PRACTICAL ASPECTS OF CENTRIFUGAL COMPRESSOR SURGE AND SURGE CONTROL Meherwoan P. Boyce i President of Boyee Engincering International, Inc., 8 major Houston-based consulting and en gineering firm. His past experience in- corporates many academic and profes slonal positons, including Director of the Gas Turbine Laboratorizs end Pro- _Jossor of Mechanical Engineering at Tex- ) ee AWM Uniecraty, whore he organized {and wos chairman of the Turbomachine ery Symposium. Dr. Boyce has authored more than fifty significant papers and technical reports on ‘fiusd mechanics and tsrbomachinery, and he ix a member of ‘SAE, ASME, ond several other professional and honorary societies, He és also a registered professional engineer in the State of Texas Dr. Boyce received B.S. and M.S. degrees in Mechanioat Engineering from the Soutk Dakota, School of Mines and Technology and the State Unicorsity of New York, respactive- Wy. aud his PhD. in Mechanical Engineering from the Unioor- sity of Oklahoma. William KR. Bohanan is Chief Me- chanical Engineer end Technology Coor. ‘inator for the Son Prancico Division of Bechtel Pol. Ino. He heads nip of engineering specialist em Fee eee ee ecaluation, application engineering, testing, and commissioning of mechan cal equipment associated with projects Nanded by this dixon. He also prides the Division's tnoaleement tn techntcal societies and committees. Much of his thirty your” experience has been involved in compressor and turbine opplicaions His precious assignments elude ceven years ttth a eampressor Inanufecturer and foe yenrs with peirolewn refining. com: panies My. Bohannen holds u BS degree in Mechanicel Engineer- tng from South Dakota Schoot of Mines ond Technology. He is 2 registered profesional enginger in the State of Calfornia, anda member of ASME and the Vibration Tnattute, and he prticiates in the mechanical equipment activites of APL ut Royoe N. Brown is a Consutting Engie neor with Dow Chemical US.4., Engi- neering and Construction Sercices, Houston, Texas. His responsibilities in clude specications, consulting and fil asristance in the area of large rotating equipment for Dow world-wide. Tor gether with his responsibilities as Con- sulting Engineer, he heads the Rotating Equiyment and Critical Equipment Ine strumentation Group. He holds a BS. degree in Mechanical Engicering from the University of Texar and a M.S. degree in Mechanical Engincering from the Univer- sity of Wisconsin, He isu fellow member of ISA; a member of ASME, SME, and API; and an associate momber of SAE. He is Active in the Mechanical Equipment subcommittee of the CRE, where he is the chairman of the APL 617 Task Force. Mr. Brown is a registered professional enginser in Tenws, Lowisi= na, Wiecontin, and California Joka R. Gaston és @ Sentor Controle Design Engineer for Drasore Industries, Ine., Clark Diciion, Olean, NY. He Joined Dresser Clark in 1956 and has held various positions in Field Sorvice. Since 1967, he hs been responsible for cepplication and design engincering of Concvut ysterss for tarbe (aenrtfaget / | Slices ele Sented papers on compressor control at warious ISA and other conferences. He received the ChemicalPetrolewn Dioison of 18's Dick Pond. ward in recognition of the best technical paper at the annual Symposium held at San Frenolsco tn 2974. Mr. Caston i member of the National Management Assocation, Senior Member ofthe Instrument Society of America and a registered preston engineer onirel systems) in the State of Cyrus Meher-Homyji is Engineering Gonsultant Manager with Boyer Eng neering International, thore he ie Solved with carious consulting work im the turbomachinery and engineering management areas His counsalting activities include de- sign, exbration and performance analy- 1s, ecamomae analysis and rlibulty em- gineering. in the past, he was a decelop- tment engineer with Boyce Engineering ‘end was involved tn the design and development of prototype ‘externally fred staan igected gas turbine for the US. Depart Imo of Bneray. His areas of tntercet are rotor dynamics, turhomachinery prognosis and diagnosis, eginecring manage> ‘ment, and reliability and maintainabil I. Meer Hol has BS. deren Mechel ng neering and an ME. degree in Industrial Engineering from ‘Tesas AGM University, and a M.B.A. degree from the Uitver sity of Houstonus PROCEEDINGS OF THE TWELFTH TORBOMACHINERY SYMPOSIUM Robert B. Meier josned Cooper Ener ay Services Company tn 1068 nd has hrad 1 long assectation with oenérifugat compressor design and development. ame) Presently, he is Manager of Design and Developmant jor Keating Product. Ts ability to reed and speak German and Branch is especially valuable in pre- semiations and specifications review. He hols a degree fram the Swiss Redsral Institute of Technology. Norbert B. Pobans graduated from UCLA with a B.S. degree in General Engineering in January, 2958. fa June, ' 198, he received a MBA, Management, from Golden Gete University im Son Francisco. During the past twenty-four tears, he has boon clesely associated with the use of computer simulation to evaluate engineering designs in mony ‘tonal fils. During the pest eight years, he has been in chorge of « simula tion and advanced control specialty group for Bechtel Petro eum, Inc. Recently, his responaibilties hove expanded to include management of a dedicated simulation facility to per- form suudics and to analyze distributed control systems. Frior 10 his years with Bachiel, he was employed by Blectronic Associates, Tne. for fourteen years. His responsibilities in cluded dynamic énalyss, echoicel support for marketing, and management of computer simslation facies, Sire ropend wat regearered projeasonat cugiueer (comet systems) in the Stato of California. He has Senior Member ster fn the Instrument Society of America and the Society for Computer Simulation SCS). Presently, hei Senior Vice Presi. dent of SCS. He has published numerous articles inthe field of simallatin, ABSTRACT ‘This paper addresses the area of compressor stability, songe and strge control and relates to the practisl aspects involved, An euplasis placed on the physical understanding cof surge phenomena and on the prectial limitations of surge ‘antral systems, Topics discussed are physical interpretation of Instability, causative factors, types of stall, machine and process design factars, surge effets and characteristics, contol fystem types and practical application aspects. Sone ease studies also are presented, The discussion pisorily relates to centrifugal oampressors, but several aspects pertain co aval fos comprestors as well, Tke papr Is spl into three sections: Seaton A const of an introduction to surge ad a discussion of eentriiugal compressor design and prowess ctor that affect foperting sabilty. Section B discusses the various types of control schemes ind surge protection devies. Several cmui- ples and common pitfalls are addressed. Section C addresses the important design tool, surge system simolation. Several refereuees are provided to enable the reader to pursue this ‘opie in greater deta A. INTRODUCTION TO CENTRIFUGAL COMPRESSOR SURGE. Centrifugal compressor systems at numerous installations have suffered scrious mechanical damage to compressor inter- nals or to other parts of the piping svstoms as a result of ‘operation in the surge conti. In other cases, compressors, usally smaller machines, have operated for long perols with intermittent or even eontinuans light siege without eechant calharm, although with significant inpsisment of aerodynamic peiformance. These conditions generally sre the product of fone oF more ofthe following deficiencies: 4, Poor matching of the compressor to the systems re- ‘quirements, 2. Inappropriate compressor design. 3. Inadequate anti-surge control system, 4. Unfivorable arrangement of piping and process com- ponents of the system. (This olten eum magnify surge) Because of the shape of thelr performance eurves, the application of centrifugal compressors frequently is considey- ably ore complex than that of reciprocating machines. A centrifugal compressor driven by an eleetrie motor offers an imposing challenge to the coctrols engineer when this con- stant-speed machine must accozamodate a wide range of vati- ables in its operating conditions. Centrifugal compressor appli- ‘ations continu to Became more complicated with inereases in the numbers of stages per casing, casings ia tandem with a single driver, and sidestream nozzies: hfeher pressures and speeds; and zsore operating conditions for a given machine involving wider ranges of ows, molecular weights, and pres sures. This ever-increasing complexity requires better under~ standing of the eauses of surge and Ite detrimental effects in ‘order that adequate control systems may be applied. It also ‘exeates tho need for surge eontrol system simulation studies, ‘which must be conducted at the design phase, This tople i addressed in Section C. ‘A twoieal pesforsance mao foc a centeeal comoressor is shown In Figure | The operating vine is the surge line ‘modified by a safety margin fo ensure troublesrce operation. Note that the otal presie ratio changes with dow, speed. molecular weight, suction pressure and temperature, Also, rote that operating at higher efficiency implies operation closer to surge. It should be noted here that total pressure increases ‘our only In the impeller. To make the curve gencr, the concept of acrodvnamic speeds and corrected mass flow rates has heen wved, ‘The surge line slope an multistage campressors can range fom a simple, single-parabolical relationship toa complex curve containing several break-points or even “notches.” The complesity ofthe surge line shape depends on whether or not the flow listing stage changes with operating speed from one compression stage to another. in particular, very closely ‘matched stage combinations frequendly exhibit complex surge lines. Ia the ease of compressors with variable inlet guide vanes, the surge line tons to bend more at higher flows than \with units which are speed controlled. Usually, surme 19 linked with exzcenive vibration and oa avdible sounds yet, there have been cases in which surge problems which sere not audible caused failures. Usually, operation in surge and. often, near surge is accompanied by Err ditions, inlusing general ao panting nie eve increases, axial shaft position changes, discharge temperature ‘excursions, compressor differential pressure fluctuations, and Iateral vibration amplitude inereases. Frequently, with high pressure compressors, operation in the incipient surge range is accompanied by the emergence of a low frequency, asynehtc- nous vibration signal which ean reach predominant am- plirudes, as well as excitation of vasious harmonics of blade passing frequencies. Besides the well-known effects of extend: fed operation in surge (Uhrust and fournal bearing failure, impeller rub), impeller hub and/or shroud failures resuliag,TUTORIAL SESSION ON PRACTICAL APPROACH TO SURGE AND SURGE CONTROL SYSTEMS Me corrected How na, (5) ALE Figure 1. Compressor Performance Map. from severe stinuation atone of thelr natural response fre Guencies ostesionally ean be found Extensive investigations ave been conducted on surge, storing a early a 1946 by Bullock [1] nd his associa, ad inter by Emmons [2] Most of the work involved experiwents soe Cons ay ae gen Io" pone qumntave univerally of aerodynante nad Species of detent aifasers sod pele, a er. next ‘owledg of boundary lsyer behavior make the exact predic ton of low urbomachins atthe design sage fit Ie However, quite evident that the underying cause of surge fevody namie stall. The stall ay cousin eter the impel or the dfiser- Recent siguiiat work in this are is revowed fh Reference 9 Definition of Suree ‘The phenomenon of surge, at pertains to @ centefagal compressor and its connected system, Is an unstable cemdition resulting in flow reversals and pressure fluctuations in the System, This coudition gecurs when ther i sufieient acrody= namic instability within the compressor thatthe compressor is table to produce adequate pressure to deliver continuous flow to the dowrstream system. The system and compressor then inteiet, causing the surge conditions with large and Sametime vitlent ag nerillations fn fhe sytem. Surge, hen ia steal stom phetomeon a mot cone othe compressor ony ‘Surge Is the voslk of an excessive ineneate i the resis tance af the sysicm while the compressor is operating at a ceriain speed. The added resistance reduces the flow 40 an unstable level. Alternatively, ifthe resistance is unchanged, bat the speed is redisced appreciably, mont sytem wil singe “Thus, surge occurrence depends ou che type ofsystemn and the shape of the vesstance curve. ‘The aerodynamic Instability is brought about by flow reduction, which enses stalling of one or more adhe elements dda sage or stages ofthe compressor. The stalling can ocer at the inducer of the impeller, in the radial portion of the impeller, i the diffuser ore the volute. A stall in on ofthese elements may aot have sufficient effect to cause the stage to be lunstable. In Tact, several elements of 2 stage can stall withoot the entire stage stalling [3]. However, if the stalling is of sulficient strength, the stage sll become unstable, and this tan lead to surge of the compressor ‘The sta af an element of a compressor stage may be compared with that of an airfoil of an aigplane. The lif of an airfoil iy related to the velocity of the afr flow and the angle af attack (imeidence). Tf the angle becomes excessive, the lift collapses, and a stalled condition results, The stall occurs because the airstream separates from the surface ofthe aifil. Airfoil stall is described by Figure 2. Japikse [3] gives a detailed explanation of stalling of compressor elements and ‘covers dynamic instability of rotating stall, pee Tange of teens, Figure 2. Lift or. Angle af Incidence and Aérfoll Flow Visual ‘ation as a Increases A typical centrifugal multi-stage compressor performance imap used in the provess industry is shown in Figure 3. The family of curves depicts the performance st varlous speeds, where N represents rpm, The ordinate may be polytropic heed "HI," pressure vatio, discharge pressure, oF sometimes differ etal pressure, The abscissa, usually ealled “Q” or "Q,” is almost always shown 35 aetual inlet volume per unit of time, such as aefin or lefm, where “a” is for “actual,” or "for “les.” Tes important to understand that the inlet flow volume or capacity is based on a gas with e particular molecular woight,150 PROCEEDINGS OF THE TWELFTH TURBOMAGHINERY SYMPOSIUM 1 vee wo se 5083 2g 8 \ i \ aot Bee gi yes ge Tose wae’ ge ee SE 2» CONSTANT Ba FH Ee 4 Go" fo” to” Who uo PERCENT INLET FLOW, Q, (ACEM) Figuee 3. Typleal Centrifugal Compeessur Pecformance Map. spceitic heat ratio, and compressibility factor ata pressure and temperature corresponding to the gas condition in the suction line to the compressor. Hany of these parameters is changed, the perfareance map is no longer exacly valid, Ifthe devia Wand are zeatively small, the ap still may be used with a fir degree of accuracy by making small adjustments, It is often impractical to make a separate set of curves for every small vapation in these erates Sige the ingele recognizes ‘nly actual volume, weight flow such as pours per minute, oF standard vale ssh ses ca oly be ese nthe sea ‘when there are absolutely no sigaifcant variations in gas Conditions at suetion “The line on the let represents the surge limit, or “pump- Inglimit,” as iis often called. Operation tothe Tet ofthis Ine unstable, resulting im unsatisfactory performance, and ts fea harmful mechanically. Notice that the surge flow in- creases as the speed increases. The surge line in many cases hhas am approximately constant valve of QIN. fe many multe lage units, and especialy’ in units which are controlled by variable inlet guide vanes, the surge line curves bead dramati cally a higher flows and speeds On, the other side of the map, the capacity lint or overload line is shown, Operation to the right of this line a more desirable location am the curve from the viewpoine of certain variable, hot this relocation will usually exeate one or more disadvantages from saother viewpoint, For instance, if the design point is located as shown, a move slightly to the left ‘would gain some effieienoy and head output and might vesult eliminating one impeller from a nmultstage application; howsv- cer, the cost could be an objectionable loss in the stable ‘operating range. A move to the right would produce opposite results, which could be desirable In some situations. Never- theless, the reduced capacity overload limit or proximity 6 stonewall could have significant disadvantages. To summarize, the combined losses force the actual cxrve to go from a ‘negative to 2 postive slope as flow decxeases; that, the actual ‘carve always has a "homp” in it, Changes in machine geometry. ‘can relocate the cvest of the hump, chenge the breadth of the Imamp, and alter the severity of the curve to some extent, However, the hump canaot be eliesinated. The surge point is Tacated at or very near the point uf zer0 slope on the compres sors chavactevistic curve, Friction lostes can be seduced somewhat by improving surface finishes and enlacging passages. Blado diffusion losses ray sometimes be mitigated by further streamlining of flow ypastages and assuring that there are no abrupt changes in the flow aroa. These techniques will improve eficioncy and tend to reduce the surge point; however, they also eost money, and theve isa point of diminishing marginal returns, THEORETICAL HEAD (BACKWARD LEAVING BLADES) cornu — EFFICIENCY Figure 7. Typinal Compressor Head Losses ‘The phenemonon of slip, which Is due to the butld-up of boundary layer on the vanes of the impeller, acears in all centrifugal machinery. Boundary layer coatrals have been used to reduce slip and increase the iow stability of the impeller, 25 seen in Figure 8, Design Factors Affecting Surge Sume elaboration on the princival machine desiga factors that affer stege will now be covered. A greater number of impellers in a ziven easing wil tend to reduce the stable ange ‘his effect 1s presented tm Figure ¥. Manuficturees have fond it physically and economically impractical to design each stage of a mltistage machine to be ‘exactly the optimom size. As the pressuve builds up cross the states, the volumetric flow is weduced. Ideally. then, each scebeding whoo! should be proportionately smaller with re- spect to flow passage area, For example, the with af che flow ext atthe tip ofthe impeller sbould actually got smaller aad smaller, Ie would be virtually impossible to achieve. perfect proportionality. This lake of optimum proportionality aggra- ‘ates the stability problem. Some larger machines are custom boil, and, therefore, better proportionality can be attained “Mechanical imitations, such as the axial length of the shaft ‘botween bearings and its relation to critical speed, may dictate the axial width of the impellers through the fob and the‘TUTORIAL SESSION ON PRACTICAL APPIOACH TO SURGE AND SURGE CONTROL SYSTEMS 50 . _ ie i wwe wseomm () ESe mm mio Figure 8. Laminar Plow Control ix Centrifugal Compressor % OF RATED INLET FLOW capacity uMIT 80 45 40 35 S & 30 a 12345 67 6 5 0 WUMBER OF IMPELLERS Crareo ~O srece ye srashuity= * YO naree (21 100%6 Spee conve) Figure 9. Effect of Number of fowpellers on Stability and Capacity Lint, impeller spacing. The shaft diameter may be constrained by torque andor eriical speed considerations, aad the inside diameter links of the impeller hub may be accordingly re- stricted. Practical eeoomics and stencardization oocasionally ‘will limit the nember of sypes and sizes of wheels in given Shachine: Por examen se-wheelrachine thee ight be only two different vbcel Gesigns. Therefore, whereas the fist impeller might be operating at optimum conditions, subae- quent impellers of the same type and size usually would be operating at less favorable flow conditions, owing to the volume reduction across the stagos, ‘There are a great many thor considerations which alfect the overall stability. The relationshio in Figure 9 is typical only, and the design can be changed to improve stability to some extent, However, more often than not, twill be at the sacrifice of seme other impor tant performance factor. The diminishing effect of the mumiber of wheels on the location of the eapacity limit i also shown in Figure 9. Just as the greater number of wheels in a casing, reduces the stability, so docs the number of sections of eam= pression, or the number of easings in series. This eff: is Iustrated by the sample ealeulation shown in Figure 10, Tecan bbe soon that individual percent stability of the HP casing is Dotter than that ofthe LF casing, However, as LP casing flow is reduced, the HP casing sarges before the LP easing because of increase in intercasing pressure, which decreases aol to that of the HP casing. Overall stability is less than the individual stability of ether easing, dew Besrces Cosy dew Presets Cay el BE Fee neon ee COND. arson Bonet nee 200 roe Fein = Poe gal 8 Foeo goon Boome 2008 sont 2 vasnceae bowie A" Br too ro Farrcteaie oP Sou S. oe Bo etnere td poe Gt tree me Aoatentie Ve aun Boo ee BS) = roses Lote Ceeause X Smanonpen mrt ctamet athe Lb Fine wise ence We re ane? Figure 10. Typieal Overall Stability vv. Irlioidhul Stability for Two Casings Operating i» Series ‘The information and curves presented thos far have beca, for the most part, based on impellers with backward leaning impeller blades. Figure LI (5] depicts cho effects of pellet blade angle on the stable range, and shows the variance in steepness ofthe slope ofthe head flows curve. The three curves sre based on the same speed, and shov actual head, The relationship af ideal or theoretical ead to inlet flow for difes~ ent blade augles would be represented by straight lines. or backward leaning blades, the slope of the line would be154 PROCEEDINGS OF THE TWELFTH TUREOMACHINERY SYMPOSIUM Apso 1°50" 5 * Ae" FORWARD LEANING RADIAL BACKWARD LEAHING IMPELLER BLADING parELEe eericreney : FORWARD, & [sins “a> 90° a va, RADIAL Ba «80° Zax {imum design entails considerably more than selecting primary flow elements, control valves, transmitters, and panel instro- ‘ments, and integrating them into a control scheme based on a wwall-defined standard for designing antisurge control systems. Unfortunately, there is no so-called “universal best mothed” of anti-surge eontrol. Before a control scheme is developed and ‘components are selected, the operating requirements of the compressor and process mnust be known and evaluated. There are many variations in compressor and process characteristics: therefore, 9 control scheme that is ideal for one application might be far from adequate for another. "8 Figuee 19, Process Operating Points, 695 (A) and 100% (D), Surge and Control Points B and C. Figure 20. Ges Flow Pathe When Process Flow Reguarement Is Lass thon the Compressor Surge Capacity. Methods of Capacity Control of Compressors ‘The two types of turbo compressors are centifogl and axial How. Methods of capacity contel vary, depending on the type af compressor and its driver. This discussion will be limited to the fllowing methods, whieh ure most commonly aed ‘Type ‘Type Method of Driver Gompressor Capacity Control Constant Speed Centeifugal Suction Throttle Variable Speed Centrifugal Vusiable Speed Constant Speed Axial Flow Variable Angle ‘Stator Vanes Variable Speed ‘Axial Flow Variable Speed ‘Typical performance curves and surge lines for the above are shown in Figure 21. Tt must be emphasized that the word “typical,” used to describe these corves, means just that. The actual surze line of many compressors is significantly different from the typica! surge line illustrated. For example, instead of the typical shape (Figure 216), the actual surge lines of many ‘centrifugal compressors are shaped more like the surge line of ‘an axial How machine (Figure Ble or 21d).188 PROGEEDINGS OF THE TWELFTH TUSBOMACHINERY SYMPOSIUM y ' 1 INLET PRESSURE | a hoes mow i 30051 Te 008 Mw. > a & oo we wo ho (eT VowUuME FLOW, % ‘CENTRIFUGAL COMPRESSOR FOAL CoMPaEs INLET CAPACITY, % wo 9 CENTRIFUGAL COMPRESSOR = - Waniaste S829 ’ 3 ep — — — 4 £0 : ’ I = ‘ be / i in | i, a oe \ | | s Ay vane roan | ® I a I Yow | . + e ee th ae Inlet capaci» water capaci, amine ANGLE STATOR VANES ON nauaieen Figure 21. Head. Capacity Characteristics of Compressors Constant Speed Centrifuen! Antsurge Control For this type machine, Figare 2 shows tht surge cours vvoren |_| covenesson at a constant falet volume flow rate, regardless ofthe inlet met * pressure Therefore, inlet volume flow rate would be an appeo- vine rate measurement for aatisurge contral. Instruments ean be pour Ine measurement oy aptsurge one Inorents nb a caoe b =¢ ere a) where Q = volume flow rate at inlet conditions, acim ihe) = head inches af water diferentil praduced by Bow “Mmsunat flemeat (FE) valve = fnlet pressure, absolute C= proportionality constant ‘Temperature compensation is uot used in the control ANTIURGE CONTROL SCHEME, MLET VOLUME FLOW schomo (Figure 22}; therefore, finlet temperature increases, Figure 22. Antisurge Control Scheme, Inlet Volume Flow.TUTORIAL SESSION ON PRACTICAL APPROACH TO SURGE AND SURGE CONTROL SYSTEMS 150 the controlled flow willbe lower than the setpoint, Actes flaw ‘can be caloulated by Rquation 2: Qua = Setpoint Flow VTaceg/Taat @ where T = degrees Rankine or Kelvin A decrease in mole weight will also resull in 2 lower controlled Dow, in accordance with Equation 3. Qua = Setpoint Flow AtWon/MWanum — O) ‘These flow measurement errors are actully desiree, because gat temperature or mole weight variations alo eanse the surge point of mast compressors to shift m the same direction. The surge point esusly shifts more then the contol point therefor, temperature oF mole weight variations nor- tually change the satety margin (iste berween surge end ‘coatzolled flow points). Small to moderate gas variadons 20%- nally cat be aecommodated by a sexpoint which provides an ‘adequate sfety margin at the lowest gis temperatere or highe tat mole weight condition. Tp handle exteiye variations ray be nocessury to chenge the controller setpoint. The Bow reasarément in Figure 22 is pressure compensated. There- {ove thesystem wil qgtiol a constant vaiume flow, regardless of the inlet pressure. This fete is needed because #urge foours atthe same value flow rate, regardless of the inlet Prose “The inlet low contol sjstem shown in Figure 22 includes some undesiable features. relatively long Inlet lise ss 2e- ‘quired to provide slits staight run of pipe between the throttle valve and flow element. Alo, the flow element (FE) is felvely expensive. and tt produces a. pressure lost that fesulsin slightly greater operating casts. These disadvantages can be eliminated hy at antisurge contol scheme based on measurements of power snd discharge pressure, es show i Figure 23. Actisurge control by thi method ss fciitated by the following characteristics of weonstant speed, nl throttled compressor: 1. At the surge point, the following parameters are una Jected by suction pressure varstlous: + Volume flow (afin, mh, ete} ‘¢ Heed. polytrope (Hp) 4 Ratio of compression (Ry R= PsP, 2. Powers proportional to weight Now (W), which in turn 4s proportional to inlet pressure (Pd 3. Pa ie proportional to Pi Le, By = Re % B, Based on the above, surge is straight line relationship of Pe vs. W, of Pus. J, as illustrated in Figure 24 As with the Inlet Bow eystem previously desorbed, this] ve, Py aystam alo F subjed to some error i inet temperatire or mole weight changes. Tor example, & tomperature decrease at constant inlet pressure will cause surge to oeeue at higher voluse flow, higher discharge pressor, and higher power. Variable Speed Centrifugal Antisurge Contra The “Flow/Deltx P” antisurge control system, is widely used wih variable speed coattfugal exmpressors. The name is dlerived from the face that contol action ls hated on measure: ments of comprestor inlet flow and. dilferential pressure (Fa — B). Two Flow Delta type contol schemes are ins tented by Figures 25 and 26. the esntrller setpoint in the ‘ppieal system show in Figure 25 fs manipulated by the "Py — Re signal, end the flow inal the process isput te the Figure 28. Centrifugal Compressor Antisurge Control, J es Ps £ £ a SY Disewanoe Prsune -14 mane vonserowen eg Figure 24, Wet Weight Flow vs. Brake Horsepower ond Dis- charge Pressure, controll. In the alternate system shown in Figure 26, the process signal tothe controller isa value which represents the Compressor operating point (Pq — Pr vs. flow signal), ind a constant controller setpoint is ted. More complete inferma- tion is provided in References 8 and 9, ‘The theory of the Flow/Delts-P system is based on the following flow element ard compressor surge line charactors thes0 PROCEEDINGS OF THE TWELFTH TORBOMACHINERY SYMPOSIUM. 1. Differential (h) produced by the flow element is pro- portional to How squared 2, Pressure rise (Ps — P.) produced by the compressor is proportional to rotative speed squared. 3. The flow rate at which surge ocers is proportions to rovative specd ‘Table 1 illustrates the above theory. For the purpose of this ilustrtion, 100% pressure rie (Pa ~ P) and 100% flow fare used for the surge point at 100% spood, Thus, ot 905% speed; % Py ~ P, = 100 (01008 = 815%; % flow sig nal = 300 (90/100 = 61%. ‘Therefore, a 90% speed the set- point will be A, ane fle wil be contre at 805% “The Plow?Deita? type control scheme fs widely used and in many eases i the most suitable method of antisurge control for contrfugals. Often, however, the system has to be modified to snit portico needs Table 1. Surge Lino Values. Speed Flow Rate BaP (5) (Signal) (5)__(% Signal) 300100 100 106 100 90 90 or 8 sD 90 80 ot a ot 76 0 6 0 0 "veces FLOWPOITENGRTIA: PRESSURE SYSIEDE Figure 5. Typical Flow! Differential Prevsure Type Anti-Surge Contrat System. Discharge Volume Flow (P and T Compensated) Antisurge Control, Variable Speed Compressor A pressure and temperature compensated discharge volume flow antisurge contral system fe shown i Fignre 27, This particular centrifugal operated in series with an axial flow compressor. Because of very large interstage coolers, separators, pipiag and space limitations, it was Geemed im” practical 10 use an igiot fow element, so alternate methods ‘were investigated, Calculated values of vacious parameters Gve., weight flow, inlet and discharge volume fave, pressure TavtesunazcorTroL SvsreEW Figure 26. Alternate lowDifferential Pressure Type Ante Surge Conteol System, OX ven vatve Rom axa ceases Sewnesson DIScHAROE VOLUME FLOW [PAT COMPENSATED) vanunble SPEED CONTA LGAG Figure 27. Discharge Volume Flow (P and 7 Compensated) Anthurge Control, Variable Speed Centrifugal zat, power, et.) atthe sorge point were examined for verious operating speeds, Culoulated discharge volume flows ior all surge points, fom 75 to 95% speed, varied less than 1%, ‘Thus, 2 carofal camination of various caleulated parameters tadicatel diseloarge voluine flaw to be aot only tie snost practical but also-a very effective variable to use for surge Prevention. Field surge tests ofthe actual machine verified the calulated predictions. Surge and contro! lines are shown in Figure 23 on the compressor performance curve. The flow measurement is temperature compensated; therefore, the con- tral line does not change with temperature variations, In this case, temperature compensation deereased the control error at speeds above 85%, but slightly increased the error at speeds Below Ot Becase of he ene at noel speeds, expe ture compensation was justied. With this ststem, the ialet flow clement was elimigated, the discharge flow element was sinaller an less expensive, the pressure loss due to the flow element was less significant at the discharge, anda convention- al flow computing system was vsed,TUTORIAL SESSION ON PRACTICAL APPROACH TO SURCE AND SURCE CONTROL SYSTEMS vet 40 ‘SURGE LINE -ONTROL LINE SETPOINT = 3200 CFM COMPRESSION RATIO, PalPt 2 3 F § 6 FLOW, DISCHARGE CFM x 10-3 SURGE & CONTROL LINES DISCHARGE VOLUME FLOW (P & T COMPENSATED) ANTISURGE CONTROL Figure 28. Surge and Control Lines, Discharge Volume Flow (P and T Compensated) Antisurge Control Discharge Volume Flow (Non-Compensated) Antisurge Control, Variable Speed Compressor A sitmiar system, bat without pressure or temperature compensation, # shown in Figuie 28. The surge aud eontre! line curves for the compressor and control system are shown in Figure 30. The system is very simple, yet i provides very effective contiel Several things account for the surge and contro lines Laving a similar slope over the normal 85 t9 105% speed range: 1. Flow units at normal conditions {ie., 1.03 kplem® abselute and 0°C) ate used. ‘et antthr = ci yfh XP 7] {in this case, the system measures only the different head {h)aeross the orifice. The controller maintains "h” ‘onstant at ils setpoint valve. Therefore, the controlled flow (nm¥fhr) varies in proportion to PT. Discharge pressure and temperature decrease with speed, On a ercentge ons, dacbrge presnure decreases uch tha temperate, 20 the slope ofthe contral line is primarily a function of 2. The capacity of the compressor relative to its compres- son rato infuces the slope ofthese ie 0 ths ‘ace, the relationship s such that the surge line slope is very similar tothe flow ws. Py slope of the contro line lin the normal speed range). By comparison, the com Dression ratio of this machine is approximately four Limes that of tse compressor used for the preceding example, but the flow capacity is lower by a factor of ight, The entisurge control scheme shown in Figure 2G is not always best suited for all watiable specd rachines, but, because of its simplicity and eflec- tiveness, it should always be considered, Pigure 29. Discharge Flow Antisurge Control, Variable Speed Centrifugal Compressor: Figure 30, Surge and Control Linos, Discharge Flow Control, Variahle Speed Centrifuea!. Flow/Delta-P Astisurge Control The results of a Flow/DeltaP type contro ae illustrated fox normal and startup conditions i Figures 41 and 82. The startup gas is more than twice as heavy asthe normal gas (12.60 MW, at startup vs. 5.97 normal). These two illustrations show the efct that a heavier gas has on the compressor, ples the lnhetent selfcompensating effect of the Flow/DeltaP control scheme, At any given speed, the machine produces a much higher discharge pressure on the heavier gas (Figure 32), and SF speed if prrchicee wore pressie Han om ormal peat 300% pin “To.compare the relative contol safety mangins for the fw) xases, consider the surge and conteol points at 12.5 kglem® discharge pressure (41 Py — Pp. Whe operating en normal fas (Figure 31}, surge occurs at 5900 mrhy and 100% speed, Gontcol line "Wat this point is 8800 m¥tbr, ora safety margin of 900 mh, ‘At the heavy gus condition (Figure 32) and at the same discharge pressure (12.5 lafer') and Fy — P,, surge occurs at 4000 m*Mhe and approsimately 68% speed. Control line “Nat ‘this point is (700 mr, or a lower safety margin of 700 mf. Contre ine "N" is prodveed by the altemate contol system of Figure 25 with a setpoint of 6500 mf. Ths setpoint could be ‘considered the normal setpoint, and a higher setpoint of TOD162 PROCEEDINGS OF THE TWELFTH TURBOMACHINERY SYMPOSIUM i ‘ne wn 21 40180 ext faagiene Te regee ts Kova [NE 5 CONTROL LANES FLOWDELTA ANTSURGE CONTRO. NORMAL GAs Figure 31. Surge and Control Lines, Plow(Delte-P Antisurge Control, Normal Ges. + (Semin Figure 32. Surge and Control Lines, FlowiDelea-P Antisurge Contrai, Heavy Gas (Startup) ‘vir esed to produce a greater sulety margin during startup {control line “S"h, Alternatively, if posible to operate ata feipoint of 7000 mtbr without eausing inefficient operation st ‘orml conditions, thea such a setpoia would be practical for all operating conltons For the eamapressor used in this example, the contol safety martin decreased asthe male weight increesed. Alb, it wwowld decrease wth a decrease im inlet temperatane. The reader should be estioned against drawiog the wrong conch- sion from this example. More often, the reverse actually is fever Le, the safety margin increases duo toa heavier gas oF lower tomiperntare. The procedure used to caleuite the con cco iges provided inthe Appendis “The FlowiDelte-P enntral is azo very practical and effec. Uve fora constant speed, inlet thytted compressor. Depend- ing on how the ssster: is calibrated, error due fo inlet presnire variations will be small or non-enstent. I the systom Calibration ie such thatthe contiol ne originates st zero Bowe and Delt P (Py — Pp, the exor willbe zero. This is exploined by the following characteristics of the How element and = ‘onstant speed compressor: 1. Inlet pressure variations do not affect the volume flow sf the compression ratio at which surge occurs. The foutte, therefore, must qaintain a constant volume ow reyardess af flee pressure . Ata constant volume How, the differential head pro- duced by the enifice varies diectly with presse, as expleined by rearranging the standard Equation 5 for Soles flow to Equation 6 Ve o or h yy r ce) 6 3. Delta P (Py — P) varies directly with inlet pressure in Nondaned with egunon Pe- R= PR -} a ‘Taking the above into account, at 50% P,, a. Ps — By = 0.5 (Ry — 1) = 50% Setpoint » n= (LY 25 5 se Cc} 2 ©. 100% volume low is required for the orifice to produce 50% differential at 50% pressure, by Equation 6. Therefore, the system will, in fact, control 100% volume Row, regardless of the inlet pressure, Modes Of Operation (Startup, On-Stream, Shutdown) ‘The complete antsurge contol system mist ffetvaly prevent surge ching #actup and shuidown, a well as ring ‘on-stream operation, Discrete signals are used to override the amalog control system and manipulate saves during startup tnd shotdown, As when designing the anlog system, i fs valy traportant © properly evaleate the compress and process requirements when designing the sastup and shot down logic system. at = For startup it has been common practic to msintain the sntisurge control valve fully open during most of the acerlera- ‘ion peri. Typical, the gic ovcrade signals temoved at 0 of 85% speed to trasfer contol tthe anelog system fr 07+ stream operation. This introduces a step control change. The Inmet setion by the atisirge controller san attempt to dive the valve filly closed. ‘The valve, moving towsed it closed position, brings the compressor operation to the point wihere the atisuge conrcller aust take aver contol ad modulate the valve in a partly open postion for sable‘TUTORIAL SESSION ON PRACTICAL APPROACH TO SURGE AND SURGE CONTROL SYSTEMS as compressor operation. IF this recovery action is not fast enough, the valve will elase too far, and the compressor wil, surge, Ifthis isa problem, the system must include aeazare to smooth the trmster action and to give the controller suicient| time to recover contro Instead of activating the antisorge controller at 80 oF 85% speed, it fs bettor to have st in control during the oxtire scceleration period whenever possible. To faite this, the steRe control fine must never intersect the surge lice. For ‘example, the control line for 4 sctpoine of 8500 m/hr (Figure ‘1 6 always on the sight side of the sure line, According to ‘general practice, Figure 3L shows surge and control lines for the normal operating gange only. bot at zero Delta-P (Le., Pa and P, = 8:19 kyfem®), the contvol point is 2247 mM when ‘aleulated by Equation Al of the Appendix. Therelore, the analog system should be in conta thronghout the startup, asc atransler a 80 or A5% speed is not necessary. However, for ‘the control described for Figure 80, a startup override to open ‘he vale at lw speed would be neces ‘Another factor that must be eousidercd isthe fact thatthe stages of a multistage compressor become mismatched when operated 2¢ less than normal speed. Under seme conditions the compressor could surge evea though the ant-surge valves folly open. Usually, surge is initiated by a stage near the discharge when the machine is operating at rated speed. At lower speeds, a stage eloser to the inet, frequently the fst stage, initiates surge. The reason for this is that at lower speeds, the gas volume is na reduced by each stage as much as itis reduced at mted speed. Consequently, at reduced speed, th inlet volumetellow rate through the final stages is wreater than their rated flow, and the volume entering stages near the inlet tends to become les than their rated flow. ‘Aavhigh flow rates, the gas velocity can reach the speed of sound at some location inthe compressor. Since the waximurn ‘elocity of any sis hited hy its soni velocity, this imposes a maximum fow limit on the compressor. This condition seldom ‘ecurs while operating at rated speed. However, when the speed ofa molt-staze compressa is reduced, a speed finaly s reached where itis no longer possible to keep the eozpressor nt of surge by recyeling or venting, gas from the final dis charge. The reason is that some stage near the discharge bi reached its choked (onic) flow limit. which backs wp the flow in the earier stages to the point where one stage reaches is surge point. To further explain what happens, assume a si stage compressor with arated inlet vohume low of 10,000 acim [Also asscihe that, strated speed, cach stege makes «compres sion ratio of 2.0, and surge occurs at approximately 7700 acim at 100% speed. Distegarding the effect of increased tempera fuze, the discharge volume from Stage 1 therefore is 5000, which is the ialet volume to Stage 2. If we carry this on through, we find che inlet volume flows te Stages 3 theough 6 ty be 2500, 1250, 635, and 313, respectively. as shown in Figure 33. Now let us assume that at 70% speed, each stage takes # compression rato of only 1 Send thet Stags 6 reaches its choke fmit sta volume flow of 600. Also assume thet each stage will surge at G04 ofits rated capacity when operating at 10 speed, the 5 volume reduction per sag meas tat tach stage will have en inlet volume flow 15 times geeater than is discharge volume. A we start with tho 600 meximam (choked) limit of Stage and multiply hack through all stages, we determine that the volume flows into Stages 5, 4,6, Band 1 are 100, 1350, 2025, 3038 and 4556, respectively, as Shown in Figure 34, The 4556 ‘volume Bow at the inet to Stage Tis considerably less than ts 900 minimum surge limit, oF, in other words, choked flv fs Stage 6 has forced Stage Ito operate below ts surge limit. This ‘condition might eecur ata speed in the veinity of 80% ko 90% 2 i OPERATING SPEED = 100% VOLUME REDUCTION PER STAGE = 2.1 VOLUME REDUCTION PER STAGE, COMPRESSION RATIO = 2.0 Figure 33. Volume Reduction Fer Stage, Compression Ratio = 20. & OPERATING SPEED = 70% VOLUME REDUCTION PER STAGE = 1.5:1 VOLUME REDUCTION PER STAGE, COMPRESSION RATIO = 1.5 Figure 94, Volume Reduction Per Stage, Compression Ruiio = 15. ‘on some compressors, and at much lower specs on others. Ja cnder to accommedate this onadition, a secyce valve at some Stage (oe startup and shutdown valve, Figure 28)we PROCEEDINGS OF THE TWELFTH TURBOMACHINERY SYMPOSIUM must be open while the compressor is operating at reduced speed during startup or tutdewn, ‘The type of gat handled is factor i the condition just dlscized, For a heavy gs, e.g, curbon dioxide, soveral valves located at intermediate stages may be required because its sone valocisy is higher. Also, a change in speed makes a reser dllrence inthe volume reduetion prodveed by eae, stage. ‘Stage mismatching also can alter the surge lineshape. The shape ean change with speed as well a8 gus variations, This 38 ‘apparent in a comparison of Figures 31 and 32. The surge line fon Figure 31 shghtly carved, bat inthe reverse direction Troan the emg line on Figve 32. The difacont clapes once hheerute Stage 5 fon Figure surge above, and by Stage 1 at 85% and lower speeds, On some. Inachines, the stages which initiate surge shit several thet ‘within the operating range, aad the reverse curvatoe i much eter than that in Figure 21, For the ame compression rato, ‘an axial flow compressor has more stages than a centrifugal, so the surge initiating stage changes more times. This accounts for the reverse curvature typical of ata How compressors. For a constant speed, axial low mochine the efflat fs less thon fora Saviable speed axial, hecause only the stages with variable angle stators are affected. Typically, approximately 40% of the feages have vaehble stator vant. Shutdown Control Action ‘A vory high potential for surge exists during an emergency shutdown, of any time the comprestor is tripped while opera ‘ing at normal conditions. Deceleration is very rapid typically vith spoeds having doercased to 80% rpm approximately 1 to Typ seconds after tripaoun. Fle vorprestur’ ain 89 proce pressure rise decreases approuimstely propotionally £9 speed squared. Thus, the need exists to lower the discharge pressure very quickly ater trip-out, otherwise, severe surging can re sult, During deceleration, the recycle valve must handle all of the compressor flow, plus any excess gos accumulation in the discharge system. W theve is an excessive quantity of gas compressed in the discharge system, the recyele valve must be ‘capable of passing a very high Row rate curing the frst several sovonds of the deceleration period in order to lower the pressure fat enough to prevent surge. The most importent lictor in preventing surge under these conditions isto have the least possible volume at the discharge, Sometimes this ear be sccomplished by loesting a check valve as close as possible to the compressor discharge, However, in many systens the roryele takeoff point rust be loested downstream of a cooler and separator, this adds a considerable volume atthe compres sor discharge. In some cases its possible to accommodate this by a fast opening (1 second or less), generously sized (125 to 200% of compressor rated flow) reeycle valve. I ather cases, it say be necessary to uso an additonal valve and check valve very close to the caraprossor discharge. If possible, this valve can vent gas to atmosphere. Hf stmaspherie venting 18 not feasible, then the extra shutdown valve coold dump hot a5 Jack to the compressor inlet. The temperature rise from inlet to dischaige decreases rapidly as speed decreases, so recycling hot gs bck toilet normaly ne a problem dung the relatively short coastdowin time. It should also be noted that ‘he high temperature effect of recycling hot discharge gs is ‘often mitigated by the temperature drop (Joule-Thounpson ellct) resulting from the expansion across the reeycie valve. ‘This effec: is significant, especially with natural ges at high presses Tf tompray hgh let temperature ia probler for a particular shutdown characteristic, then the hot gas roeyole valve could be closed spproximately 10 to 15 seconds after tipoot, For the res ofthe coustdown period, the devser- ation rate normally # slow enough that the normal reeyele Nabe is able to prevent surge System Piping and Configuration Aspects Piping and valving are important considerations fa surge control system design. A very fundamental and important aspect is the sizing of the bypass line. When startup, shut- down, and normal operation are considered. there are a am bor af variations, Field experience has shou this straighfor- ward task to be the cause of many field problems because the Tine was too stall. Another consideration i line sizing is the dyeamic or momentary requirements in the limiting type of control, such a the surge control system. Because steady state criteria would indicate a line sized to pass only the surge plus margin flow, there is a temptation to vtilize the apparent economics of a smaller line size, Tn aetual practice, the line should handle fil compressor flow witha minimom pressure drop. Correct surge control valve selection and sizing i also ob critica importance for achievement of good conte) scasitvity and a rapid response in the vietity of surge. While other components (Le. actuator, controller) also atlet the overall control system response, valve sensitivity and vosponse have the major impact ‘The control valve isa key ink inthe control chain, In most cases, the valve shovld have linear flow and fast response ‘There is very litle benefit in having a highly sophisticated ‘controller set point computer coupled to a valve either too slow ‘oF 90 small 1 handle the problem. When the surge system is ‘ued for startup, a cond valve may be required, a5 the “tuen down’ borden of a single valve may become too large. re- ‘qoontly, congo eontral selvor are olooted too Inge, ranting wot only in unecessary cost hut also ia significant loss of contre sensitivity, a 18s that eazmot be overcome Uzougb reiduced signal sampling intervals, actuator changes, ett. tri proper valve type selection ean result in insuficient sensitivity and cause excessive noise, which usually limits valve dura bility When oir is the compressed medium, the surge control valve normally can reliove to atmosphere, usually through an adequate sileneee. If, however, the compressed medium i © process aes or a refrigerant, the valve will have to bypass the Section being protected aad return the fow to the proper suetion nozale. In such eases, some form of cooling s required toremove the heat ofcompression. A shell and tube exelunger ‘oF an air exchanger, ss shown in Figure 35, may be vequired ‘The system showin in Figure 350, which is a single recycle loy system, did not work well, This was especially true for hi molecular weight operation where the low pressure (LE) casing wes in tage whe the high presage (HP) eng was ston situation was impossible to correct by openiny the recculitson valve, It vas Impossible fo state sytenh vwithont mismatching the LP and HP casings dering spood shanges. The LP and HP signals to she recreation valve also interacted end ereated a problem. The configuration showa in Figure &3b also was unvotkable, because of reverse train rotation after tip. This was duc tothe large amount of energy stored in the discharge enoler. The emergency shutdawn valve (ESD) was not fast enough to cope with this problem. Also, when dhe HP spill-back came open first, the LP casing went into surge. In addition, there was dificult iy malntaining system balance dvi stat op a cing, The peered arrangement is shown in Figure 35e. ‘the gos is sent through a condenser (as inthe ease of a refbigerant, direct coniact cooling may be possible, On smaller ‘nil, aliquid level ean be retalned in the bottom of dhe suctionTUTORIAL SESSION ON PRACTICAL APPROACH TO SURGE AND SURGE CONTROL SYSTEMS 165 : Figure &5. Equipment Arrangement. drum with normal flow into the side above liquid level. The surge control by-pass flow passes through the bottom, contact- ing the liqnd and thereby desuperheating. As the systems get Larger, this method is not feasible, due to the shaking forces| from the boiling liquid. For such service, a sparger or a modified steam desuperheater an be used, im which the liquid is spanged or sprayed into the bypass line ahead of the suction ‘drum. A ving of temperature sensors monitors the spray and is set for the desired amoent of residual superheat. Cooling a system on bypass ean make startup on process gas much easier. ‘On mnlti-nozale compressors of compressors with inter coolers, bypassing should be eonsidered for individual sec tions. This will bypass the necessary flow to keep the affected, stage out of surge, rather than waste enerny by bypassing the Entire compressor ar train (it moltibadied}- tho bypass scheme must be evaluated for both energy and stability, as there may be tendency for some raultiple loops to interact i proper eare has not been exereised. Figure 96 shows the location of a check valve in the system. Figure 36b is the requited arrangement i parallel, operation i anticipated. ‘This system ix lao required if the ‘Comprosor must staré up against a stagnation pressure, as the pressure an the downstrcam side would keep the check valve from opening. Field experience and some experimental work [10] have shown some edvantage to the control shown in Figure 3a, ifthe compressor isnot i parila operation and the davastream presstre would not prevent the check valve from opening. This configuration, however, i rather inf quently used in the petrochemical industry. The inherent stability ofsuch a scheme is much yregter and allows for more gain and responsiveness in the systexa, without the effort required far the scheme shown in Pigure 3éb. “Mieroprocessor based surge control systems can be made vvery sophisticated. Most of the variables external to the com pressor are readily measured and can be digitized without [problem or can be treated and combined in analag form. The signals may be processed either in analog form or digitally swith the latter having en advantage. For multiple component 41s, measuring the snolecular weight, "K" and "2" is still Somewhat of a problem. Sometimes # density meter can be ‘used ifthe variations are mot too great. This compensates for apparent molecular weight. but not “K” or "Z". There Is no feasy answer for this problem, ass eth @ Figure 36. Location of Check Valve, Response Time of Surge Control Systems in Upset Conditions Often, surge contro systems which are operating satisfac- torily under norteal conditions are madequate when fast re- sponse time is required. A fast rise in head would be an example. Such a fast rise can oveur fone compressor ina serios combination shuts denen beeause of malfunction or i the How is Accreated rapidly because of process malfunction, As can be seen from Flgore 37, on svt down of one compressor in a serias combination, the remaining enmpressor would expert fence an increase In Delta P, due to a back pressure effect ‘Without provisions in the surge contiol system, it would start ‘opening the surge valve when the operating point approached the surge line. Because of the response time of the system facruators, valve travel, etc.), the compressor most likely sould experience surge, This ean be avoided by adding anti cipatary action, either by adding one way derivative action 10 ‘the surge controler on # fist rse in head, or by fast decrease in flow. Surge valve opening would occur immediately after detection of such fast changes, reducing the response time and thus preventing surge.186 PROCEEDINGS OF THE TWELFTH TURBOMACHINERY SYMPOSIUM fa m= © Merating pint wih to © Speratns pit 6h ct tea Figure 81. Series Operation of Compressors Stabitey of Surge Control System ‘nd Porallel Compressors Close ta Surge arallel operation of comprossrs close to the surge line cam lea to unstable operation of the surge systems if suitable ‘control precautions are not taken. Only in rare cases are the surge lines of parallel compressors che sare. Manuicturing tolerances and piping diferences result in diferent sure lines, as shown in, Figure 38, Two types of eoatral design are possible. Tn the first type of design, the speed signal or other onteol signal is biased so that both operating points have aa ceqal distance to the serge ine. In the socond dosign, the Speed ex coneol signal to both units is the save, wich woeld result im a different location of the operating point ‘a the comoressor envelone. The surve control svstem is eauinned witha feature to detec the distance from the opersting porn 0 the surge line. Any spoed decease ofthe unt closest & surge will be interrupted at a preset distance from the surge line This preset distance is selected just to the right of the point were surge valve opening would be required. Any adeitional reduction of throughput will he achieved with the other unit tu also reaches this preset distance. This feature allows amarimutn operation without opening the neycle valves, which considerably decrease the plant effeency oa JN Go Pigure 38. Parallel Operation of Compressors ‘The fast design is more complex than the second. The preferred design depends en the location of the efficicacy: lines. A comparison of combined efficiency of the parallel compressors will most licely determine the type of design preferred. Surge system stability of pacalle! or series compressors Is Important, since oscillation of the surge control systems, be cause of their time constants, cam Tead to excursions into deep surge which are a3 damaging as the surge which is to be prevented. To prevent osefllation of compressors, their time tists she raps posible Ov he ingle sways of achieving this ts by afféeting the travel time of the recycle (surge) valve, Sines, obviously, the opening time of the valve should be as short as possible, we are left with the adjustment of the elosiag time, This adjustment has proven fictive in preventing. oscillation in most applications. Due consideration also should be given to integral aad proportional action of the surge control systems. However, itis much easier to ind a sable condition empirically with valve closing adsest- moat than to use othor methods, which often require dynamic modeling. ©. DYNAMIC SIMULATION OF COMPRESSOR SYSTEMS ‘Dynamic simulation isan engineering tol to evaluate a process design. It involves the use of a mathematical mode) {leering dhe behavior ofthe system in an equation form tht {8 then selved using a competer. ‘The equations consist of algebraic and differential relationships, whore origins come from laws of physics and from empirial data, The computer representation then can be used 10 predict hose the actval ‘peter vi behave in "what situations. The results describe both the steady state operation and the translent behavior during upsets, Because this enalysis can be performed during the design pase, the knowledge gained can be used vo reduce coveral corts of design, constriction, startup and operation of the system of interest. The development of 8 model for a specie applisation iavolves many people; 68, process eagie neers, desiga engineers, clients, simulation analysts, and ven- deer The tarkange of tabunation hetween theke groine aso serves a quality assurance function for 8 project. ‘The Need For Simulation Studies ‘With higher and higher trends in the capacity, pressure, speed and horsepower of centrifugal compressors, the design process becomes tore Involved. The eomprossar desigaors ‘and the application engineers, who are responsible for the desig of de enti prowess: are faced wih the prble of ing rnachines which are often larger and more complex. Tine cuc bebe ble “They spat sonstiow pride remo able assurances that the machine will operate a3 intended, ‘once installed. This process of satisying one's self (and the plant’s owner) thet the compressor system design will perform. satisfactorily can be very subjective. Traditional methods of analysis camot provide accurate predictions of how a compres sor and a process will intertct. Because of this, designs are often based upon the designer's experiences and opinions A tool which has been extremely valuable in evaluating the performanes of compressor system prior to installation js dynamic simulation. ‘The simulation studios result in design ‘recommendations coneoring the number and location of recy- cles required, sizing of roeyele contra! valves, and setpoint, fin and rset setting freon system nsre entation The studies also may lead to a better enderstanding of how the compressors will interact with processes. Startup procedures sre tested and amended s required. Reference 1 discusses the epplicstims of simulation to three eampressor designs, ‘The question of wilizing single vs. multiple recyele loops for malti-casing applications is dificult to answer eoneusively ‘basod on intuition, a5 past experience may not be a sufficient ‘guide, The dynamic behavior of sneh systems is nfuenced by process parameters, control systema responses and the turbine: compressor performance curves. The impractiality of manual caleulations and intuitive desig approaches ereates a nocd forAUTORIAL SESSION ON PRACTICAL APPROACH TO SURGE AND SURGE CONTROL SYSTEMS ast computer aided design, Reference 12 describes a simulation study of a compression-liguefaetion systom that rovealed the original design for a est facility 16 be severely vmdersized in ‘nwo of the five eisings. A retrofit progrees was implemented prior to operation, thus minimizing schedule delays, Compressor System Similation Ege 3 illeine the procul ste ol pio simulation study. The steps can be grouped into ve phases. Tn Phase, the boundaries of inert ae etabiched,ooesting rmodies aod limits identified, criteria of acceptable performance documented, and the types af upsets of ngeret identified. The mathematical representation is prepared and assombled, and pertinent physical data are gathered. In Phase 2, the mathe- ‘tical mode! is converted into compiter language, and the program Is checked out. 4 simolated steady stato 0 Condit is acheved and compared vith Snow/epectel operating conditions. In Phase 3, the simulation model is evaluated at transient conditions. ‘rhe moxie) behavior Is an lyzed and validated as representing the system performance validation involves comparison of past knowledge fom similar systems, review of behavior by specialists, comparison with expoceed behevior of physical las, and a review by personnel with many years of dynamic simulation experience, In Phase 4, the simulation model is used to analyze “what if” scenarios, series of test runs is perfarmed to document hos the actuel system is expected to behave, and the results axe analyzed, In Phse 5, conclusions and recommendations te ‘incing system performance arc documented and submitted mm {final report, The date from the test vans are assembled, cited and presented. Tho final report also dosurnents impor tone aspeets of the simulation acevity DerNTON OF sioTeM ATHEMATICAL MODELING. (OF DATA SENULATIOW STUDY. BIACRAM Figure 39. Simulation Study Diagram. Figure 40 Wustrates how the mathernatical model for a typieal compressor study 1s organized [11]. The model for 3 ‘compressor system can be considered to consist of three sepa rate modules, among which information is passed: 1, Proves Module—The proves module consists ofalge= braic and differential equations describing the accu lation af mass and energy at various places im the ‘stem such as vessels or piping jumetions (nodes), the transfer of mass and energy belween the nodes, and ‘thermodynamic characteristics such as enthalples oF phase equilibria. The process model can be relatively simple, such as the madel for an injection systorn 45 ‘which Variations in temperature or molecular weight 4re not significant, and therefare only mass balances ‘need be considered; or may be highly complex, such as refrigeration systom models in which hoat transfer, phase equilibria and temperature variations have to be Accoanted for, in addition to mass balances. 2, Compressor Mogule—The compressor module con- fists af the empirical heasilow characterises ofeach section as provided by the manufacturer, thermody: ‘namie corapression equations, and a torque balance between the compressor and its driver. Reference 1] provides a typical compressor model block diagram ‘The compressor module receives suetion and discharge pressure and temperature from the process module. It feu caleubates enipressun flow si beget ature as outputs to the process module. In the ealeula- tion process, compressor head, speed and volumetric low are determined. The eomipressor model ray be a simple representation of a constant speed machine, or 2 very complex one in which variations in speed, efficiency and polytropic coefficients are considered ‘The degree of sophistication requived is ¢ function of ‘ott the type of system under study and the objectives of the study. For example, a model to study startup would have to be more commplex than a model used &@ study constant speed operation, 3. Gantrol Sestem Madule—The eontel system module represents the ichavior of the process control in: Stramentation and fsa contral elements. Contelles, transmitters, computing relays and contol valves tsu- ally are redcled machematally, A tie-in #9 actual comerol barchware, eg, uistbuted contra systems, ‘aay be requlzed to evalvate the response of software tleordhme: Gontasl vale demasnior fesootally the Salve stoking period) are modeled. a5 oe sgnicant Teanement [ag F]——amoncason rows ——nenurce Terenas Jn a ORG PRESSURES SUCTION TEMPERATURES. Figere 40, Compression System Model Example of a Simulated Compression System Figure 41 is a simulation flow diagram far one train of a ‘parallel train configuration that wes used to handle gas from aa ‘ofl eld separation process, The incoming gaswaterll fow from the field at ow pressure conditions was separated and the {a8 routed to the suction hescer of the compression system. ‘The gas was compressed ftom 140 poig to 700 psig for distribu: tion to a central compression plant, which further inereased the pressure for injection back into the oil reservoir. “The maior objectives of the sizulation study were to evaluate alternate process configurations during both tip and. normal process upsets, to evaluate the ellects of valve lures, and to determine the sensitivity of valve paramoters « the ‘prevention of surge, The methodelogy desettbed in the prev ‘ous section was followed. The model was subjected to a series168 of text runs, OF particular importance were the “Emerkency “Trip” and the blocked compressor discharge upset; c.g, clo- sure of Block Valve 5 in Figure 4 ‘The simolation model results were instrumental in the deoision 1 add a second check valve hetwoen the compressor ciscbarge and the discharge cooler. and the addition of « hot bbypass loop arouad the compressor, Also, control system pe raters, eg. valve parameters, controller parameters, te, needed for the prevention of surge were determined. In & fallow-ap study, the performance of pavalel trains during normal operation, upset conditions, and startup was analyze iguces 42 and 43 illustrate typical results for the revised design undes “Emergency Tip" eoulitions, Piguie 42 shows the time history of important compressor variables, while Figure 48 shows that the compressor did not surge at high nergy levels. fn Figure 45, the variable SURVOLI reflets the flow (ecm) shat would indicate the boundary of surge. The variable ACFML reflects the acwual low through the machine if ACEMT is greater than SURVOLL, 0 surging ocxurs ‘Without the process changes identified above, this condition ‘could not be met. Once a machine eaters surge, the simulation ‘iodel does not deseribe the resulting penomena. Figures 44, 45 and 46 illustrate results from a blocked discharge upset. igures 44 and 45 show the variables previ- ‘ously described, As Figures 45 and 46 indicate, the compressor ‘went nto the reeycle mode end aperated onthe control line et by the FlowDelte-P controller Contral Hardware Realuation ‘Once the parameters for the antissurge reeyele loop, 0:8 valve capacity, characteristics, response time, and controller “gouithan and settings, hove bevn determined, the nent gues tion centers around the type of eontral hardware to be used. A ecision har to be made whether the control algorithm he PROCEEDINGS OF THE TWELFTH TURBOMACHINERY SYMPOSIUM PHASE 2 COMPRESSION SIMULATION ‘TURBINE 1 TRIPS FROM STEADY STATE 1 pS F—A—~J | CS | Figure 42, Simulation Results implemented with « coatiouous systom, e.g, electronic analog, or witha sampled-data systere, e.g, distributed digital contra, Chin and Pobanz (13) discuss the applicability of sazzpled data systems to fast response loops. The simulation here involved the actual tie-in ot hardvae for the antsurge conteoller deserved in the simlation example above ‘The simulation model was vun in eal time, It provided “eld ite” tansater ougputs for she low and dela pressure measurements and accepted "feld ie” transducer iapats rom the actual control hardware. Figure 47 illustrates the setup. “The ntoreonnettone hotven the iapetontput (UO) of the compater and the contol hardware seore the same as feld Signals (£20 ma, 145 vats). When the train discharge block ra cot pte wot ¥ A a S An petyaio . “ facil ee eR . Sa] | Preven eek? Op The ” Ne . . reemore we "aA tetra, a ri peteennee COMPRESSOR SYsTEm PHASE 1 SIMULATION FLOW DIAGRAM Figure 41. Compressor Flow Diagram.TUTORIAL SESSION ON PRACTICAL APPROACH TO SURGE AND SURGE CONTROL SYSTEMS 163 PHASE 2 COMPRESSION SIMULATION TURBINE 1 TRIPS FROM STEADY STATE 1 PHASE 2 STEADY STATE FULL CAPACITY BLOCK VALVE § FAILURE AT TRAIN 1 2 ~ LN! 1 - et Figure 43. Simulation Results valve (Valve 5 in Figure 41) closed, the discharge pressure fnereased and the compressor moved toward the sugo region, “The response ofthe ant-surge recycle loop put the compressor in reeytle flow ata flow set by the control line. PHASE 2 STEADY STATE FULL CAPACITY BLOCK VALVE 5 FAILURE AT TRAIN 1 Figure 46. Simulation Results REAL TMKE SIMULATOR oF sven pysannie SHULATION constren busty NODE) ve. - 1 | : }. foe 1 TDSTRIBUTED CONTROL SYSTEM ce ~ ACTUAL CONTROL Figure 44. Simulation Results PHASE 2 STEADY STATE FULL CAPACILY BLOCK VALVE 5 FAILURE AT TRAIN 1 = [ere —— > ” Figure 45. Simulation Results (CONTROL ALGoRITHD) Figure 47, Distrbured Control System Interfacod With Real Tire Simulator. ‘The performance of this contol loop depended wpon the harawareltoPware implementation of te contol sgorthm. A comparative analysis for three difereat distributed conteol systems (DCS) and an clecrouic anaog contaler was made Figuves 48 through 51 summarize the results Four time Iistory aces are provided in each figure. The fst one shows the black valve position. To fntiace the upset, twas closed fa fone second The socond trace shows dhe compres SUrse ‘margin actual Now less surge Aw). When the trace exossed the center line, the comprossor entoved the surge region. Although some of che aces show the compressor coming vit Of surge, there is n0 assurance thatthe actal compressor wwonld behave ia the same manner. As previously stated, the Simulation model does not select true surge behavior The as two aces are from the output of the actual control DCS oF alg hardvare10 PROCEEDINGS OF THE TWELFTH TURBOMACHINERY SYMPOSIUM Jn Figure 4, the sun was made using an eletronle analog, como, ths Weck ao bop cr Ta) he compressor di casure rises, From the compressor tna, this ess dereaseofflow trough the compressor ‘snd seduction ofthe surge maigin (Trace 2), Thi redwetion prompts the analog Pl contoller to ae (Trace 4) and brings the ompresor’s operating point, before it reaches che surge tne, Uusck to the control lie. The ants-sorte controller acts before the operating ptae reaches the set point (eotral ine), enue os antiveset whadu feature, From Trace 2, it can be seen Uhat the controller provides & comfortable surge margin. This run is the basis fo comparison ofthe remalning runs. ‘Tha throe DCS syetoms involved sn ths eoanparzon had the fallowing sampling Kequency System A 9 samplessecond ified) System B 2 sumples/socond (fixed) System © 10, 9 and 1 samplesiseoond and 1 samplets seconds (selectable) ort BLocK VALVE Position tae hog. uta re MARGIN gouge. cece BURGE Fests lee! “peo! outeuy eae 3 on {Wot ANALOG, OuTPUT oes 8) oh ANALOG CONTROL Figure 43. Analog Control. ‘The vesults are showin fa Figures 49, 50 and 51, respectively. In Figure 49, tho run was made with System A, which had «sampling rate of 3 somples/second, I¢ showed that the. com- ‘pressor will surge for a shart time (Trace 2). In actual opera ‘on, once the compressor is in surge, the final outcome is unknown, The reason it came out of suige in ths study is ‘because ofthe extrapolation of the-eomprossar map beyond the surge lige, Therefore, the results must be used only to deter- ‘mine whether the compressor surged and to determine the margin from the surge line fit did not surge). In Figure 60, System B, with 2 samplesisecond, was nut fast enough to preventthe compressor from surging. In Figure 52, System C, ‘sith 10 samples/seoond, was selected. Run results shovsed that itdid not surge the compressor, however, the surge margin was small you ortu—e] BLOCK YALVE POSITION swe Lee autee fn ARG gy = ae 100 % ‘pos! ute ine on 00h ANALOG, OUTPUT of ‘pcs’ CONTROL GAmpLts/CbCOND Figure 48. “DCS” Control, 3 Samples!Second. In this evaluation, a simple proportional plus integral {PD) ‘control algorithm was used. ‘The controller parometers were the same for all four cases. This comparative analysis applies to aspecile gas compression systern. No atternpt should be made to gencrilize these conclusions Also, some DCS systems alford great Mesibiity in the algorithm calculation; this might result in an algorithm that performs better than the one used. Reference 13, ako discusses other characteristics of a DCS system which contabute to its performence. CONCLUSIONS Compressor surge is a phenomenon of eonsiderable inter- t itis not fully understood. Te is a form of unstable ‘operition and should be avoided, Tt is a phenomenon that,TUTORIAL SESSION ON PRACTICAL APPROACH TO SUNGE AND SURGE CONTROL. SYSTEMS m ope Block VALVE POSITION close: ‘pos’ outeuy {eae 9 on 100% ANALOG, outPuT Sieee 0 ‘069! CONTROL 2 GAMPLES/ SECOND Figure 50. "DCS" Controt, 2 Samplea!Second. unfortunately, oceurs frequently in the process industry, some- times with damaging results. Surge has been traditionally defined as the lower limit of stable operation in x compressor ‘vd it Involves the reversal of low, This reversal of law oceur Ibecause of some kind of aerodynamic instability within the system, Usually, a part of the eampressor is the eause of the sscrodynatnic instability. although it fs possible forthe systern arrangement to he capable of augmenting this instability Corapressors are usually operated af a working line, separated bby some safely margin from the surge line. Surge is often symptomized by excessive vibration and an audible sounds, Ihawever, there have been cases in which surge problems that ‘were not audible have eaused failures, Extensive investigations have been conducted on surge. Poor quantitative universal of aerodynamic loading capacities of different diffusers and Smpellars, and an inemict knoveledge of boundary-layer hehav- Jor make the exict prediction of flow in turbomachines at the esign stage diflcut. However, i¢ is quite evident that the underlying eause of surge is aerodynamic stall. The stall may secur inether the iipellecar he eifuser. When the impelier seems to be the case of surge, the inducer section is where the flow separation begins. decrease in the mass low rate, an Increase in the rotational speed of the impeller, or both ean ‘cause the compressor to surge. Whether surge is caused by a decrease in flow velocity or am increase in rotational speeds, BLOCK yAlve Position sukae MARGIN “pes! ouTeuy ANALOG, OUTPUT - “068' CONTROL 10 SAMPLES /SECOND Figure 51, “DCS” Control, 10 Saraples!Second. cither the inducer or the dius can stall. Which stalls ist is ‘ificat to determine, but considerable testing has shown that, fora low pressure-ratio compressor, the surge usualy itates in the diffuser sect. For units with single-stage pressure ratios above 5:1, surge ts probably initiated in the inducer ‘The antisurge methods disenssed, as well at most other proven surge prevention schemes, utiliza conventional in- Strumeatation to measure and contol certain parameters (€.& flos, pressure tse, ee.). Compressor and process applications vary 30 much that it would be diffult, i not impossible, to devise a universal standard antisurge contro) sebeae. Fach application must be evaluated in order to determine the re- ‘iced contol functions. This requires not only a knowledge of Instrumentation, but expecially good understanding of the compressor and load characteristics, The advent of the te roprocessor has enabled consolidation of many contol fane- tins into one unt, but ithes ot decreased the amount of basi engineering that must he performed prior to programming A rebable contzal whichis aot inlluenced by surge point predic~ tion errors and other eonditions such as compressor foulingand variations fale weight, temperature, ete.) certainly 1 eseable, and shovld reduce the aplistion engineering re Gqiced for each machine. Efforts to develop such « control Package continue, based on the theory that « measurable parameter exists inthe region near surge. Additional research needed to determine ifthe threat of strge can he detected carly enough 0 facilitate control action in time to actuallymm PROCREDINGS OF THE TWELFTH TURBOMACHINERY SYMPOSIUM provent it. The surge control system enggneer has to take inte secount system valving, response rate, valve selection and ‘ocation, bypass line sizing, and a host of other factors in order to develop a snecessful desigr. ‘Dynarsie simulation of compressor systems is 2 tool used to evaluate process design and system behavior at the design stage. It involves computer besed mathematical modeling. The sionlation procedure can provide a better understanding of how the system will behave during process operation, emergency shutdown, and process startups and shutdowns Dynamic simeiation also peraits an evaluation of catrel hard- ware, Itis a design tool that, if properly used, can avoid costly probleme sith comprossor opotation. APPENDIX: EVALUATING THE ERPECTS OF VARIABLE GAS CONDITIONS “The following shouldbe done to evaluate the effects of gas aviation: 1. Drow compressor pevarmence cures for normal and alternate operating conditions 2, Select an appropiate antisurge contol scheme, and determine the calibration equited for normal opera: 3, Plot control tines ou the compressor perfomance curves for normal and alterate operating conditions Peréormance, curves ate produced by. the compressor Ianuactorer. References 8 and 8 include information on bow calnnation constants ae deteemined. The follwing explains how to ealelte the data nacesry for ploting the cootrol lines. Perforrimee euryrs for norial ene lar, conatias {Figures 31 and 39, and « Flow Delt Peontrol scheme (Fi ‘ve Bi) wal be used as 99 example. The flowing date spol 1. Contos at Compressor Inlet Condition Normal «Startup Py, bavom? 815 819 me su sit Zz. 1.006 1.000) Mwy 537 1260 2 Defiaition of Symbols B, — Pressure at compressor inlet, absolute T, — Temperature at compressor inlet, absolute Z_——-Compressibility factor MW. — Molecular Weight Q = Flow rate, m'Mhr, at inlet conditions hh Head, inches of water column iffontial produced by the flow element S Spa FU — Flow transmitter PDT — Pressure diffevential hansmiteer ASC — Antisorge controller FY — Suraming relay — — Gan, summing relay Input B K — — Bias, summing relay outpat ©! — Constant used In flow equation Subscripts m — Maximum, sp — Setpoint pat — Pressure diforential transmitter 4 — Inlet dd — Discharge 1) — Normal conditions 8. Calibration Valves Gq — 10,000 he Sqas— 10 kgiom* differential 5 0.81 K —0si Calculation Procedure ‘The procedure consists of ealeulating Q vs Py data, for ‘plotting control lines, using equstion AV = yf MBER Fx MW, al) Px MW. c= yee ° a (aa) He royo00, | SES 5 955.24 (311) (1.006) Equation AB is used to caleulate “Wh,” for each discharge pressure pos. we [fone Eaquation A3 is evived by rearranging Bqvation 4. Equation defines the izputloutpat velatiowship of stmming zelay FPY. ‘When the antisurge contaller (ASC) is operating, the output fiom FPY is miatsined eqoal tothe Sctptnt. The setpint, Py and F, are known values, so v:th Equation 4 reananged to Eguation A3, the valuc of "hh cane calulated for Use in Equetion ALA value of to 1 represents fll ange of the ow od pressure rise terms ia Equation AS. The equstsn ts written inthis manner sa thatthe acto value of “hy” 1808 nieeded forthe calculations O-A-gBtK as) where O= Output A= Flow sigal, fe. “hth” B= Pe ~ Pe signal §§— Cain, egal line slope adjustment K= Bias With Equations Aland AS programned into a end held caleulator, che data for plotting @ contol lie ean be calulated very quickly Any’ gis condition ean be readily changed and ¢ control line quickly euloulated 9 evaluate the effect. Likewise, ltr: nate lines for ditirentsefpoints esa be produood, cin Figures Sand 82 Foquation Al is used fr all eleulasions, but Equation A vwvies, depending on the particular contrl scheme andthe valves Pj, Re, ete) On the “Y” axis of the compressor perforan- tnce curve Since Equation Al is used for all eleulation, itis Drograiamed to periorm a eonteal ine sbroutine labeled FCLSR™ Other programs, applieble tthe particular mstni>TUTORIAL SESSION ON PRACTICAL APPROACH TO SURGE AND SURGE CONTROL SYSTEMS 1 ment scheme, ete., are used t» calculate “ih,” and the Subroutine is used fo complete the oaleulation, Table 2 shows the Hewlett-Packard 41CY program that was used to caleulate the control lines in Figures 22 and 23. REFERENCES 1. Bllock, B. O., Wileox, W. W., and Moses J.J, "Epes mental and Theoretical Studies of Surging in Continuous Flow Compressor,” NACA Report No. 861 (1945 2 Emmons, Hl W., Pearson, C. E., and Grant, H. Py “Compressor Surge snd Stall Propadation,” Trans. ASME ‘Say 1955) 3. Jailae, D., “Stall, Stage Stall and Surge,” Proceedings of 10th Annual Turhomachinery Symposium, Tesas A&M University (December 198), 4, Desa, BC. and Young, LR, “The Tine Domata of Centifogal Compressor and Pump Stability end Surge,” 1876 Joint Gax Turbine ée Fluids Enginesring Divisions Conference, ASME (March 22-25, 1976) 5 Falleman, J. “Centrifugal Compressors, Chapter 8. In Advances in Peroleure Chemistry and Refining, Inters fence Publishers (division uf John Wiley & Sons), New YoriLonden 1902). 6 Hallock, D, C., “Centrifugal Compressors. the Cause of the Curve,” Ait and Gas Engineering, 1.1 (1963) 1. Royce, M. P., "A Practical Three-Dimenstonal Flow Visvalization Approach to the Complex Flow Characters= 1s ina Genteftgal lepeller,” ASME Paper No. 66-67. cs 8, White, M. HL, "Surge Conteol for Gentil Gompces- sors," Chemical Engineering (December 23, 1973) 8 Gaston, J, R, "Antivurge Control Schemes for Torbocom prossors,” Cemiea! Enginecring (Apall 19,1082) 10, Brown, R. N,, “An Experimental Investigation of Pacutmaie Close Loop Ant surge Control for Cantifgal anid Axial Flow Compressors.” Masters Thesis-— University of Wisconsin, Modizon, Wisconsin 1965) 11. Stanley, R.A. aad Bohannan, W, Ry “Dyoamie Simla tion of Centritigal Compressor Systems,” Proceedings af the 6th Tuntomactinery Sympostem, Teas ABM Univer> si 097. 12, Eby, R$. and Pobaas, N. E., “Dynamie Simulation and Verification of a Compression-Liquefaction System for Material Withdravm Froma Uranium Bncichment Plant,” Proceodings forthe 1983 Summer Competer Simalation Conference 989), 15, Chit, K. © and Pobana,N. E.,"Applialty of Distitb- tted Control toa Fast Reopoase Loop.” Proceadings ofthe 18a 89 International Conference (1083). ACKNOWLEDGEMENTS “This paper was intended to be a comprehensive tutorial om sunge and sarge control, oreated by incorporating and editing {Inputs rom all contributors into one dactument, WB, Bobs ran contributed a major portion af Section A, along with contributions from other authors, J. R. Gaston contributed heavily to Section B, once ggnin with significant material from the other authars. Scetian C was contribited by N. E. Poban2. ‘The neorall paper ceanilinating and nditing were Aone by MP Boyce and C.B. Moker-Homji, who would lke ro thank all authors for their eonteiimion and te, Thanks ar also dive to Mrs. Bille Burton, Boyee Fngineering International, ne., for typing the final manuscript Table 2. Calculator Program (for HP A1-CV) to Coloulate Control Lines Storage Registers Step No Programm Subroutine 00 SP, MPFR, 6500 & 7000 a Lab, curd LBL, CLsR 01 Sa, KgCMD = 10 e KCL cy RCL 07 02 Fy KCN ABS = 6 19 03, x 5 Qa, MIMR = 10,000 ot RCL a ROL oe OT." = SLL 08 i ; 05 MW. = 5.97 & 1860, 08 RCL 08 RCL or 05 C= 3,953.21 w x x 07 Z = 1.008 & 1.000 % Rel. 0 RCL 6 08g = 0.951 co RCL 08 ; OK — 02 10 i sont u x KCL 06 ry + x B RCL cy BND “ ~ 1B To cLsR 16 END Assign "LBL CPA" to key "LN." To run program, key iv Py, then press “LN” to calculate Q.