You are on page 1of 149
Multiphase Flow in Wells James P. Brill Floyd M. Stevenson Endowed Presidential Chair in Petroleum Engineering Executive Director, Fluid Flow Projects U. of Tulsa and Hemanta Mukherjee Manager, Production Enhancement, West and South Africa Schlumberger Oilfield Services First Printing Henry L. Doherty Memorial Fund of AIME Society of Petroleum Engineers Inc. Richardson, Texas 1999 I rc Table of Contents Chapter 1—Introduction .....s6.06++4+ LI Scope .. 1.2. Objectives of Monograph . 1.3 Organization of Monograph 14. Historical Background 1.5. Nomenclature and Units Chapter 2—Single-Phase-Flow Concepts . 2.1. Introduetion 2.2. Conservation of Mass . 2.3. Conservation of Momentum 24 Pressure-Gradient Equation 2.5 Flow inan Annulus ... 2.6 Conservation of Energy Chapter3—Multiphase Flow Concepts no Introduction _— 33 Definition of Variables 3.4. Pressure Gradient 3.5. Flow Patterns . 3.6 Liquid Holdup 3.7 Pressure-Traverse Computing Algorithm 3.8 Dimensional Analysis Chapter 4—Multiphase-Flow Presure-Gradient Prediction Introduction 42 Preste Grains reicon 7. occooceco 43. Evaluation of Wellbore Pressure-Gradient-Prediction Methods 56 44. Pressure-Gradient Prediction in Annli.....2..-0.ee00e00000001 vee 58 4.5. Evaluation of Annulus Liquid-Holdup and Pressure-Gradient-Prediction Methods... oo 66 4.6 General Observations ....... eee 6 Chapter S—Flow Through Restri 5.1. Introduction 5.2. Description of Restrictions . 53 Flow Through Chokes 54 Flow Through Piping Components ‘Chapter 6—Well Design Applications 6.1 Introduction. 62. Vertial-Flow Performance 63. Inflow Performance 6.4 Production-Systems Analysis. 65 Anificial Lift ...... 66 Gas-Well Loading 627 Erosional Velocity 68. Special Problems... Appendix A—Nomenclature and SI Metric Conversion Factors ...+.+++e++0++ Appendix B—Fluid and Rock Properties ... Appendix C—Vapor-/Liquid-Phase Equilibrium Appendix D—Tubing and Casing Properties . Author Index ions and Piping Components Subject Index .. Chapter 1 Introduction 1.1 Scope The accurate design of ol and gas well ubing strings requires the ability to predic flow behaviorin the wells Wells normally produce ‘a mixture of gas and liquids, regardless of whether they are class: fied as oil wells or gas Wells This multiphase flow is significantly ‘more complex than Single-phase flow. However, the technology (0 ‘predict maltiphase-iow behavior has improved dramatically inthe past decade tis now possible to select ubing sizes, predict pressure ‘drops and calculate flow rates in wells with acceptable engineering accuracy. This chapter sets the stage forthe monograph by describ ing the nature and occurrence of multiphase flow, andby presenting important historical events that have impacted on the development ‘of modern multiphase-flow concepts ‘The common occurrence of multiphase flow in wells can be di ‘cussed with the simplified production system shown in Fig, 1 Fluids entering the wellbore from the reservoir can range from an tundersaturated ol toa single-phase ges. Free wate can accompany the Muids as a result of water coning, water looding, or production of interstitial water. Alteratively,a free gas saturation in an oil res- ervoir can result in a gasiquid mixture entering the well. Retro grade condensation can result in hydrocarbon liquids condensing ia gas condensate reservoir so that a gasliquid mixture again eners the wellbore. Even when single-phase gas or liquid flow exists near the bottom ofa well, multiphase low can oceur throughout most of ‘the wellbore This is a result of evolution of gas from oil or con- sigh aie ‘Values of fare estimated. fu, and then calelated, fe, until they agree 10 within an acceptable tolerance. A direct substitution procedure {hat uses the calculated value asthe next assumed value results in convergence in only two or thre iterations, The inital assumption, can be obtained from one of the explicit smooth-pipe equations, or from explicit approximations tothe Colebrook equation Numerous explicit approximations to the Colebrook equation have been proposed. Zigrang and Sylvester! have given one ofthe ‘most accurate and simple to use. ain aa) SINGLE-PHASE FLOW CONCEPTS 005 of sf ace 001 008 2005 2003} ‘ove 001 edge 20003 0009! 2000396} 00004} 000,03} ocooce| ‘200001 ‘00.08 0co.ee! e $456 80 20 3000 28010 Fig. 23-Pipe roughness? In most cases, Ea, 2.19 can be used in lieu of Eq 217. Fig. 22 isa graph that shows the variations of fiction factors with Reynolds number and relative roughness asedon Eas. 2.12 and2.17, Ttisimportant to emphasize that eis nota physically measured propery. Rather, itis the sand-grain roughness that would result inthe same friction factor. The only way this ean be done is to ‘compare the behavior of anormal pipe with one thats sand-rough- ‘ened, Moody” has done this and his results, given in Fig. 2.3 are sill accepted values. However, these values should not be consid ceredinviolate and could change significantly asa result of paraffin ‘deposition, hydrates, erosion, or corrosion. If measured pressure eradients are available,a friction factor vs. Reynolds number rela- tionship can be established and an effective relative roughness ob- tained from Fig. 22. Until itis again updated, this value of e/d should be used for future predictions Initial values of roughness often are needed for design calcula- tions. The recommended value for new tubing is€=0.00005 ft. A ‘common value used to generate pressure-gradient curves is (0.00015 For tubing exposed 10 an environment that causes sig- nificant changes in roughness, "very dirty” pipe can have rough- ness values of 0.00075 ft! For most wells, the friction component ‘of he pressure gradients small compared withthe potential ener- ‘gy component. Consequently, approximate values for absolute roughness normally are sufficient. 219) Example 2.1—Single-Phase Liquid Pressure Drop, Calculate the pressure change in a water injection well. The following data ae known.* 8,000 ft 20,000 B/D 90° 62.4 Ibm 50in Hu = 1060 ‘= 0.00006 f ‘The average velocity inthe pipe is 9.532 fusee, om 3 210, he Retold mune wosasoal§ hy = ip as Beco Me >208, ow tube Te ae oghes for coo = 3.688 x 10% 0.000148, From Eqs. 2.17,2.19, or Fig.2.2,f= 0.0155. Calculate the pressure sradient from Eqs. 2.5 and 2.9, neglecting acceleration effects dp. _ = (0.0155)662.4)(9:531)? _ (62.4)(32.2)fsin(- 90) ake Ga = = 3274 + 62.400 = 59.126 pstit ~ 00227 + 0.4333 = 0.4106 psi ‘The pressure change is then ‘Ap = (— 00227 + 0.433378, 000) =~ 1819 + 3,466.4 = 3,2845 pai [Note thatthe pressure change consist ofa loss owing to friction of = 181.9 psi anda gain from an clevation change of +3.466.4 pi 2.4.3 Single-Phase Gas Flow. When a compressible uid, such as 225, flows in a wel, the density velocity, and, consequently, the pressure gradient all vary with pressure, Cullender and Smith! de- veloped the most widely used method to calculate flowing bottom- hole pressure in gas wells Neglecting kinetic energy, Bq, 2.5 can be written 3s se _ oY sina 220) For gases, (P= PM/ZRT, v= 4/A, q=4nBy. and B,™ peT2/T ap. ‘Combining these expressions With Eq, 2.20 and separating vari= ables gives ean arin i nosh bgt come cv = 322 oy z i where Eq. 2.21 is applicable for any consistent set of units Substituting field units and integrating the left side of Eq. 2.21 gives "1 teas | non " ee , ex) 5 o001(fj) ano + and p= psia, 7=°R, auc = MMsc{7D, d=in, and. ‘The right side of Eq. 2.22 cannot be integrated analytically with- ‘out making assumptions about Z, 7 and f However, the trapezoidal ‘ule for numerical integration can be applied by determining the vi te of [for each of any numberof increments in p between py and ‘ag. Assuming the well can be divided into two halves for integra- tion purposes—where ony the intermediate value of pressure, that atthe mid-depth, py is considered —Eq 2.22 may be expressed as, (ea1~ Palle + ta) 2 18.757,L = (loa ral > Eq. 2.25 can be separated into two expressions, one for each half of the well (228) Upper hat 1875y41/2 = (bay ~ py) - 026) Lower ba lag 1B.1574L72 = (Pay ~ raj ttets) en ‘While his metho canbe used with any number of steps, Callender and Smith! demonstrated thatthe equivalent of foursep aecuraey an be obtained wit a two-step calculation if the Simpsen's rule! ‘umerieal-imtegraton approach s used. The resulting equation is wars, tm (EPL + ary tH). -- (228) Use of ine Callender and Smith method to calculate owing bot tombe pressures in gas wells can best be explained with an exam Ble problem Example 22—Single-Phase Gas Pressure Drop. With he follow- ing data, calculate the lowing botiombole pressure in a gas wel us- ing the Cullender and Smith method with two increments. = 075 tL Gc 4915 MMsctD -MULTIPHASE FLOW IN WELLS B= 0012 ep o= 90° Preliminary Calculations. Calculate f assuming a completely turbulent love (fully rough wall). If flow is not completly tubu lent, a Reynolds number calculated at surface conditions can be 3s sumed to exist throughout the well. From Eqs. 2.16 and 2.24, f=0.015 and pe x (066730.015)4.915 es? Calculate fy Eq, 2.23): Atpy= 2.000 psia, T= 110°F, and. p _ 2,000 12 ~ Grovo7H ~ 49? and ly 4.342 181.60. (0014 942)" + 0.00279 Upper Half of Well Estimate p3y(First Tal) iy = pl +25 x 10°8L/2sin8)} = 2,000[1 + (2.5 x 107*K5,000sin90") 250 psia Calculate I (E9. 2.23) 250 psia, T= 110-+67.5°F, and Z= 0.797, Be 2250 72 Gayore7 ~ 45 and ey 4.425 = 19781 (@.001¢8.425)* + 0.00279 Calculate Py (Eq, 2.26): IB.75y_L 18251075100.000) Teal Po = Pet = 2,000 +" 97.81 + 181.60 2,000 + 371 + 2,371 psa (no close enough to py) Estimate ply (Second Tria) Select piy=2,371 psia Calculate I (Eq. 2.23) At pig 2,371 psia, T= (78°F, and Z=0.796, pre eeas7 TZ ~ GBx0.796) and 669 4.669 (0.001514.66)° + 0.00279 Calculate pay (Eg. 2.26) {(18.75)10.759(10, 000) 189.88 + 181.60 189.88, Poy = 2,000 + ‘= 2,000 + 379 = 2,379 psia (not close enough to pi). Estimate py (Tied TH: Select po 2379 psi Calculate Ig (Eq 223) At p= 2.39 pia, T= 178 °F, and opie e237) TZ * CRx0795) 4684 SINGLE-PHASE FLOW CONCEPTS and 4.684 (C.o0iys 684)* + 0.00279 CCateulate pay (Ea. 226) (18:75)0.75\10, 000) Ta9.41 + 181.60 ‘Therefore, the pressure atthe mid-point ofthe well is 2379 psia. ‘The value of paris now calculated = 189.41 000 + Pay 2,379 psa, Lower Half of Wel Estimate py (Fest Ti) Pod +25 x 10°9L/2sin0) = 2,379[1 + (25 « 10-4}, 000sin90%)] = 2.676 psa. Calculate iy (Eq. 2.23): At ply= 2.616 psia, T= 245°F, and Z=0.867, Phy p _ _2.676 72 ~ Gsnosay ~ 4378 and in ae 19939. (00144378) + 0.00279 Cateulate pus (Eq. 2.27) (18.75410.759(10,000) Pos + [99.39 + 189.41 Pay 9.379 + 362 = 2,741 psia (not close enough to pp. Estimate pi (Second Teal): Selet p= 2741 psa Caleulte fy (Ea. 223) At ply=2,781 psia, T= 245°F, and Z=0.868, B= ht 72 ~ Goswo.s68) ~ +47? and ty 4879 = 19600, * @o00n.479)' + 0.00279 Clout py (Eg, 227) (18.75X0:754(10,000) 196,00 + 189.41 ‘Thisisclose enough tothe previously calculated value of,741 pia Therefore, the flowing bottomhole pressure is 2,744 psia, From Bq, 22, ifthe more accurate Simpson's rule aumerical-in- tegration approach is used, the flowing bottomhole pressure is pre- dicted tobe Pay = 2,319 + = 2,744 psia “4 —(6218.750:75)110,000) 6+ TRL60 + ISAT + 196.00 Por 2,743.2 psi 2.44 Non-Newtonian Fluids. The material presented previously is, Valid only for Newtonian fds, However, ids encountered in the Petroleum industry often act as non-Newtonian fluids. These in- ‘lude many drilling muds; uid, such as cement sluries, fracturing, fuids, and spacers used during well-completion activities; poly- mers injected during EOR projects crude oils at temperatures ap- proaching the pour point; and many oil/water mixtures. ‘The design of piping systems for non-Newtonian fluids becomes ‘complicated because the use of conventional fiction-factor corre tions isnot directly applicable. These non-Newtonian fluids may as- sume any type of rheological behavior, depending on such factors as shear rate, temperature, and fluid composition. Fig. 2.4, from ° HEAR RE seman Fig2.4—Rheological models. Knudsen and Katz! describes the types of rheological behavior that ‘can be encountered, ‘Twomethods commonly are used to design piping systems forthe transport of oilwater mixtures. The first method teat the mixture 1 @ Newtonian fluid with an apparent viscosity that can vary with water fraction, This method is covered in Chap. 3 because it can in volve the combining of the viscosities from each phase to obtain a ‘mixture viscosity. The second method teats the mixture as a non "Newtonian fluid and is based on the following assumptions 1. The mixture is homogeneous 2 Slippage between phasesis neglected. Thus in-situholdups are the same as their respective input volume fractions 3. The rheological behavior ofthe oi/water dispersion system is Suitably described by the Ostwald-de Waele power-law model! Power-Law Model. The power-law stress strain relationship can be expressed as reky - ea (229) fn" isunity, Bg. 229 will describe Newtonian behavior, and K" will be equal the constant viscosity. , For atypical oilwater mixture, ‘usually iss than unity, and Eg 229 will describe pseudoplastic {shear thinning) behavior. i also possible for an oil water mixture tohaven’ greater than unity, resulting in dlatant (shear thickening) behavior. The type of behavior that a uid system will follow nor rally is unknown but can be determined from laboratory exper ‘ments with an appropriate viscometer. ‘These tests must be conducted fora specified set of operating conditions: input water fraction, temperature, and droplet site dis tribution ofthe dispersed phase (or, indtecty: mixing speed). Once the fluid system is correctly characterized, te frictional pressure _radients for pipe low inthis particular oil/water dispersed system readily can be determined, Generalized Reynolds Number. Metzner and Reed! introduced the concept of a “generalized Reynolds number” for non Newto~ sian flow. The usual Reynolds number definition is given in Ea 2.10, which also can be written as, ped (ay «= eG) ™ 230) where ty is the laminar wall shear stress, while the quantity, 8d, isalso the true shear rate atthe pipe wall fora Newtonian fluid Thus, ‘he ratio, (4/(80/a)], corresponds 1 the Newtonian viscosity, For non-Newionian flow, the relationship between the true shear ‘ate and the apparent shear rate can be expressed as 0 ous tye) o @3y where n’=the slope of a logarithmic plot of ty vs. Sv, ofien called the flow-behavior index. rom the same plot the lationship ofthe tangential ine to the laminar-fow curve can be writen as n= «(ey oes OR Combining Eqs. 2.30 and 2.32 gives (-2), ponds Meng RE ecccccscsesscssescses ax aed a «(sg (%) 34) here Nang “tbe generalized Reynolds aumbet, sd = apparet viscosity. Eqs. 2.53 and 2.34 clearly reduce to the normal Reynolds ‘number and the Newtonian viscosity for a Newtonian fluid. when n= and K’ ‘Non-Newtonian Friction Factor. In general, theprocedire for calculating friction factors for non-Newtonian fui ssa that for Newtonian ud Laminar Flow. From Metzner and Reed,"6 the Fanning friction factor for non-Newtonian laminar flow can be writen as fag. 238) Turbulent Flow—Smooth Pipes. Dodge and Metznet” proposed this implicit riton-faio equation, i. 49 nsonia) — 04, [p= oan, S27] ~ 98. 236) Turbulent Flow-Roush Pipes: Goviee and Azie suggested this fiction factor for power iw pseidoplai Huds osing in rough pipes: sa) ee where fyyq =the friction factor calculated from Eq, 2.36; f= Newtonian flow Fanning frietion facto for rough pipe, calculated at the same generalized Reynolds number; and j; = Newtonian MULTIPHASE FLOW IN WELLS flow Fanning fietion factor For smooth pipe, calculated atthe same generalized Reynolds number. Szilas er a. obtained a similar = Sut for pseudoplasie Muids from the following equation, = 400g] —_10-#2 i lee ne ar where (238) p= ssi (0292 5 212) 915-087... a0 Note that Eq 2.38 is analogous tothe Colebrook equation for New- tonian fluids given by Eq. 2.17, and also requies a teal-and-eror solution procedure, Fully Rough Wall Turbulence, Govier and Azia'¥ recommended, For Newtonian fluids, Ea. 2.40 essentially reduces othe von Kar ‘man equation,!® 3 Frictional Pressure Gradient. Once the Fanning friction factor, {fs determined for the particular sysiem, the fetional pressure ‘gradient can be calculated from (@) 2 2.65 4.06 ogd/2e + 60 — 240) O6logd/2e + 3.36. seseeees QAl) 2a) andi ar clearly the two important parameters that wll affect the fictional pressure-gradient calculation for a dispersion system, ‘These two parameters can be determined experimentally with an ap- propriate viscometer Example 2.3—Non-Newtonian Oil/Water Mixture Pressure Drop. A horizontal pipe discharges an oil/water mixture a atmo- spheric pressure. Determine the inlet pressure required to maintain ‘constant volumetric ow rate ofthe dispersion for an input-water fraction of 20%. The Following also are given: 30 see 2vin O01 k 10,000 f 53.74 Ibe 62.23 bavi? o8s89 7.1475 x 10° Tht sect? Calculate the mixture density, using n= pafe+ Pf = 554 Ibm {3 Forthe given flow rateand water fraction, calelate the general ied Reynolds number, pasta Mees BER 9.97 x 10 ‘The criterion for turbulent flow is considered to be Npa,,> 1,500, For this example the low regime is turbulent. ‘Using Eq. 2.37, compute the friction factor for turbulent Now of pseudoplastic fuids in rough pipes 1 = fas (F) = 00010006 00708, where fxg is given by Eq. 2.36, is given by Ba, 2.14, and is iv cen by Bq, 2.17. Note that if Eq. 2.38 were used, the trial-and-error procedure would have given a value of 0.00676, ecanticiy, 620A) one an Fig. 25—Annuli configurations 2° Calculate the pipe inlet pressure, py P= Pst Yat 105.9 psa, 2.5 Flow in an Annulus Inthe petroleum industry, ow in wells normally occurs in tubing. string. However, many oil wells with high production rates produce through the casing/tubing annulus. This trend is dictated by eco- nomics, multiple completions, and regulated production rates. Al- though few in number when compared with all producing wells, these “casing flow" wells account for a significant part of the world’s oil production ‘Other easing-Nlow applications are found in wells under various types of artificial lift. In sucker-rod pumping wells, a rod sting is installed inside the tubing sring to connect the prime mover unit on the surface tothe pump at the bottom of the well. The fluids are pumped upward through the tubing-rod string annulus. Casing low so can occur in gas wel production. To remove or “unload” unde- sirable igus that can accumulate atthe bottom of these wells, si- phon tube often is installed inside the wbing string. The permanency ofthe siphon tube in the tubing string requies the Nuids to flow up ward through the tubing-string/siphon-tube annulus. Inthe past, annuli have been treated based onthe hydraulic diam- eter concept. The hydraulie diameter is four times the area for flow divided by the wetted perimeter. For annulus configurations, d= de dy coerce @43) However, the hydraulic diameters not always the most representa tive characteristic dimension for flow in an annulus. To determine appropriate characteristic dimensions, a lear understanding of flow inn annulus must be achieved, ‘Anannulus is characterized by the existence of two ctcula pipes, where the flow occurs through the area bounded by the outer pipe inner wall and the inner pipe outer wal. Two geometrical parame- ters identify these configurations: the annulus pipe-diameter rai ‘=i, anu the degree of eccentricity. The degree of eccentricity Accounts for the displacement ofthe inner pipe center from the outet pipe center and is expressed by Dae aa where d= outer diameter ofthe inner pipe (tubing) de = inner ameter of the outer pipe (easing), and Dgc = distance between the pipe centers, Annul can have eccentricity values varying from zero toone. Fig. 2. shows cross sections of annul with the same pipe-di- ameter-ratio value, K, and for eecentricities of 0.0, 0.5, and 1.0, (24ay 2.84 Friction Facto for Single Phase Flow. For noncircular con- dis, single-phase-low friction factors often are determined through the aplication of the hydraulic-diameter concep, Howe ex, this concep is better sited for high depres of turbulence For laminar low, rigorous eatment of te flow fields possible forany annul configuration, Bide gave analytical olwions forbath the veloity profile and fiction factor for aconcentic ann lus, Combining the developments by Heyda™ and Snyder and Goldstein" analytical solutions also can be found for these low Parameters for an eccentric annulus. These solutions are presented ina later section, In turbulent flow, even for circular pipes, the mechanisms of ture bulence are by no means fully understood. However, various semi ‘empirical and analytical models have been used successfully to pre dict the velocity distribution and pressure gradient. In noncirctlar configurations, where the transport phenomena are intrinsically ‘mote complex than for circular pipes, the formulation ofan analy «al model is even more complicated. Three ways have been used to Predict the Flow behavior ofaturbulent-flow field in an annulus: em Pirical correlations, semi-empirical corelations, and application of ‘Universal velocity distributions Empirical corelations involve the application of curvesfiting techniques to experimental data to predict an overall flow quantity Such as friction factor. The resulting frition-factor correlations nor, mally take the form ofthe Blasius-type expression, as given in Eq 2.15 where C and n are determined empirically. Examples include the correlations of Knudsen and Katz,! and Dodge?" Winkle? Successfully used the Dodge correlation for flow in an annulus; ‘however, he note tha this procedure didnot take into account the annulus pipe-diameter ratio, ‘Semiempirial approaches involve the use of experimental data for turbulent flow in combination with characteristics of laminar flow in the same noncicular configuration. The Gunn and Darling” proce 0.3. Te relationship between tue shea rate and apparent shear rate atthe wall is expressed as 258) ar), = a here, n' = flow-behavior index. ‘The generalized Reynolds number concept alsocan be applied for ‘an annular-flow geometry. The generalized Reynolds number is Mange 259) where K's a parameter related the consistency index and is ex pressed by Eg. 2.60 for an annular geometry. - 06) Then, fiction factor ean be writen 3 fait. esr ‘Therefore, with Ba, 2.62, the fictional pressure gradient fora con- centric annulus can be calculated as (4) - 2% om and; Eccentric Annulus. Haciislamoglu and Langlinais® developed a correlation to predict fritional pressure losses of power-law fluids inan eccentric annulus. A correlating parameter, R, was defined as the rato of fictional pressure losses in an centri annulus to those Jnaconcentric annulus. The empirical correlation, whichis valid for evcentrcities ftom 0 10 0.95, pipe diameter ratios of 0.3 100.9, and flow-behavior indices, n’, of 0-410 1.0, is given by one (t ‘) R= ~ssea (fy ase 08 ‘This corelation has a claimed accuracy of + 5%. To predict he fric- tional pressure gradient in an eccentric annulus, the factional pressure ‘gradient in a concentric annulus is first calculated with Eq. 2.62. The Corrected value for an eccentric annulus i then determined fom (@),-(@),e- ‘Non-Newtonian Turbulent Flow. There is no documented model to predict non-Newtonian turbulent low in an annulus. However, an approach analogous to regula pipe flow canbe postulated. The gen eralized Reynolds number fora concentric pipe can be used inthe ‘non-Newtonian pipe flow friction-Factor correlations. Ten, the Bla- sius equation canbe used to calculate the fictional pressure gradient Foran eccentric annulus, one might try to use the R correlation (Eq, 2.63) developed for laminar flow with a specific value of did. =0.01, as suggested by Sas-Jaworsky.*? The frictional pres- Sure gradient then can be obiained by use ofthe modified R for tur- bulent flow and a concentric anus in Ea, 2.64 268) 2.6 Conservation of Energy Application of energy conservation to fluid low in pipes requires ‘that ina given pipe segment the energy in, minus the energy out, plus the heat energy transfered to or from the surroundings must equal the rte of energy accumulation! afoleest) Qn 0 For steady-state flow, Eq. 2.65 reduces to a +e Ord slob] sary when dealing with customary units where mechanical energy and thermal energy have different units. 266) FLOWING FLUIDS . & a: a ey 3 WR ee ¥ So es %3 Se Ry 255 RR ey 3 Be Fig. 27—Cross section of typical wellbore.2 (Reproduced with permission of the McGraw-Hil Cos.) ct en 2 sled mri(eoa8s) +(e+ un In Eqs, 2.66 and 2.67, es the intrinsic specifi energy and is de- fined by (268) Combining Eqs. 2.67 and 2.68 with Eg, 2.2 from conservation of mass principles yields wd (ehsind _ = Ona Pra gd *2 rat Be pel) A ad = 2.69) Because specific enthalpy is defined as neue kh + 00) pst ‘ 4.269 can be expressed as £5, om ge Bene, Om a en Finally, solving forthe enthalpy gradient yee dh = Oxd_y wy gsi g-o we am at The eat x, Qs dened terms of over beat raat cof ficient and temperate dference bincen the Dds and ie sor roundings Thus, o=ur,-1) em a, 272 clearly shows thatthe steady-state enalpy- gradient equations made ip of de component Ths, @),-@),+ en = ), as Because of the strong dependence of enthalpy and heat transfer fon temperature, Eq. 2.72 is used to determine temperature change ‘when fuids flow through pipes. Normally, the kinetic energy tern is negligible. Therefore, fora horizontal pipe, an increase in uid ‘enthalpy equals the heat transferred tothe fluid from the surround: ings. Also, if no heat transfer oceurs, an increase in elevation cases decrease enthalpy anda corresponding decrease in temperature 2.6.1 Wellbore Heat Transfer. When hot reservoir fuids enter & wellbore and begin to flow to the surface, they immediately begin losing heat to the cooler surrounding rock. The surrounding rock _radualy heats up reducing the temperature difference and the heat ‘wansfer between the fluids and the rock. Eventually fora constants ‘mass low rae, the earth surrounding the well reaches a steady-state temperature distribution. Prediction of fluid temperatures in the ‘wellbore as a function of depth and time is necessary to determine ‘the flud’s physical properties and caleulate pressure gradients, Because ofthe high thermal conductivity and relatively smal ra it distance between the flowing fluids andthe borehole wal, heat transfer inthis region normally can be considered steady state. All heat ost by the fluids instantaneously flows through the welloore and into the surrounding rock, An axial cross section of 2 typical \wellbote is shown in Fig 2.7. The following description of steady. State heat transfer in a wellbore would have to be modified for other "pes of completions. Heat transfer within the tubing and in a ud-flled annulusis pi ‘marily a result of convection. Heat transfer through the tubing nd casing walls and though a cement-filled annulus between the cas- ing and borehole wall primarily results from conduction Heat transfer resulting from conduction ean be descibed by Fourier’s equation in radial coordinates! a= = 2a, @75 Where q isthe amount of heat flowing radially through a solid with thermal conductivity k Integration of Eq, 2.76 gives en MULTIPHASE FLOW IN WELLS “Heat transfer resulting from radial convection can be described byl q = 2xrALhAT, where h=local convective-film coefficient If steady-state heat transfer occurs in the wellbore, q is constant Expressions for temperature change through the wellbore can be de veloped ftom Eqs. 2.77 and 2.78 as follows. For convection in the tubing, = 078) -tyesfe de. 279) Foe conduction hugh the tubing wal, ese) 280) = SAE he : For convection through the casing/tubing annulus, cee Ta ~ Ta ™ sot he ean For conduction tough the casing tol tanta shy on aT BG pe ec teee eee Forconduction through the cement inthe casing/borehole annul, sles) a Me Ta Te = 5h 8) Heat wransfer ino the surrounding rock is by heat conduction and {satransient process. The transient radal-heat-conduetion equation is identical tthe diffusivity equation encountered in transient well- test analysis! The infinite-reservoir, line-source solution is = a fo) mT at Be. 84) where Tz =the undisturbed geothermal earth temperature, f). is aiven by =le(st @. : ; ‘To monitor temperatures at the wellbore, the logarithmic approxi. imation tothe E solution is valid for times greater than 1 week >= “Thus, for x<0,0025 and, + G86) £(~ 9) ~ Ine) + 0572 es and $10) = 0405 + OSes sseeserseeecessrses 88) where ty = e239) Hasan and Kabir showed that for typical reservoirs Eg 2.88 an ‘cause significant erorsif applied to times less than 250 hours. They presented Eqs. 2.90 and2.91 as simplified equations that, when used together are valid for al times. It, < 15, so) 1.1281 Jf ~ 03 Ji) ss 90) Wipe > 1S, fo fos06s + ostnioaI(1 +98)... @9D Hasan and Kabir} also stated that, in most cases of oil produc- tion, the temperature difference across the annulus is usually small ‘and convective (natural) heat transfer becomes important. Unfortu nately, the lterature reports no work on natural convection in verti cl annular geometry, Hasan and Kabir recommended using 0.0491 aA) tae an) ‘where the Grashof number, Ngo reflects the extent of motion ofthe annulus fluid resulting from natural convection 292) (a — to) abn Nee 293 “The density ofthe ested Mid nexttothe tubing walls es than the {id nex fo the easing, creating a buoyany fre. The peed of B coeticient of thermal expansion) andthe temperature erence iva measure ofthe density diferene. The viscous force working gaint the buoyancy generates circular mation ofthe Mud inthe tml. The Prandt mumbo, Np smear ofthe interaction b= tween te hydrodynamic boundary layer and he thermal Boundary layer and is defined as are Ny = ass Combining Eqs. 2.79 through 2.84 determines the total tempera- ture change between the fluids and the undisturbed geothermal tem perature of the surrounding rock, + 295) {A simple expression forthe total heat loss from the fluids in the tubing can be estimated from Newton's law of eooling,!9 4 = der ALUAT, 2.96) ‘where Un overall hea-transfer coefficient. Comparing Eqs.2.95 and 2.96, itis evident that 7 U)~ isthe bracketed term in Eq. 2.95, 2.62 Temperature Prediction, Prediction of temperauresin wells ‘eauies application of conservation of mass momentum, and ener £2 princiles, This canbe accomplished by coupling the pessre- dient and entalpy-ratien equations given by Eqs. 2.5 and 72. The complet ol these equations prevents argoreus anal cal solution, but a numerca-soluion procedure is given in Cha $S Mowever Sagar e a2 Alves ete and Hasan and Kbit? present appreniate analytical sohuions, The Alves ef. fr- tmulation was for flow in pipes at any inclination angle, Tee su. tho degenerates to equations presented by Ramey forthe ease of| ction welsandby Coulter nd Bardon" forthe ase of orzon- telppelioes. Because the Alves eal method involves fewer reste. tive ssumprions, it should yield more accurate predictions, Al though numerous other publcaons have appeared, primaily pertaining to scam Injection or producing wells al are modifca- tons tothe Ramey method or tothe calculation of heal-ransTer eo effciens. A summary ofthe Alves et a slution follows Because enthalpy isa state property, A= Hp), a change in en- ‘thalpy can be calculated by considering effects of temperature and pressure separately. Thus, on = (9) ar (2) scars en Comite nents ps ms tnoncare() ),- -6(@) = -en em the Joule-Thompson coefficient and represents isentha- Bic cooling (or heating) by expansion. Combining Eqs. 2.97 and 2.98 gives Oh = CAT ~ Condy. eee csees 299) Combining Eqs. 2.99 and 2.72 gives ar, Toy aL” Mar ~ EE og — adn, — ny 2100) 4. 2100 canbe simpli to this fermi equation ar, 1 % BEB oo eececcccceee 10D whee anSe, nm w= oncvi ~Pegee — a on & (2.103), a 1 the surounding temperature varies linearly with depth, Te = Ty golsind, 2.108) where 7 =surrounding temperature atthe inlet of the pipe and is ‘often taken asthe reservoic temperature. The geothermal tempera, ture gradient, gc, typically varies from approximately 1.) to 2.0°F/100 ft of vertical depth, depending on the thickness of the ‘arth crust, presence of voleanic activity, and other such favors ‘Combining Eqs. 2.101 and 2.104 yields a generalized dtferenvial uation that incorporates both the enthalpy and pressue-gradient ‘equations with no limiting assumptions Thus, q_t, ai, Tu _acbsind, _1 ap aa RO + oe ar a, 2.105 canbe integrated assuming constant values fr U, Gu. 0.8, ¥.dv/aL,and dp /d. The resulting solution is 2.105) T= (Pq gobsind) + (7, ~ TyerH * sosingalt —em¥t) 4 he Beals ~ «1 e106) [Bg 2.106 degenerates toall the more restrictive approximate analyt- ical expressions to predict temperatures of fluids flowing in pipes For the ease of horizontal flow, where 8=0°, and neglecting ec, ccleration effects, Eq, 2.103 simplifies to = sone, 2107 and Ea, 2.106 degenerates to Ty Cy ~ Te-UA + Bal ~ et) 2.108) g,2.108is equivalent tothe Coulter and Bardon equation ope. dict temperatures in horizontal pipelines For an ideal gas, =0, and neglecting acceleration effects, Eq 2.103 sinplifies to — eesind) E 1 2 2.105) a nd Eq. 2.106 degenerates to 1, = (Tq ~ goLsind) + (7, Tye-H 1) = 850040) — un * gesindalt eH) BEBE — ga, 2.110) whichis equivalent tothe Ramey equation for an ideal gas, For the case ofan incompressible liquid, 1-5 ean and ? eur Neglecting fiction, =0, and Bq. 2.106 degenerates to 1) (Ta ~ acksind) + (T,— Tape + sosindalt ~ eH) a3) Which is equivalent tothe Ramey expression for incompressble- liguid fle ‘Comparison of Eqs. 2.106 and 2.113 shows thatthe Alves era. Solution is actually the Ramey equation for single-phase gud. pls {correction term. The corection term is function of the total pres. sure gradient andthe dimensionless parameter, . Analysis of tos Aimensionless coefficient ean show when consideration of the correction term becomes important Calculation of lowing temperatures as a function of depth and time can be very tedious because ofthe complexity ofthe oveval beattranstercoeficien in Eq 2.96, Shiv and Beggs proposed an empirical correlation for A that was developed from a broed set oF Mowing temperature surveys. The resulting equation iindependent of time A= D014 (Ia) Py My Ay 3 ui ‘where w is in pounds pe second, dy is in inches, and py isin pounds per cubic foot Eq, 2.106 is recommended when calculating. temperature changes in wells lowing multiphase mixtures. Iti ist necessary {0 define several physical properties in terms of mixtures, These mixture properties are discussed in Chap. 3, -MULTIPHASE FLOW IN WELLS Example 2.4—Temperature Prediction Comparison. A single- phase ol flows up a tubing string that is cemented from top to bot tom. Using Eq. 2113 andthe Shiu-Beggs correlation for A, calew Jatethe wellhead temperature ater the well has been flowing for 2 ‘weeks. The following is given. Keen = 082 Brahe oF 25 Bur fF 14 Blue fF 0.08 Bewhr-h.°F 0.65 Buu/lbm-°F 0.08 fe 5,000 BBD Tay = Ty = 200°F 10 cp=6.72 x 10 bit = see 2.750 in, 2.446 in 40 os = 0015*Fit 30°API 10,000 # cr 1. Determine i. The Ditas and Boelter equation” gives an ac- ceptable correlation for the Nusselt number for turbulent low (Wiee> 10%) in the tubing when cooling is taking place Nagy = O.O2BNBENQ?, Reynolds Number: 000 B/DI5.615 = 2.489 fuse. oan isn woo) % 0876, Po and _ pnd _ (54.7)2.489)(4,892/12 . ne = SAD = 8259 x 108 Because Mge> 2,000, flow is turbulent From Ref. 15, the Prandt! number is 13, 60010.65) _ 19.66 0.02398.259 x 104/"19.66 = 4823 482,30.08) ‘y= "[as92/12) 946 Blume 62°F 2 Determine reservoir te function fi), fom Fg 289. cat, (008 bee wooks(168 /wk) “ ? 12096, a (a0/13} SINGLE-PHASE FLOW CONCEPTS Because x <0,0025, Eq, 2.87 gives ) $uwoo02) + 05772) = 2794 3, Determine heat-transer coefficient using Eq. 2.95. HOT 0.0519 + 0.0047 + 0.8921 + 1.9957 2.9444 he F/B, 4. Determine A From Ea, 2.102, (6.398 an( 2282) ae mayne 0 19,488 f From Shiu-Beggs (Eq, 2.114), A = (0.0189(17.77) (5.892) a0)"" x 0.8)'*54.7° = (0.01499(4.534)(0.631)(2.428)(0.373)(123, 829) = 4,785 fe 5. Determine Tya from Eq. 2.113, Ty = [200.0 ~ (0.0153(10,000) sin 75°] + 0 aoa) + (0.015)sin75°(19,488)(1 — 00 - 1449 +1133 68.47. From Shiu-Begas (Eqs 2.113 and 2.114), Ty = (200.0 ~ (0.01510, 000) sin75°} + 0 + (0.015)sin 75%9, 783)[1 ~ «1900047 00 — 1449 + 60.7 IS. References | Knudsen, 1G. and Kat. DLL Fluid Dynamic and Heot Transfer ‘McGraw-Hill Book Co ine, New York City (1958), 2. Moody LF: “Faction Facto for Pipe Flow” Tran, ASME (1944) (6,NO.B 674. 3. Allen Te and Ditsworth RL: Fld Mechanie, McGraw-Hill Book (Co Inc, New York City (1975), 4. Poise. L Compie Rens (1840) 1, 961 apd 10815 (1841 12,112 5. Drew, TB., Koo, EC, and McAdams, WH: Tran, AICHE (1930) 28.55, 6. Blasi, Hs Z Math Phys (1908) $6 1 1. Nikurase, 1: Forechungehef (1933) 301. r 8. Colebrook, CF: “Turbulent Flow in Pipes With Panculr Reference ‘othe Transition Region Between the Sinooth and Roush Pip Lawes 4 dna Cv Eg. 1939) 1,13, 9. Bri JP and Beggs, HD. Two-Phase Flow in Pipes, Uf Tulsa Tl 3, Oxlanoma (199). 10. Zigrang DJ. and Syivester,ND.:“A Review of Explicit Fiction Fac tor Equations."J. Energy Res. Tech. Ue 1988) 107,280, 1M. Theory and Practice ofthe Testing of Gas Wel thie eon, Energy Resouces Conservation Board, Calgary (1973) 12,Cullendr, Mand Smith, RV. “Praca Solution of Gas Flow Equations for Wells and Pipelines Wits Large Temperature Graeme 4PT (December 1956) 281; Trans, AIME. 207. 13, Buringion. RS: Handbook of Mathematica! Tables and Formulas fit eon, McGraw-Hill Book Co ne, New York Cy (1973) 14, Marines, A.Ee al: “Prediction of Dispersion Vscony of Old Wa- {erMixture Flow in Horizonial Pipes,” paper SPE 18221 presented a the 1988SPE Annual Technical Conference and Exhibition, Houston 2-5 October 15, Bind, RB. Stewart, WE, and Lightfoot, EN. Transport Phenomena, hin Wiley & Sons, New York City (1960). 16 Metane. A.B. and Reed C."“Flow of Non-Newonian Fids--Cor: ‘elation ofthe Laminar, Tanston, aad Turbulent Flow Regions AICHE. (1985 1, 34, 17 Dodge, DW. and Metzner, A.B: “Turbulent Flow of Nom Newtonian Systems." AICRE J (1958) 8 Te, 18 Govier, G W.and Azz, K: The Flow of Complex istaresin Pipes, Van Nostrand Reinhold Co, New York Cty (1973). 19, Sula, AP. Bobok, E. and Navel “Determination of Turbulent Pressure Loss of Non Newtonian Oil Flow in Rough Pipes” Rheoloy, fe Acta (1981) 20, No.5. 20, Heya JF: "A Greens Function Solution for he Cae of Laminar In § = (22208) = o.o0012 From Fig. 22 oF Fg. 2.17, f=0.0135. 6, Determine pressure gradient neglecting kinetic energy eects: rom Eq. 42, de _ (00135)27.04)°7.837 & * BATOHGTTHOS) 2.174) (32.178) + 27.08) = 0.70 + 27.04 = 27.74 pst 0.193 psi. 17. Ifcalculations had been made with the incorectliquid-holdup value of 0.3, the pressure gradient would have been 0.14 psift Gray Method. Gray developed a method to determine the pres- sure gradient ina vertical gas wel that also produces condensate Auids or watecS A total of 108 well-est-data sets were used to de- velop the empirical corelations. OF these data sets, 88 were ob- tained on wels eporedly producing fee liquids. The authors eau- tioned use ofthe method fr velocities higher than SO fuse, noni diameters greater than 3. in., condensateliquid loadings above SO DBOVMMscf, and waterfiqui loadings above 5 BbUMMscf Gray proposed this equation to predict the pressure gradient for two-phase flow in vertical ges wells, Oe, 19 — pyr £2) GBP + oe -omeE le Liguid-Holdup Prediction. From a dirvensional analysis and se- lected laboratory tests, three dimensionless groups of variables were selected to correlate liquid holdup, au Div ft a) wib.= Pl 4.13) ay ‘The resulting correlation developed to predic iguid holdup was Ret Hate (sy 31 ! L where B eon aossn( «Pato Lis + 061704 Ceti ain Gray stated that the liquid holdup in condensate wells often tends to be smaller than in oil wells producing at comparable pas/iquid tatios. Tis probably arises from natural ime lags inthe ashing and ‘condensing processes, and also from the lower average interfacial surface tension in condensing systems compared with Mashing black ol systems. The formulation of Eq, 4.15 such thatthe liquid holdup is ordinarily very near to the no-slip holdup, Friction-Factor Prediction, The effect of liquid holdup on fic- tion loss can be interpreted. asa variation in wall roughness om that ‘ordinarily experienced in single-phase, dry-gas flow. Gray pro posed that friction factors for wel-gas wells were wholly dependent ‘on «pseudo-wall-roughness factor because Flow is normally inthe fully-developed turbulent region, Friction factors can be obtained from the Moody diagram (Fig. 2.2) from Eq. 2.16 for completely, turbulent flow. The pseudo-wall-oughness €, was correlated with modified Weber number, similar to that used inthe Duns and Ros? ‘method for mist ow. Defining @ roughness variable, 2850, oe 4.18) Pave o then, for 80,007, (4.19) and, for R<0.007, Re’ ~ e,) eet TE: cee sisteeeees 420) ‘With the restriction that ¢ 22.77% 10-5, ¢g is the absolute wall roughness for single-phase gas flow. It is noteworthy thatthe cor ‘elation finally adopted includes both the effects of unusually high apparent roughness observed in cern laboratory test conditions And the eects of low apparent roughness in certain field systems, Asheim Method. The Asheim® method, or MONA, is indepen- {ent of flow patter, but does petmit the selection of thee empirical Parameters for either the bubble- or slug-flow pattems. Once se lected, however, these parameters must be used for all multiphase. ‘Now caleulations. The method can be usedin conjunction witha his. tory-matching procedure to adjust the three empirical parameters {and minimize computation errors, ‘The pressure gradient is determined from APE psi coccccseesesne AD Liguid-Holdup Determination, The gas and liquid velocities can ‘be determined from wae it oO seeeeeees 422) and “ - 42) Eqs. 422 and 423 canbe combined 10 give A na e Nama: - 424) 2 Eq, 4.25 gives the functional relationship assumed between the {- and liquid-phase velocities, Me + a. exo (425) Using Es. 4.22 and 4.25 to eliminate the gas- and liquid-phase velocities in Eg. 4.24 gives [lv +a 0) + teara] as mn ~ tetova ee . 2) Two limiting cases must be Uefined to make the holdup expres- sion complete. When the constant -slip term, a, (buoyancy) is 2er0, the holdup becomes Hae, aan ‘When the liquid superficial velocity approaches zero, the flow situae tion may correspond to gas bubbling through stagnant liquid. The holdup for this case is found by rearanging Eq. 4.26 to obtain Higa (428) Frition-Factor Determination. This two-phase fretion-factor correlation was used in Eq, $2 Pa b exxeccom cesceeses 429) ‘where 9, is defined in Eq. 3.24. The wal friction factor. fis deter ‘mined from Fig 2.2 fora no-slip Reynolds number defined by Pata Nae, = Pate (430) Parameter Values. For slug flow, the normal values of the three parameters area) =1.2,a)=0.38 Jed, and ay 1,0. For homoge. ‘neous flow the normal values of the three parameters are ay = 1:0, 420, and a3~ 1.0. Asheim showed that, by selecting either slug flow or homogeneous low, the tree parameters ean be adjusted for an optimized fi of Field data, Category “c." The methods considered inthis category differ in how they predit flow pattern and how, foreach flow pattern, they predict liquid holdup andthe fiction and acceleration pressure-gra- ‘dient components. For vertical flow of a homogeneous slip mixture, Eq. 3.26 can be expressed as $+ (8) ra e(¥). a ‘Duns and Ros Method. The Duns and Ros? method isa result of an extensive laboratory study in which liquid holdup and pressure gradients were measured. About 4,000 two-phase-flow tests were Conducted in a 185-fehigh verical-flow loop. Pipe diameters ranged from 1.26 10 5.60 in. and ineluded two sanulus configura tions, Most ofthe tests were at near-atmospheric conditions with ait for the gas phase and liquid hydrocarbons or water as the liquid ‘Phase. Liquid holdup was measured by use of a radioactve-tracer fechnigve. A transparent section permitted the observation of flow pattem For each of three flow paterns observed, correlations Were are given in Fig. 4.7. They are functions ofthe liq- uid velocity number, Nz. F ean be obtained from Fy Ny where Fy and Fy also are obtained from Fig 4.7 ‘The friction pressure-gradient component for bubble flow is tiven by (43 Bar 438 (8), Pitutn 39) 4 ‘. ra ‘e Fig. 4.6—Duns and Ros? bubbla/siug tranaltion parameters. 2 eeteezeng 5 Fig. 4.7—Duns and Ros? bubbie-low,sip-velocity parameters. From experimental data, Duns and Ros developed this equation for f: =hf FES B cece cesses (440) ‘The friction factors govemed mainly by f, whichis obtained from ‘Moody diagram (Fig. 22) as function ofa Reynolds number for the liquid phase. aan corection forthe in-situ gasiguid ratio andis given in Fig. 4.8. The factor fis considered by Duns and Ros ‘sa second-order correction factor for both liguid viscosity and in tu gasfiquid ratio. k becomes important for kinematic viscosities ‘greater than approximately 50 cSt and is given by aay ‘Duns and Ros considered the acceleration component ofthe pres sure gradient to be negligible for bubble flow. ‘Slug Flow. Slug flow exists if Nevyy < Nev € Nyy For lug flow, the dimensionless sip-velocty number is (we + Fe (+ FM) where Fs, Fe, and F7are given in Fig. 49 as functions of the liquid viscosity number, Nz. and S= (14h) 43) TAR OU On carreras (444) ‘The fiction pressure-gradient component for slug flow is calcu lated exactly the same way as for bubble flow. Also, the acceleration ‘component for stug flow is considered negligible Mist Flow. Mist flow exists if Nev > Nyrpy Duns and Ros 35- sumed that, at high gas flow rates, the liquid transported mainly as small droplets. The result is nearly « no-slip condition beeen the phases. Thus. $= 0, v, = 0, and H, = 2,, The mixture ‘density for use in the elevation component of the pressure gradient then is calculated from Eq, 3.23. Friction in the mist-flow pattem originates from the shear stress between the gas and the pipe wall, Thu, the friction component of the pressure gradient is determined from Divi, 445) Because there is no slip, the friction factor is obtained from a Moody diagram (Fig. 2.2) asa function ofa Reynolds number for the gas phase “ (446) Duns and Ros noted thatthe wall roughness for mist flow is the thickness of the liquid film that covers the pipe wall. Waves on the film cause an increased shear stress between the gas and the film that, in tum, ean cause the greatest part ofthe pressure gradient, ‘These waves result from the dag of the gas deforming the film in ‘opposition tothe surface tension, This process is affected by liquid Viscosity and also is governed by a form of the Weber number ewe Mu = oe aan) ‘This influence was accounted for by making Ne function ofa d- ‘mensionless number containing liquid viscosity, we = pale (448) Fig. 4.10 shows the functional relationship, where the coordinates ate Nye #5. Nwe Ne ‘The value of roughness may be very small, bu the relative rough- ress never becomes smaller than the value forthe pip itself. At the transition o slug flow, the waviness ofthe film may become large, with the crests of opposite waves touching and forming liquid bridges. Then eld approaches 0: Between these limits, eld can be ‘obtained from equations developed from Fig, 4,10 e _ 207480, Nua $ 0.005; 5 = Pid (4.49) at Mud, > 0005; = STE ny). 4.50) Ped Where dis in feet, is in fet per second, pis in pounds per cubic foot, and a, is in dynes per centimeter ‘Values off forthe mist-flow pattern can be found for eld > 0.05 from this extrapolation ofthe Moody diagram 4 asi [+>eu(0279)] Aste wave height of the film on the pipe wall increases, the actu al area available for gas low decreases because the diameter open to low of gas is now d~. Duns and Ros suggested thatthe fiction component of the pressure gradient could be refined by replacing d -MULTIPHASE FLOW IN WELLS. - Fig. 49-Duns and Ros? slug-tiow,slip-velocity parameters. with de and 5p with vspl'd—)? throughout the ealeulations “Tis results in atnal-and-ror procedure to determine & Tn mist flow, acceleration often cannot be neglected ait was in bubble and slug flow. The acceleration component ofthe pressure radial can be approximated by (@),.- =" (@) ss 459) ‘Beggs and Brill! provided a derivation of Ei, 4.52. If we define a dimensionless Kinetic energy, Ej, a8 Vag a. 453) ‘the total pressure gradient can be calculated from fg -@ ‘The dimensionless kinetic energy term, Es elated tothe sonic velocity of a two-phase mixture. Bis similar to the Mach number for compressible flow. Sonic conditions are reached when the Mach ‘number becomes 1.0, When this occurs ina pipe, a “shock” is estab- lished, across which the pressure gradients infinite. Eq. 4.54 il predict the same result if & becomes 1.0. Unfortunately, Eq. 4.53 ‘can incorrectly yield Ey values greater than 1.0. (asa) Transition Region. The transition region exist if Noy, < Ng < Meryae this region is predicted, Duns and Ros Sus gested linear itepoation between the low-pattem boundaries, Ney aNd Neyo oblain the pressure gradient. This wl equre a calglation of Pressure gradients with both sluglow and mist fow comlations. The pressure gradient inthe tansiton region then is calutated fom dp ‘P) ay 2 (8), -4(8)_+0-0() «as we Nenu ~ New ae Nein Nein 880 Increased accuracy was claimed in the transition region ifthe gas {density used in the mist-flow pressure-gradient calculation was mo- ified to be sn PO Re, where y= gas density calculated atthe given conditions of pres- sure and temperature, This modification accounts for some of the liguid being entrained inthe gas. Fig. ¢.10—Duns and Ros? mist-flow, fllm-thickness correlation. ‘Modifications. Two proprietary modifications of the Duns and Ros ‘method have been developed but are not availabe in the literature. ‘The frst, known as the Ros field method, involved modifications based on carefully obtained daa from 17 high-GOR vertical oil wells. Ta joint Mobil-Shell study undertaken between 1974 and 1976, ‘4 modification resulted in the Moreland-Mobil-Shell_ method (MMSM).In this study, 40 vertical oil wells, including the 17 used in te Ros field method, and 21 directional wells were selected as thebasis forthe modifications. The MMSM method includes liquid- holdup correlations derived from the data for bubble and slug flow thatare simpler in form than those used inthe original Duns and Ros ‘method. Possible discontinuities at flow-pattern-transition bound- aries also were removed. Example 4,3—Using the Duns and Ros Method, Calculate the ‘Vertical, Multiphase-Flow Pressure Gradient for Example 3.2. Given! yy =0.97 cp, a,=8.41 dynestem, 4g =0016 cp, and €=0.00006 1 1. Determine the flow patter: From Fig, 45, Region I (froth or bubble). Verification From Fig. 46,£)= 1.0; 021.1 Nay = bi + baNees = 1.0 + G.AG187) = 14.06 and Nye = MSA < Nery bubble ow exists. 2. Determine liquid holdup: From Fig. 312, F)=1.2, F;=024, Fy From Eq, 4.38, _ 265) RoI Tas From Eg. 437 for bubble flow, 3,and F4=265. 6. sit +a2niian +0.116( 9) T+ in 4.946, From Bq. 435, 559. 3s 3. Determine friction factor: From Eq. 441, 810769109) _ 545 1s eABHUTENOINOD «145 5 1 md = Dons § = 200006 — pore From Fig 22 Ea 217./;=0015 omg a8 _ 01791386043. oo = tare 0°! “he i0 Fro Eq 42, fon 0008 = 10006 From Bq, 440 0) (00175,52 = aos 4, Determine pressure pradient neglecting kinetic energy effects rom Eqs. 4:31, 3.22, and 4.39, dp _ (0.01759047.6113,9710.83) a anty03) . . 63 (ar6110 859 + (5.88) - 0559379 = 080 + 29.21 = 30.01 pst 208 pst Orkiszewski Method. Orkiscewski® tested several published cor- relations with field data and concluded that none was suiicienty accurate forall flow pattems. He then selected what he considered to be the most accurate corelations for bubble and mist flow and Proposed 3 new correlation for slug flow. The shug-flow corelation ‘Was developed with the Hagedorn and Brown® data, OrkiszewsKa Selected the Griffith and Wallis!2-¥ medhod for bubble flow andthe Duns and Ros? method for mist flow. Flow:Pattern Prediction. Orkiszewski used the Duns and Ros ‘low-pattem transitions forthe boundaries between slug flow and mist flow. including the transition region between them. Eqs. 4.32 1nd 4.32¢ defined these, For the boundary between bubble How and slug flow, he chose these criteria established by Griffith and Walls, Bubble/sug transition: A is (asa) where 4, = 1071 ~ 022182, (459) and rp = USEC, vyp = FUSE, ically to be 20.13, Bubble Flow. Bubble low exists if Ay = 1 liquid holdup for bubble flow is determined from worlfiet- ) #) 460) fi, and Lp is constrained algebra AS days The 36 on a3 + C02 4 os 4 ° ao ry =o Maas: Fig. 4.11—Gritith and walis3C, correlation, which s equivalent to Eq 4.36 forthe Duns and Ros correlation. r- Kiszewski adopted the Grifith'* suggestion that 0.8 fsec isa good approximation ofan average vy. In See. 4,2.2 we will show that» isa function ofthe gas and liquid densities and surface tension. The liquid holdup determined from Eq. 4.60 then isused to calculate slip density with Eq, 3.22, which in turns used ocalculate the elevation, ‘component ofthe pressure gradient The friction pressure-gradient component for bubble flow is aiven by (de) 2), ‘The fiction factoris obtained froma Moody diagram (Fig.2.2) as, ‘8 function of relative roughness and Reynolds number forthe lig. id phase, Prubsh tty an Mg = Palate (462) ‘The acceleration pressure-gradiem component for bubble flow was considered negligible, Slug Flow. Slug Now exists if dy > Ayr 8M4 Nex < Np The slip density is calculated from bliss + ¥4) + Posy eth tet (463) Orkiszewski developed Eq. 4.63 by performing mass and volume balances on atypical slug unit consisting ofa Taylor bubble'® and 8 liquid slug. A similar Griffith and Wallis development neglected the presence of aliquid film around the Taylor bubble and the possi bility of liquid droplets being entrained in the Taylor bubble. Conse. ‘quently, Orkiszewski proposed the lst term in Eq. 4.63 to account forthe distribution of liquid in these regions, This modification was ‘meant to extend the Griffith and Wallis work to include the high-ve- locity-flow range. Griffith and Wallis correlated the bubble-rse velocity, vp, by the relationship w= CC fed, (4.64) where Cy and C2 are expressed in Figs. 4.11 and 4,12 as functions Of May, 2d Ne (465) MULTIPHASE FLOW IN WELLS, Fig. 4.12—Grittth and TABLE 4.1—ORKISZEWSKI® LIQUID DISTRIBUTION ‘COEFFICIENT EQUATIONS Continuous Value Equation Liquid Of Mn Yor Phase see) r Water 10 472 Water >10 473 ou 10 44 oi 210 475 and = Poin Mes = Te - (466) Fig 4.12 was exrapolteaso that could be evated at higher liquid Rol numbers, When Cy cannot be ead fom Fig 12 veean be aula rom hi sto euaons. When Nag, = 5,00, v= (0586-4878 10-%%)VEdcceeeeee OD) When Nas, = 8.000, va (035 +874. 10°, ) 6d 468) When 3,000 < Np, < 8,000, Lot meee Pee ae where v= (0251 + 874 x 10°*%q,) 6d +. 70) Because vp and No, are interrelated, determination of v requires an iterative procedure when using Figs. 4.11 and 4.12 or Eqs. 467 through 4.70. The procedure follows 1 Estimate a value of vp, A good First guess is w=Osied. .. ary 2. Calculate Nps, using the value of » from Step 1 3. Calculate vy using Eq. 4.64 or Eqs. 4.67 through 4:0. ‘4 Compare the values of, obtained in Steps 1 and 3. Ifthey are not ‘sufficiently close, use the value in Step 3 a the next guess and go to Step I. Continue until convergence is achieved. This procedure will converge in one step if the new valve of Nr, doesnot result in ‘a change among Bqs. 4.67 through 4.70. MULTIPHASE-FLOW PRESSURE-GRADIENT PREDICTION Wallis'® C, correlation. (Orkiszewski used the dat of Hagedorn and Brown to calculate and comelte the liquid distribution coefcient, P: Orkiszewski did not de- fine criteria to esublish which liquid phase is the continuous phase ‘when both oil and water are present. Fig. 33 canbe used fortis pure pose. Table 4.1 identifies liquid distribution coefficient equations, De- pending onthe continuous liquid phase and the value ofthe mixture ve- locity the value of Tis calculated from one ofthese expressions, P= SOBs 9681 + 0232104 = 0408topd, am = 20551084 9709 - 01621084 = ost gd, a) govzTiti +) 9264 + a167logrn ant r= 2077s + 9161 + ossot0ga + x, ars) x = gr SUH Ys 207 «ost ne] a7) and i is in centipoise, dis in fet, and vq isin feet per second ‘The value of Fis constrained by these limits Ive < 10, Pe ~ 0.065% a and, if vm > 10, these constrains are supposed to eliminate pressure discontinuities be- tween equations foc I. Fig. 4.13 shows the discontinuities can be sig- nificant. A method to eliminate them is described inthe next section. a — =| bees Fig.4.13-Liquid distribution coeticient discontinuities. 16 ‘The ftition pressure-gradient component for slug flow is ‘dp\ _ few] (vs. + vp) (2) -42[(eee) «¢] a seston ie Monty apo 22) ote Hejl a ee hea 22 by te Secmmen Bia et ace mean is predicted. Modifications, The Orkiszewski method can cause aconvergence ‘problem in the computing algorithm, dessribed in Sec. 3.7, for cal. culating pressure vs. depth traverses in wells. The problem results from discontinuities between Eqs. 4.72 and 4.73 for water as the continuous liguid phase, and between Eqs. 4.74 and 4.75 for ol a8 the continuous phase. Brill!® showed tht the constraints given by ‘Eqs. 4.77 and4.78 are ineffective in eliminating the pressure discon ‘inuites that result from the use of either pair of equations. Triggia” suggested that coefficients in Eqs. 4.73, 4.75, and 4.76 be modified 0 thatthe slopes of these curves are retained but the discontinuities are eliminated. This solves the convergence problem but also may. affet the accuracy of results. The resulting equations for water and Oil, respectively, were 0.013 logue, oa ~ 0428log¢ and (0287 ~ 0.162108 ¥~ (4:80) 00127 ogbes + ae + CU ~ log ra), where a 201 oglu + 1) C= Gt + 0.397 + 0.63t0¢d. ...... 4.82) O17 + 0.113 ogd (aan ‘The liquid distribution coefficient also can become too large & ‘negative number for high flow rates corresponding to lange values ‘of vm. When this occurs, Eq 4.63 ean become less than the no-slip ‘mixture density. A second modification in the Orkiszewski corel tion isto replace p, with p, when this occurs. I also becomes nee ‘essary 0 replace Eq, 4.79 forthe friction pressure-praient compos nent. A logical replacement is ee) _ Bev az), - "2a * here fis obtained from the Moody diagram (Fig. 22) by use of a Reynolds number given by a =P ae (483) Pura snmancaon win A Trg Peabo aere (184) 38 Example 4.4—Using the Orkiszewski Method, Calculate the ‘Vertical, Multiphase-Flow Pressure Gradient for Example 3.2, Wen: = 0.97 cp, y= 0.016 ep. G,=8.41 dynesiem, and 00006 1. Determine flow pattem: From Eqs. 4.58 and 4.59, 7.83? days = LON ~ (02218) TSE = ~ 26.1 Because ~ 26.1 < 0.13, st 2, = 0.13 = 0507 = 0.493 > 4, flow pattem is not bubble flow, From Eq. 4.32, Mery, = 50+ (36) 11.84) 476. Because Nyy Ls4 < 4762, Flow pattem is slug low. 2, Determine slip density: From Eq, 4.66, Fig. 4.12 cannot be used, Guess v, = 0.5 Jed = (0.5) (217805) = 20 We. From Eq. 465, Ng, = {1:488047.6112.0)0.5) rn From Eq. 4.68, = 73x 108, v= [(035) + (874 x 10-8286 « 105) GLAS 1143 fusee, From Bg. 4.65, 488)(47.61(11.43)0.5) ., Nye, = ERO ENADOS 4.15 5 19 Because Ny, > 8,000, vy = 11.43 faces comect. From Bq. 474, F = 0012106097 + 1.0) _ 954 (05) + (0.167) 0p(7:83) + (0.113) 10g(0.5) = 0159, From Eq. 477, Is — 0.159 = ~ (0.065)(797) = ~ 05187 Yes. Therefore, T= ~ 0.158. MULTIPHASE FLOW IN WELLS, From Eq. 463, (47.61)3.97+ 11.43) +(5.88)3.86) p= Os + (47.61) 0.159) = 3925-757 = 31.68 bau 3. Determine fiction factor: Nae, = 288 X 108 00006 20006 — 0.00012. 22 oF Bq, 2.17, f= 00188, 4. Determine pressure gradient From Eqs. 4.31, 4.63, and 4.79, La 7ay+ 1143 dp _ (00158(47.61)7783)"| (ae ~~ WELITHOS, + 1.68)(334174) ) on] = 092 + 31.68 32.60 pstft = 0.226 psi. ‘Adi etal Method. The Aziz et a9 method uses many ofthe funda ‘mental mechanisms that form the basis of modem mechanistic models. Flow-Pattern Prediction. Fig. 4.14 shows the low-pattern map ee Sey prc senna a w= valsoie) (G25) 485) ant (72) 2. \)""* w, = val BE)] - 486) " _ % 4 % os Me Fig. 4.14—Aziz of a!9tlow-pattern map. (Reproduced with per- mmission of the Petroleum Soc) -MULTIPHASE.FLOW PRESSURE- GRADIENT PREDICTION ‘The bracketed terms in Eqs. 4.85 and 4.86 are attempts to validate the flow-pattem map for fluids other than air and water. For an aie! water system at atmospheric conditions, the coordinates are the st= perficial velocities of each phase. ‘These equations represent the flow-patterm transitions in Fig. 4.14, er 1, = 051(1008, (437) Ny = 864 38M. ccc cesses (488) and Ny = To{100N,) °", ses 89) ‘where vz isin feet per second, ys is in feet per second, ps in ‘pounds percubic feet, p, isin pounds per cubic feet, and isin dynes per centimeter. Bubble Flow. Bubble low exists if N_- Aziz eta state that (495) (496) sn (498) » and m is determined from N m 2250 0 250 >N,> 18 69N;% 18 25 ‘The fiction pressure-gradient component for slug flow is deter mined from dp’ Hw, (2) 28 nr lm “The fiction factor is obtained from a Moody diagram (Fig. 2.2) asa function of relative roughness anda Reynolds number given by Pera Ne = GE + (4.100) ‘The acceleration pressure-pradient component was considered neg- ligible for sug flow. ‘Mist Flow: Mist flow exists when N, > Ns for N, <4 or N,>265 foc N,>4. Aziz etal recommended the Duns and Ros mist-low ‘method be used to calculate pressute gradient for this low pattern, Transition Region. The transition region exists when N2 4, When te transition regions predicted, the Imistflow equations. To obtain the pressure gradient, linear inter- polation is performed, similar to the procedure deseribed in the Duns and Ros method. Thus, 101) 4.102) Modifications. AL-Najjar and Al-Soot! showed that improved results could be obtained with the Aziz etal method if the flow-pat- tern map in Fig. 4.14 was replaced with the Duns and Ros map in Fig. 45. Their conclusion was based on a comparison ofthe pre dicted and measured pressure drop fr 80 tess on 15 flowing well, in lrag, as well as on data of Poetmann and Carpenter! and Orkis- zewski® The wells from Iraq were flowing primarily through acas- ingtubing annulus, Example 4.5—Using the Azizetal. Method, Caleulate the Vert cal, Multiphase-Flow Pressure Gradient for Example 32. Given: 4)=0.97 ep. 4,=0.016 cp, p= 841 dynesiem, and £=0,00006 f. 1. Determine flow pattern: From Eqs. 4.85 and 4:86, en ei) [(FANBAN] ~ 225 and sol BE] = 65 From Fig. 4.14, sug flow exists, Verification, M, = 0.511(100)6.35)" = 1.55 for N, 24, 1.55<2625-<265, «0 Fr rman Wat? « ‘00 vaaoe "3000 Mee, Fig, 4.15—Chier et al*® C, correlation, 2. Determine liquid holdup: From By. 498, Yose2.178)47.61)47.61 — 5.88) : (0.97/1,488) = 131 x 10 m= 10 From Eq. 4.97, wy, = 32174005) (47.61 - 5.88) 881/858) 112 x 10, From Ba. 496, cx osashi —ecoomif — 34s, From Eq. 495, (0.345), /OHATS a7) = 1295 fee From 491, (1.2)(783) + 1.295 = 10.69 see. From 63.490, 21-386 Hy = 1 33S, = 0639, MULTIPHASE FLOW IN WELLS 3. Determine friction factor: (4, 488)647.61)0.8300.5) _ Nye = GET NOD = 2.86 x 10° and = 9.00006 _ § = 9.00006 - o.00012. From Fig 2.2 or Eq, 2.17, =0.0158, 4. Determine pressure gradient dp _ (0.0158)47.61)(0.639)7.83)" a> aan. t74y03) + 4+ Ga0K1 — 0629) 2179 1(47.61)(0.639) + (5.88)(1 9.639) F174) = 092 + 32.55 = 33.47 pst = 0232 psi, Chierii etal, Method. The Chiericiet al! method to predict ‘low pattem and o aleulat liquid holdup and pressure gradient ia the bubble. transition, and mist-flow regions is identical that of (Orkiszewski. The only difference isin the slug-flow region, Thus, only slug-flow treatment is presented, ‘Slug Flow. The Chierci eral. slug-flow treatments almostidenti- cal to that of Aziz eral Liquid holdup is determined from (4.103) where w= Cased (4.108) CC is obtained feom Fig, 4.11 as a function ofthe bubble Reynolds number and is essentially 0.35 for all cass, Values of Cz ean be ob- tained from Fig 4.15 for liquid Reynolds numbers less than 6,000. For his range, Fig. 415isidentical to Fig. 4.12 used by Orkiszews- i. Chierci eral. show in Fig. 415 that the extrapolation equations proposed by Orkiszewski in Eqs. 467 through 4.70can resultin ds- continuities. From earlier work by Nicklin er al.2° Chirici eral proposed that for liquid Reynolds numbers greater than 6,000, (4.108) Curves based on Ea. 4,105 for extrapolation also are ploted in Fig 4.15 and show no discontinuities. Combining Eqs. 4.104 and 4.105 gives vy = 02, + 035 fad 4.106) “The resulting prediction of iqud holdup essen i identical wo that Aree "The pesur-graient fiction component is identical that of Acero andthe aecleraion compare isneplsed os fom Example 4.6—Using the Chierii etal. Method, Calculate the ‘Vertical, Multiphase-Flow Pressure Gradient for Example 3.2. Given: pp =0:97 cp, ay =0.016 cp, 0,841 dynesiem, and 1£=0,00006 f. 1. Determine flow pattern: Orkiszewski predicted slug Now. MO TIPHASE. OW PRESSUIRE-GRADIENT PREDICTION 2, Determine liquid holdup: From Eq 4106, 5 = (0240.8) + 035) (GRITHOD = 297 hee From Bg, 4103, esos eee He= 1 — Gaye 2a = 0.682 3. Determine friction factor: Me™ pot 2.86 x 10° ana = 0.00006 _ 5 = 200006 - oonora From Fig22 or Fa. 2.17, = 0.0158 4. Determine pressure ardent dp _ (00158%47.61)0,642)7.83)" @ (@32.178N03) 0217) + [476110642 + (S88) — O64) EETD = 092 + 32.67 = 33,59 pst 0.233 psi. ‘Beggs and Brill Method, The Beggs and Brill" method was the firstone to predict flow behavior a al inclination angles, including Airetional wells, Their test facility was 1- and 1.5-i. sections of acrylic pipe, 90 long. The pipe could be inclined at any angle from the horizontal. The fluids were air and water. For each pipe size, liq- uid and gas rates were varied So that, when the pipe was horizontal, all low pattems were observed. After a particular st of flow rates was established, the inclination ofthe pipe was varied through the ‘ange of angles so that the effect of angle on holdup and pressure sradient could be observed. Liquid holdup and pressure gradient Were measured at angles from the horizontal of 0°, 5°, 10%, 15%, £20", £35°, 455°, £75", and 90°, The correlations were developed from 584 measured tess ‘Beggs and Brill proposed the following pressure-gradient equa- tion for incline pipe P+ pissin ¢ EL) whore Eis given by Ba, 4.53 and Ps = Pun + Pll ~ Hus): 4.108) Flow-Pattern Prediction. Fig 4.16 tustrates the horizontal flow pattems considered by Beggs and Brill. On the basis of observed. Flow pattems for horizontal flow only, they prepared an empirical ‘map to predict flow pattern. Their original flowpattemn map has been modified slightly to inlude a transition zone between the seg- regated- and intermitentflow patterns?! Fig, 4.17 shows both the original and the modified (dashed lines) flow-patera maps. Beggs and Brill chose to correlate flowpattem transition bound ares wth no-sip liquid holdup and mixture Froude number, given by seeeeseeees G09) a Segregated ‘nnutar Intermittent Fig. 4.18—Beggs and Bril!t horizontal-flow patterns, ‘The equations forthe modified flow pater transition boundaries are Ly = 31612, (4.110) 0.000925", any Ly = .0a7 4, 2) and L, = 05175 (uy ‘The following inequalities are used to determine the flow pattern that would exist ifthe pipe were horizontal. This flow patter is a correlating parameter and, unless the pipe is horizontal, gives no in formation about the actual flow pattem Segregated. Ay < 001 and Np < Ly Froude Numbet. Neg Input gu Canton, Fig. 4.17—Beggs and BrilP horizonta-fow-pattern map. Hoy guid Holdup, ration {Angle of Pipe From Horizontal. degrees Fig. 4.18—Etfect of inclination angle on liquid holdup.** Ay = 00L and Ny, < Ly Transition, Ay = OOLandLy = Nyy = by Interment 001 $A, < O4andL; < Ny <= Ly Ay 2 Oand Ly < Nyy S by Distributed. Ay < OsandNy = by Ay 2 Osand Ny, > Ly Liquid: Holdup Prediction. Different comlations for liquid hold up were developed for three horizontal-low patterns. The liquid holdup that would exist fthe pipe were horizontals frstealeulated and then corrected for the actual pipe- inclination angle, Fig, 4.18 shows the variation of liquid holdup with pipe inclination for thee oftheir tests. The liquid holdup wasa maximum atabout + 50° from horizontal and a minimum at approximately ~50°. For high low rates corresponding to what now is called dispersed-bubble flow, Tiquid holdup essentially was independent of inlination angle ‘The same equations are used to calculate liquid holdup for all flow pattems. However, the empirical coefficients are different for each flow pattern The liquid holdup that would exist if the pipe were horizontal is calculated first from aa Fe au MULTIPHASE FLOW IN WEIS. i ‘TABLE 62—BEGGS AND BRILL" EMPIRICAL (COEFFICIENTS FOR HORIZONTAL LIQUID HOLDUP ‘TABLE 4.3_BEGGS AND BRILL" EMPIRICAL COEFFICIENTS FOR C Flow Pater 2 > e Segregated 0980 © 04886 © 00868 Intermitent oss 0535100173, Disirbuted 1.065 05624 0.0609, ‘with the restriction that Ho) 2A The coefficients , 8, andc are de- termined from Table 42 for the appropriate horizontal flow pattem. Eq, 4.115 gives the liquid holdup corrected for the effect of in clination angle Hoy = Hy ais) ‘The factor to corect liquid holdup fo the effect of pipe inclination is given by Y = 10+ Csi. 80) ~ 033355088), (16) where @ isthe actual angle of te pipe from horizontal and Cis de- fined by c 10 ~ a ileti¥M). ay ‘with the restriction that C20, The coefficients ef and h are de- {termined from Table 4.3 for the appropriate horizonal-low pattem. ‘When the flow pattern falls inthe transition region, the liquid holdup must be interpolated between the segregated and inermit tent liquid-holdup values as Hay, = MHaidyy + OU ~ Aags cats) 2.119) Frietion-Factor Prediction. The two-phase friction factor is culated from P= Sab s+ 4.120) ‘The normalizing tition factor, fis determined from the smooth pipe eurve on the Moody diagram (Fig. 22) or from Eqs. 2.12 oF 2.13 by use of a Reynolds number defined as = Pata Mae GE ee a0) and jg is obtained from Eq, 3.21 ‘The ratio ofthe two-phase friction factor to the normalizing fic- tion factor was correlated with the Beggs and Brill experimental data, resulting in Me G12) . Iny 0.0523 + 3.182 Iny ~ 0.8725(In)* + 0.018530iny)* (4.123) .124) Ea, 4.123 contains discontinuities for y values of about 2.63 x 10~4 and 1.016. It is unlikely thet the discontinuity atthe ‘smaller value of y would ever be encountered, However, itis neces- sury that 5=0 for y= 1.0 0 ensure thatthe correlation degenerates MULTIPHASE FLOW PRESSURE GRADIENT PREDICTION Flow Paton 6 r h Segregated 0011 3.7680 3.5390 1.6140 ‘phi Inermitent 2.960 0.3050 0.4473 0.0978 ‘uphl Distroutes No corecton phil Alpatioms 4700-03692 0.1244 — 0.5086 downhill, to single-phase liquid flow. For s that is used when 1, Ayn = 0924 yyy 4.126) 1a <0, Hug) = 06854 ey aim However, the resulting liquid holdup for 8>0° should not be less than 2c. The original Beggs and Brill method has been found t0 ‘overpredict pressure drops in producing wells, Consequently, im proved resulls should be obtained ifthe Payne ea. liquid-holdup Correction factor is applied for well. Example 4.7—Using the Beggs and Brill Method, Calculate the ‘Vertical, Multiphase-Flow Pressure Gradient for Example 3.2. Given! yp =0.97 cp, ay =0016 cp, 0,=8.41 dynesiem, and = 0.00006 f 1, Determine flow pattern: From Eq. 4.108, From Fig 4.17 ford =0.507, the horizontal-flow paternsinter- mitts 2. Determine liquid holdup: From Eq. 4.114 Fy, = QEASHOSODE 9 5g Gai) 6 c From 84 40, € = oxynfa9soson?™u1s9-™a50™] = ~ 0048 < 0. Therefore, C= 0, W = 10, and Hygy = Hy) = 0574, Apply Payne et al correction factor: Hag = (0924N0.578) = 0.530, 3, Determine fection factor: From Eq. 321, Ha = (97050 + (0.016)0.493) = 0.50.¢p. From Ea. 4121, (1, 488)(27,04)7.83N0.5) __ : SD = 31s x 10% From Fig. 2.2 fore rom Eq. 4.124, = 507) (0530 From Eq, 4.123, Ne 00012, f, = 0.0188. (0391-0500 + 3.1820391) — og725(0591) + 0.0185300591)4 = 03873, From Eq. 4.122, foe aun ‘Therefore, 1.473) (0.0155)=0.0228 4. Determine pressure gradient: ip _ (0.0228%27.04)7.83)" a ~~ @G2.178)03) (2.174) + 147.61,0.83) + 5:88N047)] SET sin(90%) = L17 + 28.00 = 29.17 patie. = 0.203 psi ‘Mukherjee and Brill Method. The Mukherjee nd Brill! method as initiated in an attempt to overcome some ofthe limitations of the Beggs and Brill method and o take advantage of new istrumen- tation developed to measure liquid holdup. Their test facility con- sisted of an inverted U-shaped, 1.5-in, nominal ID steel pipe. The ‘closed end of the U-shaped pipe could be raised or lowered to permit flow at any angle from 0° to + 90° from the horizontal Each leg of the U was 56-f long with 22-f entrance lengths and 32-f test sec- tions on both uphill and dowahill sides. A transparent section in cach leg allowed for flow-pattern observations and permitted the use of capacitance sensors to measure liquid holdup. Fluids were ait and Kerosene or lube oil. Approximately 1,000 pressure-drop mea “ ‘TABLE 4.4 MUKHERJEE AND BRILL’? EMPIRICAL, ‘COEFFICIENTS FOR LIQUID HOLOUP: Upail___Down &% =oxe01s ——1.380282 0516644 Ce 0.129675 4.808139 0.789805 Cy -o.19768 4.171504 051627 Cy 2349227 © 56.260268 15519214 0.475686, 0.079051 oari771 % _0.268687 0.504887 0.399052 urements and mote than 1,500 liquid-holdup measurements were obtained fora broad range of gas and liquid flow rates. FFow-Pattern Prediction. For each ol phase tested, flow-pattern ‘maps were drawn on log-log scales with dimensionless gas and i uid velocity numbers asthe coordinates. The transition curves were «drawn on each map. These curves then were fited with nonlinear regression equations. Two transitions were fited for upflow, The Dbubble/sug transition was found be lineata 45° withthe axes and was best fit by Eq, 4.128, Margy = 10% : 28) where 2 = log Ny, + 0.940 + 0.074 sino = 0.855sin*@ + 3.695M,. seeseeeees G29) ‘The slug/(annularimist transition was identical for horizontal and all upflow and downflow angles, However. liquid viscosity had a Significant effect on ths transition, Increased liquid viscosity accel tates the transition from slug to annular/mist flow. Eq, 4.130 de scribes this transition olter-reeeosnng) (4.130) Indownflow and horizontal flow, the bubblelsug transition is de- seribed by aay 431 = 3.003N, ~ 1.138{log.N,)sind = 0.429)l0gN,,)’sin® + 1.132sin8. ses 4.132) ‘This transition generates a family f curves for different inclination angles and liquid viscosities. In horizontal Now, this transition be- comes a function of viscosity only and isa vertical straight line Eq, 4.133 gives the downflow and horizontal stratified flow boundary. ee ee sess 4.133) where 2 = 0321 ~ 0.017Ny ~ 4.267sin8 — 2.972%, = 01033(l0g.V,)° — 3.925 sin®9. (4.134) Indownftow at higher liquid rates, bubble /slug-flow transition oc- ‘urs as gas velocities increase ata fixed liquid velocity. At lower liq uid rates, bubble~/stratified-flow transition occurs below inclination angles of ~30°. From 0° to ~30° at lower rates, however, bubble flow did not occur forthe range of flow parameters considered, Fig. 4.19 gives a flow chart to predict flow pattems by use of these flow.pattern transition equations. The subscripts BS. SM. and ST in Eqs. 4.128, 4.130, 4.131, and 4.133 and in Fig 4.19 rep- resent the bubble/siug, slug(annulae/mst), and the stratified tan- sitions, respectively. [MULTIPHASE FLOW IN WELLS rc Fig. 4.19—Flow chart to predict Mukherjee and Brill! fow-pat- tern transitions. Liguid-Holdup Prediction. The Viguid-holdup data were corre- Tated with an equation ofthe form. H anni + Coane + cyanto + caplgn] - G35) ‘Three different sets of coeficiens were developed for Eq 4.135. ‘One sets for uphill and horizontal flow: one sets for downtil stra tified flow: and the third set is for other downhill flow patterns. ‘Table 4.4 gives the coefficients. ‘Bubble and Slug Flow. The pressure gradient for bubble flow and slug flow is determined from ove sind ay SFP + 0 gee ses 4136) where Pins, an ‘The ftietion factor i obtained from a Moody diagram (Fig. 2.2) fr Eq, 2.17 for an appropriate relative roughness and a Reynolds number given by Ng, = Batll oe = 4.138) where iy is obtained from Eq. 3.21 Sears, we Tr seatne ae Mukherjee and Brill developed an empirical expression forthe fric- tion factor that depends on liquid holdup. Liquid holdup firsts cal- ‘ulated from Eq. 4.135 and appropriate coefficients from Table 4.4, ‘A ratio of holdup values then is calculated as dg Waa Fp eee cece eeeeeeeesecees (140) and a frction-factor ratio, fe is interpolated from Table 48, A n0- slip friction factor, fis obtained from a Moody diagram (Fig. 2.2) -MULTIPHASE-FLOW PRESSURE.GRADIENT PREDICTION Fig. 4.20—Control volume for stratified flow.'2 ‘TABLE 4 5—MUKHERJEE AND BRILL’? ANNULAR-FLOW FRICTION-FACTOR RATIOS fa Saaaeaaas 4.00 007 098 020 120 030 125 040 4.30 050 125 070 1.00 1.00 1.00 10.00 ‘or Eq, 2.17 for an appropriate relative roughness and a Reynolds ‘number given by Eq, 4.138. Then, Sta. aun Stratified Flow. lis tkely that stratified flow will occur only in highly deviated or horiaontal wells. Mukherjee and Brill chose a separated-flow or two-fluid approach to develop a pressure-gradi- cent-prediction method. Fig. 4.20 gives acontrol volume that defines all variables for this approach, {A steady-state momentum balance on the gas and Tiguid phases yields Eqs. 4.142 and 4.143, respectively, SP 2 (5B, + 2M) — pAgssin Ange = = (toaPe + 1) ~ pels sind (a.ta2y and EY Po +e) = in AGE = = lr aePa + 1) ~ peAce sind. (4.143) Either equation can be used to calculate pressure gradient, Muk- herjee and Brill chose Eq, 4142, To eliminate the effect of intefa- cial forces in large-diameter pipes, they suggested the edition of gs. 4.142 and 4.143 to obtain Eq. 4.144, — (eeiPo + FoxPe) ~ rds + PeAsle sind. aay “Mukherjee and Brill concluded that, for most cases involving stratified flow, liquid holdup is suticiently small. W, is then small compared with Py. Consequently, Eq. 4.142 becomes cy a Pe = = rage = paasind. . (44s) From simple geometrical considerations based on Fig 4.20, itcan bbe shown that 8, Az. Ay. Pe, and Py are all related to hy and d. The relations are : ato ny a A en) where is in radians. Hydraulic diameters for the gas and liquid phases also can be defined by x = (8 = sind)) a 2 = 4150) (0 = sind) 328 (as Govier and Aziz? suggested thatthe wall shear stresses can be evaluated approximately by assuming single-phase flow to oecurin the cross section occupied by a given phase. With these assump- tions, these relation result. fipwt ra = et (4.152) and Lebo ta 4.153) where fz and fg ~ Moody friction factors based on Reynolds num- bers defined by Nae, (assy eet Nine = PE = 155) + 4.156) (asy Use the following steps to obtain the pressure gradient for strati- fied flow. 1. Calculate, on the basis of Eq. 4.135 2 Solve Eq. 4.147 iteratively ford. A value of 0.001 is good ini- tial puess for 3. Ay, also is calculated from Eq, 4.147 ‘3 Solve Eq. 4.146 fr yd. Calculate dg and dy from Eqs. 4.150 and 4151 ‘4. Knowing 6 and P,P, and Py are caleulated from Bs. 4.148 and 4.148. “6 5. Solve Eqs. 4.152 and 4.153 for tar and Top 6. Calculate the pressure gradient from Eq 4.14 for large-diam- exer pipes oF Bq. 4.145 for smaller diameter pipes Example 48—Using the Mukherjee and Brill Method, Caleu- late the Vertical, Multiphase-Flow Pressure Gradient for Exam- ple 3.2. Given: s4y=0.97 cp. ty =0.016 cp, op =8.41 dynesiem, and 00006 f. 1, Determine flow pattern: From Eq, 4.130, = 1o(t401-2eugons osu 9) Mer (4.130) 3508, Nyy = H1.$4< 350.8, Therefore. flow patter is not annularimis. ‘Angle > O°, Therefore, from Eqs. 4.138 and 4.129, x = log(11.54) + 0.940 + 0.074sin(90") = 0.855sin*90") + 3695100118) 2648. Margy = 10 18.40, ‘Nyy = M87 < 18.4, Therefore, flow patter is slug flow. 2. Determine liquid holdup: From Eq, 4.135 for uphill eveffcients, = 0.3801 + (0.1299) sin(90") + (= 0.1198 sin'(90) + sone 34300118)" ave = 0560. 3. Determine frietion factor: From Eq. 4.138, 88 10\7.83)0.5) 3) From Fig. 2.2 for efd = 0.00012, Nee = BUS x 10% 013s, 4, Determine pressure gradient: From Eq. 4.136, dp | (O01SS){(47.6)0.56) + (5.8810.44))0.83)* “ 2.17403) yy 32178) + (47.640.56) + (5.88V0.44)\6in90 SETH = 0.864 + 29.249 = 30.113 psfift 0.209 psi. 42.2 Mechanistic Models. Most mechanistic models that predict two-phase-flow behavior in pipes ae for an isolated mechanism, such as flow patter, film thickness, or rise velocity of gas bubbles in liquid columns. Although several methods predict flow behavior fora single flow patter, in this monograph we discuss only those that precict flow behavior fr all low patters. Specifically, we de- setibe the methods of Ansari et al and Hasan and Kabir.!527 -MULTIPHASE FLOW IN WELLS t pusale = SLUG. CHURN FLOW FLOW FLOW FLOW ANNULAR Fig. 4.21—Flow patterns in upward two-phase flow.20 (Repro- ‘duced with permission of the American inst. of Chemical Eng neers.) “The models first predict the existing flow pattern and then calculate the flow variables by taking into account the actual mechanisms of the predicted low patter, ‘Two mechanistic models2829 curently in the review process for possible publication in journals, are not described in this mono raph, In addition, OLGA and TACITE are commercial computer programs that include proprietary steady-state mechanistic models. Because oftheir proprietary nature, details ofthe models are notin cluded inthis monograph Ansari etal. Method, Ansati et al formulated a comprehensive ‘mechanistic model for upward, vertical two-phase low. This study ‘was conducted a partof the Tusa U. Fluid Flow Projects (TUFFP). ‘Alhough the pressure-pradient equations imply tht the model can be used for deviated wells, no attempt was made to aocount forthe effect of inclination angle onthe mechanisms considered, Because fof the model's complexity and because it was developed with SI units, the development and examples also ae presented in ST unit, Flow-Patern Prediction. Tate et al presente the basic work on mechanistic modeling of flow-pattem wanstons for upward ‘worphase flow. They denied fou dstinet lw paterns and then formulated and evaluated the ianstion boundaries among them Fig. 421 showsthefourflow pattems bubble low sug flow, chum flow, and annular flow. Bamea tal. modified thes transitions to extend the apliabil ity of the Titel el. mode to inclined flows. Barnea™= combined flow-patem-prediction models applicable to different inelination- angle anges into one unified model. On the basis ofthese studies, flow patterns can be predicied by defining transition boundaries among bubble, sig, and annular flows, Bubble/Slug Transition. Tate eral. gave the minimum diameter for which bubble low aceuts 901] 2.= Pb ris For larger pipe sizes, the basic transition mechanism for bubbly to slug flow is coalescence of small gas bubbles into large Taylor ‘bubbles. Experimentally, this was foundteoccurata void fraction, (4.188) -MULTIPHASE FLOW PRESSURE. GRADIENT PREDICTION ‘Transition ‘Slug oF Churn t D L oor ar 1 10 100 ‘Superficial Gas Velocity, mis Fig. 4.22—Typical flow-pattem map for wellbores.2* of about 0.25. With 0.25, the transition can be expressed in terms of superficial and slip velocities as, 25y, + 0.333%. cesses GIS9) lip or bbble-rise velocity given by Harmathy® as sees (6.160) This is shown as Transition B in Fig. 4.22. ‘AU high liquid rates, turbulent forces break down large gas bubbles into small ones, even at void fractions greater than 0.25. This yields the Bamea?® transition to dispersed. bubble flow as fpctsg] GO"(G)" ew” ons + 415(5285) « where fis obtained from the Moody diagram (Fig. 2.2 fora n-slip Reynolds number. This s shown as Transition Ain Fig, 4.22. ‘Athigh 2s velocities this transition governed by the maximum packing of bubbles 1 give coalescence, Scott and Kouba’ con- ‘luded that this occurs ata void faction of 0.76, giving te transition {or no-slip dispersed-bubble flow as (4.161) vag = 3.15. (4.162) “This is shown as Transition Cin Fig. 422. ‘Transition to Annular Flow. The transition criterion for annular flow is based in parton the gas-phase velocity required to prevent fallback of the entrained liquid dropletsin the gas stream. This gives the transition as (416) and is shown as te left Transition D in Fig. 4.22. ‘Bamea proposed addtional transition eriteria that considered the effects of film thickness. One effect isthe bridging of the gas [— core by a thick liguid film at high liquid rates. Another effec isthe instability ofthe liquid film, which causes a downward flow ofthe film at low liquid rates. The bridging mechanism is governed by the ‘minimum liquid holdup required t form a liquid slug Myr > 012 eS seteeeeeeeees IGA) ‘zp is the fraction ofthe pipe cross section occupied by the liquid film assuming no entrainment in the core.To account forthe effect of liquid entrainment inthe gas core, Eq. 4.164 was modified as ia (hues) >on wt where, from geometric considerations, Hyp canbe expressed in terms of the. dimensionless film thickness, = 8/4. as Huy = 48(1-8). The dimensionless im thickness canbe calc lated by solving a dimensionless combined-momentum equation 2. OH te here Zs detennined from either Ea, 4.221 0° 4222. Xy and Ya the modified Lockhart and Martinelli?® parameters,*® are Lys Xe Yu (4.166) a nee (SE san Yu = oe) (4.168) W/aDsc and B = (1 ~ Fp)" f/f). Other parameters are defined in the annular-flow model. The film instability mechanism also can be expressed jn terms of the modified Lockhart and Martine? pa 2-1 5Hy te" AT 13 a enocoococetc, (4.169) ‘Annular flow exists ifthe two Barca criteria are satisfied and if ‘8 greater than that athe transition given by Eq 4.163. First, Eg 4165 i used to determine the first riterion for annular ow. Ian- nular low is predicted, then Eq, 4.169 must be solved implicitly for Spur If 3s greater than 3, the second erteron for annular Now is Satisfied. Eq, 4.169 can be solved for 3. by use of a Newton- Raphson approach. Eq 4169 is first expressed as 2-1 Aba) = Yo ~ BETS (4.170) aed 1st Fn) = a= Tt = 15H HAD ~ 55d eee. aim yA ~ 15H) ‘The minimum dimensionless film thickness then is determined it. eratively from Fema) am) A good initial guessis By = 0.25.InFig. 4.22, theright Transition Disa result of applying the two Barnca film thickness criti Flow-Behavior Prediction, Ansari et al presented physical ‘models to predict flow behavior ia each flow patter, Ths resulted in separate models for bubble flow, slug flow, and annular low. (Charm flow was not modeled because ofits complexity, but was treated as apart of slug flow. Bubble-Flow Model. The bubble-flow model is based primarily ‘on the work by Coetano’? for flow in an annulus. The two bubble flow patterns. dispersed: bubble flow and bubbly flow, were consid «ered separately in developing the model Because of the uniform distribution of gas bubbles inthe Tiquid nd the lack of slippage between the two phases, dispersed-bubble Tow was approximatedas apseudo-single phase. With this simplii- ‘ation, the two-phase parameters can be expressed as Pre = Pact dl Ads G12) Mr = Made + Hdl Ade amy and Vie = Yn = Van + og airs) For bubbly flow, slippage is considered by taking ito account the bbubbie-rise velocity relative to the mixture velocity, Assuming a turbulent velocity profil forthe mixture, withthe rising bubblecon- ‘centrated more a the center than along the wal ofthe pipe, the slip page velocity can be expressed as 7 12m 5 ceeeeeeees 176) Eq. 4.160 gave Harmathy's expression for bubble-ise velocity. ‘To account for the effect of bubble swarm, Zuber and Hench? mo- died the expression as va ve[ siesta i. _ e where the value of n' varies from one stay to another. Ansari et al ound thata value of 0.5 forn’ gave the best results. Combining Eas. 4.176 and 4.177 yields rofiecell PL ‘This isan implicit equation forthe actual holdup for bubbly flow. ‘Two-phase-flow parameters then can be calculated from 12vqs s+ (4.178) Pm = puts + 90 ~ 4)». eves GtT9) aod be = mite + dt ~#1) ses (4180) ‘The tworphase pressure gradient for bubbly and dispersed- bubble flow is composed of two components. Thus, (2) = (#). = (2) ay a) ete congoen i (dp fale (ey) eee. eusy where fre is obtained from a Moody diagram (Fig. 2.2) fora Rey- nolds nbmiber defined by Prev ae Because bubble flow is dominated by a relatively incompressible liquid phase there isnosignificant change in the density ofthe flow Neng (4.134) rc DEVELOPED TAYLOR, BUBBLE Fig. 423-Slugtlow schematic # ing fluids. The Nid velocity is nearly constant, resulting in essen- tially no pressure drop from acceleration. Therefore, the acceera tion. pressure gradient is negligible compared with the other pressire-gradient components, ‘Slug-Flow Model, Fernandes eral presented the first complete physical model for slug flow. Sylvester developed a simplified version using a correlation for slug void fration. Both studies as> sumed a fully developed slug flow. Fora fully developed slug unit, as shown in Fig. 4.23, overall gas ‘and Tiguid mass balances, respectively, are given by vag = Boga Hard + (1 Biel ~ Hg) (185) and vag = (= Arustlus ~ Brartlrn 4.186) tere p= (4.8) “by Mass balances for steady-state liquid and gas transfer between the liquid slug and the Taylor bubble, respectively, are (vse ~ YeaslHase = Won ~ (~ vero IHore (4.188) and [10 — vast ~ Has) = (re ~ rye) — Md 4.189) ‘The Taylor-bubble-rse velocity is equal to the centerline velocity plus the Taylor-bubble-ise velocity ina stagnant liguid column, dp. ~ Ps) Se (4.190) va tan 03 -MULTIPHASE-FLOW PRESSURE-GRADIENT PREDICTION Similarly, the velocity ofthe gas bubbles in the liguid slug is vys = 120m + 153) $2e2L= 20) ps asp mi where the second term onthe ight side represents the Bubble-ise telociy a8 defined in Eq. 4.177 The velocity vp ofthe falling film around the Taylor bubble canbe correlated wit he film thickness dy by ase the Brot? vm = (19678; where isthe constant film thickness fr fully developed sug low. From geometrical considerations, vzrp can be expressed in terms of the Taylor-bubble void fraction 1 give (am) ven = 9916ed1 ~ ‘The liquid slug void fraction canbe obtained fromthe corelation | developed by Sylvesier9 who used data from Femandes etal? and Schmidt 4.193) TS + Te, gs. 4.185, 4.186, 4.188 Brough 4.191, 4.193, and 4.194 canbe solved tratively to obainthe following eightunknowns tha define the fully developed slug-flow model: B, Hure. Hs. vera. M18: ts. ¥1s. and vg. Voand Shoham’ showed that these eightequae Hons can be combined algebraically to give igs (a9 (0916 feat - T= Fi = vp — Hixe) +H = 0. where « 4195) He Ayslrmrtys) + Ya. (8196) With vy and Haus given by Eqs. 4190 and 4.194, respectively, X can be determined readily from Eg, 4196, Bg. 4.195 an be solved for Haze by use ofan eratve-soltion method. Defining the let side of Eg, 4.195 as Fira) FH) = (9916 /ed)(1 — J = veal = Ha) + «497) “The derivative of Ea. 4.197 with eespet to Here yields Fae) = v0 + (9916/82) x (1 - =A] 4.198) Eqs, 4.197 and 4.198 suggest thatthe Newton-Raphson approach ‘ean be incorporated easily t determine Hera, the root of Eq 4.197. ‘Vo and Shoham also showed tha, if root exists in the interval of (0.1 the root is unigue. Thus, An) F(t) Hum,,, = Hur, (4199) — _ ENTRAINED LIQUID DROPLET y, Fig. 424—Annular-low schematic 2 ‘To determine al slug-flow variables, use this step-by-step procedure 1. Calculate vy, gus. and gus from Eqs. 4.190, 4.191, and 4.194 2. Using Eqs. 4.196 though 4.199, determine Hira. A 00d ini tial guess is Hira =0.15. 3. Solve Eq. 4.193 for vera. Note that Here 4. Solve Eq, 4.188 for vuzs. Note that Huy 5, Solve Eq. 4.191 for sus 6 Solve Eq. 4.189 for very 7, Solve either Eq, 4,185 oF 4.186 for. 8. Assuming that Lys = 30d, calculate Zgy and Lg from the def nition of. Fr fully developed slug flow, the elevation component of the pressure gradient occuring across a slug units given by i = Hire Hs. (= tous + Aovesina, (4200) were Pus = Puts + pdl - Hus). (4.201) Eq. 4.200 assumes thatthe liquid film around the Taylor bubble ‘doesnot contribute tothe elevation component. Friction losses were sssumed to occur only across the liquid slug and are neglected along, the Taylor bubble. Therefore, the friction component of the pressure sradient is (i), fbesray _p, = 4202) ‘where fs is obtained from a Moody diagram (Fig, 2.2) for a Rey- rolds number defined by Past oo OE Nae (4203) so For fully developed slug flow, Ansari era assumed that the ac celeration component of the pressure gradient can be neglected. ‘McQuillan and Whalley introduced the concept of developing slug low during their study of flow-pattem transitions, In develop- ing slg flow, the length ofthe cap atthe top ofthe Taylor bubble isa significant portion ofthe total length of the Taylor bubble. The Film thickness varies throughout the Taylor-bubble region, Ansari ‘etal presented a detailed description of the complex equations to predict flow behavior for developing slug flow. Barnea®® contends that failure to consider the developing slug region can result in sig. nificant pressure-drop-prediction errors. However, pipe in produc- ing wells isso long tht only minor enors should result if the entire slug-flow region is treated as fully developed flow. Consequently, this monograph does not address the developing slug region. Annular-Flow Model. Walls!8 discussed the hydrodynamics of annular flow and presented the classic correlations for entrainment and interfacial rition as functions of film thickness. Hewitt and Hall-Taylor® gave a detailed analysis of the mechanisms involved ‘nannularflow. All subsequent models were based on this approach. ig 4.24 shows a fully developed annular flow. Conservation of| _momentam applied separately tothe core and the film yields, forthe core and the film, respectively, ‘@) os a) n8.—ncdrsna 4209 and Ae ‘The core is considered a homogencous mixture of gas and en- ined liquid droplets flowing atthe same velocity. Tus, Pc = Ptkic + Pall ~ duct (4.206) wae he = att a) ‘isthe fraction ofthe total liquid entrained inthe core. Wallis!® determined it empirically as Fea 1~ exp 0125 = 15) a wer Yow ® 1.00" (20) (4200) ‘The shar ses ne Sn cane oxen - a0 ‘where fr is obtained from a Moody diagram (Fig. 2) fora Rey: olds number defined by Purdue ne, = Bui aan nee eM Fo val = 9 : yt amy = day = #41 ~ 04 aan) Tig of aa -MULTIPHASE FLOW IN WELLS. Eq, 4214 reduces to 4 ll-Fe fe (#),. * 4 pagar = op ha ae where the superficial liquid-fiction-pressure gradient is given by (2) BE cess (215) where fi. =the friction factor based on superficial liquid velocity, tnd can be obtained from a Moody diagram (Fig 2.2) fora Reynolds ‘number defined by a2 For te shear sess al the inet, lh icc aay where ve = F 219) os a2 and WA lies coscssssccessssssssssecceeessees 4220) Where Za correlating factor relating interfacial fiction to the film thickness. On the basis of the performance of the model, Ansar et ‘al, found that the Walls expression for Z works well for thin films ‘righ entrainments, whereas the Whalley and Hewitt expression Works wel fr thick films or low entrainments. Thus, 2= t+ 00HtaF, > 09 aan wa 2-1 +2(0t)"sore, <9 any ominig Bs 4218 tough 220s ‘ “ cevceeeveeee (42 ‘lt, ‘The superficial fiction-pressure gradient in the core is given by (8), Sucbetig a, ans ‘where fic is obtained from a Moody diagram (Fig. 2.2) for a Rey nods number defined by Potacd Hang = St, (4228) Ysc = Feta + Ysa 4226) and bse = Mibac + Hdl ~ Ac) a2 ‘The pressure gradient for annular flow can be calculated by sub- stituting the above equations into Eqs. 4.204 and 4.205, Thus, (2), - Aap) rane and (@),-arcteld), MULTIPHHASE-FLOW PRESSURE-GRADIENT PREDICTION (4208) z 2a = + pigsind. a2) ‘The basic unknown in Eqs. 4.228 and 4.229 isthe dimensionless film thickness, 6. The pressure gradients inthe core and film must be the same. An implicit equation for 8 can be obtained by equating Eqs. 4.228 and 4.229, giving (4) O aa li), roe (4230) ‘Tosimplify this equation, Ansari eal. used the dimensionless ap- proach developed by Alves etal This approach defines the fl- Jowing dimensionless groups in addition to using the previously de- fined modified Lockhart and Martneli parameters, Ny and Yi. lp/able~ spcsind we 4230 Wiis. and » _ Plt), ~ eoesind 8 all, (4232) Bq, 4.250 then reduces to one ee 4901 = B11 = 4601 = BY © [4s — By 4233) Eq, 4.233 can be solved iteratively by use of the Newton-Raphson approach to obtaind. IfEq 4.233 is FO), the derivative of Eq, 4.233 with respect to yields ra = 2-2) [49t1 ~ oy ~ 480 - BY -—___2_____ 40(1 = HL = aga = HF 2szi4il - 29) 491 — Ut — 48 — BP? _ 3x40 = 291 ao ser = a 3) acti 4235) (Once disknown, the dimensionless groups and ic can be obtained from te folowing forms of Eg. 4.228 and 4.228, respectively. 2. z (4.236) a-i5 4236 sot 2 = Fee fr a-ak [r-a-ay s Zz a (4237) 45 - ft — 0 ~ 25)" (= 25) ‘eecause the pressure gradient in the film and core mustbe the same, the total pressure gradient can be obtained from ether Eq, 4.231 of 4.232. Thus, Tr (2), ~ (8), =a), eee an (#), (:2), = (G2), + ssn (4.239) ‘Ansari eal neglected the acceleration pressure gradient for n= rular flow. Lopes and Dukler*® found that, except for a limited ‘ange of high liquid low rates the acceleration component resulting, {rom the exchange of tiqud droplets between the core and the film is negligible Example 4.9—Using the Ansar e culate the Vert 83x10 m, 1504 m, 208 ml, 9p =3.86 see 1.173 mp = 4761 LT kgm, pp dynesiem=841 10-9 bys Flow-Pattern Determination, 1. Check dispersed-bubble transition boundary (Eq, 4.161) af arn. x 10-> ( 1617 )" 149 x 10-2)" (210.1525) 761.7 ~ 94.1981 | (Bai x 28 +L wan + as X L144 3.636. "Then, flow pattem is not dispersed bubble. 2. Check annular-flow transition boundary: From Fa. 4.163, = 3] 9841 x 10-061.7 ~ 94. | 4)? = 087 mk. Although vg = 1.173 > 0.87 gives the annular-flow pater, also check the Bamea transition eriteria. From Eg. 4.209, f 73)0.016 > 10: (eal x 10>) You = 10,000 (BL) = net ‘The entrainment fraction i calculated from Eq, 4.208, Fy = 1 ~ expl~ 0.12507.84 — 1.5] From Ba, 4.207, __(05471.208) W347 208) + LIT 547. duc = 036 2 ‘The core density and viscosity ae obtained from Eqs. 4.206 and 4227. respectively, as Be = (761.710.36) + (94.1)(1 ~ 0:36) = 3344 kg/m? and = 0970.36) + (0.01611 ~ 0.36) ee 7,000) = 036 x 10-*Pavs ‘The superficial core velocity from Eq. 4.226 is se = (OS47V1.208) + 1.173 = 1.834 ms From Bg. 4.225, (334.11.83400.1524) (0.36 x 10-3) and relative roughness is Nase 260,000 fd = WBN 12008 x 10-4 ‘Then, the flow is turbulent From Fig. 2.2, Sac = 0.0152, Similarly, from Eg. 4.217 the film Reynolds number is ‘gg, = (1S1-TA1.2080.1524) tes = (097 10 144, 565 and the friction factor from Fig 2.2 for a relative roughness of 1.2008 x 10" is Sus = 0.01708, ‘The superficial gas-core and liqui-film pressure gradients are de- termined from Eqs. 4.224 and 4.216 as |, 0.01708)36.4).1.8344) alse 7210-1524) 63.13 Pode and ao} _ (0.0152N761.7)1.208)* a = ase = 88 ‘Assuming that ff, ~ 1, the modified Lockhart and Marinelli pa meters can be estimsted fom Eqs. 4.167 and 4.168 as 66.34, Yu Because Fe 057, wing By 4.222 yields z= 1+ 00(tl2)"9 ting Xi. Ya, Z, and Hp 1+ 8.195 441 ~ Stl = 480 — 8)? MUT-TIOHASE FLOWN WEIS ( - r- Solving the above equation for 8, = 0.1268, Therefore, 8= (0.1265) %(0.1524)=00193 m and Hz =(4N0.1265) (1~01265) = 0.482. the criterion from Eq, 4.165, [ose + osol0uss= 202200 245) ena 0.1524)" 072 > 0412, ‘Thus, the liquid film holdup is sufficient to bridge the pipe, and an- ular flow does not exis. 3. Check bubbly flow existence: ‘With Eq, 4.158. the minimum pipe diameter at which bubbly low dag = 1901] 251-7 = 94.1084 x 10-4 (61.981) = 00189 m < 0.1524 m= ‘Then, bubbly flow can exist 4. Check bubblefstug transition: From Eq. 4.160, 1.53] @810761.7 ~ 94.0) (761.7) (sat x 107 r From Eq. 4159, the superficial liquid velocity atthe transition is Ysus = MIAT3 = (0.25)0.151)6sin90°)] = 3.4 mvs and = 0151 ms, Mus = 34 m/s > vgs = 1.208 ms The flow pattem is then slug low, ‘Slug-Flow Modeling. ‘The step-by-step procedure described in the discussion of the slug-flow model is followed to determine al sug-flow variables, 1. Calculate wpa, gus, and ves from Eqs. 4.190, 4.191, and 4198 22) (9.81)(0.1524)(761,7 ~ 94.1] v= dase + sf annus? — sun] 3.258 mls =o Be Aus = Faas + Q.65ya.881) = 914% an Yqus = 1.202.381) + (0.151)(0.826)"* = 2.994 mis 2. Using Eqs. 4.196 through 4.199, determine Hire. {A good initial guess is Hera = 0.15 A = (0.1742)3.258-2.994) + 2.381 = 2.427, Ft) = [9916 OaOIsTD|t - T=)" 01s = 32581 ~ 115) + 2.427 1658, -MULTIPHASE FLOW PRESSURE GRADIENT PREDICTION = 8.4383, and 0.1658 Hara = 015 ~ S41SSE = 0.13, With the calculated Hzrp asthe new guess for Hira. = [p916 @an.1535|(1 ~ = 3.258(1 = 0.13) + 2.427 = 000132, Fm) x|(1- 073)" 4 f= o1{t — 08) = 8031, 238+ 16 ORO] ani ooo1s2 ‘3031 ‘The difference between the guessed and the calculated valves of Hara is very small. Therefore, Hera =0.13 3. Solve Bq. 4.193 for sera [Note that re = 1 Here. Hy = 1 = 013 = 087 and Hany = 0.13 ~ = 0.1298, va = 991e{9.8190:1520(1 ~ (O87) = 3.184 4, Solve By, 4.188 for ns [Note that igs = 1— Hgts. Hus = 1 ~ 0.174 = 0826, (3258 ~ v,)0826 = 3.258 + 3.144503, and vuus = 2.25 mv 5. Solve Ea, 4.189 for vera: (2.258 ~ 2.998)(1 ~ 0.826) (8.258 ~ vyra)t 0.13) 3 —- and 205 mis, 6, Solve Bq. 4.185 for: 1.173 = (1 ~ 8)2.994)01 ~ 0.826) + B(3.205\1 ~ 0.13) and 8 287. 7. Assuming Las = 30d, calculate Lsy and Lp from the definition off Lag = 300.1524) = 4.57 m, (02874457) = (1 ~ 0287XLy). and Lay = 1.84, 8, Elevation component of pressure gradient: rom Eq. 4201, Pus = (761.7)0.826) + (94.1)(1 = 0.826) = 654.4 kg/m? From Eq. 4.200, [0 ~ 0.2874654.4) + (0.287)(94.1)19.81)sin90° 4,779.2 Palm, 9. Friction component of pressure gradient us = (097 x 10-4}0826) + (0016 x 10-91 ~ 0826 108 10-4 Pars From Eq. 4.203, ‘Turbulent flow exist. From Fig. 2.2. e/d= 12% 10 Nang = 291,284 f “Ses © 00166. From Eq. 4.202, (0.0166,645 42.381) 2OIs2) 987) 142.0 Palm, 10, Total pressure gradient: 4.7792 + 1420 = 4,921.2 Pum = 4921 Pum = 0.218 psi Hasan and Kabir Method. Hasan and Kabir?" also developed ‘a mechanistic model to predict pressure gradients in wellbore. low-Partern Prediction. To model flow-pater eansitons, Hasan and Kabie adapted an approach very similar to that of Tate ea. Hiasan and Kabir? dented the same four flow pats shown in Fig. 4.21; bubble flo, sug flow. churn flow, and annular flow. Bubble/Slug Transition, Transition from bubble flow the condi- tion of small bubbles dispersed theoughout the flow cross section) toslug flow (when the bubble becomes large enough to fill the entire ‘ross section) requires agglomeration or coalescence, Bubbles oth certhan very small ones, generally follow a zigzag path when rising through aliquid. This results in collisions among bubbles, with the ‘consequent bubble agglomeration and formation of larger bubbles, hich increases wth an increase in the gas flow rate. Hasan eral. reported that a transition o slug flow is expected a avoid fraction 00.25. By use ofa drift-fux concep, the transition then can be ex- pressed by Eg. 4.240, sind w a + ¥) (4240) Gis the Now coefficient given by Eq, 4.241 L.2ifd < 0.12 m or ifvg > 002 m/s ©) 20ifd > 0.12 m and ifvg, < 0.02 m/s 2a “The Harmathy?® expression, Eq. 4.160, is used for the slip or bubble-ise velocity. Transition to sug flow takes place at supert- cial gas velocities greater than that given by Eq, 4240, "The terminal rise velocity of smal bubbles given by Eq. 4.1605, dependent on uid properties but independent of pipe diameter. However, the rise velocity of a Taylor bubble 2° vm = 035 Sin B (1 + c0s6)"?, (4282) i dependent onthe pipe diameter. When vrg > vs, the nose ofthe Tay- lor bubble sweeps the smaller bubbles ahead of it. However, when ‘1g < vy, whichis possible in small-diameter pipes, the rising small tr bubbles approach the back ofthe Taylor bubble, coalesce with it increase is size, and ultimately cause a transition to slug flow, Dispersed-Bubble Transition, When large bubbles are dispersed {to small bubbles at high liquid rates transition to slug flow isin hibited, even though the gas void fraction exceeds0.25. For the ran: sition to dispersed:-bubble flow, Taitel etal proposed wfc] gga (4243) ‘When the gas void fraction exceeds 0.52, bubble coalescence can- not be prevented and transition io either slug, churn, or annular flow ‘must oveur. Slug/Chum Transition. The Barnea and Brauner) model pre- diets the slug/churm transition. They argued that the transition fom Slug to chum flow cccuts wen the Iigud sug traling Taylor bubble auains the maximum posible voi ration of 0.52, a5 given by Eq, 4.243. In other words, tis transition boundary i a locus of constant mixture velocity values, whee the trbilet intensity {s maintained atthe same level as thats dspersd-bubble flow. Annular Flow Transition, The transition cxiterion given by Bq 4.163 ao is sedi the Hasan and Kabir mechasistic modelo pre dict the transition to annular low-E, 4.163 suggests that thistranst tin terion is independent ofthe liquid low ate. MUUTIPHASE FI OW IN WFLIS, [- i - Flow-Behavior Prediction. The wtal pressure gradient in two- phase flow canbe wnttenas the sum of the gravitation or hydrostatic Inca (dp friction (dpid.y and acceleration (djvL ee compo- rents. Thus, (a), ~ (),* (),* (@), FO | die BF + Bain GE = pagsind + a2 where i= pil — Hy + pill, (4245) In general, the acceleration component canbe neglected during all but the annulaeflow pattem. Eq, 4.245 suggests that an accurate estima- tion ofthe liquid holdup is essential when computing the elevation ‘component. This component accounts for most ofthe pressure drop ‘cecurringin the bubble- and stug-low pattems. Because of the tft ‘ent hydrodynamics in each flow pattem, estimations of holdup, F. insta mixture density, p., and friction factor, fare made separately Bubble and Dispersed-Bubble Flow. In bubble and dispersed bubble flow the expression for holdup, Hi, Ys mh Gin (4.246) where C, and w ae given by Eqs. 4.241 and 4160, respectively. To estimate total pressure grodient, Eq. 424 can be used withthe mix ture density calculated from the liquid holdup estimated with Eq. 4.246, The friction component can be computed by treating the mul tiphase mixture as.ahomogencous uid. Friction factor, f.can be de- termined from the Moody diagram (Fig. 22) fora Reynolds number defined as = Peta 5 Meg = PS (4247) ‘This was recommended by Govier and Aziz? because pyle would not be 100 different ftom py/d. and the contribution ofthe Frition component tothe total pressure gradient i very small Slug and Churn Flow. The drift-flux model of Eg, 4.246 also was applied in slug flow, but with different values for C, and y, given by C= 12and wae twas ecommended that the equations developed for slug flow also bbe use forthe chur-flow patter, However, Because ofthe chaotic flow nature that tends to make the gas concentration profile fst is suggested that a Value of 1.15 be used for Gy rather than a value ‘of 1.3 used for the slg-flow pattern. As in bubble flow, the tot pressute gradient can be obtained by Eq. 4.244 by use of Eqs. 4.246, and 4.248, The estimation of the friction component presents some ilficuty because some of the liquid flows downward in a film around the Taylor bubble while most ofthe liquid flows upward in the liquid slg. Wallis! suggested thatthe wall shear stress around, ‘the vapor bubble be ignored. With this assumption, the fiction pres- sure gradient bec (), ‘The product pis very nearly equal to pq for low-pressure sys- tems, indicating the similarity in evaluating te friction terms in slug and bubble flow. The contribution ofthe friction component is no Tonger negligible but is il small (typically 10% of the total gradi cent). Acceleration, however, is small and can be neglected. 28) "nb + cos)". (4248) Pe (4249) ‘Annular Flow. In annular flow, it is appropeiate toestmate the lig- uid holdup in the central core rather than forthe entire pipe eross section. The liquid holdup forthe core, dz. is given by Eq. 4.207 Steen and Wallis? suggested that, when the liquid fl i fully tar bulent, the iguid enrsinment can be obtained as a unique function, of the critical vapor velocity from, Fe = 000SSv28ity, <4 Fe = 0887 l0B¥.u ~ ODDIE vey > 44" 4.250) here vei is defined by Eq. 4.209. Tnannular flow, fine liquid droplets flow inthe gas core witha ve- locity the same asthe gas phase while a thin liquid film ereeps up the pipe wall. Thus, the friction pressure drop pertains tothe gas in- teraeting with the Wavy liquid film. The friction pressure gradient, Which ia large component of the total pressure gradient for annular flow, can be written as (2) -2(s,): aasn te sy en bet abe «as ‘The liquid holdup inthe core can be given by this simple equation fom Wallis!® (4258) Where X: the Lockhart and Martineli’ parameter, is defined in terms of the gas mass faction, x, and fluid properties as Lan)" Beas)” x= (4) iG) Inthe eat of annular Now the gs oi fraction i forthe channel volume not occupied bythe liquid film. Hence, the gas mass fraction, calculation should include the entrained liquid droplets, (4255) ‘Example 4.10—Using the Hasan and Kabir Mechanistic Model, Caleulate the Vertical, Multiphase-Flow Pressure Gradient for Example 32. Given: ¢=0.00006 ft vse 3.97 fuse i P=4761 lat? 94.1 ken? 01, =841 dynesicm=B41 x 10~} kg/s? jy =0.97 cp=0.97 x 10-3 Pars. tg = 0.016 ¢p=0.016% 10-2 Pas, 83% 1075 m, d=600 in, =0.1524 m, Flow-Pattern Determination. 1. Check bubble/stug transition (Eq, 4.240): 1 sal (281N8.41 x 10->,761.7 — 94.1) (761.7) 1151 avs, C= 12, and va” (ecbyg) ana) + 01801 = 0572 - r i Nomina Diameter Sous fo) (1a TUFFP Databank* 18 Gover and Fogarasi™ 2104 Asheie® 2ig108 CChierii et at10 2Nig108 Prudhoe Bay 85107 2, Check dispersed-bubble transition boundary (Eq. 4.243): x (84Lx103)"/_761.7 7617) \oa7 x 10%, oF vane = 4.401 >2.381. Therefore, neither dispersed-bubble nor ‘chum flow exist. The flow pater is sug, (9819061. ~ 94, oe ‘Slug-Flow Modeling. For slug low, C= 12. 1. Calculate slip velocity (Bq. 4.248): 61.7 ~ 94.) 7617 oss[@anioases x {Sin 50° (1 + €0590°)'? 0.401 mis. 2, Calculate holdup (Ea. 4.246) and slip density (Eq, 4.245); and a = (T61.7V0.64) + (94.10.36) = 521.4 kg/m? 3, Blevation component of pressure gradient From Eq. 4.244, (@),, = 62149080 sn90" = 5.115 Pam aL, 4. Fiction component of pressure gradient: From Eq. 4.247, (261.7)2.38190.1524) ews = (097 x 10) ‘Turbulent flow exists From Fig. 22, 284,942, afd = 12x :10-* Nau = 284,942 fF * 00166. From Eg. 4249, do) 0.0166)(2.381761.7)0.64) , 200.1524) at, 150.5 Pat, "TABLE 4.6—RANGE OF DATA IN TUFFP WELL DATA BANK™ yeast otra nan a? Barra? Onsen Oi Rate Gas Rale OiGravity STB) (Msc) _CAPI) DieI0,150 “1S t010567 8510705 Bto1,600 114192740017 to 112 7201027,0007401055:700 951086 oa69 61027814 BSI04B 6001023,000 20010 110.000 241086 5. Total pressure gradient: (@) sus 10s 5,265.5 Palm 5.266 kPalm 0.233 psi. 4.3 Evaluation of Wellboro Pressure-Gradiont Prediction Methods “The pressure drop ina wellean be calculated with the computing al gorithm described in See. 3.7. This requires that the profile of the ‘well is known, that a pressure-gradient-prediction method is speci fied, that suficient information is available to predict pressure/vol- tumefemperature (PVT) relationships, and that adequate informa- tion is available to calculate fluid temperatures as a function of, depth. Each requirement sa potential source of error when calcula ing pressure drop. "A common error in deviated wels is to not use accurate wel pro= files in the marching algorithm. Many times the profile is unknown and vertical flow is incorrectly assumed. Back-oi-mode! PVT calculations, which include the prediction ‘of mass transfer and fluid physical properties ofthe gas and liquid ‘phases, can be highly inaccurate under many real applications. For ‘example, bubblepoint pressures calculated from different black-oil, models can have errors in excess of 50%, Depending on the depths ff wells. this can leadto serious errors in pressure-drop calculations. PVT property correlations often incorrectly predict the partof a well ‘nwhich single-phase flow occurs. Tis in tur can lead to very con- ttadicrory results when evaluating. pressure-gradient-prediction methods with different PVT correlations. ‘When using empirical correlations or mechanistic models to pre- dict pressure gradient, another common source of erors is the fre- {quent absence of reality checks, such as ensuring thatthe calculated Tiguid holdup always exceeds the no-slip holdup for upward flow. Correlations that have been modified normally include reali checks, Correlations that ave not been modified, such as Duns and Ros,? Mukherjee and Briand Aziz eral? may be vulnerable to such programming inadequacies, ‘With this introduction, an evaluation of six commonly used cor- relations and two mechanistic models was performed by Ansari et 1213) The pressure-drop predictions from applying each correla tion and mode! by use ofthe computing algorithm described in Sec. 3.7 were compared with measured data in a well databank devel ‘oped at TUFFP. The databank contained 1,712 well eases with a ‘wide range of data, as shown in Table 4.6. 43.1 Criteria for Comparison With Data. Evaluation ofthe cor- ‘relations and models with the databank was based on these statist cal parameters. MULTIPHASE FLOW IN WELLS | Soe ts Past ~ Pram (4.257) ; (4257) (4.256) £ indicates the overall trend ofthe performance: sured pressure drop. Be (Se : rD Ez indicates an average of how large the errors are, lative tothe mea (4258) (4259) Ey indicates the degree 10 which the erors ate scattered about theit average percent error, (4260) ©. BP ~ MP ae: =. (4261) Es indicates the overall tend independent ofthe measured pressure drop. += (i) (4262) -Esalso is independent ofthe measuted pressure drop and indicates the magnitude of the average errr. (426) indicates the scattering of the errors about their average ero, 43.2 Comparison Method. The comelations and models used for the comparison were a modified Hagedom and Brown,* Duns and Ros,’ Orkszewskit with he Tggis® comection, Beggs and Brill! with the Payne et al2? correction, Mukherjee and Brill! Aziz et al.9 Ansari etal. and Hasan and Kabir232° The evaluation was accomplished by a comparison of the statistical parameters. The tape aa Sow cra Rea Aa ats eavaton nur Rane Hasan anda es Bog bog be aban TABLE 47—RELATIVE PERFORMANCE FACTORS IES Fag Vl oe A gS pV GOO SS VENA Toa a 1712 4086 6267554381 a0? was a? 70 Ansa 07001214978 0.081 0.000 0.143 1.295 «ast att? 0442 0.000 Hag6 05850600 0818 087 © O77 20230386 0488057 0009.8 Adz 151216082085 © 0@09 1.062» 0621708 °«17e4 «1314488 ote DunsRos «171916783678 711792 1.108« 2058 20881882 22051218 Haske 1940 2005220168 «1.780 000825752500 oe 1903 104s BeyBr 2.082 2.008 Dads aaa aate «aes ned 2885281 3.282 1972 Onis 428452732020 Sens 4658 «1.208 3.108 3189881 44036000 Mabe evaluation also involved the use ofa relative performance factor Fp: which was defined by ei - Ef (426) ‘The minimum and maximum possible values for Fr are and. in- dicating the best and worst performances, respectively. Table 4.7 sgivesan evaluation ofthe methodsin terms of Fr withthe best value Tor each column being boldface. 433 Overall Evaluation, An overall evaluation first was per- formed by use of the entire databank, resulting in Col, lof Table 4.7. “The performance also was checked for vertical well cases only (Col 2) and for deviated well eases only (Col 3). To make the evaluation unbiased with respectto the methods, asecond database was created that excluded 331 set of data fom the Hagedorn and Brown study. For this reduced databank, Col 4 shows the results forall vertical Well cases. Col. 5 shows the results for combined vertical and de- Viated well eases 43.4 Evaluation of Individual Flow-Pattern Models. Petfor- ‘mance of individual flow-patter models is based onsets of data that are dominant in one parcular flow pattern, as predicted by the tran- sitions described by Ansari er al2* To have an adequate number of ‘cases for the bubble-flow-pattern database, cases with bubble flow predicted over 75% of the well length were considered. These re- sults are shown in Col 6. Cols. 7 through 10 give results for cases predicted to have slug flow over 100% ofthe wel length. The cases used for Cols. 7 and 8 were selected from the entire databank, Wheres the cases used for Col. 9and 10 were selected from the r= ‘duced databank. Co. 11 gives results for those cases in the total da- ‘abank that were predicted o be in annular flow for 100% ofthe well, length. The performance ofthe Ansari eral. model for annular Now is significantly better than all other methods. ‘Ansari eral. concluded that the overall performance of their mod- el was better than all other methods. However, the overall perfor- smances ofthe Hagedom and Brown, Azizeral, Duns and Ros, and Hasan and Kabir methods are comparable, Forthe latter three meth fds, this can be attributed to the use of flow mechanisms. The excel- lent performance of the Hagedom and Brown cortclation can be ex- plained only by the extensive data that were used inits development “Peninalconmuncson wih A Tegpa Patz SA, de rao (86) [— ti and modifications that have been made to the correlation. Infact, ‘when the data without Hagedorn and Brown are considered, the An. sar etl, model performed best (Cols. 4and 5). Although the Hage- ¢dorn and Brown correlation performed better than the others for de viated wells, none gave satisfactory results (Col. 3) (Only 29 cases satisfied the bubble-low criteria, The Hasan and Kabir model gave the best predictions for these cases. When the Hagedom and Brown data ae included (Cols. Tand 8), their meth- fod exceeds the performance ofthe Ansari eal. slug-flow model. ‘The Ansari et al. model performed best when Hagedom and Browndataare excluded forbothall well cases andall vertical well ‘eases (Cols. 9 and 10), For annular flow, the Ansari er al. model is significantly beter ‘than all other methods (Col. 11) Pucknell er al’? compared predicted pressure drops with mea- sured pressure drops for 246 data sets collected from eight fields, in- cluding agas and gas-condensate field. None ofthese data Was used in the development of the pressure-gradien- prediction methods. ‘They reached four specific conclusions 1. Despite the development of new mechanistic models, no single method gives accurate predictions of bottombole Mowing pressures inal elds, 2. Traditional methods to predict pressure deops, such as that of ‘Duns and Ros, give good results in oi wells, but can give very poor results in gas Wells. The new mechanistic models give reasonable ‘results in both oil and gas wells. 3. Overall, the Ansar etal. mechanistic model gives the best re: sults of al the methods evaluated. In oil wells, 62% ofthe pressure drops were predicted with errors of less than + 6%. In gas wells, {68% ofthe pressure drops were calculated within + 15%. These conclusions were derived from deviated wells with larger tubing sizes, typical of North Sea wells “4. The Ansari et al. and Hasan and Kabir models behave no beter ‘than many ofthe rational methods in predicting ewroncous discon nutes. Reliable convergence is desirable in any multiphase-low ‘model. The Ansari er al model performed only as well a existing ‘methods, but the Hasan and Kabir model performed somewhat beter. Salim and Stanislav® compared methods that describe the low of gasiguid mixtue in wells with more than 189 data sets collected from five different sources. Most of thei Field data were taken Irom wells that exhibited the annular/mistflow patter. They concluded that the empirical comelations by Orkiszewski and Duns and Ros appeared to be less efficient than mechanistic models Clearly, evaluation of pressure-gracient-prediction methods to calculate pressure drop in Wells isa difficult task. In this context, we strongly recommend that users exercise careful judgment based on evaluation of actual measured data to select a particular method. 4.4 Pressure-Gradient Prediction in Annul In the past, corelations were the most common procedure to predict liquid holdup and pressure gradient for multiphase flow in an annu- lus, These comelations were developed empirically from exper- imental data. Most correlations make predictions without taking into account the existing flow pattern. The predictions either have ‘applied correlations originally developed for low in pipesby use of the hydraulic diameter concept or have applied correlations devel: coped from data obtained from two-phase flow inan annulus. Baxen- el. Gaither eral Angel and Welehon,® and Winkles® pres ented empirical correlations for the prediction of presssure gradients in annul. Except for the Winkler correlation, which pi ‘marily is for slug-flow conditions all are low pattern independent Ros*$ proposed a method to prediet both liquid holdup and pressure gradient on the basis of dimensional analysis and experimental daa, [None ofthese empirical corelations can be considered general oF ‘comprehensive, nor do they capture the complex flow behavior that ‘occurs when gas and liguid flow simultaneously in an annulus. ‘The use of mechanistic mode to predict multiphase-flow behav- {orn annul isa recent end, similar to their use in predicting regu lar pipe flow. Caetano et a®87 and Hasan and Kabie presented ‘mechanistic models to predict flow behavior when gas and liquid si- ‘multaneously low upward in an annulus. Fach is described in fol- lowing sections 4.441 Caetano etal Method. Caetano et al %® conducted acom- bined experimental and theoretical study of upward, gosiqud, ‘two-phase flow in vertical concentric and fully eccentric annul, The ‘experimental pa ofthe study involved flowing ar with ther water ‘or kerosene up a vertical annulus, The experimental study included the establishment of ow-patter definitions, determination of flow-patter-transition boundaries, development of flow-patern raps, and measurements of average volumetric igud holdup and average pressure gradient for several ests in each flow pattern. The theoretical pat included development of flow-patten-prediction ‘models ad the formulation of models fr each Flow patter o pre dict average volumetric liquid holdup and pressure gradient, Cats ‘no etal used Fanning fiction factors throughout the study. They are preserved here to rein consistency wit the orginal work Flow-Pattern-Transtion Prediction. The models proposed 10 predict flow-pattem transitions in an annulus are similar t those proposed by Ansari et al for circular pipes Bubble-/Slug-Flow Transition. The minimum equiperiphery di ameter for which bubble flow occurs is ea dn = 197 | 285 wi bead were dep ithe eqiprghery diameter defined by dep = d+ 3286) and. and are the inside casing diameter andthe ouside tubing diameter, respectively For eqipenphery diameters lager than this, transition to slug ‘tow was foundexperimentallytoccurataverage gas voidfraction values f 0.20 fr flow trough a concentric annulus and 0.15 for ow through fully eccentric annulus. Thus, the Tite eral mod cl was modified forthe bubbles transition in an annulus by Use of values ofthe gas void action measured ahi transition, There: Suling equations forthe wanton then canbe expressed by ¥8t 4 396 2x = Palen - = 50 one 3 | (4267 and vu = pio: v5, = Gs + 0230} | (4268) Eqs, 4.267 and 4.268 constitute the criteria forthe bubble/slugtran- sition boundary at low liquid flow rates in concentric and fully ec- ‘entre annul, respectively Transition to Dispersed-Bubble Flow. At high superficial iguid| velocities, the Bamea™ eriterion for transition to dispersed: bubble flow was modified withthe hydraulic-diamecer concept. The transi- tion From either bubble or slug flow to dispersed-bubble flow then becomes, fetttg] ee” =ons+4n(-) where the hydraulie diameter is defined by (4269) d= dem dy (4.270) AIOUTIPHASEFL OW IN WELLS,

You might also like