GAS-LIQUID FLOW THROUGH CENTRIFUGAL PUMPS.— CORRELATION OF DATA by dim L. Turpin Professor, Chemical Engineering Department University of Arkansas Fayetteville, Arkansas dames F Lea Research Associate ‘Amoco Production Company Tulsa, Oklahoma and John L. Bearden Manager of Mechanical/Hydraulic Engineering Hughes Tool Company, Centrilift Hughes Division Claremore, Oklahoma son L. Turpin i @ Professor of Chem cal Engineering at the University of Ar- ransas at Fayeteuile, Arkansas, He is responsible for undergraduate and grad uate teaching and vesearch in the area of enhanced oll recovery. He received BS.ChE. (1960), and MS.ChE (1961) degrees from the Uni versity of Arkansas and a Ph.D, (1966) degree from the University of Oklahoma, Dr Turpin has industrial experience with Reynolds Metals Company and Exon Company. He has done consulting for Monsanto, S&B Engineers, Amaco Produc: tion Company and the U.S. Army Corps of Engineers. He isa ‘member of AICKE, SEI _ James F Lea ta Research Associate in Fine Production Mechanes Grow of Amoco Poducion Company In Tusa } | He orks on compucer implementation + of exon design and eyes methods | ferret and improved appeaton | techniques. Prenoushy he worked with Prt de Whiney ait and Sun Oi! Campany and taught engineering scence ot the tinventy ley Dr Lew olde 8 S86 and MME. degrees ond a PhD. degree in thermelfiud Science from Southern Methods Univers 13 John L, Bearden joined Centr. Hlughes in 1976 os a Project Engineer and currenty seres as Manager of Me- chanteaFlydrauie Engineering. He for: mmerly worked os a Project Engineer for RC. Ingersoll Research Center, Borg- Warner Comordion, Des Planes, lino, asa Project Engineer Mi. Bearden is @ Texas notive and holds B.S. and MS. degrees n Mechanical Engineering from ‘Texas A&M University He is a reisered professional enalneer in the State of Okiahonia and is a mem ber of ASME and SPE. He hos published a numberof articles ABSTRACT ‘The head capacity characteristics of three diferent submmers- ible centrifugal pumps were correlated as a function of the toriquid rato and pressure at the pump suction. The exper mental data were from two independent laboratory invesiga- tions. One investigation used air and water as the working fluids, while the second study was conducted using dlesel fuel and carbon dioxide with suction pressures up to 400 psig, The results indicated that at low pressures, gas volumes exceeding fen percent began to cause a serious degradation of pump performance. Satisfactory performance was achieved at high gas ratios as ‘the suction pressure was increased INTRODUCTION Its known that the presence of free gas in a liquid afects the performance of a centrifugal pump. Test work as early as 1929 ‘was concerned with the possibilty of controling pump output and, thus, conserving power by the admission of air to the pump suction ((1] as reviewed in [2), although this idea has long since been abandoned (3) Published studies ofthe effects of entrained air on the performance of a submersible centrifugal ump are presented [4], and effects of gas on staged submers- ible centrifugal pumps ate presented [5]4 PROCEEDINGS OF THE THIRD INTERNATIONAL PUMP SYMPOSIUM Analysis of the charactistcs of a pump under the loss of ‘coolant accident for nuclear reactors is studied by Runstadler [6}. Patel and Runstadler present results of visual observations of aithvater flow in a 1/20 scale model pump and show the similarity of a pump and rotating channel (7]. Sekoguchi pre- ‘sents experimental data using resistivity probe techniques and found as air increases the low changes from bubble to shy flow [8]. Then a large air pocket forms near the inlet of each flow ‘channel ofthe impeller and the pump fal to function effectively when the end of the sug reaches the impeller periphery: Sip ‘measurements show gas slugs move slowly, compared 10 the liquid, efetively reducing the area for liquid flow. ‘A program to define the effects of free gas on the perform. ance of electric submersible centifugal pump stages is de. scribed. The effects ofthe free gas show up as a deterioration of, the head-capacity curve, such as areas of unstable head pro- duction, and effects similar to cavitation at higher flowrates, Depending on the amount of free gas through the pump, these effects may vary from slight interference to gas locking ‘Gas interference is indicated on the surface amperage chart ‘by rapid vatlation of the motor loading, Gas locking occurs when the pump ingests too much gas and actually stops pump- ing because its head (or pressure) production is drastically decreased, This causes the motor to unload and to shut down, because of low ampere surface control protection ‘When designing an electric submersible pump for a gassy application, its desirable to know the amount of free gas the pump can’ tolerate and to compare this to downhole gas Conditions. Thus, the objectives of this project were (1) to generate experimental data relating pump performance ie., head-capacity) tothe gas-iquid rato atthe pump suction and to the pump suction pressure, (2) to correlate these data, and (3) to develop a model which would predict head-capacity per formance of a submersible pump as a function of gastiquid ratio, suction pressure, pump type, and any other pertinent parameter Experimental data examined here were collected by two Independent research efforts. Two test faites. and programs are described and resutts rom each are presented and discussed 5] ests conducted by Amoco Production involved using water > tewPenarunt _ =, lh H the | OR) | Mt 6 B] OL @ nana fret all coe] vette a Pe aries ro aT fe honeaear rs aE Comers Fre 1 Schemas lage of Tet Fea. and ait A separate program was conducted for Centilift Hughes, Incorporated at the R. C. Ingersoll Research Center, using diesel fuel and CO;. In both programs, the puxpose of the tests was to define the performance of typical submersible centrifugal pump stages, when fee gas volumes were in troduced af the pump suction under various flow and pressure conditions FACILITIES AND PROCEDURES FOR AIR/WATER TESTS “The test facility at Amoco was an above-ground installation in hich the flow patterns and distibution of air volumes could be Viewed. For simplicity, water and air were chosen as working fluids. A schematic diagram of the test loop is shown in Figure 1 ‘The pump housing, containing fve pump stages, was installed ln an 8 in outer diameter (OD), 7 in inner diameter {ID} plexiglas tube. Because of the plastic container, the annulus backpressure was restricted to low pressures, with the majority ‘of the tests being run at 25 psig to 30 aig ‘The system was designed so that the water flowed through the system in a closed loop, but the air was directed to make a ‘single pass through the pump ot annulus before separation and ‘venting. The injected air that did not enter the pump continued up the annulus and out through a backpressure regulator. The air entrained into the pump was separated from the fuid steam ina gas separator downstream from the pump and was vented through floating ball rtometers for measurement, See Lea and Bearden for additonal details [5] DATA REDUCTION AND RESULTS FOR AIRWATER ‘TESTS The data collected were seduced to compare them with published perlermance curves, including the head, efciency, and horsepower performance values. The motor horsepower ‘delivered tothe system was calculated. To obtain this value, the amperage through each of the three motor leads was measured ‘at 440 V supply and used to calculate the motor horsepower published curves. ‘The average head, h, produced for one stage was calculated from the measured delta pressure across the pump and a calculated bull density forthe air/water mixture based on the following expression p(144) Favaalnd a where ‘Op = delta pressure across the pump, ps! umber of stages “av ww = average weight density of the atiwater mixture through the pump, Ibi, Fons wXGu-+Fan 20s Fa aw (ete) 2 where Faw w = weight density of water, ibftt’, Fava = weigh dent of ag bff, a the averoge pressure between sage inlet and out pressure, Itt du = volumetric fow rate of water, glinin, and GL = volumetic fow rate of ai atthe average pressure between stage inlet and outlet presi, galmin,GAS-LIQUID FLOW THROUGH CENTRIFUGAL PUMPS.~CORRELATION OF DATA 15 ‘The efficiency: E, of the pump for one stage was found ftom the ratio of the output hydraulic horsepower to the input horsepow: ras calculated from amperage and voltage data and the motor ApKisayiqa: Trp 7-48N60)(550)00) 3) where hp is horsepower, Results ‘The resulls (data points) for head only are shown versus published performance curves (smooth line) on Figute 2, re- peated here from a previous study [5]. The data colleced were focused on the recommended operating range. Each data set shows a diferent value of the percent by volume ol gas atthe pump intake. The results show the beginning of serious depar- ture from the head curves at about seven percent fre gas by total volume and intermittent gas locking at about eleven percent. The bands of calculated head, shown when surging begins, indicate the head oscillated From hig to low values with a period of a second or two. Although there i stil the periodic head produced for greater than eleven percent gas volume flows at the pump intake, the pump is not performing near ‘published head values once the percent by volume of gas ‘exceeds some point between seven parent and eleven percent by volume at the intake. 20 wl —rrC~—™——C 10 20 30 40 50 60 70 Gallons Per Minute Figure 2, Deterioration of Water Head Curves with Gas (5) FACILITIES AND PROCEDURES FOR CO,/DIESEL TESTS. A simple closed Joop system was designed to test various submersible pumps. Simplicity was destable because of the problems encountered with two-component flow 'A schematic dagram ofthe tes faciy is shoun in Figure 3. ‘The casing (1), which housed the submersible pump (2), eight inch pipe andthe discharge few tine (3) i to inch pipe. The casing was designed so that various flow inlets (4 an 4a) could be used, and two windows (5) were mounted so that fow atthe pup inlets could be observed. A sheet metal bafe (6) was Placed between the easing and pump. The bafe has the same {nside dimensions asa seven inch eating and forces the Mud to follow a path similar to an actual Seep wel ating, The pump was belt crven by @ 7) hp vertical surface motor (7) mounted below the casing base fange. The nonelectcal instrumentation consisted of a positive displacement flowmeter (8), a diferent pressure gauge (9} betwoan the pump intake and discharge, a pressure gauge (10) for casing static pressure, and temperate. thermocouples and gauges. 42 Whe, Tati 13" Bet changer Figure 3. Research Pump Test Facitty. ‘The test fulds were diesel fuel and COz. These fluids were chosen to test with a gas soit effect present similar to actual cll well production fuids. Additonal discussion may be found in the Lea and Bearden article [5] Test Program For each test, the loop wae filled with diesel, and then a percentage of the total enclosed volume was displaced with CO; After the desired amount of disel was displaced for a Particular test, the pump was started and the static system pressure was stabilized with an additonal amount of COs. Once the systern wes stabilized at the desied test loop pressure and liquid volumetric percentage, the pump performance tests were started, "Numerous tests were conducted on several diferent types of pump stages: a racial pump stage designed for 42-gal‘min ‘optimal fow (1-428), a radial pump stage with an optimal flow rate of 73 gain (C72), and a mixed flow pump stage with an optimal low rate of 80 gal!min (K-70) (Figure 4). The last two stages provided a comparison of two diferent design styles with the same approximate design flowrate. DATA REDUCTION AND RESULTS FOR CO,/DIESEL TESTS ‘The data collected forthe various fluid mixture and pressure tests were plotted as preseutellow rate curves and compared against the purp’s pressure/lowrate performance with 100 percent diesel.16 PROCEEDINGS OF THE THIRD INTERNATIONAL PUMP SYMPOSIUM 1-428 C12 K-70 eA Za Figure 4. Pump Stages Tested. ‘The tests (Figures 5, 6, 7) run on the 142B pump were redyced from pressure to feet of head so they could be com- pared with the airwater tests run on the same type stages. The Dressure was reduced to head by using an equation similar to Equation 1. The average fuld gravity, Equation 2, was modified to account for the COzidisel volume factor (two percent/100 sig) Results ‘The results of the COz/desel tests are shown in Figures 5,6, 7,8, 9,10 and 11. The results of varying gasfiquid volumes at '50-,100-, and 400-psig pump intake pressures are shown in Figures 5, 6 and 7 for an eight-stage I-42B pump, the same type as used in the alsvater tests, Several conclusions can be reached ftom these three curves. Fist, pump performance isa function of both the percent of free ‘gas by volume and the pump intake pressure. In general, the head measuted to the right of the performance curves drops as the Intake pressure decreases. The higher intake pressures 50 nso 60, by Volu 40}. 022 By Nolume . © 10% : £30; © 159% < ae ~ af 2 Sse sn 2 we 1 7498 fe 10 20 "30 40 50 60 Flowrate, gpm Figure 5. DiesellCO,, 50-psig Intake Pressure, I-42B Pump. Head, 5 60, by Volume 40 10% = gq) > 15% gs 20% 8 20; 2 30% OTe ao% 10} --- 0% o 10 20 30 4% 50 60 70 Flowrate, gpm Figure 6, Diesel!COz, 100-psig Intake Pressure, -42B Pump, : 2 ~ et Jaq + 40% gO] oie T20- --- 0% 10. (sO 20m 30240) Flowrate, gpm Figure 7, DieseiCOp, 400psig Intake Pressure, 42 Pump, require lower percent gas compression through the pump. Second, at values equal to or greater than 15 volume percent ‘2s, the pump enters an area of severe head oscilation at Flowrates less than the volume flow atthe design point, which in this case was about 50 galimin. Finally, at higher flowrates, the head takes a nearly vertical dip, which is similar to the effect of cavitation. It is noticed that the calculated head values for the lowest gas volume present (ten percent at 50 and 100 psig) are above the diesel-only base curves. This may be explained as caused by gas holdup in the system, where a small percentage ‘of gas holdup isa large percentage ofthe total gain the sytem. ‘Therefore, uid through the pump can be approaching a pure liquid and pethaps should not be conected uniformly to ac- count for gas presence as was done here. Although the gas presence decreases pressure output, the use of mixed density ‘equations may simply not be adequate to describe head pro: duction over some ranges of performance ‘The effects of gas on the performance of ferent types of stages (mixed and radial) are shown in Figures 8, 9, 10 and 11. ‘The mixed-flow design, K-70, i represented in Figures 8 and 9,GAS-LIQUID FLOW THROUGH CENTRIFUGAL PUMPS—CORRELATION OF DATA 7 and the radial design, C-72, is shown n Figures 10 and 11. Both stages are designed for the same approximate design flowrate. ‘Therefore ary dlllernes in the ability to handle gas would be attributed to their design differences. From the test results, note that even though both stages have a surging region, the mixed: flow design has less deterioration in its pressurellowate curve. ‘Therefore, a8 a pump siage tends toward a highly mixed or axal flow design, its gas-handling capability should increase. Even ‘hough the K-70 pump had less deterioration in head than the 20} Intake Pressure to * SO psig © 100 psig Pressure, psi 0 2040608010020 Flowrate, gpm Figure 8. DieselCOz, 10 vol% CO, K-70 Pumo. Intake Pressure © SO psig © 490 psig 2 200 psig 0 20 40—60—COCT0DS A Flowrate, gpm Figure 9, DiesellCOp, 15 vol% COz, K-70 Pump. : surging $10 eo 3 | take Pressure = | + sopsig [P eee © 100 psig 0. oy 2 200 psig oS es Te o 2 4 60 80 100 120 Flowrate, gpm Figure 11. DieselICO,, 15 vol% COz, C-72 Pump. ‘C-72 pump, both types entered a pressure oscillation or surging ‘area at flowrates equal to or less than the optimal design Towrate CORRELATION OF THE DATA Eighty-six experimental head-capacity curves were utilzed 10 develop the correlations. There were twelve sets of data fot the 1-42 pump operating on aiviwater, 24 sets of data for the I-42 with COaidiese, 24 sets of data for she C-72 with COpidiesel, ‘nd 26 sets of data for the K-70 pump with CO,/diesel Fora gven pump, the controlling factors used to describe the deterioration of head were the gasliquid ratio, the suction pressure, and the capacity That s, Huy (& ¥ (GP. 7 where H = head with gastiquid flow Hg = head with single-phase Haid ow indicates a function of following quantiies 4g, = volumetric ow rate of gas at pump suction, Q = volumetric ow rte of guid at pump suction, , = pressure at pump suction 20|---—Havi Lear singe Pose —««-GENTRUFT #25 5a 0 :: oes 4 eaas se : coe “ 9-985. a i * Head, 90 i - ee, Intake Pressure oo re © 80 psig B to mn : © 100 psig o ‘ol a rome ee 0839046 8070 Bo Flowrat, gom ee Figure 10. DiesellCOp, 10 vol% COp, C-72 Pump, Figure 12. Pump Performance Deterioration for I-42.18 PROCEEDINGS OF THE THIRD INTERNATIONAL PUMP SYMPOSIUM Establishing the functional form of Equation 4 required con- siderable tial and enor. Crossplts indicated a general expo- renal decay in performance with ajQ. The resulting corela- tons for each ofthe thee pumps ae included in the APPEN- DIX The conelatons prac the head-capacty curve fy well for low ges volumes at low suction pressures and for higher gs, volumes at higher suction pressures. The prediction fal of in the diecton of higher gas and lower pressure. conditions. However, the region of poor predcve capabiity ofthe corla- tions coincides with the region of unacceptable pump periorm. ance. The correlations only hold in general to the ight ofthe best eficeney point ad, s the data show, stable head produc- tion with gas only occurs in this region as the gas percentage Increases ‘A second relationship was developed to quantify the region of unacceptable pump performance. This was done by noting the pump performance at the various combinations of gasfiquid fatioand suction pressure employed A quantity 6 was defined 2000 (aa) 3P, 6) ‘ith P, in units of psia. Plots of data with various @ values are shown in Figures 12, 13, and 14 against the correlations shown in the APPENDIX. Note that, in general, =1.0, then the Conelations are fally accurate, and, of perhaps more impor- tance, the head produced is sill a substantial fraction of liquid only ead, Values other than I eauld be chosen tobe more or les conservative { “The following example illustrates how 4 could be used to predict whether or not a gassy well could be pumped with a downhole centrifugal pump or not Example Given the following data: WOR (water to oil ratio) = 1.0 calculate the gas volume a, (in bbVA) at py K-70 °@ = 051 2 = 093 ob = 158 cliterature Single Phase + @ = 3.82 50} se Correlation 40) = 51 Head, 39| Ftd, Tm 20 10 0 20 40 “60 * 80 100720 Capacity, GPM Figure 13. Pump Performance Deterioration for K-70, 00 50| 2 4d : re Corretation 20) pont = 10 10 30 50701 80 Capacity, GPM 110 730 Figure 14. Pump Performance Deterioration for C-72, o-c.conn 5) 88) -(8) where Q, = ol flow rate, bod R, = soluton GOR, SCF T = pump intake temperature, F C = 5.61 fr'/bbt Calculate the liquid votume at pump suction in bbU¢: QQ, (WOR+B,) a where Q = total liquid volume rate at py, bbls BB, = formation volume factor (oil swellage factor), BbIbbL Using Standing’ conelations for R, and B, [9], the following procedure can be used to calculate p, (minimum) from & (2000(4/Q)3p, where = 1.0. Procedure 1. Guess py 2 Calculate q,, Q, P, given the relationship and previous ‘expressions ‘3. Compare p, to quessed P, 4, Return to Step 1 until old ~ new P, calewlated Once p, is found as a function of GOR, it can also be calculated as a function ofthe gas percentage at the suction: x00 a ard ® Intake pressure as a function of gas percentage atthe intake ‘and also versus various possible well GORS is shown in Figure 15, In Figure 16, a family of hypothetical inflow performance telationships (IPR) curves for a well are plotted. A straight line relationship is calculated above a calculated bubble point. The IPRs are constructed with the shut-in well pressure of 3000 pst and atest point of 300 bid at 1500 psi bottom hole pressure Vogels IPR curve is used to complete the curves below theGAS-LIQUID FLOW THROUGH CENTRIFUGAL. PUMPS—CORRELATION OF DATA, » 600; 500| é 8 Gor. 950 152035 3035405 Percentage Free Gas At Intake “N24 Free Gas 8 8 8 8 Intake Pressure, PSI 8 0 50 100 150 200 250 300 350 400 Gor, SCF/STKBBL Figure 15. Acceptable Free Gas Correlation for ESPS. bubble point [10]. Note thatthe five IPRs are shown with GORS ranging from 0 SCF/bb| to 400 SCF/bbl. Distegarding net positive suction head (NPSH) requirements, the 0 GOR well ‘could be pumped off, However, as the GOR reaches 400, the ‘well cannot be pumped off below P, = 520 psi without the corelation indicating the pump woukd net perform saisactorly ith free gas present, Flowing Pressure, Gor — abo 360 zb0 tho bs 00-380 400480 500 550 600 otal Liquid Prod. UD Figure 16, Example of Pump-Off Limits for Reservoirs with Various GOR. 250 ‘This conelahon slows 2 potential pump user to predict whea ‘gns becomes a problem. Then a gas separator can be used or less drawdown of the well can be expected. Some additional stages should be added to acount for head deterioration with ‘5. The additional stages required could be estimated from the Correlations in the APPENDIX or from examination of test data presented. The pump should be designed to operate tothe right Of the head curve to avoid surging when gas is expected through the pump. CONCLUSIONS ‘An empiical model wes developed to predict the head capacity curve for electric submersible centfugal pumps at 2 funeion ofthe gaslquld rato and sucton presure. Approx ‘mate conelations for different pump designs are given by equations (A-) to {A} of the APPENDIX. The comeatons Show the exponential decay of head with increasing (@y/Q). “The regions of reasonably good pump performanced coin cide with the regions of applicability ofthe corlations. These regions are identified by a parameter, 200 (a/Q} 3P, °) For 6 < 1, conelations axe eppicabe and good head produc ton uth dbs presenti possible. For 6 > 1, constant ae not ppleabe, but pump performances In general 30 poor tht operation in these region sto be avoided Representative vals fiom Eaton (9) indicate that each of the pump sages tested wil tolerate about 13 percent ge (by olume) at 100 pia suction presute and up to37 percent get 38400 psia‘scton pressure. uth no serous deterioration of head performance. Test ress show stable head production only at fowratessigher than the best ecleny pont asthe gas Solum at suction increases unt more gs Sealy completely gos Tock the pump. APPENDIX Approximate conelation of the data for 1.42 and the K-70 pumps is achieved by. (an) with “2 given by: (a2) CConlation of the data for the C-72 pump is by: H Fy 92) [1 ~0,0258 (Qo) +0,00275 (Q—Qo/?~0.0001 (Q-Qo}*(A-3) with Q5=983-383.4 (a4) and 285940 2 (as) The conelations are designed for flawrates higher than the best efficiency point. Accuracy is rapidly ost a begins to exceed 1, but the exponential drop in head with increasing q/Q is a general characterise.2» PROCEEDINGS OF THE THIRD INTERNATIONAL PUMP SYMPOSIUM REFERENCES L 2, Seibrecht, W,, “Untersuchungen uber Regelung von Keeiselpumpen,” Z. Ver Deut. Ing., pp. 74, 87 (1930), Stepanoff, A. Centrifugal and Axial Flow Pump, 2nd Edition, p. 230, Wiley & Sons (1957), Stepanoff, A., Pumps and Blowers - Two-phase Flow, Krieger Publishing Company, p. 264 (1978). ‘Murakami, M. and Minemura K., “Effects of Entrained Ai ‘on the Performance of a Centtllugal Pump," Japanese Society of Mechanical Engineering, pp. 17, 110, 1047- 1055, and 17, 112, 1287-1295 (October 1974), Lea, JLF and Bearden, JL, “Effect of Gaseous Fluids on Submersible Pump Performance,” Journal of Petroleum Technology, pp. 2922-2930 (December 1982) Runstadler, P W., Jr. and Dolan, F X., ASME Proceeding ‘Spmposium on Polyphase Flow in Turbomachinery, San Francisco, California, (1978). 7, Patel, B. R. and Runstadler PW, Je, “Investigations into the ‘Two-Phase Flow Behavior of Centrifugal Pumps,” ASME Proceedings Symposium on Polyphase Flow in ‘Turbomachinery, San Francisco, California (1978) 8 Sekoquehi, K., Takoda, S. and Keemori, ¥, "Study of Air/Water Two Phase Centrifugal Pump by Means of Elec- ‘tic Resistivity Probe Technique for Void Fraction Measure ment," Bullets of Japanese Society of Mechanical Engi neering, 27, (227) (May 1984), 9, Frick, T.C., Petroleum Production Handbook, If, New ‘York: McGraw-Hill Book Company (1962) 10. Vogel, JV, “Inflow Performance Relationship for Solution Ges Drive Well,” Journal of Peteleum Technology, Pp. 83.98 (January 1968)