A NEW METHOD FOR COMPUTING MINIMUM FLOW by 8. Gopalakrishnan Byron Jackson Products BWIP International, Los Angeles, California S. Gopalakrishnan ts Manager of Re- search and. Development for Byron Jackson Products, BWHP International, The.,m Los Angeles, California. Dr. Gopalakrishnan obtained the Doe- toratein Engineering degree tn Mechant cal Engineering from the Massachusets Institute of Tchaology in 1969. Dr. Gopalakrishnan has been responsi ARR bic fo edeanced research in the areas of Sluid mechanics and rotordynamies. of turbomachiner. His work at AVCO Lycoming (1968-73) cul nated in the deelopment ofa numerical technique for transonic flow calculation in turbines and compressors. His work athe Borg,Warner Research Center (1973-78) was related fo paps, compressors an torque converters. ‘Heh been with Byron Jackson since 1978, ntally nthe c- pacity of Senior Hydvculte Engincer wet responsibility in the evelopment of a computerized hydraulic design system. At present, eis responsible forthe development of new products “pplicabl to nuclear and fosil power plants and tothe petro leur refinery industry. He is a member of the International, Pap Users Symposium Advisory Committe ABSTRACT Appropriate specification of minimum flows is important to avert pump rellabiliy problems. A method is presented for the falculation of minimum flow by allowing a satisfactory margin below the onset of flow recirculation in the impeller. A semi analytical method is proposed forthe calculation of recieulation ‘onset. The margins depend upon site, speed, specifi pravty of pumped fluid, NPSH available, intermittence of operation and ‘mechanical design. Empirical charts axe provided for estimating these margins INTRODUCTION Many pump reliability problems in the field are being attri- buted to the operation of the pump at flow rates well below ts best efficiency point, Generally, the vendor and the user try to avert such problems by specifying a minimnm flow, below which the pump should not be operated. Unfortunately, the eriteria for specifying sueh a minimum flaw are not clearly understood by all parties concerned. The user normally would expect the ‘minssnum flow to be the lower limit for reliable operation ofthe ‘whole pump. This tequirement includes reliability of bearings, seals, auxiliary systems, etc, On the other hand, the vendor's specification of minimum flow is typically uot as broad-based, and reliability problems ean, therelore, arise. The vendor nor ‘mally represents minimum flow from a hydraulie, or cavitation view point, For example, historically, minimum flow was based ‘upon the point at which the pump inefficieney raised the uid temperature by a specified amount. Generally, ths flow rate is 4 sital fraction of best elfcieney point flow (BEP flow) and ex tended operation near this vakue would lead to premature pump faslore in most pumps, ASTRATEGY FOR DETERMINING MINIMUM FLOW, Through the entire head es low operating range of a pump at constant speed, several important flow rates can be established. ‘This is shown in Figure 1, with the positioning of the various flow points being somewhat arbitrary. These concepts are briefly described below. More detailed description and calculation will be given in the next section, HEAD TEMP.RAISE CAVITATION SURGE} SATISFACTORY BRG. ‘AND SEAL LIFE LONG LIFE-IMPELLER, SUCTION RECIRC. MO14 DISCH. RECIRC. SHOCKLESS, Figuee 1. Typical Head-Flow Charucterstic of a Centrifugal Pump Indicating Seceral Important Flow Rates, ‘The best efficiency point (BEP) is the flow rate at which the combined hydraulic losses of the pump unit are the lowest. The individual component lasses in general are not all minimal at this point. The low rate at which the relative low te the impel ler is perfectly aligned withthe impeller vane is called the shock: less entry point. For impellers designed to have high suction ca- pacity the shockless entry point is usualy to the right of BEPa PROCEEDINGS OF THE FIFTH INTERNATIONAL PUMP USERS SYMPOSIUM At flows lower than BEP, recirculation flows ean commence 35 shown in Figure 2, For a given NPSH available, an impeller starts undergoing cavitation damage at some flow rate lower than BEP (assumingthatat BEP the available NPSH isadequate to prevent damage). The bearing and seal life begin to be im- plred at some low low rate because ofthe lage dynamic forces ‘generated by the hydraulic phenomena in impellers and volutes or diffusers. At very low flow rates, large scale axial and trans- verse motions occur because of two-phase flow regiines arising from severe cavitation. Also, as mentioned earlier, high pump ing inefcieneies at low flows enuse significant temperature rise of the pumped fluid Figure 2. Recirculating Flow Pattern in the Meridional Plane of 4 Centrifugal Impeller Specification of minimum flow must recognize the purpose for which tis done. For example, foraireraft fuel pumps, vapori- zation ofthe inlet flow isthe main concern rather than seal or bearing life. Consequently, temperature rise is the criterion for setting up minimum flows. For low energy pumps operating ‘with marginal NPSH, the eriterion could be impeller life from ‘cavitation damage point of view. For high energy pumps, even ‘modest recirculation may be intolerable because of severe vibra- tions, in which case recirculation onset points will establish sinimum flow. Apart from these special eases, for abroad range of centrifugal pomps, overall pum reliability is perhaps the best criterion for Specifying minimum flow rates at which hydranlie interactions cause forces which compromise seal and bearing reliability. The basic assumption in this strategy i that bearing and seal lie is alfected by flow rates. The calculation outlined below demon- strates the validity ofthis assumption The radial load exerted by an impeller varies with flow. The statie radial thrust coefficient K defined as Jo H Deby {is about 0.025 for double volute pumps at the best efficiency point, The value of K increases as the flow rate is reduced. For single volute pumps, the ratio of K at shutof' to that at BEP can bbe as high as 20/1), For double volutes, this ratio ean be much sumaller (as low as 1.5 to 2.0). However, the dynamic forees in- exease rapidly with veduetion in mass flow. The work of Kanki [2} shows that near shutof, strong subsynchronous and vane ‘passing force components exist. K values of about found for dif- fuser and double-volute pumps. Based on these data, the var ation of K with flow fora 4.0 in discharge process pump with an ‘overhanging impeller hasbeen hypothesized as shown in Figure 3. From these impeller forces, beaving loads can be calculated and the average bearing life expectancy ean be estimated. Since bearing life varies inversely as the cube of load, life decreases dramatically with flow as shown in Figure 3. At BEB, the ex pected life is indefinitely large, but asthe flow goes dosin, itde- creases rapidly. For example, at 40 percent flow, expected life is only about six months 03 }30,000 K 02 20,000 O41 10,000 = g 2 £ 0 02 04 06 08 1.0 1 Qsee Figure 3. Reduction in Anticipated Average Bearing Life Due to Inereasing Radial Loads at Low Flows. Seal life depends strongly upon shaft deflections. Even ‘though no empirical rules are available to relate seal life to shaft deflection, its evident that increasing forces tend to decrease seal life Since the onset of recirculation signals the generation of strong forves, the strategy of determining minimum flow i volves providing an allowable margin belov the onset point ya tehalgues for cleulating these fits ave desertbed low. min = Qe Ky Ke ky ka by Qn = recirculation onset point (suction or discharge, Whichever is larger) factor to account for power density factor to account for liquid specific gravity factor to account for’ margin of available NPSH over re quired NPSH ky = factor to account for intermittency of operation k= factor toarcountfor specific mechanical design margins The techniques for calculating these factors are de- seribed laterANEW METHOD FOR COMPUTING MINIMUM FLOW 48 COMPUTATION OF MINIMUM FLOW Recirculation Onset ‘The phenomenon of recirculation has been extensively re- searched and many publications dealing withthe theoretical as pects as well as control may be found in the literature. Reeirew lation flow patterns are complex. The view of the meridional plane is shown in Figuee 2. In the blade-to-blade plane, there isa strong convection ofthe fluid particles toward the pressure side, asthe suetiem side i covered by separation cavities. When the convected bubbles reach the pressure surface, they implode because of the higher ambient pressure and cause damage an the backside ofthe vane. Thus, recirculation damage often isnot seen easily on the impeller unless care is taken to look for damn- age on the backside using mirrors. In extreme cases, the back side damage is large enough that a hole i poked through and then it is visible on the suction side. In general, the onset of recirculation is gradual, and the low rates at which it begins ate different for suction and discharge. Ofallthe publications dealing with the recieulation phenomena ‘that may be found in the lnterature, the work of Fraser [3] pro- vides the simplest method for calculation. The discharge recit- colation onset point Quy is given in as: [3] Qon = 7 DF be # pn a ‘The quantity Cpp was empirically found in Reference 3 to be a fonction ofthe ext vane angle B ‘The suction recirculation onset is given in (3]a: en with Cx given asa function of the vane inlet angle By Fo DP 0-y)TH e Dee ey For Dy 05. Gsn™ Son Using these relationships, suction and discharge recirculation points can be caleulated. ‘The suction recirvulation onset predicted by Equation 3s in good agreement with test data for normal vane angles. For B valves less than about 15°, this equation appears to predict rather low values for the onset. To iamprove this situation, the present study has utilized the following equation: oq = tan 8 (1 ~ 0.2001 (8 - 9.5)°4) “ Even though the above equations are simple, their use is lim- ited because they require a knowledge of vane inlet and dis- charge angles. An approximation to the suction recirculation ‘onset point can be made inthe following way. ‘eis genarally understood that suction specific speed and sue- tion recirculation onset are linked together. The higher the S,, the higher the onset points expected tobe. The slassteal expres sion for NPSH at shockless entry point is 2g NPSH = 8, cyua® + ag WiE ® where a and a, are constants In the absence of prewhirl Wi? = Uj? + ei? Also at shockless entry point equ = U; tam 8 Substituting these in (6), we get 2gNPSH = (a, +a) U3 [tanh + ‘The flow rate at shockless entry poiat is: Qe = Fw DEG - 9 tan By a 5, = NG * INPSHP =386/1- 7 fh 8 tt, + 2S Dividing Equation (8) by Equation (7), after using (4), we get Sse = 0.90016, -9.5)°* Se =1- 0 s00108,- 9.5 Since Equations (8) and) ae functions ony of, itis clear that the ratio Qswuosels uniquely related to 5, The constant a represents the non-uniformity ofthe velocity prfle at the eye: A good discussion oft effects may be found in Spring’ article [4]: For approximate ealeulation, an a8suimp- tion ofa; = 1.2 is generally adequate “The constant ag represents the dynamic depression ofthe static pressure and ita function of the detals of the vane 0.28 0.20 os] o.10 0.05 °. os 10. 18.20, 25.60.85, a0. DEG. a Figure 4. Variation of NPSH Coefficient ay with Vane Inlet Anete 20. — 16. eo x To ) 8. — 4 10. 15. 20. 25. 30, 35. DEG. B, Figure 5. Variation of Suction Specife Speed with Vane Inlet Angle,“4 PROCEEDINGS OF THE FIFTH INTERNATIONAL PUMP USERS SYMPOSIUM ‘geometry. Spring states that a varies from about 0.05 to 0.15 (4). In the present author's experience, a depends strongly om By. A. typical variation Is shown in Figure 4. Using a, = 1.2 and ay giten by Figure 4, suction specific speed can be plotted as a funetion of By using Equation (8), as shown in Figure 5. Equa tion (9) plotted in Figure 6. Using Figures and 6, the relation ship between suction recirculation onset and suction specific speed ean be graphically plotted as shown in Figure 7. This fig tre demonstrates the expected tendency that high suetion spe- cifie speed machines tend to recirculate closer to the shockless entry point, 10 08 0.6 Qs, SE 94 02 1015. 20. «25. 30. 35. DEG. 8 Figure 6. Variation of the Ratio of Suction Recirculation Onset Flow to Shackless Entry Flow with Vane Inlet Angle. 10 o| os oa] Asp lOse 02 10 o| os oa. QcniQoee. o2| 6 8 10 12 14 16 18 20. 8, x 10° Figure 8. Variation of the Ratio of Suction Recirculation Onset Flow to BEP Flow with Suction Specific Speed. sich san vege betwee the wo vas may be desi le. If information isnot avalable, Figure 8 can be used as & reasonable estimate, = Example A single stage end suction process pump of 1750 specific speed was tested for reciculation onset. The static pressure athe inlet pipe was measured just atead ofthe impeller for various flows, ‘The absolute low angle was also measured near this loeatisn as close tothe pipe wal as practical. These variations are shown, Figure 9. The sudden increase in pressure is indicative ofthe beginning of recirculation ax explained by Fraser (Sl. The pres ence of preroation als shows that recieulaton has commenced. Tfmust be pointed out that there is considerable uncertainty in the measurement offlow angles near the wall, rariculariyas the flow is quite unsteady, The recirculation onst shown in Figive appears to be about 1000 gpm, when one considers the stati pressure increase. It sppears 0 be abost 1200 gpm, ifone ob- Serves the beginning of inlet prewhir!. The discrepancy is due to both the uncertainty in angle measurement (as much as 15 de- 0 6 8 10 12 14 16 18. 20. Sq x 10° Figure 7. Variation ofthe Ratio of Suctional Recirculation Onset Flow to Shockless Entry Flow with Suction Specific Speed. ‘Tosimplify the caleulations stil further, ifwe assume that Qse. 2 Quen the recirculation point can be plotted as a ratio to BEP flow against §,. This variation is presented in Figure 8 ‘When vane angle information is available, the onset pointean, be caleulated in two ways: one by using Equations (2) and(3)and the other by using Figure Gand Equetion (7). In general, these results may not agree because of the assumptions involved. In STATIC PRESSURE JUST AHEAD OF IMPELLER (in of HO) a NY 05 oe SBS Tad Boo" shoo FLOW RATE (GPM) ABSOLUTE FLOW [ANGLE (OE). Figure 9, Experimentally Measured Suction Recirculation ‘Onset — Variation of Absolute Flow Angle and Static Pressure at Impeller Inlet with Flow Rate.A NEW METHOD FOR COMPUTING MINIMUM FLOW 4s rves near the wall) and also tothe fct that the state pressure tap as located slightly upstream of the wedge probe. "Yor this impeller By = 21 degrees giving Csy = 0.171 fiom Equation d). Substitingsbisia Equstion (9), with Dy = 6.875, in, w= 188.5 Vsec, sod = 0.418 gives Qoy = 880 ap, Using the second method requires knowledge of 8, which from test data, i 1150. From Figure 7, QgwQse = 0.46. From Equation ©, Que = 1950 gpm. Therefore, Osx = 910 gpm which is else to the B80 gpm caleulated by the fst method. Thus, the predic tion ofabout 900 gpm agrees reasonably well with the measured value between 1000 and 1200 gpm "To check the validity of the calculation for higher specific speed, comparison is madefor svertial pumpofN, = 7000. The ‘aration of absolute angle with radius at th inl is shawn in Figure 10 for two flow rates, With BEP at 9500 gpm, it can be seen that minor reirelation is evident at 6000 gpin, while at 4000 gp, recieulation dominates nearly the entire ialet area. 0 6000 GPM 20 | 1 4000 PM PRE WHIRL ANGLE iS & SINGLE STAGE MIXED FLOW PUMP) 800 APM, 20 INIA a= 80 g 4 100 . 2 4 6 8 10 D-Dy D,- Dy Figure 10. Variation of Absolute Flow Angle at the Inlet for a 7000 Specife Speed Mixed Flow Pump. For this impeller, 8) =22.5" giving Coq = 0.173. With D, = 12.195 in, y= 0.289 and w = 158.1 Ler, we get Qgn = 5000 ‘gpm. Using the second method with the known 8, = 12650, the results Qo Qge = 0.51. Ose iscalculated tobe 11970 gpm from Equation (7). Thus, recirculation onset flow is 11970 x 0.51 6100 gpin. Between the two methods, a value fairly close to the observed onset is achieved. Tn the above examples, 5 has been cxlculated at BEP, Strictly, the eumpulation method is intended to be used with S,, calculated at shockless entry point. Fortunately, in most cases, the error caused by this discrepancy is small. For example, 8 the second ease, 8, calculated at 11,970 gpm is 11,000. Then Qay/ ‘Qse = 0.46, and onset low is 5500 gpm. In the frst case, $, at shoekless entry’ is 10,900 giving Qsq = 890 gpm. Observations of recirculation in practice are not precise. The position of the pressure tap, or depth af immersion of velocity probe below the pipe wall, strongly alfect what is recorded 95 recirculation onset. Since tiny areas of recirculation may exist immediately ahead of the impeller &t neatly every flow condi tion, care mast be exeresed in identifying recreation. The present cilulations ean be expected € agree with measuted data within about 10 to 20 percent Allowable Margin below Recirculation Onset The minimus Now to be specified wil be less than or equal tothe recirculation oset pata. The allowable margin wil be de termined bythe actors shown in Equation (}) These betorsand their determination are enumerated below. ower Density: The fores ean be expected to sale as N® D* with N being speed scaling factor and D size sale. Since N* Dis N 0.68 = 1088 gpm. 1k, = 0.85 for 1600 gpm pump at'3600 rpm from Figure 11 ky 0185 and ky = 0.87 Irom Figure 12. Thus, Qua = 1088 * 0.85 % 0.85 X 6.87 = 680 gpm which is very close tothe value previe ‘ously caloulated by Heald and Palgrave (5 380 X 0.8 x 0.8 x 0.87 190 gpm. MINIMUM FLOW FOR IMPELLER LIFE Covitation damage may restrict impeller Ife in. margisal NPSH cases. The method of specifving minimum flow for such applications involves determination of the required NPSH for long life. Analytical methods for such ealevlations do not exis. However, for approximate specifications, the method of Gopala- ‘rshnan (6) may be applicable, Here, NPSH required to avoid damage is calculated as function of lowe The flow rte at which the required NPSH exceeds the available NPSH can then be specified as the minimum flow st which long impeller life can beexpected CONCLUSIONS In the majority of industrial pump applications a strategy for specifying the minimum fow ean be based upon obtaining satis: factory bearing and seal life. The method presented herein it volves calculating the recirculation onset point first and then ap- plyinga margin below this point ‘A method is provided for calculating the recirculation onset points, For suction recirculation, the method isslight variation fof Fraser's report [3]. Since this fechnique requires some de- tailed information about the hydravlic design, a simpler method requiring knowledge of suction specie speed only is also given. The allowable margins depend upon size, speed, specific gravity of pumped liquid, available NPSH, intermittence of op- eration and mechanical design. Empirical charts ae provided to compute these factors. Ittmust be pointed out that in spite of the directness of the proposed method, allowances must be made in practice to ac Count for specifi pump designs and applications. NOMENCLATURE, 41.82 coeficients used in NPSH calculations 1b” * smpeller width © on-dimentionsreireulation onset velocity fa meridional velocity D, eye diameter Dy hubdameter F, radial force § acceleration due to gravity H pump head ks. coefficients for ealeulating minimum flow K radial force coefficient N pm Q flow R_NPSHavailable/NPSH required for 9% head drop S, suction specife speed U___petipteral velocity at impeller eye W_ velocity relative to impeller 8 vane angle ratio Dy/D, 9 density angular velocity subscripts 1 impeller inlet 2 impeller exit BEP best efficiency point DR discharge recirculation SE shockless entry SR suction recirculation Recirculation (either suction or discharge whichever is larger)ANEW METHOD FOR COMPUTING MINIMUM FLOW a REFERENCES, L Agostinelli,A., Nobles, D., and Mockridge, C.R., “An Bx perimental Investigation of Radial Thrust in Centeifugal Pumps,” Transaction of ASME Journal of Engineering of Power (Apr! 1960). >. Kanki, H., Kawata, ., and Kavsatani, 7, “Experimental Re- search on the Hydraulic Excitation Force on the Pump Shalt,” ASME Paper 81-DET-71 (1981), Fraser, WH., “Recirculation in Centsfugal Pomps,” pre- sented at the ASME Winter Annual Meeting, Washington (0380). ‘Spring, H.. “The Impeller Inlet Diameter Methodsto Justify Froper Selection fr Pure,” ASME Fuper No. 64WAIFAL 09s 5. Heald, C.C., and Palgrave, R., "Backflow Conteol Improves Pump’ Performance,” Otl and Gas Journal (February 25, 1985). 8, Gopalakrishnan, S., "Modern Cavitation Criteria for Cen- trifugal Pumps.” Proceedings of the Second International Pump Symposium, Turbossachinery Laboratories, Depart- ment of Mechanical Engineering, Texas A&M University, College Station Texas (1985) ACKNOWLEDGEMENT ‘The author is indebted to the management of BWA Interna tional, Ine, fr permission to present this paper.