is presented. Write PublicationsManager, SPE, P.O. Sox 833S36, Richardson, TX 760634636. Telex, 73C989 SPEDAL.

ABSTRACT INTRODUCTION

A scale-up correlation derived from laboratory Using an on-site computer system to perform
experime?.tationak calibrated with field tubing real-time fracturingpressure analysis of the type
data provides improved real-time estimatea of proposed by Nolte and Smith requires reliable
bottomhole treating pressure (BhTP) during estimetea OY measurements of bottomhole treating
displacementof delayed-crosslinkHPG fluids with or pressure.1“ Analysis of treating pressure using
without proppant. Current real-time hydraulic this technique requires that a logarithmicplot of
fracturing pressure analysis requires knowledge of net pressure versus time be prepared. Net
BHTP throughout treatments. Measurement of this fracturingpressure ia defined as follovs:
pressure indirectly at the surface through a live
annulus or directly by a downhole gauge is not Pn = BHTP-pc. . . . . . . . . . . . .(1)
always feasible, practical, or economically
justifiab?.e.In these situations,this pressure is where:
estimated by adjusting meaaured surface treating
pressure to downhole conditions. = net fracturingpressure, pai [kPal
BH% = bottomhole treatingpressure,psi [kPa]
Hydrostatichead is added and friction pressure P= = fracture closure pressure, psi [kPal
losses subtracted from surtace treating pressure to
adjust to downhole conditions. Friction pressure Unfortunately,indirectmeasurementof BHTP at the
loaa estimates typically depend on laboratory,yard surface through a static annulus/tubingfluid col-
teat, or field annular fiow based, semi-empirical umn, or a direct measurement by a downhole gauge
correlations. Correlationsdeveloped from data col- is not always feasible, practical, andlor
lected by these methods are shown not to be economically justifiable. In the majority of
applicable to friction loss predictions for flow of cases, it is necessary to estimate BHTp by
these particular fluids down lor.gvertical tubing adjusting measured surface pressure to bottomhole
atrings. Verificationof thts point is presented in condjtionawith the followtngequation:
the form of comparison of directly measured values
to values predicted by correlationsbased on these BHTP=pw+ph-Ppf-Pf . . . . . ...(2)
methods presented in this paper and in the
literature. Hydrostatic head, p , is added and both fracture
entry, and tkeating strings Pf) friction
A laboratory correlation developed in this pressurep%;a valuea are subtracted from measured
paper waa found to possess some unique wellhead (surface)injectionpressure,PW* to ~ke
characteristics which lend themselves readily to an adjustment to bottomhole conditions.
field calibration with actual tubing friction loss Hydrostatic head determinationis straightforward
data. Field calibration, combined with a capacity and requires only knowledge of formationdepth and
to consider various tubing diameters, flow rates, tubular fluid density. Fzac:ure antry pressure
and gel/proppant concentrationscreates a powerful leas can be determinedfrom an ins~mtaneous shut-
tool for estimating tub%ng friction pressure loss in pressure obtained while pumping & flutd having
and hence BHTP when Injecttng these fluids on a vior. Typically,
a well defined friction leas bctr.
real-timebaais. a shut-downduring injectic~OZ 3 nonproppant-
 2 REAL-TIME TREATING PRESSUREANALYSIS AIDED BY NEW CORRELATION SPE 15367
I
iaden pad fluid serves this purpose. The d!:ference tas~ without application of compu~,arized
between BHTP determined just prior to shut-in by stat stical methods such as those provided by
1! For example, the SAS stepwise ?inear
~.ccountingfor treating string friction loss~ and Sf$.s,
I,HTP determined from the instanta~.eouashut-in regression procedure implemen~s a multiple iinear
pressure is equal to fracture entry fricticm loss. regressionanalysis according to the model,
The value thus obtaine+ is often erron ously assumed
5
to be constant throughouta treatment. Y= ~o+~~x1+B2x2+---
+Bmxm, . . . ● . . . . .(3)
Treating string friction loss estimates depend
entirely upon laboratory, yard test, or field which relates the bet,ti.ior
of a dependentvariable
derived semi-empirical correlations. Laboratory Y to a lines- function of the set of independent
faciliti~aoften permit only short, small dia~eter, variablss x , x , - - -, xm. The key to achieving
horizontalor vertical test pipea to be used. Yard a satisfact~rycorrelation with the model lies in
tests typically u~e larger, field-size tubular the selection of appropriate, forms for both
goods, dependent and independent veriab?es, Variable
horfzont%y~;Y’ ;;~;;-d;;;ve~~;;;~;;ti;;;~ ;; fo?mls selected should exp101t any lineaz
the most part, have relied more on data coll~cted dependence of y on x , x , - - -, Xm. Whenever
~2
from injectiondown annulus rather than tubing. there are a large num er of potential independent
ti exception to this latter generalization is the variables, it is desirable to resort to filst
fiel -derived correlationpresentedby Cooper et principles to provide a reduction in number and to
9 s,~ekout any potential linear dependencies.
al., which is based on limited data collected from
injecti~n down open-ended 2 7/8-h. [7.3 cm]
tubing. AS a firat step in the current analyais,
available data were cast in a form aug$~sted by
It ia shown in this paper that delayed Savins and Seyer for scale-up purposes. Their
crosslink fluid fri~cion pressure correlations method consists of plotting drag ratio versus
developed from laboratorymeasurements,yard tests, frictton velocity for the base fluid (water) on
or field annular flow situations may not be semi-logarithmiccoordinates. Drag ratio, o, is
applicable to friction loss predictionfor flow down defined as,
long vertical tubing strings. In particular, we
show that a correlationof laboratory friction loss u= ApG,p/Apo. . . . . . . . . . . . .(4)
measurements made with fracturing gels with or
without proppant overpredictsfriction loss of these where APG PIAPO are frictional pressuze drop
fluids when they are injected down field tubing values forsgel or gel-proppant slurry and water,
strings. A hypothesis is proposed, but not proven, respectively. Each of these values correspond to
in an attempt to explain why this lab correlation, equivalent pipe diameter, length, and flow rate
as well as other correlationsbased on the methods conditions. Friction velocity for watvr, v-~, is
.-. u
discussedearlier, do not provide adequate predic- aerlneo
“ aa follows:
tions. The proposed lab correlation was found to
possess a unique characteristicwhich lends itself VO* = (To/Po)? . . . . . . . . . . . .(5)
readily to field calibration. Field calibration,
combinedwith a capacity to consider various tubing wherez
diameters, flow ratess and gellproppant
concentrations, creates a powerful tOGl for frictionvelocity for water, ft/sec2[m/sl
estimating tubing friction pressure loss and, ‘?: : wall shear stress for water, lbf/ft [kPa]
subsequently,pn from Eqs. 1 and 2. . 3dAp /L
d . pipe”internaldiameter, in. [cm]
Ap = pressure drop for water, psi [kPal
LABORATORYDATA CORRELATION E= length, ft [m]
P. = water density, 62.4 lbm/ft3 [1000 kg/m3]
The authors chose to use data presented
previously by Shah and Lee to develop a new, A Blasius-typ~l equation describing turbulent
laboratory-based correlation that w uld be more Newtonian flow,
%
amenable to real-time application. The data
comprise pressure drop/flow rate measurements made f= 0.046 (NRe)-002,. . . ● . . . .(6)
in three different vertical pipe sections (nominal
3/4, 1, and 1 l/4-in. diameter [1.9, 2.54, and 3.18 where:
cm]) using water gelled with hydroxypropyl gu~r
(HPG) and containing O to 8 lb/gal [0 to 959 kg/m ] f . Fanning fr~ction factor
20/40 mesh sand. Although our desire was to model . 6dApo/LpV
friction loss behavior of a temperature-delayed N= Reynolds”number
Re .
croaslinked HPG fluid, we felt comfortable using 124 dvp IV
data collected with an uncrosslinked fluid since v= average”fl~idvelocity, ft/sec [m/s]
‘previous lab experience had shown no observable .
11 = water viscosity, cp [pa*sl
difference in friction loss for the two fluids at dYApo,L,Po = as previouslydefined in Q. 5
ambient conditions.
was used as the source of Apo values for use in
The large number of independent variables and ~s. 4and 5.
the broad range investigated for each constitute a
large data base in excess of 1000 experimental Representativeplots of log o versus VO* are
observations. Because of the vast quantity of data shown in Figs. 1 through 3. All the data, with
collected,analysiswould have been a fermidable exception of some of the higher prouDant concen-
SPE 15367 DAVID L. LORD and JAMES M. MCGOWEN 3

trations, showed diameter invariant behavior when independentvariabl. forms to describe dependence
}~lotuedin this manner. Exceptions noted at the of . on v, d, L, G, and E’!
highest proppant concentraticilsare attributed to
data scatter (or inability to acquire reliable in (1/0) = in al + a2 (l/v) + a3 ‘G{~~ . .(8)
laboratory data at high proppant concentrations) +a41n G+a5Pe
because there is not a uniform ~~ariationwith pipe
diameter. We were surprised that proppant-laden where:
fluid data produced diameter invariant plota since
there was no precedent set for this behavior in the constantsarising from
al’a2’---’a5 =
literature. In fact, the apparent universality of the individualcurve-fit
the correlation car.sed some suspicion. Kenney operations.
reports that a spuriGua self-correlationcould arise
wheli a parameter A (a sum, ratio, or product) is By comparison, we find that Eq. 8 has the same
correlatedwith a second parameter B (a sum, ratio, form as Eq. 3, therefore, we have defined an
pro3uct, or a single variable) ~d when both appropriate dependent variable, In (l/u), f7g
parameters have a common element. He further independent variables l/v, G/v, in G, and ?e
states that logarithmic transformations tend to for application in SAS linear repression
enhance apurioaa self-correlation of ratioa and procedures. A linear regression analysis using
products. The Savinti-Se: r correlation haa all these variables with 1049 experimental
these features. With it, we are att~mpting to observations produced the following cor elation
correlate a logarithmic transformationof a ratio, having a coefficient of determination,Rf , equal
with v * which has the common element to 0.981:
% P’@o;s ~
Ap ! %est for possible spurious
se?fcorrelation, Ap was eliminated as a common 1.1(l/u) = 2.1505 - 8.024/v - 0.23651~&v-
element by correlating T directly with T . 0.1639 In G - 0.05266 Pe . (9)
Figures 4-6 ahow an excelf“ ent diame~er invaria~t
correlation. Although these results demonstrate where:
absence of a spurious self-correlation,appltcatian
Of T versus T correlations at higher proppant u = drag ratio defined in Eq. 4,
conceGdLprationa(F~g. 6) did show that the equivalent dimensionless
log u versus v * plot (Fig. 3) does tend to mask v= average velocity, ftisec [m/s]
some data scatt&. G= HPG gelling agent concenr:r~::ion,
lb/Mgal [kg/ms]
At this point, our experience with the Melton P = proppant concentration,
and Malone modification of the Bowen scale-up lb/gal [kg/ms]
procedure led us to try c~~;~~ting log o versus
average fluid veloctty, v. Figures 7-9 ahow
diameter Invariant correlattonawere also obtained APPLICATIONMETHOD
by this method. Since this latter method appeared
to offer greater simplicity in application, it was Dependentvariable in (1/0) obtained from Eq.
selected as the beginning point for developmentof a 9 can be used to predict friction loss behavior
SAS-based data correlation. The hyperbolic over a wide range of treating conditions. From a
appearance of these data plots suggested that knowledgeof these conditions,various independent
reciprocal relationshipssuch as I/o and l/v could variables required for solution of Eq. 9 are
possibly be used to provide data linearization. calculated. The first of these is:
When these two parameters were used in a
semi-logarithmiccurve-fit of data correspondingto v= 17.156Q/d2 . . . . . . . . . (10)
velocities greater than 10 ft/aec (mostlytransition where:
and turbulent flow data), they were found to produce v= average fluid velocity,
the desired linear relationships: ft/sec [m/s]
:= flow rate, bblfmin [m3/min]
ln(l/u) = lnbGp+mGp(l/v) . . . . . (7) = tubing internaldiameter, in. [cm]
s *
where: Second through fourth independent variables
in Eq. 9 each contain HPG concentration, G. A
lnbGp = interceptvalue obtained for data nominal value based on the chemical quantity
s with gel concen+rationa,G, and mixed may be used in absence of the more desirable
proppant concentration,P. on-site viscometer or rheology ~4100p for
= slope value obtained for data with determining effective concentration. With G
‘G,P
gel concentration,G, and proppant defined, the second dependentvariable becomes:
concentration,P.
G/v = 0.0583Gd2/Q . . . . . . . . .(11)
Slope values were found to be independent of
proppant cone.mtrationand could be correlatedwith where:
gel concentrationonly. Interceptvalues were found
to be dependent on both gel and proppant G = gelling agent3concentration,
concentration and required a number of curve-fit lb/Mgal [kg/m ]
steps to properly define their behavior. Results d = tubing internaldiame er, in. [cm]
5
from all individual curve-fit steps were combined Q = flow rate, bbl/min [m /orlnl
into one equation by substltutton into Eq. 7 to
provide an Indicationof appropriatedependentand The thtrd independentvariable is simply the
L REA1.-TINE
..—.— —- NEW
.—-— TREATING PRESSURR ANALYSIS AIDED BY -.—.. CORRE1.ATTON
---——-------- ------- .
SPl??5’+67

natural logarithmof gel~ing agent concentration,G, B). BHTP and, hence, p determined from an
expressed as lb/Mgal [kg/ms]. The fourth instantaneous shut-in presnsure (ISIP) following
independent variable is a product of proppant flush la independentof friction pressure loss and
concentration, P, expressed as lb/gal [kg/ms] and depends only on hydrostatic head exerted by a
the exponmtial of reciprocal gel concentration,G, static fluid column.
expressed as lblkfgal[kg/msl.
The magnttude of pressure rise at Point B in
Once values for all four independentvariables each figure is a measure of friction loss
have been determined by the preceding methods, they overprediction for the pr,)ppantatage just prior
are substituted ir.toEq. 9 to determine in (1/u). to flush. Point C on Fig. 10 represents a short
Drag ratio, o, can be obtained from in (1/aj aa time period In which no proppant was being pumped
fellows: and provtdes yet another indication of
underestimationof p occurring when proppant is
~,e in (l/a)
a= . . . . . . . . .(12) present. Figure 12 is a similar plot for
treatment down 11,000 ft [3,353 m] of 3.~2-in.
Frictional pressure loss for gel with or without :9.96 cm] ID casing at 12 bbl/min [1.91 m /rein]
>roppantcan be obtained as follows: with3proppant concentration to 15 lb/gal [1,797
kg/m ]. This plot shows similar trends at Pointa
= u“Apo . . . . . . . . . . . .(13) A and B but with less severity because overall
‘PG,P
friction loss is much lower in larger diameter
Equation 6 is ●he source for water frictional casing. Also shown on these plots and others to
pressure loss vf.lues,Ap , used in Eq. 13 and during follow are results from a field-calibrated
data correlat~on. Rearrangement of Eq. 6 and correlationto be discussed later.
substitution of water physical properties at 80”F
[27”C]yields: Data were also available from a few
treatments performed down tubing with a static
Apo = 0.40429 d-’’”8Q1”8L. . . . (14) annular Cclumn. TMa wellbore configuration
provided a means for obtaining an indirectmeasure
where: of actual BHTP and friction 10SS, Figures 13 and
14 demonstratean overpredictionof proppant-laden
d = tubing Internal dlame ers, in. [Cd fluid frictton loss which was expected based on
5
Q = flow rate, bbl/min [m /rein] our previous analyaia of logarithmic p versus
L = length, ft [m] time plots. Wnat was not expected was annapparent
overpredicttonof nonproppant-ladenfluid friction
It is not necessary that Eq. 14 predict wi:er loaa shown in these figures. Figure 13 shows
friction loss with extreme accuracy because it friction leas response to a claasical stepwise
servea only as a correlationvehicle and its effect increase in proppant concentration. Figure 14 is
cancels out when ApG p is calculated according to Interesting because it shows a situation similar
Eq. 13. 1 but not identical to that ahown in Fig. 10. In
Fig. 14, proppant is shown be g cut entirely
9
followir.ga 12 lb/gal [1,438kgfm ] stage ~nd then
COLLABORATIONWITH FIELD DATA being returned to a 10 lklgal [1,198 kg/m ] level
after a period of time. Prior to discussing
When the laboratory-based correlation method remedies taken to provide an improved correlation,
described in the preceding section was used with we wfah to report observations of others and
computer-collectedfield data to prepare logarithmic speculate concerning a potential source for
net pressure versus time plots for numerous discrepanciesnoted here.
treatment (>20), overprediction of friction
pressure gradient during proppant-ladenfluid stages
waa evident. Each of these treatment involved OTHER OBSERVATIONSAND CORRELATIONS
tnjection of a delayed crosslink HPG fluid
containing 20/40 ~and at concentrations up to 16 An earlier laboratory correlation baaed on
lb/gal [1917 kg/m ]. Treating conditt s for each more theoretical considerations than presented
%’
were typically 12 bbl/min [1.91 m /rein] down here, but Guaing the same lab data for ita
approximately9,000 to 11,000 ft [2,743 to 3,353 m] derivation, showed similar results when applied
of 2.441-in. [6.2 cm] ID tubing. Since these to each fracturing treatment in the preceding
treatments did not have a means for BHTP discussion. fiis earlier lab result was alao
measurement, it was necessary to analyze each plot shown to satisfactorily describe yard test and
to detect potential problems with BHTP estimatea. field annulus injection data. However, as
mentioned previously, tt is not only laboratory
Figures 10 and 11, whtch are typical plots correlations which have problems In predicting
prepared for these treatments, show evidence of field tubing friction losses, but yt.~dtest and
friction pressure overpredictton in two different field annular flow-baeedcorrelations,aswell.
ways. First, the dip in each log p versus log At
plot, which begine at point A (~oinciding with Swanson and Meeken reported extensfve
proppant start) and continues downward with friction leas measurement made during yard teats
increasing proppant concentration, is evidence of while pumping a croasltnkedderfvatizd wfr fluld
BHTF un~eresti~tion or friction loss conta ning up to 10 lb/gal [1,198 kglm ] 20/40
$
overprediction. Second, friction leas sand. A centrifugal pump was used to disp@ce
overpredictlon la evident from a large rise in pn these fluids at rates up to 35 bblfmin [5.56m
which occurs when each treatment ie flushed (Pol~t
SPE 15367 DAVID L. LORD and JAMES M. MCGOWEN 5
Iminl through 1.75 and 2.875-in. [4.45 and 7.30- woulci produce flows which would be considered
cm] ID tubinq laid out horizontally. They pseudo-homogeneous,i.e., having proppant uniformly
expressed their friction loss measurement as distributedthroughoutthe fluld. An exception may
psi/1,000 ft [kPa/304.8 m] and also in terms of occur with thin fluids and exceptionally long
the same drag ratio that we correlated earlier horizontal pipe runs where it is conceivablethat
based on lab data. Following these yard tests, they proppant stratificationalong the pipe bottom could
alao had an opportunity to use the same type of develop at low flow rates. However, these
fluid on two fracturing jobs having open-ended conditionsare not usually present In laboratoryand
tubing. This wellbors configurationpermitted them yard teat setups. Laboratoryvertical ptpe aetupa
to monitor BHTP and friction loss directly from are usually too short to permit developmentof a
surface pressure of the static annular column. They nonhomogeneous flow pattern. Proppant-laden
found a field tubing drag ratio value, 0,24, which fluids would also potentially behave differently
was smaller than values (0.27 to 0.32) obtained for in flow down annuli than tubing. Couplings
nonproppant-ladenfluids from yard tests. However, present at each tubing joint and a natural annulus
they did not report any differencebetween friction eccentricity WOUAU tend to disturb any
loss of gelled fluids with or without crosslinkerin preferential proppant segregation in the flow
either yard or field tests. They did discover fie~d stream.
tubing friction loss with a 10 lb/gal [1,198 kg/m )
slurry wz~ only one-third to one-half that expected It appeara ‘hat a long vertical tubing string
based o.t yard rest results. Hather than suggest offers the onl+~ substantial possibility for
that this difference could possibly be a result of development of a heterogeneous flow pattern that
some proppant effect, they credited it to additional could aubatantially influance friction pressure
shear experienced by fluid being pumped down long loss. What is the Iikeliinoodthat such a flow
tubing atrtngs In contrast to short pipe runs of a pattern could develop? Excellent arguments for
yard test. However, Cooper et al. later adopted cha existence of non-homogeneous flow of
these open-ended tubing results as being crosslfnked flui
representativeof fri~ion loss to be expected with been Published‘%. ‘i~~ literature,
~~ho~~eprc/Pant have
proppant-ladenfluids. evidence is reported that solid particles tend to
migrate away from the wall in laminar viscoelast?.~.
Walser reported tubing friction loss, based on fluid flow at low Reynolds ~qumber and low
field results with a teuerature-delayed crosslink~d volumetric solids concentration. Investigation
HPG fluid containing up to 7 lb/gal [839 kg/m ] of th$s phenomenonwas extended to higher Reynolds
20/40 proppa~:l~id not follow trends predicted by numbers and soltds concentrationsby McMehon and
Hannah et al. For exampl~, injectionof a fluid Parker, who used microwave Doppler techniques to
containing 7 lb/$al [839 kgim ] 20/40 proppant at 12 chara~~erizethe conditionof the pipe core during
bbl/min [1.91 n /rein]down approximately 10,000 ft flow. They found that increasing solids
[3,048m] of 1.995-in. [5.067 cm] ID tubing produced concentration and/or Reynolds number of a
only a 1.1 to 1.2-fold increase in friction leas Newtonian fluid (water) would cause the central
over a fluid without proppant. Hannah et al. would core of the t-lbeto bacome more consolidated in
have predicted a 1.4 to 1.5-fold increase over plug flow. Moreira and Felder, from their
nonproppant-laden flutd friction loss for these investigation of resxdence time distributions in
conditions. Can this difference be attributed to vertical lamfnar flow of paper pulp slurries,
the Hannah et al. correlationbeing developedwith a concluded that a model consisting of a central
conventionalcrosslinkedHPG fluid while Walser used core in plug flow circumscribed~~ a clear arinulua
a delayed crosslink fluid? Swanson et al. also used fluid beat described their data.
a conventionalciosslinkedHPG fluid and did not see
a proppant effect of thts magnitude. It is possible If these heterogeneousflow phenomenawere to
that this difference can be attributed to the Hannah develop during diaplacamentof crosslinked fluids
et al. correlationbeing developedand verified with (with or without proppant) down tubing, they could
an apparently heavy reliance on field annular flow account for the observationof lower than expected
data. friction loss. However, experimentalverification
of these or other phenomena responsible for
Another interesting observation reported by reduced friction loss in these situations is
Walser was a measurable reduction in friction loss outaide the scope of this paper.
when crosalinkerwas added to a 40 lb/Mgal base gel
prepad. As reported earl$er, our laboratoryresults
disclosed no measurable difference in friction loss FIELD CALIBIUTEDCORRELATION
of an HPG gelled fluid with or without delayed
croasltnker. Walser’s obaenation could explain our We recognized at this point that our current
lack of success describing delayed crosslink fluid laboratory data correlation, Eq. 9, was not
friction loss with a laboratorycorrelationbased on capable of describing field tubing friction loss
uncrosslinkedgel flow data. because a nc-i-homogeneousflow phenomenon is
present which was not modeled in the laboratory.
We also recognized that tt might be possible to
NON-HOMOGENEOUSFLOW adjust two of the ftve correlationcoafficientato
obtain a match to ffeld tubing data. The
Unsuccessful application of proppant-laden intercept coefficient,80 in Eq. 3 and equivalent
fluid friction pressure correlations based on to 2.1505 in Eq. 9, determines the overall
laboratory, yard test, and field annular flow magnitude of the correlation prediction.
measurements suggests physical phenomena are Therefore, thts coefficient can be adjusted to
occurring in tubing flow that cannot be modeled by create ~.new, lower baaelina correspondingto the
these methods. In all likelihood,these methods reduced friction loss observed for
.k ----- .-. ——.
REAT.-TTMI7 ----------
TRF.ATTNC -- —------
PRF.SSIIRI? ANALYSIS AIDED
-—---—-———
--————RY
——NEW
--—--CORRELATION
----—.——-——
—-. _-— 15367
SPE

nonproppant-laden,delayed crosslink HPG flcids in NOMENCLATURE

tubing flow. An implied assumption for making this
adjustment is that the functional dependence on symbols
gelling agent concentrationremains the same as that
determined previously. An intercept coefficient al, a2, - - -, a = empirical regression
5
value of 2.38 was found to best describe field cc~fficientsin Eq. 8
tubing friction loss for nonproppant-ladenfluid. b = interceptvalue from first
order linear regression
Using this new tntercept coefficient formed a BHTP = bottomhole treatingprsasure,
new baaeline correlation from which further psi [kPa]
adjustments could be made for ~roppant d = pipe internal diameter, in.
concentration. Only one independent variable is a [cm]
function of proppant concentration,therefore, its f = Fanning friction factor
coefficient.can be easily adjuated to obtain the G = gelling agent3concentration,
best match to field tubing friction loss of lb/Mgal [kg/m ]
proppant-laden fluid. An implied aaaumption for in = natural logarithm,baae e
making this second adjustmentis that the functional log = common logarithm,base 10
dependence on proppant concentration remains the L = pipe length, ft [m]
same aa previously determined, but at some reduced In= slope value from first o~der
level of activity. A correlation cueffictentvalue linear regreaaion
of 0.028 for the fourth independent variable was N = Reynolds numb~z
Re
found to best describe field tubing friction loss P = pressure, psi [kPa]
for proppant-laden fluid. With these two P = proppant concentration,
adjustments, the “field-calibrated”form of Eq. 9 lb/gal [kg/m ]
becomes: Q = flow rate, bbl/min [m3/min]
t= time, min
in (1/u) = 2.38 - 8.024/v - v= velocity, ft/sec [m/s]
0“2365 fj? ● ~15) V* = friction valocity,
- 0.1639 in G - 0.028 Pe
ft/sac [m/s]
Improvement in friction 10.ss prediction xl, X2, - - -, x = general independentvariables,
m
provided by Eq. 15 is shown by comparisonsgiven in Bq. 3
Figs. 13 and 14. Application of Eq. 15 using Y = general dependentvariable,
methods describedearlier also provided improved log Eq. 3
p versus log At plots as shown in Figs. 10 through
12. 8., 81, - - -s f9m = general regression
coefficients,Eq. 3
A = difference
CONCLUSIONS P - viscosity, Cp [y*s] J
P = density, lb Ift [kg/m ]
1. A statisticallybased model with four u= drag ratio ~eflned in Eq. 4,
independentvariableswas developed and found to dimensionless
represent 1049 laboratoryf$ow ratalpressure T= wall spar stress,
drop observationswith an R value of 0.981. lbf/ft [kPa]
These experimentalobservationscomprise
transitionand turbulent flow data from three
different pipes, HPG conce~trationaof 30 to 80 Subscripts
lb/Mgal [3.59 to 9.59 kglm 1, and sand
conc~ntrationsof O to 8 lb/gal [0 to 959 c = fracture closure
kg/m ]. f = pipe friction loss
G = gelling agent concentration
2. This correlation,along with other published h = hydrostatic
laboratory,yard teat, and field annular n= net
flowbasedcorrelations,is inadequate for o = base fluid (water)
predicting frictton loss of croaslinkedfluid pf = fracture entry friction 10ss
with or without proppant during their P = proppant concentration
displacementdown long tubtng strings.

3, Field calibrationof the laboratorydertved ACKNOWLEDGMENTS

correlationpresented in this paper provided
improved real-timelogarithmicplots of net The authors thank Halliburton Sarvices for
pressure versus time for analyais purposes. permission to prepare and present this paper.
Special thanks are due Mr. Steve Henry for his
4. Additionalwork needa to be done by the Industry assistance in application of the statistical
to determine the physical mechanf.amthat procedures dtscussed In this paper. We also wish
produces less than expected friction lees when to thank those customers and Halliburton Services
croaslinkedfracturinggels (with or without field personnslwho were responsiblefor supplying
proppant) are pumped down tubing. treatmentdata.
------- .
SPI?15367 DAVID L. LORD and JAMES M. MCGOWEN 7

REFERENCES 140 Shah, S, N., Lee, Y. N., and Jensen, D. G.:

“Frac TreatmentQuality Improvedwith Field
1. Nolte, K. G. and Smith, M. B.: Rheology Unit”, Oil & Gas J., (Feb. 4, 1985),
llInterpretationof Fracturingpressures”? 47-51.
JPT, (Sept. 1981), 1767-1775.
15. Walser, D. W.: “Field Study of a New
2. Conway, H. W., McGowen, J. H., High-TemperatureFracturingFluid in South
Gunderson,D. W., and King, D. G.: Texaa”, SPE 13813, presented at the 1985 SPE
?Ipredictionof Formation Response from ProductionOperations Sympoaj.um, Oklahoma City,
Fracture Pressure Behavior”, SpE 14263) (March 10-12, 1985).
presentedat the 60th Annual SPE Technical
Conferenceand Exhibition,Las Vegas, (Sept. 16. Nolte, K. G.: “Survey of BottomholePressure
22-25, 1985). Predictionfrom Surface Pressure While
Fracturing”,Final Report prepared for Gas
3. Crump, J. B. and Conway, M. W.: “Effect of Research Instituteunder P,O. P-4614-92220.
PerforationEntry Frictton on Bottomhole
Treating Analysis”, SPE 15474, presented at 17. Gauthfer, F. J., Goldsmith,H, L., and Mason,
the 61st Annual SPE Technical Conferenceand S. G.: “PaiticleMotiona in Non-Newtonian
Exhibition,New Orleans, (Oct. 5-8, 1986). Media. 11. Poiseville Flow”, Trans. Sot.
Rheol., 15, 297, (1971).
4. Shah, S. N. and Lee, Y. N.: “Friction pressure
of Proppant-LadenHydraulic FracturingFluids”, 18. McNahon, T. A. and Parker, 1..R.: “Particles
SPE 13836 to be published in SPE Prod.Eng., in Tube Flow at Moderate Reynolds Number”’,
(NoV. 1986). Trans. Sot. Rheol., 19, 445, (1975).

5. Swanson, G. S. and Maeken, R. B.: “An Analysis 19. Moreira, R. M. and Felder, R. M.: “Residence
of FracturingPressures in the South Belridge Time Distributionsof Paper Pulp Slurries in
and Loat Hills Field”, SpE 99359 presentedat Vertical Laminar Flow”, AIChE J., 25, 131,
the 1981 SPE CaliforniaRegional Meeting, (1979).
Bakersfield, (March 25-26, 1981).

6. Hannah, R. R., Barrington,L. J., and

Lance, L. C.: tr~e ~al.Time calculation‘f
Accurate BottomholeFracturingPressure from
Surface Measurement Using Measured Pressurea
aa a Baae”, SPE 12062, presented at the 58th
Annual SPE TechnicalConferenceand Exhibition,
San Francisco, (Oct. 5-8, 1983).

7. Cooper, G. D., Nelson, S. G., and

Schopper,M. D.: “ImprovingFracturingDesign
Through the Uae of an On-Site Computer System”,
SPE 12063, presented at the 58th Annual SPE
Technical Conferenceand Exhibition,San
Franciaco, (Oct. 5-8, 1983).

8. SAS is the registeredtrademark of SAS

Institute Inc., Cary, NC, USA.

9. Freund, R. J. and Littell, R. C.: “SAS for

Linear Modela”, SAS Series in Statistical
Application, SAS Institute Inc., Cary, NC*
(1981).

10. Savina, J. G. and Seyer, F. A.: “Drag

Reduction Scale-Up Criteria”, Phya. Fluids,
Vol. 20, No. 10, 78-84.

11. Bowen, Jr., R. L.: “DesigningTurbulent F1OW

Systems”,Chem. Eng., (July 24, 1961), 143-150.

12. Kenney, B. C.: ItBewareOf spd.ow

Self-Correlationa!”,Water Resources Research,
(Aug. 1982), 1041-1048.

13. Melton, L. L. and Malone, W. T.: “Fluid

Mechantcs Research and EngineeringApplication
in Non-NewtonianFluid Systems”, SPE Raprint
Series, Well Completions,Vol. 5S ~91.
 SF% 15367

1
❑
❑
u
a 0

0 0
0
0 •1

0
❑
0
u o

0
0%

>, 1,,, !,,

0

-Q ~
Ssaluo!suaw!p ‘f-)

-b 0
, c 0
1 0
(IY 0
0

IQ

m
o

i’

00
0“
.4

Ssquo!sual.1.up ‘0 Ssquo!suaul!p ‘fl

 0
f
II .-c 1 0
0

0
❑
0
c1
0
0
❑
0
00
0
00

-n

% al
❑
I @ ‘a
❑
0
0“”
111 1,,
i
I ‘Q -Q >

M71 ‘*’9: Ssa!do!suaul!p ‘a

——
o
L

t
 SPE 15?67

-i

10-’! 1 1 1 1 , ,
0 10 20 Sc 60 o
v~ft/seco;d
60 lb HPG/Mgai, 6 /b/gal 20/40 mesh sand
F@. 9-Compsrlscm of kbomlory data to cormktlon predktlons of drag ratio vs. velocity

2.875 6.5#/FT
9200b
5DAHD4Gt4HfG 1000 GAL
DELAVRt CROSSU4 I(EO FLUIO
AVG. RAIE 12.2 EPM
MAX. SANoCoHc.S @AL

a Laboratory Correlation
o Fidd Collbfotod Correlation

o
At, mhwtos
Fig. 1O-FIM CaSO of IabOrmoqandW
COmpdsotl c@lbntd COfrelatlenpmd~ of Ml mm -. ~~.
 &7# ‘&s #/n
SO AHD 40 L6 HPG 1000 GAL
WE 15367
DEIAYEO CRos31i t( KID FWID
AVG. RATE 12.S WM
MAX. SAW CONC. 16 #/GAL
n

Fig. 11—FlcldC8CCcompwfwn of Idromtory ●cd field ccllbratcd corrdctlan prcdktlons of ne! preccure vs. time.

4,500”13.5
#/n
11000 F&
SO AW 40 LB HPG 1000 GAL
DELAYEDCR033114 K FLUID
AVG. RATE 12.S 6PM
MM. 3AN0 CoNc. 13 #/GM

-tr
.. tee ft
At, minutes “--
Fig. 12—Fiefdcccc comparfcorrof Ieborctorywrd ffefdcclibmtcdcorrelationprcdlctiOnsOt~t pfe~re ~s. tlmefl
 SIT 15367
MINIMUM STEPSIZE MAXIMUM
$.&5’~{ LB/FT
Friction - Labratory Corr. II 4000.0 600.0 ~~~~~.~
50 AND 40 LB HPG 1000 GAL Friction - Calibrated COW. Q 4000.O 500.0
DELAYEDCROSSU4 K FLUID Actual Friction Pressure A 400$~ 60~~ 10000:0
AVG. RAlf 20.1 BPM Sand Concentration v . . 10.0
MAX SAND CONC. 8.0 #/GAL
—

,..
10:25 10:34 10:43 10:52 11:01 11:1’3 11:19

F@. 13-Ffefd w88 c=omfw’kon of OCIUSI frkfbn loss to hlwafwy ●’)d ffofd wllbmtad comekflofI ~wodkfkns for
cfwglng propfmt conconlmffon with the.

MiNiMUM STEPSiZE MAXIMUM

&WJ’&6~#/FT 4000.0
Friction – Labratory Corr. g O.O 400.0
50 AND 40 LB HPG 1000 GAL Friction - Calibrated Corr. O 0.0 400.0 4000.0
Actuai Frictio ‘. Pressure A g.g 400.0 4000.0
DELAYEDCROSSUF(K FLUID 15.0
AVG. RATE 15.5 BPM Sand Concentration O . 1.5
MAX SAND CONC. 12.0 #/GAL

+“”””’” ”’ s’””””’””””’ l“s””’ ””’’’”’’”””’”’’” i

4
1