Você está na página 1de 7
PELLET ‘Number styie SPE 28510 Tite style | Scioty of Protez Mechanical Stability of Propped Hydraulic Fractures: A Numerical Study Cases | Mi. Asgian, SPE, and PA. Cundall,tasca Consuting Group inc, and B.H. Brady, SPE, Schiumberger Dowell Crore 1, Sy ot Penton omen te] COPYFIght StyIO Troan open peers ue So maa agama en Para style Proppant is sometimes produced along wityZtrocarbansinhy- 75 psi/fiaresufficien wo destabilize the ch and to cause progressive failure of the propped fractures. introduction ian ‘Proppant backproduction can significantly affect well econom- ics because of (1) loss of fracture conductivity [ung ‘= tions of fractures tgnd to close under stress (Figs. 1 and 2)], @) the damage it inflicts on equipment (i... abrasion of pumps. ‘casing. and wellhead). and (3) downtime and expense required toclean up the wellbore. ‘Refs. through 3 address technical issuesinvolvedin the flow- {ack Uf proppant Ref. 1 preseuts expersental dats lsat show ‘how various properties and conditions (ep closure tress, fac- ‘tare width, proppant embedment in the fracture walls, grain-size ‘istribution, relative inclination of the fracture walls, and draw down) aflect the mechanical stability of cohesionless, unbonded ‘roppant packs. It also raises fundamental questions regarding production (os why docs the emai ‘occur over anarrow range of fracture Para’ style |proppant packs in the laboratory fail at much ‘widths than are known to exist in the field). ‘A nitmerical study presented in this paper addresses some of these fundamental questions. The objectives of the study are t help understand the mechanical process of proppant backproduc- ton and o show how various field conditions and proppant prop- erties affect the propensity for proppant flowback. This is ccom- plished by means of numerical simulations of proppant on a ‘grain-size seale. ‘A distint-clement-ype numerical code is used to compute the forces and motions of several thousand individual proppant ‘grains near the proppant face. The simulations take into account the effects of the following factors on stability of propped hy- ddraulc fractures: net closure stress (net closure stress acting on «4 propped fracture is equal tothe total compressive stress minus the fluid pressure), drawdown, ratio of rock modulas to prop- pami-grain modulus. and fracture width. The predicted resus are ‘Compared io laboraiory and field data. Numerical Analysis Technique: Distinct Element Method (Cundil* developed the distinct-element method (DEM) to ana- Iyze the time-dependent mechanical behavior of a large number ‘of individual panicles of various shapes and properties. The DEM-type TRUBAL code*® (the original version of the TRU- [BAL code isa nonproprietary code written atthe U. of Minneso- 1a) was used in this study to analyze the mechanical interaction ‘of spherical. elastic particles (.c...proppant grains). The DEM ‘technique can also apply to angular panicles. In TRUBAL and ‘other DEM type codes. the particles defarm in th shear and nor- ‘mal directions when they come into contact with each other (Fig. 3). The particles may also slide past one another when the shear ‘sess exceeds the shear strength. A Hertz contact law is used in the TRUBAL code. For such a law, the contacts become stiffer ‘ith an increase in compressive force (because ofthe increase in outact area) despite the fact thatthe mates ial is Lively elms ‘The DEM technique uses the explicit fnive-difference method 1 sole the equations of motion of each panicle fora series of Fi Mu. ASGIAN, PA. CUNDALL 8H. BRADY” ence ‘SPE 25810 successive small imestops. During each imestep the forces at- Ingoneachparicleare aun Uc constan. Than the particle tmotions ean be eomputed directly from Newton's second law (Ghesum ofthe fores equals the mass tans the acceleration) In- tegration of te acceleration, with respect to time, yield the par- ticle velociry a the end ofthe timestep. Also, integration of the ‘velocity with respect to ime yields the displacement. Arte be- ‘inning ofeach new timesep, the new forces can be calculated {rom the "just-computed displacements by means of the contact laws (by means ofthe force-dsplacement relation forncigh- bonng parucies). Res, 9 trough 9 provide dell of we com putational scheae and other TRUBAL features. “To use TRUBAL to sul proppan' flowback, code modifica- ‘ions are required to account for the following conditions in propped hydraulic factores: compaction during fracre closure, proppant embedment, fracture roughness, and drag forces in- Aoced by drawdown, “The firs sip in each TRUBAL simulation i 10 place several thousand proppant panicles in the fracure (Fig. ). Next. the Pormant celle byte scr wals ute at aonee ‘Sess isatained (During early stages of compaction, acompxes- ‘ve stress equal in magaitnde tothe net closure sess is also sp led to the proppant face. This srestis removed atthe endof he ‘Compaction afr serving its feaction, which isto proven aifi- ‘ial propa production during very rapid fracture closure.) The tet closure stress isheld constant for the duration ofthe simala- thon by continual adjusment ofthe fracture wall displacements. “After compaction is compete fractures that are stable under net closure sess ape subjected wo drawdowns. Fluid flow is not smedeled explicitly in TRUBAL. Insiead. drag forces ending 10 Dash the proppant in the direction of flow are applied to each Droppant grin. The magnitude of drag force depends on the {ansize es wellasthe drawdown, eccording wo the solution giv- nin the Appendix. “The above procedures of compaction to net closure stress, ad- {justment of net closure sass, and application of drawdown were ‘aed to simulate the mechanical stability of propped fractures un- derthe vargas condions described i the next secon. From series of preliminary ess to determine theinfluence of various numerical parameters on the prediced results, we found 1hat50 grain diameters was more than adequate to represent the temte fracture lengh-Anintinitely high fracture i modeled with the periodic boundaries in Fig. 4b. The section between these boundanes is representative of all sections along the fracture height (Le the section is bounded by identical neighboring sec- ‘ions. Thus, the paricls on the right side ofthe boundary intr ‘ct with ones identical those onthe lft side. By numerically taking this interaction into account, an infinitely high fractures modeled. (A 30-t seca, or 36.300 grain diameters of No. 40 {siginte in height compared with its width of $1 in.) Note thatthe distance between penoic boundaries is continually adjusted daring compaction so that the compressive load acting tn tese boundaries fs equal toa specified value (in this study, the net closure sires). Afir compaction, each periodic boundary {s "fied" in apace, butellowance ie made fer paicles 8 PPOs ing penodic boundaries to imeract mechanically. Eight numerical simulatione chowhow the following parameter: ‘can affect fracture stability: net closure sess, fracture width. ‘mean grain size and drawdown (Fig. 4 and Tables 1 and 2) Cases I through 3c pertain to fractures of various widths with anct closure stress of 940 psi. Cases 4 and 5 pertain to fracnures 5.6 prain diameters wide with anet closure stress of 2.150 psi. In ‘all eases, 20/40.size proppant has been used, with a mean grain diameter of 0.028 in. for Cases 4 and $ and 0.023 in. forall other ‘cases. Drawdowns ranging between 75 psift and 150 psift are considered. The gravtayonal force on 2 single proppam grain (mass equal to 2.4% 10- g for No. 16 size, assuming a specific ‘gravity of 2.7) is =1% of the drag force owing toa drawdown, of 75 psif. Hence, gravitational forces are not important and ‘were not considered inthe simulations. ‘The last column of Table | suramarizes the stability of each propped fracture. A comparison of Cases | through shows that fractures wider than =5.5 mean grain diameters are inherently unstable under net closure stress. This is somewhat dependent on Aaveragenetclorure stress and fracture wall modulus, Case4 (soft fracture walls but relatively high stress) and Case 5 (stiff walls and high sess) are anstable at 5.6 mean grain diameters, where- 2s Case 3 (soft walls and low suess) is sable at 5.5 mean grain diameters. ‘Compariton of Cares 3a through 3e shows that drawdowns of | ‘more than * 75 psift are suficient to destabilize fractures that are otherwise stable under net closure stress. Fora drawdown of 75 pevft (Case 3c), the average radial (compressive) grain-to- ‘grain force is 0.478 Tht, whereas the drag force is equal 10 7.9% 10°5 Ibf(0.016% ofthe average radial force). Inspection of ‘Fig. 5 shows how sucha small drag force can destabilize a frac- ture. The figure shows “snapshots” for Cases 3 (1=0) and 3c (© 0) of the proppant-grain velocities (vectors emanating from ‘the grain centers) and of the radial grain-to-grain forces (bars ‘consecting grain center; the thicker the bar, the preaierthe com- pressive force between grains). A geometrically irregular arch of Proppa goalie fs forte beso te propypant face ana reaul of ‘et closure stress. The arch is at incipient failure under net clo- sure stress; we know this because the columns of proppant in front of the arch have buckled and carry virtually no losd. When Ihydrocarbon production begins (1>0). the loose particles in front of the afch are transpored imo the wellbore. As Fig. 5 shows, a75-psi/f drawdoun has no noticeable effectan the arch or however, a 100-paiffi drawdown (not [-neasaae fren stoner va Interpretation. The proppant pack in fractures propped with co- ‘esionless,unbonded proppantis inherent!y unstable for fracture widths greater than or equal to =5.5 mean grain diameters, which s consistent with published laboratory data" The numer ‘cal simulations go one sep further than the laboratory tess; they show failure occurs through arch collapse. The predicted fracture ‘width at the onset of such proppant flowback (equivalent to (02-in for No. 16 mesh-tize proppant) is apparenty inconsistent ‘sPe25810 with eld data. Stable propped fractures 0. in. wide exist in the {icld, One possible explanation is that proppant lowback occurs vera mal fraction of the length asin Fig. 1)butisbridged open atthe wellbore by a single layer of widely spaced proppant rains. An open fracture is exremely permeable (an 0:008-in.- wide gap. equivalent toa No.40 proppant radius gap bas afrac- ture conductivity of 2.650 md-f) if it approximates parallet plates; therefore, it has no noticeable effect on productivity. ‘Another possible explanation is that proppant in the field is “onded” by gel residue or by capillary forces. This explanation ‘ncannintenl with labuiainysunbconiiol experiments tha show tension prevented 20/40 Onawa sand la “ierin. circular opening, provided that ‘coough and that the water was immo ‘cal results and field data isthe rat of proppent production. Inthe ‘numerical tests, mechanical failure is catastrophic (i. proppant dein the simulations (cg. pres- that degrades with time). ‘Numerical Experiments. Toe numerical experiments show that <¢raucowns of less than = 75 psi are not large enough to affect the stability of propped frcturesbutare large enough to ransport loose proppant (ic. proppant in front of the arch athe proppant ace) wo the wellbore. In the field. operational drawdowns typi- callyareon the order of 1010 30 sift, well below the =75-peit ‘valu: at which proppant flowback occurs inthe numerical tess, It's possible that pressure transients above the operational (de- sign) drawdowns may be responsible for proppant production in ‘the field Such transients also may explain why proppan: produc- tion in the field can be gradual rather than instaniancous. Ifthe drawdown exceeds =75 psf. then the drag forces are sufficient to destabilize the propped bed. Tis is true despite the ‘act that the drag forces are <0,02% of the average radial (com- ressive)erain-to-erain forces induced bv the net closare stress. ‘This implies that bonds with very low shear and tensile strength ‘would provide sufficient resistance to overcome the destabiliz- ing effects of fuid flow. This is consistent with laboratory data ‘where resin-coated proppant with an unconfined compressive strength of only 0.6 psi was able to withstand 25 cycles of sess (net closure ses varying hermeen 650 and3 900 si) at gas flow rales equivalent 1010 to 30 MMsef?D before proppant produc- tion began? Conclusions 1. Hydraulic fractures propped with cobesionless. unbooded ‘proppant fail at acrtical mean-grain-diameverfractre width ra- io, 2. For widths les than the critical value, such propped frac- tures developanarchat the proppent face. Infront ofthe arch, the columes of proppant buckle and carry virtually no load. Hence. they can easily be tansponed by flowing hydrocarbons 10 the ‘wellbore. The arch remains stable provided thatthe drewdevns are les than =75 psf for 2040-size proppant. 3. Drag forces induced by drawdowns Jess than = 75 psi are very small (1% of the compressive grain-to-grain forces) but _aze sufficient to cause failure of the geometrically imegular arch neacern MECHANICAL STABILITY OF PROPPED HYDRAULIC FRACTURES: ANUMERICAL STUDY 3 of proppact grains. Once destabilize, the arch migrates inward from the proppant face along the agth ofthe fracture 4. Distine-element-ype numerical models can provide valu- able insights into mechanical-failure processes in propped hydrau- li fractures. Thus, in conjunction with laboratory and field test = cohesion of proppant pack, mvL1, psi ‘De mean grain diameter, Ln. = Young's modulus, m/L, pi Fe force, mL bf y= Sacre beh, Lf ‘be permeabiliry, C2. md LL fracture length, L, Len= lenath of fowback 2one.L. ft ‘= numberof grains in vohume dx 2 wy radius of proppant grain L, in. Ye = farfild velocity offi flowing around an isolated sphere. Lt. fvsec fracure width, Lin. {re distmce in direction of fracture length, Lt y= distance in dretion of fracture wid. L ft = istnce in direction of fracue height, Cf fiction angle of proppant pack. degrees dynam viscosity. m/Lt. ep ‘Acknowledgments. ‘We thank Roger Card, Jean-Pierre Ferand, Paul Howard, Ken "Nolte, and fan Walion of Schlumberger Dowell for valuable dis- cussions on the interpretation ofthe numerical tess: David Rus- sell of hasca G are modifiea- ‘ratory Stuy.” paper SPE 2482] presented the 1997 SPE Anaaal “Techneal Conference and Exhibiban, Washingtes, DC, Or. 4-7. 2, Vieeburg Red. ere: “Proppan Bactpreducion During Hyérabc Fractunag—A New Failure Mechanism for Resin-Coutd Prop- pans,” JPT (Oct. 1994) 884, 3. Maris, eral “Devited Well Fractring an Proppnt Contd in the Prudboe Bay Fk” paper SPE 24858 presented atte 1992 i ML. ASGIAN, PA. CUNDALL BLH. BRADY SPE 25810 ‘SPE Annual Technical Confereace and Exubiton. Washington, Oe 27. 4 Cancall, PA. “A Compas Model fot Simulating Large Seale ‘Movements in Blocky Rock Systems." Prec, Symressm of ts Im Soe. fr Kock Mechanics, Nancy. France GB7D 1. 5, Soack,O.D.L and Cundal, PA: “The Dist lemen! Method at ‘Tool for Research Granalar Media, Par "report, Nall Science ounduion NSF Grant ENGTS.20711, Depo Civil & Mineral Ea ‘Boeerng, U.of Minnesos, Mineapole (Nov. 1978). 6 Cundall PA and Stack, ODL: "The Disine Elemeat Method as 2 Tool for Research in Granular Media Par T."repo Nal Seence Jes (es) Elsevier Scieace Publishers, amseréa (1988) 113. '. Cuadall PA. and Suack. OD L-"A Discrete Numerical Model for ‘Granular Assemblies." Getechnique (1979) 29, No 1,47. 8. Candal. PA. Jenkins, 3-7, and Ieibahi 1: “Evolston of asic ‘Modul in a Deforming Granalat Assembly.” Powders ond Grois, ‘Blarez and Goures (eds), A.A. Balkema, Rover (1989) 319, 10. Mall, CD de aa WF. “Sabiliy of St Arches: A Key to Sand Contol.” JPT Guly 1970) 821 11. Asgian, M1, Cundal. PA. and Brady BL. “The Mechanical “Sui oF Poppe Myorasac Pacts: A Nematea Sach pa ‘er SPE 285 0 presented at the 1994 SPE Anal Tecnieal Confer ‘nce and Exhibition, Sept 25-28. 12. Bad, RB. WE. Stewart and Lightfoot. EN Transport Phenome- ‘na Job Wiley & Sons. New York Cny (1960) 56. ‘Appendix—Drag Force Magnitude ‘The drag force tending to push a proppant grain inthe direction ‘of ud flow has two components: pressure and shear The pres. ‘sure component results from the fac tha the pressure onthe up stream face ofa proppant grain i greater than that onthe down ssream face because of drawdown (dp/ds) The pressure ‘component varies with varies linearly with drawdown given Lexatonene (avalyie) solu Feo |W/3ierep/as}. a1) [Note thatthe varahles in all equations inthe Appendix have & Self-consistent set of units. For the shear component, suppose thatthe following analytic solution for laminar flow aroundan isolated gran applies toflow {in propped fractures!2, Fi = Gaur (82) Ive istaken tobe the average poe velocity in a propped frac- ture... Darcy velocity divided by effective porosity), then Ea, ‘A-2 bas te form a3) Forte shear component tobe greater than, for example, 105 of the pressure component, then, teens > aingyrme 2s ea o> 0109/97 ve For most practical situations, the effective poroiy. 4. 6 > lO Randle pppactratentt > COURS Ne wee eo Pant). Thus, Eq. A-5 implies that #> LS x 10°? in? = 100 darcies «=... (4-6) ‘Such apermeabilty sat the upper range of permeabilities usu- ally found in propped fractares. Thus, the shear component of drag force is effectively zero if the low regime around an 1s0- lated sphere is applicable to pore flow in a propped fracture and, therefore, the lower bound forthe shear component of drag force An upper bound forthe shear component of tag force can be ‘computed from limit-equlibriam considerations as follows, (Consider the overall force balance on the fluid contained inthe ore space of the propped fracture of Fig. 4. Ifthe walls and trans are fixed and a pressure is applied on the fuid only, thea the net applied fore is equal to the pressure difference between += and x= dx multiplied by the cross-sectional are of flow: Fay * (Ap /8x\dx{wjdz). ..... 7) ‘This resultant applied force i resisted by shear forces that de- ‘elop on the fracnae walls and by forees that develop on each of ‘he proppant rain. Thus, the sam of the forces onthe grains can ‘ever exceed the net applied force Shee. ‘the particles are all the same size, then Wy < Fae coesees ces AS) ‘But, by definition of porosity n can be expressed in tems of the ‘otal volume andthe particle volume: me [0 ~ grertow YAP) ee et) Substitution of FqA-10 for m and of ig. A 7 for Fapp into Ey Ad yields F, < G/3uerhap/axy/a ~ 9). sel) ‘Note that this expression must include all components of the force (reactions, shear and pressure drag force components. et.) because it was derived from an overall force balance. The actual particle force must be less than this because pat ofthe applied {ore is resist by the fracture wall Inpractice.9 will be < 50%, otha the total drag foree (shear and pressure componens) will lays be les than te following quantity: For S 26/9 PP/AEL ecco (12) (Comparison of 3s A-12andA-1 shows thatthe shcarcompo- Det of ag frei iss han the presure component For the apn fh sh sea compenn en ob ‘othe pres Secate this isconservatve for design Duper Fis. Al anes tl (rere pes ae fort range of drawdown end grin sz, se AS) ‘Matic style }WORELLIC FRACTURES: A NUMERICAL STUDY 5 ‘Tab Set: 0.2R, 0.851, 1.90R £2.03, ‘81 Metric Conversion Fa Delo EeRZ Rxaoe E-01 ff 9290 308" E-02 =m? fi 2831 685 E-02 =m? in x25a" | E+00=em Wf x4448222 E+00=N md x9:869 233 E-08 myn? si X6894.757: E400 “kPa erease10 ne t ft ft i asi wo ong (Chasen armen) can fea 100 tas of eons eee aee se SRT Mu, ASGIAN, PA. CUNDALL 61H. BRADY Mechanical device Mechanical device ‘to represent grain torepresent grain shear deformations compression Fig. Features of wstinctslement¢ype medals fer simulating ‘mecharucal ‘benor of spherical paricls (.,Proppant and (gravel pects) ons grametze seal Co Fig, Four views of prepped nydraute trctue simulated by a etinctsement code. White Dries represeat propper graina. shased pares, except those slong the y axis, represent ‘cure wale. ol the ieckse wall parbcies are acjusted conanal 20 that hat ‘Secure sien: reesina comatan. Pats how isin the Tenec¥or= ‘spe esto SPE 25010 MECHANICAL STABILITY OF PROPPED HYDRAULIC FRACTURES: A NUMERICAL STUDY ,

Você também pode gostar