e Fiow RECIRCULATION IN CENTRIFUGAL PUMPS. by Warren H. Fraser Recirculation, a potentially damaging flow reversal at the inlet or discharge tips of the impeller vanes of a centrifugal pump, has tong been a source of consternation 10 centrifugal pump designers and users. Until recently, the pump in- dustry had little recourse other than to cope with the effects of recirculation after they had already begun wreaking havoc on a pump- aystem, Now, after years of analysis and testing, Worthington has presented its findings on the ‘mathematical relationship which makes it possible to determine the flow patterns which exist when recircula- tion occurs. This research makes it possible to design a pump for specified flow ranges to avoid or minimize the impact of recirculation. Analysis and testing has also led to the determination that the symptoms associated with recirculation are very spectfic. This article, ex- tracted from a paper presented at a 1981 ASME ‘meeting, and another pre- sented at Texas A&M in 1981, discusses some of these symptoms, diagnoses their causes, and, based on Worthington test and field experience, explains how ‘many of these problems can be avoided. Warren H. Fraser is Worth- Ington’s chief hydraulic: engineer in Harrison, New Jersey. The pressure field produced in a centrifugal pump impeller at a flow cortesponsling to peak effcieney is more uniform and mote symmetrical than at any other flow At flows less than that at peak efficiency, the pressure field becomes in creasingly distorted until at some point the pressure ara dient reverses and a localized reversal of the flow takes place. ‘This is the point of recirculation, whieh ean oceur at the dis charge or the suction of the impeller, or at both the suction and the discharge. Recirculation characteristics ‘aie dependent on the design of the impeller. It is inherent in the dynamics of the pressure field that every impeller design must reeirculate at some point-it can: not be avoided. It is important for both the designer and the operator of centrifugal pumps to realize that the capacity at which discharge recitculation ‘eccurs ean be reduced through design procedures, but only at a reduction in the rated elfciency of the purnp. Similaily, the capacity at which suction recirculation oe- ‘curs can be reduced, but only with an accompanying increase in the rated net positive suction head (npsh) required, Optimiza tion of efficieney requires a reduction in the margin of safety between the rated capac- ity and the discharge recircula- tion capacity: Also, the higher the dlesign suction specific speed, defined as: = fhm / apa npshiity the narower is the margin of safety between the rated capac: ity and the suction recirculation capacity,Causes of recirculation. Why does a reversal of flow ‘occur at reduced flows? The answer to this question seems to be elated to the fact that the pressure field not only increases from suction to discharge, but also that the total head pro- duced is the sum of the centrifugal head and the dy. namic head, The centrifugal head for any given impeller diameter and speed is independent of the rate of flow. The dynamic head, however, is a function of the absolute’ velocity that is related to the rate of flow. At some point on the head capac curve, the dynamic head will exceed the centrifugal head. At this point, the pressure gradient reverses, and the flow is from the discharge to the suetion of the impeller. Because the pres- sure field is now not sym rmeiical, and because the vanes themselves distort the pressure field, the reverse ot back flow takes place in the vicinity of the vane itself The condition news exists where a small portion of the total flow has reversed its ditection, and the shear face between the two flows pro duces vortices that are “locked” info the vane system and rotate with it ‘This is the point of recrcula tion, and this isthe eapacity at which noise, vibration, and eav- itation damage to the pressure surface of the vanes are most likely to occur, The severest damage at this capacity m often occurs locally Effects of recirculation, is now well established from field observation and from tests in the laboratory that surging and cavitation at the inlet vanes of the impeller can be caused by suction recitcula- n the impeller eye tion, despite the fact that a Figure 2 shows a more detailed wide margin exists between the view of the pump with strearn nnpsh required and the npsh ers atfached to the inside of the available, transparent pipe to show the flow patterns of the vortex peep Fon. As the suction reereuetion eee Figh velety aus of fu protuced ae wal whe a the same time, fluid is ap- proaching and entering the eye Bie impeller though tne exe pee then Cetuwen the flow thgh the core and the ning a ne eee ie Som chp cracking ose “The aueton vortex the Se ee eee tharincceuringat theinletel @Q)) Loe eg erie Similarly, extensive damage to the pressure side of the impeller vane at the discharge hhas been observed in many’ ‘pumps operating at reduced low tates. These effects are the mote obvious resuits of reci- culation, In addition, there are less obvious and more subtle symptoms and operational dif celles that ate associated with ‘operation of the pump in rec culation zones, The most ‘common symptoms and a diag: nosis of the causes of each are shown in Table Ion page 10, Suction recirculation. The reversal of flow in the themselves, Between the shear eye of the impeller at the point face of the flow entering the cf suction recirculation has been impeller vanes near the hub, observed in laboratory tests, and that ejected at the impeller Figure 1 shows a Worthington eye diameter, a fixed vortex is laboratory test arrangement of produced that travels around. 2 eich cndesuction pump ins with the seston of the vane stalled in a test loop with a system. This vortex will cvitate transparent suction pipe that at its core and attack the metal permits visual observation of surface of the pressure side of Figure 1-Laboratory test pump.@ the vane in the area approx: Jmately midway between the hub and the shroud. Figure 3 shows schematically the fow of the impeller eve ding recirculation, Discharge recirculation. The reversal of flow at the discharge of the impeller is more difficult to examine directly than is the suction recirculation, One technique is to reeord a trace of the pressure pulsafions in the discharge cas: ing as the out flow of the pump is reduced. At some point, that magnitude of the peak-to-peak pressure pulsa- tions will increase at a very steop rate. This is the point of discharge recirculation, ‘The mechanics of the attack from discharge recirculation are very similar to that in the suction, At some point on the head capacity curve, the flow reverses on the pressure side of the vane and produces a vortex that rolates with the vane system. Figure 4 shows schematically the flow at the impeller ls charge during recirculation, If igure 3-Suction recirculation at the impeller eye. the velocities of the reverse flow are of sulficient magnitude, the vortex will cavitate and attack the metal surface of the vane. Figure 5 shows the damage to the pressure side of the dis. charge vane from operation in the discharge recirculation zone. Predicting recirculation, Figures 6 and 7 show the approximate values of suction recirculation as a percentage of Figure 4-Discharge recirculation at the impeller discharge. the best efficiency flow for various suction specific speed designs, The suction recitcula tion values also inewwase as the specific speed (N,} increases. N= Pm epm (ibe p* This effect is shown in the form of two prediction chats, Figure 6 should be used for specitic speeds in the 500 to 2,500N, tange, and Figure 7 for Specific speeds from 2,500 to 10,000N,. While these predic- tions are based on the analysis ‘of hundreds of pumps in actual ‘operation, as well as pumps tested in the laboratory, it must be recognized that these are guidelines for a wide range of pumps and services, While some designs may be better or worse than the chats indicate, the predictions do represent the average or the norm for good commercial designs A more precise prediction of recirculation values can be determined if the design parameters of the impeller are available. The appendix provided on pages 15-17| | presenis a more detailed analy- 55 of this procedne While Figures 6 and 7 can be used to predict the suction recirculation values, the recom: mended minimum flow in actual operation will depend upon the size of the pump, as well as the fluid pumped. For ‘example, in pumps of 2,500 ‘gpm oF less and heads per slage of up to 150 feet lapprox- imately 100 bp), the energy levels may not be sufficient to cause damage or operational problems even though the pumps are operated in the re- circulation zone, Asa general rule for pumps of 2.500 gpm oF less and heads per stage up to 150 feet, the minimum flaw values can be set at 50 percent of the recir- culation values for continuous ‘operation and 25 percent for Intermittent operation. For higher flows and heads, the minimum flaw should cone. spond to the full suction recirculation flow On the oiher hand, for pumps in that larger size range but handling hyxtrocarbons, the minimum flow value can be set at 60 percent of the reciecula- tion flow for continuous operation and 25 percent for intermittent operation. Problems result from misuse of guidelines. There are a number of cases where appreciable deviation from the guidelines has resulted in very serious operational problems. For example, a double-sue- tion pump on cooting tower sevice was selected with a spe- cic speed of 2.750N, and a suction specie speed of 10,0005. Most of the operation Figure 5-Damage tothe pres- sure side ofthe discharge vane from recirculation. ‘of the pump occurred at 75 percent of the best efficiency flow, and after only a few ‘weeks of operation in suction recirculation the inlet vanes of the cast iton impeller were cavi fated to the point where they broke away in pieces. A steel impeller was substituted for the ‘east iron, anid the life was ex- tended to approximately six months before repair welding of the impeller was required. In another case, a high-pres sure boiler feed pump of 1,500N, specific speed was op- erated lor extended periods at Jow loads during the start-up of a plant. During this period of operation in the discharge rei culation zone of the pump, the fuser vanes were subjected to severe eyetic loading, and sev eral of the vanes failed in fatigue and broke away. Once the plant was operating at or near full load, however, the pump operated above the dis charge reciculation capacity, and no further damage was experienced, Mechanical damage. Experience has shown that many centrifugal pumps oper- ale either continuously or intermittently in suction or dis charge recirculation, o both, Under these conditions, me- chanical damage may or may not develop, depending upon the size, horsepower rating, the head developed, the charac- Figure 6-Suction recirculation-500 to 2,500N .teristics of the fluid pumped, ‘and the materials of construction, For example, a 2,500-gpm pump operating at 100 feet of ‘otal head may not exhibit any damage, but a 25,000-gpm, pump of the same design and specific speed will produce high noise and vibration levels in the recirculation zone with pro- agressive cavitation damage to the impelier and to the casing Similarly, while the 25,000- ‘gpm purp handling water will have the symptoms associated with recirculation, the same ‘equipment pumping a hydro- carbon wall exhibit lower levels of noise, vibration, and cavita- tion damage. Evaluate specific speeds; know your flow rates. The potential effect of suction specific speed on the point of suction recirculation, and the cifect of discharge recirculation on efficiency, should be evalu- ated for pumping installations of 2.500 gpm or laiger. As a gen eral rule, suction specific speeds in excess of 9,0008, for pumps 9 Inspection of pressure side of inlet vane for recirculation damage. Inset: Damage to the pressure side of the inlet vane from suction recirculation. designed for specific speeds of 2,5B0N, or higher should be evaluated very carefully to avoid recirculation within the ‘operating range of the systern, ‘The analysis of suction recir- culation has revealed that the higher the design suction spe- cic speed, the closer will be the point of suction recirculation to the rated capacity. Similarly, Figure 7-Suction recirculation-2,500 to 10,000N,, the closer the discharge recir- culation capacity is to the rated ‘capacity, the higher will be the efficiency, ‘There isa great temptation to design for the highest possible efficiencies and suction specific speeds, but this may result in designs that are very limited in their range of operation, Both the designer and user of cen: trifugal pumps must know the flow rates at which suction and discharge recirculation occur ‘These values are as rich a part of the specitied perform: ance of any given design as are the head, capacity efficiency, or sh With complete performance characteristics in hand, a more realistic evaluation can be made as to the risks associated with operation at oF near recircula- tion as against the anticipated savings in power costs and lower npsh requlrements.ec vege of the icy is cinected clowns a veuriace ofthe aluouds al the impetler : oc | 1 | | 1 | | | |Frow {RECIRCULATION IN | CENTRIFUGAL PUMPS: FROM THEORY TO PRACTICE. by Igor J, Karassile The lack of @ mathematical ‘model has for years left cen- {rifugal pump users ond designers to thelr own devices twhen attempting fo specify acceptable values of mink ‘mum flows for various “S." or suction specific speed, vale ues. This, frtunatel, is no longer the situation. There dloes now exist @ mathem cal made! which till permit users and designers to estab- lish sensible limits for centrifugal pump minimum operating flows, as already discussed in tis fssue of Power & Fluids, in “Flow re- circulation in centrifugal pumps.” But before this theory can be put into practice effecx tively, it is essential that users and designers frst gain a thorough understanding of the critical relationship be- ftaeen minimum flows and recireulation flows. This arti- le discusses the nature of this relationship, ts effect on pump operation, and practi cal approach to applying the 1 mathematical method of pre- slicting recirculation es Igor J. Karassik is chief consulting engineer for Worthington, I should be understood that when utilizing the data pre sented in “Flexw recirculation in centrifugal pumps" (page 5), S must always be calculated for the conditions corresponding to the capacity at best efficiency Values calculated for any other flow do not have any relevancy and do not deseribe the charac- teristics of that pump. The guaranteed conditions of service may or may not correspond te this best efficiency low; in fact, they seldom do, What may be less obvious is that the $ value for any pump ‘must be calculated on the bas ‘of pump performance with the maximum impeller diameter for which it was designed, This stricture becomes apparent when we consider that the internal recirculation at the sue tion occurs because of certain conditions that arise in andl around the inlet of the impeller, conditions that are not neces sarily affected by decreasing the impeller diameter, Decreasing impeller diameter moves the best efficiency point to a lower flow value, but does not reduce the capacity at which suetion recirculation wall occur, As a matter of fact, in certain cases where the onset of suction recircllation may be lriggered by the discharge side rweairculation, decreasing the im. peller diameter may actually increase the actual flow at which suetion recirculation takes place. Pump and pumping system distress. Operation of a pump at flows below the recirculation flow cre ates an anay of events which leads to unfavorable effects, or distress, on the performance and integrity of the pump, and sometimes on the system in 4 Which itis operating. The de- gree of distress depends on a vatiety of factors such as: size of the pump (capacity, total head, and horsepower); liquid characteristics; materials of con- struction; length of time the pump operates below certain critical flows: and the degree of tolerance exhibited by the pump user to the signs of dlisiress exhibited by his ‘equipment Size of the pump. It was stated in “Flow rect culation in centrifugal puraps" that, “For pumps of 2,500 gpm 36 and heads per stage up feet, the minirourn flow values can be set at 50 percent of the recirculation flows for continuous operation, and 25, percent for intermittent opera- tion.” This indicates that pump size and energy level directly alfect the degree of damage that can be experienced when a pump is operated within the zone of internal recirculation, A 25., 50-, or 100-hp pump will not Severely damage itself when operating in full recireula- tion, A 500-hp pump may suffer considerable damage to its impeller, and a 5,000- or 10,000-hp pump has enovigh ‘energy under certain conclifons. to shake its piping loose. A sense of proportion is important when sizing a pump. Lig Tes clear that liquid charac teristics do not affect the flow at which recirculation takes place, It's only the symptoms and the accompanying damage which ate affected by these charac tetistics, Inthe case of cavitation caused by recirculation, liquid properties which mitigate eavita tion damage in “classical” cavitation are equally effective2 jin the case of recirculation caitation. Inthe case of flow force effets, a lower fluid den sity may reduce the force level of pressure pulsations. Con- Sequentiy minimum flows for pumps handling hydrocarbons: need not be selected as con seratvely as for water pumps, Materials of construction. Materials of construction are important factors to consider ‘when establishing minimum flow limitations. The erosion effects of intemal recirculation cavitation on pump parts are Increased or decreased depend- ng upon the relative degree of resistance to this kind of damage of the materials of construction. Type 316 stainless Is a superior material which is better than bronze, which in fum is better than cast iron, However, the use of better ‘materials will not help very much if the effects of internal recir- culation manifest themselves not only in material degrada: tion, but also in the form of pulsafions and vibrations, ‘The time element. ‘There is a certain degiee of ambiguity when discussing the time element contribution to pump distress, We dllferentiate between “continuous” and “in- termittent” operation, but these two terms have never really ‘been defined. It would be im- practical to specify exactly the number of houts that demar cate the boundaries between these fwo conditions, but itis possible to be more specifi “There are two separate peri ds in the operating life of a pump that should be reviewed in the application of any cen- tuifugal pump. The fst is the period where the system is in full operation, and the range of ‘capacities and heads encoun- tered can be predicted with reasonable accuracy. We can define this continuous operation as that range of capacities ‘which will be encountered at least 75 percent ol the time. Intermittent operation during this period would cover the Femaining range of operating conditions 25 percent or less of the time if they differ substantially “The second period refers to the star-up conelitions of the suster, It's dificult to predict what the duration of pump operation at any capacity may be inthis period. Unfortunately, this segment of pump hfe Is seldom given any consideration in the design stage of the pumping system, while the pump may operate over very ‘wide suings in capacities for extended periods of time, The results that frequenily there is significant damage to the pamp caused by reciculation, even before the system goes info full operation Tolerance of user. Another important factor to be considered is the degree of pump user tolerance to the signs of distess exhibited by his equipment. This is what makes the choice of minimum flow guidelines such a subjective ‘one, For instance, while some users will aceept the fact that an impeller may have to be re placed every year, others wil be dlsatisfied if an impeller on exactly the same service lasts only thiee or four years. Noise and pulsation levels which ate perfectly acceptable to one user are totaly unacceptable to another What confuses the issue even further is that recommendations (of minimum flow to the user by pump manufacturers frequently seem to bear no relation to the suction specific speed of a particular pump offering, In- dicative of this is the manu. facturer with a pump offering with a higher $ value who quotes a lower minimum flow than another manudacturer with a more conservative design; hhence, a lower recirculation flow: These differences merely reflect more or less conservative ‘estimates on the pat of a designer as to what constitutes tolerable distress. We can expand on this. Starting with the proposition that the requited npsh, and, therefore, the suction specific speed of a pump, is determined bby the conditions of service and bby the geometric configuration of the impeller, it follaws that ‘two pumps with the some S value will have the same suc tion Tecirculation flows Tt also follows that if an equal amount of distress is to be considered acceptable, the minimum flow to be recommended should be the same for equal values of S ‘Therefore, the higher the $ value, the higher should be the value of acceptable minimum flew. Probability of low-flow operation, ‘There are applications where a pump will never be expected to operate at flows below its dlesign flow. This is the case, for instance, with constant-speed condenser circulating pumps operating in parallel, since itis not normal practice to throttle their discharge. ‘The situation can best be understood by reference fo Figure 1, which describes the ‘operation of thiee cooling tower ‘pumps operating in pall, 9Figure 1-Operation of three cooling tower pumps in parallel. superimposed on the system- hwad curve. Note that where only one or to pumps are operating, the head-capacity curves intersect the sustem- hhead curve at flows in excess of the design capacity. As a matter of fact, this situation is even further accentuated where the system head curve has been drawn up pessimistically high. In a case such as this, there ‘would be few objections to using S values considerably higher than in a case where a pump may be called upon to ‘operate over a wide range of ‘capacities to the left ofits ‘design coneition, However, this approach shout not be applied if the pumps discharge through sev- ‘eral heat exchangers in parallel, some of which may, on veca- sion, be taken out of semvice. ‘The system-head curve under these circumstances would steepen, and the operating ¢a- pacity reduced significantiy below the design conditions. A typical case of a pump in such a cooling system is ils ‘rated in Figure 2. Here, the pump may discharge through cither 10 or 5 heat exchangers, In the latter case, the intersec- tion of the system-head curve with the head-capacity curve occurs at a flow sufficiently reduced to cause the pump to operate in the recirculation zone. To avoid this, a bypass must be provided, increasing the capacity just beyond the recitculaton flow whenever only 5 heat exchangers are used, Application guidelines. Based on the previous dis- cussion, the following cen- trfugal purnp application guidelines are suggested as ‘ways fo help avoid problems resuling from internal recirulation © Unless there is a compeling reason to do so, do not specify npsh values which result in S values much above 9,000. * When dealing with relatively small pumps, the effect of recirculation is not apt to be as significant as for larger ‘pumps; hence, you can aiford to be less conservative. ‘* Pumps handling hydrocarbons ‘can be operated al lower flows than equivalent pumps handling cold water. * The risks of operating at flows much below the recivculation flow, and the user's teaction to4 S the resultant problems, can be better determined after the ‘pump is in operation. To avoid expensive “alter the- fact” changes, provide for the possibilty of increasing the ‘minimum flow bypass if there is a suspicion that too optimis- tic a decision has been made | at the time the pump was | selected, ‘#1 pump is never expected {0 operate at flows below its design condition, higher S values can be used without concer for unfavorable effects from intemal recirculation. With the publication of “Flow recirculation in centrifugal pumps.” another brick has been added to the edifice of knowledge on centrifugal pumps. By no means, however, does it mean that our knowl- ‘exige is complete, With continuing research efforts, we ‘an hope that we may manage a better coexistence with inter- nal recirculation, even to the point of regulating its occur- rence to lower flows, or mitigating its effects after onset ‘without sacrificing the advan- tages of higher suction specie speed values, ie. lower re quired npsh, Figure 2-Pump installation discharging through ei or 5 heat exchangers in parallel.Discharge Recirculation: 1. Caleulate discharge vector diagram angle Bi Sin py ADR (ler to Figues A & B) 2. Determine = from Figure € {page 16) U, 43, The flow in gpin at discharge recirculation is equal to: DB. x wpm. C, me, Gu a Suction Recirculation: Cave where: Ds _ 9.5 D, 1. Calculate inlet weetor diagram angle 6 L278, Sin py = 2273E) (efor to Figues A & Bi Br = ee (ler to Figy ] 2, Determine Y* from Figure D (page 16) u, 3. The flow in gpm at suction recirculation is equal to: (DE = hi v. DWDM a ase 2 where Pt = 0.5 Case . Whenever D/D, is equal to or greater than (0.5, the suction recitculation is either equation Tor 2, whichever is qeater, Nomenclature B, _-wieth of the impeller waterway atthe discharge diameter, Cog) ~meridional velocity atthe Impeller discharge D, eye diameter ofthe Impeller Dy ~tischarge diameter of the impeller F, -a1ea between the vanes al the impeller let normal to the average meridional velocity F, area betuwen the vanes atthe imple discharge normal io the average meridional velocity Z ~shaft diameter through the Impeller eye peripheral velocity of the impeller eve peripheral velocity of the Impeller discharge diameter axial fluid velocity in the impeller eye ~dlscharge vector dagram angle-degvees (see Figure B) inlet vector diag angle-cegrees {sce Figure B} net postive suction head “Anpenslcisexractd from “Recreation in Centifugal Pung.” peserted as paper atthe 1981 onal ASME most6 nan - - 12h = iu IH | ! 4 Suetion Ds at U; Ws Discharge Vector Diagram . Inlet Vector Diagram O Sin fi W, he = ra Y) ol 04 Hj ¥ o20 : "a8 i 7 ® 0 15 20 28 30 Discharge Diagram Angle @, Degrees Inlet Diagram Angle Degreesa Examples 1. Double suetion pump where: Pump rating at peak efficiency is: D.= 155i hy 3500 gpm B, =175 in. 22004 thd F, =30in? 1750 rpm D, =7.185 in, 17 fe mpah Discharge recircul snp, - = ____- asso aney FADE, “aes Se «9084 tom Fague¢ sw = EB. 25 ‘Suction recirculation: = 0.44 (260°) 217.185° 28757 TANS (TABS 28759 51750 9 172 = 1060 gem x 2 = 2000 gpm 2. Single suction pump where: Pump rating at peak elfictency is: D, = 11875 in. hy = 0 1750. apn B, =1875 i, F, =17 in? 100 A tha F =19in? 1750 rpm q Dos D=65in p08? 17 ft npsh Discharge recireulatio E, 9 ee Soh = Spee, ~ WeLaeeTe 7 0271579 Ga - 652fom Pau € DixB,x rpm, Cy _ 1.875" 1875 x 1750 4.469 — 1990 gpm Bex om Gos a x 0.062 = 1220 gp aw 27x17 A277 — ste (305 85 (6.520) x 1750 Se x08. 125 apm. 2) In this case, B exo 0.3: therefore, the fat suction recreuston i the gator of equations fan 2, oF 1220 gpm