First published in The International Journal of Hydrocarbon Engineering, UK, 1997

IntroductionWhether for lube oil, fuel oil, or general fractionation,vacuum columns utilize ejector systems to maintaindesign vacuum levels within the column.Noncondensibles, cracked gases, hydrocarbon vaporsand steam are removed from the column by the ejectorsystem. Extraction of these fluids from the column is key toa proper vacuum level within the column and consequently,design charge rates and specification quality product areachieved.Refiners do have lengthy operating experience with ejectorsystems. Ejector systems have been the mainstay forrefinery vacuum distillation. Whether a crude vacuum toweroperates as a ‘wet’, ‘damp’ or ‘dry’ tower, an ejectorsystem is the vacuum producer. Different tower operatingpressures and overhead load characteristics of wet, dampor dry operation affect only the configuration of an ejectorsystem but the basic operating principle remainsunchanged.Even with lengthy operating experience, refiners viewejector systems with hesitation and uncertainty. Thisuncertainty results from an incomplete understanding ofthe basic operating principles of ejectors themselves and

their interdependency with any vacuum condenser itsupports or to which it discharges. There is only limitedinformation in technical journals or books addressingoperating principles of ejector systems. On a positive note,ejector systems are quite reliable and performanceshortcomings are not a common problem. However, whenoperating problems do occur, they appear as a dramaticchange in performance rather than a gradual loss ofperformance. Vacuum tower crisis is always critical and animmediate remedy is necessary. The purpose of thisarticle is to offer a concise and complete overview ofejector and condenser fundamentals, system operationand troubleshooting.

EjectorsComponent partsIt is important to know the proper nomenclature for internalparts of an ejector before beginning to discuss how anejector works. An ejector is a static piece of equipment withno moving parts (Figure 2). There are four majorcomponents to an ejector, the motive nozzle, motive chest,suction chamber and diffuser.

Ejector system troubleshootingJ. R. Lines and R. T. Smith, Graham Corporation examineejector systems and provide troubleshooting experience

with reference to case studies.Figure 1. Three stage twin element ejector system

Operating principleThe basic operating principle of an ejector is to convertpressure energy of high pressure motive steam into velocity.High velocity steam emitted from a motive nozzle is then usedto work on the suction fluid. This work occurs in the suctionchamber and diffuser inlet. The remaining velocity energy isthen turned back into pressure across the diffuser. In simpleterms, high pressure motive steam is used to increase thepressure of a fluid that is at a pressure well below motivesteam pressure.Thermodynamically, high velocity is achieved through adiabaticexpansion of motive steam across the converging/divergingmotive nozzle from motive pressure to suction fluid operatingpressure. The expansion of the steam across the motive nozzleresults in supersonic velocities at the nozzle exit. Typically,velocity exiting a motive nozzle is in the range of Mach 3 to 4,which is 3000 to 4000 ft/sec. In actuality, motive steam expandsto a pressure below the suction fluid pressure. This creates thedriving force to bring suction fluid into an ejector. High velocitymotive steam entrains and mixes with the suction fluid. Theresulting mixture is still supersonic. As this mixture passesthrough the converging, throat, and diverging sections of adiffuser, high velocity is converted back into pressure. Theconverging section of a diffuser reduces velocity as the cross-sectional area is reduced. The diffuser throat is designed tocreate a normal shock wave. A dramatic increase in pressureoccurs as flow across the shock wave goes from supersonic, tosonic at the shock-wave, to subsonic after the shock wave. In adiffuser diverging section, cross-sectional flow area isincreased and velocity is further reduced and converted topressure.

The performance curve

Ejector manufacturers summarize critical dataon a performance curve. Figure 3 shows aperformance curve for a single stage ejector.On the y-axis of this curve is suction pressurein millimeters of mercury absolute (mm HgA).On the x-axis is the water vapor equivalent load(Ib/hr).Equivalent load is used to express a processstream, which may be made up of manydifferent components, such as air, water vaporand hydrocarbons, in terms of an equivalentamount of water vapor load. Figures 4 and 5,from the Heat Exchange Institute Standards forJet Vacuum Systems, show the curves that areused to convert various molecular weightgases to the appropriate vapor equivalent at areference temperature of 70°F.The performance curve can be used in twoways. First, if the suction pressure is known foran ejector, the equivalent vapor load it handlesmay be determined. Secondly, if the loading toan ejector is known, suction pressure can bedetermined. If field measurements differ froma performance curve, then there may be aproblem with either the process, utilities orejector.

Motive steamMinimum motive steam pressure is importantand is also shown on a performance curve.The manufacturer has designed the system tomaintain stable operation with steampressures at or above a minimum steampressure. If motive steam supply pressure fallsbelow design, then a motive nozzle will passless steam. When this happens, the ejector isnot provided with sufficient energy to compressthe suction fluid to the design dischargepressure. The same problem occurs when thesupply motive steam temperature rises aboveits design value, resulting in increased specific volume,and consequently, less steam passes through the motivenozzle.An ejector may operate unstably if it is not supplied withsufficient energy to allow compression to its designdischarge pressure. Unstable ejector operation ischaracterized by dramatic fluctuations in operatingpressure. If the actual motive steam pressure is belowdesign or its temperature above design, then, within limits,an ejector nozzle can be rebored to a larger diameter. Thelarger nozzle diameter allows more steam to flow throughand expand across the nozzle. This increases the energyavailable for compression. If motive steam supplypressure is more than 20 - 30% above design, then toomuch steam expands across the nozzle. This tends tochoke the diffuser. When this occurs, less suction load ishandled by the ejector and suction pressure tends to rise.If an increase in suction pressure is not desired, thenejector nozzles must be replaced with ones having smallerthroat diameters or the steam pressure corrected.Steam quality is another important performance variable.Wet steam may be damaging to an ejector system.Moisture droplets in motive steam lines are accelerated tohigh velocities and become very erosive. Moisture in motivesteam is noticeable when inspecting ejector nozzles.Rapidly accelerated moisture droplets erode nozzleinternals. They etch a striated pattern on the nozzlediverging section and may actually wear out the nozzlemouth. Also, the inlet diffuser tapers and throat will havesigns of erosion. On larger ejectors, the exhaust elbow atthe ejector discharge can erode completely through.Severe tube impingement in the intercondenser can alsooccur but this is dependent upon ejector orientation. Tosolve wet steam problems, all lines up to the ejectorshould be well insulated. Also, a steam separator with atrap should be installed immediately before an ejectormotive steam inlet connection. In some cases, a steamsuperheater may be required. Wet steam can also causeperformance problems. When water droplets pass

Maximum discharge pressureThe maximum discharge pressure (MDP), also shown on theperformance curve, is the highest discharge pressure that anejector has the ability to achieve with the given amount of motivesteam passing through the steam nozzle. If the dischargepressure exceeds the MDP, the ejector will become unstableand break operation. When this occurs, a dramatic increase insuction pressure is common. As an example, when a systemdesigned to produce 15 mm HgA pressure breaks operation,suction pressure sharply increases to 30 - 50 mm HgA. Thisoften causes a tower upset. Therefore, it is of paramountimportance to make sure ejectors do not exceed their MDP.Since increasing the discharge pressure above the MDP causesa loss of performance, it seems logical that lowering thedischarge pressure below the MDP should have the oppositeaffect. This, however is not the case. Ejectors with acompression ratio, discharge pressure divided by suctionpressure, higher than 2:l are called critical ejectors. Performanceof a critical ejector will not improve if its discharge pressure isreduced. This is primarily due to the presence of the shock wavein the ejector diffuser throat.

CondensersComponent partsCondensers are manufactured in three basic configurations:fixed tubesheet, U-tube or floating head bundle.Thermodynamically, these units perform identically. They differonly in ease of maintenance and capital cost. The fixedtubesheet unit, typically TEMA, AEM, BEM, AXM or BXM styles, hasa bundle that is not removable from the shell. This unit isgenerally the least expensive to build. The major disadvantage ofthis type of unit is that the shellside of the condenser is notaccessible for normal cleaning methods. The U-tube exchanger,TEMA, AEU or BEU, is the next most economical type ofconstruction for a removable bundle. Since the bundle iscompletely removable from the shell, it allows thorough cleaningof the shellside as well as the tubeside. The major drawback tothe U-tube unit is that the U-bend section of the tube can make

through an ejector nozzle, they decrease theenergy available for compression. Furthermore,water droplets may vaporize within an ejector astemperature increases. Vaporized water dropletsact as an additional load that the motive steammust entrain and compress. The effect is adecrease in load handling ability. With extremelywet steam, the ejector may even becomeunstable.

difficult cleaning of tube internal surfaces. Floating headunits, TEMA type AES, AET, AXS or AXT, are generally themost expensive. The floating head adds complexity andmaterial to the return end of the condenser. These units areadvantageous because they allow complete access forcleaning of both the shellside and the tubeside. Figure 6indicates typical TEMA nomenclature for condenser designs.

Operating principleThe primary purpose of a condenser in an ejector system isto reduce the amount of load that a downstream ejectormust handle. This greatly improves the efficiency of theentire system. Often condensers are analyzed like shell andtube heat exchangers which are common throughoutrefineries. Although vacuum condensers are constructedlike these exchangers, their internal design differssignificantly due to the presence of two phase flow andvacuum operation.Vacuum condensers for crude tower applications generallyhave the cooling water running through the tubes. Thecondensing of the water vapor and hydrocarbons takesplace on the shellside. Generally, the inlet stream entersthrough the top of the condenser. Once the inlet streamenters the shell, it spreads out along the shell andpenetrates the tube bundle. A major portion of thecondensibles contained in the inlet stream will change

phase from vapor to liquid. The liquid falls bygravity and runs out of the bottom of thecondenser and down the tail leg. Theremainder of the condensibles and thenoncondensibles are then collected andremoved from the condenser through thevapor outlet.Vapor is removed from the condenser in twoways. In larger units, approximately 30 in. indiameter and larger, a long air baffle isused. The long air baffle runs virtually the fulllength of the shell and is sealed to the shellto prevent bypassing of the inlet streamdirectly to the vapor outlet (Figure 7). Thisforces the vapors to go through the entirebundle before they can exit at the vaporoutlet.Similarly, smaller units use an up and overbaffle arrangement to maximize vapordistribution in the bundle. In thisconfiguration, the exiting vapor leaves thecondenser on one end only. The vapors areforced through a series of baffles in order toreach the vapor outlet. Figure 8 illustrates atypical AEM cross-sectional drawing.Both the long air baffle and the up and overbaffles are normally located in the coldestcooling water pass in order to guaranteecounter current flow, and cooling of vaporsand noncondensibles below exiting watertemperature and optimal heat transfer.As mentioned previously, a condenser isdesigned to limit the load to the downstreamejector. In many cases, the load to acondenser is ten times the load to theejector. Consequently any loss in condenserperformance will have a dramatic affect on

the downstream ejectors. This makes the performance of ejectorsvery dependent on the upstream condensers.The first intercondenser is the largest and most critical condenserfrom a design and operation standpoint. The pressure that the firstintercondenser is designed to operate at is directly related to themaximum cooling water temperature for which the system isdesigned. The pressure inside the condenser must be high enoughfor condensation to occur. For instance, with 91 °F cooling water,an initial condensing temperature of approximately 115 °F isreasonable. This corresponds to a first stage intercondenseroperating pressure of 76 mm Hg.The equation for design of a vacuum condenser is the classic heattransfer relationship:

Q = U x A x LMTDwhere:

Q= Amount of heat transfer required (btu/hr)U = Overall heat transfer rate (btu/hr ft2 °F)A= Surface area of the condenser (ft2)LMTD = Log mean temperature difference (°F)

During the design phase, all of these variables are fixed. Q is fixedby the amount of steam being used by the upstream ejector andthe amount of load coming over from the tower. The amount ofsteam that an ejector uses is directly related to the compressionratio. Therefore, a high design cooling water temperature results ina high minimum first intercondenser pressure which results in ahigh steam usage for the first stage ejector.The heat transfer rate is a function of cooling water flow, processside condensing characteristics and tube material. Normally theheat transfer rate is determined for the tubeside and shellsideseparately and then combined into an overall heat transfer rate.The overall heat transfer rate is then used in the above equation tocalculate the required surface area.The surface area is set by the number of tubes in the condenser.The tubes in most crude vacuum system condensers are 3/4 in.diameter tubes and the surface area is calculated based on theexternal surface area of the tube.The LMTD is a thermodynamic quantity that is used to calculate theamount of heat that is given up. The LMTD is set by the coolingwater inlet temperature, cooling water temperature rise and theshellside inlet and outlet temperatures.

Cooling waterWhen cooling water supply temperature rises above its designvalue, ejector system performance is penalized. A rise in coolingwater inlet temperature decreases condenser available LMTD.When this occurs, the condenser will not condense enough andmore vapors are carried out as saturated vapors with thenoncondensible gases. As discussed in the preceding ejectorsection, this increased load to a downstream ejector cannot behandled by that ejector.Similarly, if cooling water flow rate falls below design values, agreater temperature rise across the condenser occurs. Even ifcooling water is at its design inlet temperature, a greatertemperature rise reduces available LMTD. Condensation efficiencyis reduced and additional load is passed on to a downstreamejector. Losses in cooling water flow occur over time as moreprocess equipment is added to a cooling water loop or systempressure drop rises and reduces capacity of cooling waterpumps. Furthermore, reduction in cooling water flow lowers theheat transfer rate.Lower than design inlet cooling water temperature does not havea negative affect. Actually it often removes system performanceproblems. Typically summer months place the greatest strain on anejector system. It is at this time that cooling water is warmest anddemands on the cooling tower are the greatest. During wintermonths, the lower inlet cooling water temperature increases thesafety margin for condenser operation as LMTD is greater than thedesign value.

FoulingIntercondensers and aftercondensers are subject to fouling like allother refinery heat exchangers. This may occur on the tubeside,shellside or both. Fouling deters heat transfer and, at some point,may compromise system performance.Cooling tower water on the tubeside is prone to biological foulingor fouling due to corrosion products. Vacuum condensers arealways designed to include a margin for fouling. Over time,however, fouling deposits continue to accumulate and exceed thedesign value. When this occurs, condensation within thecondenser is reduced. A good rule of thumb for tubeside fouling inthe condenser is if you are unable to see the tube material, thenthe tubes are fouled.On the shellside, hydrocarbon vapors, steam andnoncondensibles are handled. Depending upon tower fractionationand the type of crude handled, a hydrocarbon film may develop ontube external surfaces. Also, during tower upsets, hydrocarbonliquids are carried over from the tower. During this type of upset itis common for hydrocarbons to bake on to external tubingsurfaces. This hydrocarbon film on external tube surfacesreduces condensation efficiency and results in carryover ofadditional vapors to a downstream ejector.Routine refinery procedures should include periodic cleaning ofcondenser bundles. These procedures must include a provisionfor cleaning both tube and shell-sides. A noticeable impact offouling is increased cooling water pressure drop across thecondenser or an increase in process side operating pressure. Forease of shellside cleaning a removable bundle should be used,TEMA, AXS or AXT.

Steel tubingWhile steel tubing may be compatible with process vapors,noncondensibles and cooling water, periods of extendedshutdown for routine maintenance, revamp, or even startup are aconcern. It is during this period that steel tubing is exposed to airand moisture. This permits rust to develop and form a scalebuildup. When the process is eventually started, the condensersmay be severely fouled. Experience has shown that on occasionthe fouling is so severe that the operation of the ejector system iswell below design values. Modest savings in initial investment arequickly lost to reduced unit charge rates and/or product quality. Itis for this reason that vacuum system manufacturers often cautionagainst the use of steel tubing and suggest a nonferrous orstainless material.

Rating programsComplexity of vacuum condenser design is of critical importance.Thus proprietary designs are developed and offered by vacuumequipment manufacturers. These proprietary designs musteffectively manage heat transfer requirements and at the sametime, be of proper internal configuration so as to minimize pressuredrop. Another important aspect of design and internalconfiguration deals with assuring adequate noncondensibleremoval and eliminating the potential of noncondensible blanketingor pockets.The proprietary design discussed here, has evolved and wasdeveloped from research, as well as ongoing evaluation andperformance monitoring of condensers during operation. Avacuum system is very unforgiving to poorly designed condenserswhich will have a dramatic negative effect on vacuum levelsmaintained and fractionation achieved by the distillation tower.

Proprietary design procedures incorporate the followingconsiderations:• Condenser vapor inlet location and distribution area

above the tube field so as to insure proper vapor entry tothe shell and penetration into the tube field.

• Tube field layout and penetration areas to guarantee thatflow distribution into the bundle is well maintained andpressure drop is held to a minimum.

• Noncondensible gas cooling section, where bulkcondensate is separated from the vapor and finalcooling to design saturation temperature is achieved.

• Bulk condensate and noncondensibles exit the shell atdifferent locations and temperatures. In this way,noncondensibles and vapors are cooled below thecondensate temperature to maximize condensationefficiency without contending with excessive condensateloading and associated thermal duty.

• Support plate spacing and bundle penetration areas toinsure velocities are well below those necessary toestablish vibration.

• Process vapors assessed to properly ascertainvapor/liquid equilibrium (VLE) conditions throughout thecondensing regime.

• Condensing profile broken down into as many as fiftysteps to properly determine the effective LMTD and VLEat each step.

Often proprietary designs are compared to thosedetermined by computer programs available frominstitutional organizations, research companies orsoftware companies. These generic programs do notproperly model flow configurations typical of vacuumcondensers. A number of organizations put forth excellentsoftware to reliably predict performance of process heattransfer equipment, however, that same software shouldnot be applied to exchangers designed for vacuumcondensation. The software is unable to model internalconfigurations typical of vacuum condensers and theytypically force condensate and noncondensibles to exit thesame connection and be at the same temperature.

The ejector systemType of towerAs mentioned above, typical operating modes for a vacuumtower are classified as wet, damp or dry.Wet towers have overhead loading characterized bysubstantial amounts of stripping steam plus typicalamounts of coil steam to the fired heater. Operatingpressure for a wet tower has a range of 50 - 65 mm HgAbs at the tower top and a flash zone pressure ofapproximately 65 - 75 mm Hg Abs. With such moderatevacuum levels, often it is possible to have a precondenserbetween the vacuum tower and a two stage ejector system.The precondenser reduces loading to the ejector systemby condensing substantial amounts of steam andhydrocarbon vapors, thereby reducing energy demands tooperate the ejector system.

• A damp tower operates typically in the range of 15-25mm Hg Abs at the tower top, with flash zone pressure ofapproximately 35 mm Hg Abs. Stripping steam isappreciably reduced and the ejector system is a threestage system.

• Dry towers operate between 5-l5 mm Hg Abs at thetower top, flash zone pressure at 20 mm Hg Abs, and donot utilize stripping steam. Here again, it is customary toutilize 3 stage ejectors. It is not possible to operate atthese pressures and utilize a precondenser. Theoperating pressure is below a level where cooling wateris cold enough to induce condensation. There are casesof deep-cut operation where the pressure may be below5 mm Hg Abs and a 4 stage ejector system is used.Here two ejector stages are in series ahead of the firstintercondenser (Figure 9).

Ejectors/condensersFrom the figures referenced above, it is understood thatejectors and condensers are staged in series with eachother. Process vapors and noncondensibles flow in seriesfrom the tower to an ejector, then to an intercondenser,followed by another ejector, then to an intercondenser, etc.The purpose of an ejector is to entrain tower overheadvapors and noncondensibles, and then compress them toa higher pressure. Ultimately, via a series of stagedejectors, process fluids are brought to a pressureequivalent to atmospheric pressure or greater. Forexample, a vacuum tower is maintained at 10 mm Hg:• 1st stage ejector compresses process fluid from 10 - 80

mm Hg.• 2nd stage ejector compresses from 80 - 250 mm Hg.• 3rd stage ejector compresses from 250 - 800 mm Hg.The purpose of intercondensers, as mentioned previously,is to be positioned between ejector stages to condense asmuch steam and hydrocarbons as possible. Bycondensing steam and hydrocarbon vapors, the loadhandled by a downstream ejector is reduced. Thismaintains energy usage (motive steam consumption) fordriving the ejectors, to a minimum.

Process conditionsThese are very important for reliable vacuum systemoperation. Process conditions used in the design stageare rarely experienced during operation. Vacuum systemperformance may be affected by the following processvariables, which may act independently or concurrently:• Noncondensible loading. Vacuum systems are

susceptible to poor performance when noncondensibleloading increases above design. Noncondensibleloading to a vacuum system consists of air leaking intothe system, lightened hydrocarbons, and cracked gasesfrom the fired heater. The impact of higher than designnoncondensible loading is severe. As non-condensingloading increases, the amount of saturated vaporsdischarging from the condenser increases. The ejectorfollowing a condenser may not handle increasedloading at the condenser design operating pressure.The ejector before the condenser is not designed for ahigher discharge pressure. This discontinuity inpressure causes the first ejector to break operation.When this occurs, the system will operate unstably andtower pressure may rapidly rise above design values.

• Noncondensible loadings must be accurately stated. Ifnot, any vacuum system will suffer performanceshortcomings. If noncondensible loadings areconsistently above design, then new ejectors arerequired. New condensers may be required dependingon severity.

• Condensible hydrocarbons. Tower overhead loadingconsists of steam, condensible hydrocarbons andnoncondensibles. As different crude oils are processedor refinery operations change, the composition andamount of condensible hydrocarbons handled by thevacuum system vary. A situation may occur where thecondensible hydrocarbon loadings are so different fromdesign that condenser or ejector performance isadversely affected. This may occur in a couple ofdifferent ways. If the condensing profile is such thatcondensible hydrocarbons are not condensed as theywere designed to, then the amount of vapor leaving thecondenser increases. Ejectors may not tolerate thissituation, resulting in unstable operation. Anotherpossible effect of increased condensible hydrocarbonloading is an increased oil film on the tubes. Thisreduces the heat transfer coefficient. Again, it may resultin increased vapor and gas discharge from thecondenser. Unstable operation of the entire system mayalso result. To remedy performance shortcomings, newcondensers or ejectors may be necessary.

Tower overhead loading. In general, a vacuum system willtrack tower overhead loading as long as noncondensibleloading does not increase above design. Tower toppressure follows the performance curve of the first-stageejector. Figure 3 shows a typical performance curve. At lighttower overhead loads, the vacuum system will pull towertop operating pressure down below design. This mayadversely affect tower operating dynamics and pressurecontrol may be necessary. Tower pressure control ispossible with multiple element trains. At reduced overheadloading, one or more parallel elements may be shut off.This reduces handling capacity, permitting tower pressureto rise to a satisfactory level. If multiple trains are not used,

recycle control is another possible solution. Here, thedischarge of an ejector is recycled to the system suction.This acts as an artificial load, driving the suctionpressure up. With a multiple-stage ejector system,recycle control should be configured to recycle the loadfrom before the first intercondenser back to systemsuction (Figure 10). This way, noncondensible loadingis not allowed to accumulate and negatively impactdownstream ejectors.

• System back pressure. Vacuum system back pressuremay have an overwhelming influence on unsatisfactoryperformance. Ejectors are designed to compress to adesign discharge pressure (MDP). If the actualdischarge pressure rises above design, the ejectors willnot have enough energy to reach the higher pressure.When this occurs, the ejector breaks operation andthere is a sharp increase in suction pressure. Whenback pressure is above design, possible correctiveactions are to lower the system back pressure, reborethe steam nozzle to permit the use of more motivesteam or install a completely new ejector.

InstallationSufficient clearance should be provided to permit removalof the motive chest which contains the motive nozzle whichprotrudes into the suction chamber. The ejector may beinstalled in any desired position. If the ejector is pointedvertically upward, a drain must be present in the motivechest or in the suction piping to drain any accumulatedliquid. This liquid will act as load until it is flashed off,giving a false performance indication. The liquid could alsofreeze and cause damage. The motive line size shouldcorrespond to the motive inlet size. Oversized lines willreduce the motive velocity and cause condensation.Undersized lines will result in excessive line pressure dropand, thus, potential low pressure motive to nozzle. Themotive fluid lines should be insulated.The suction and discharge piping should match or belarger than that of the equipment. A smaller size pipe willresult in pressure drop possibly causing a malfunction orreduction in performance. A larger pipe size may berequired depending on the length of run and fittingspresent. Appropriate line loss calculations should bechecked. The piping should be designed so that there areno loads (forces and moments) present that may causedamage. Flexible connections or expansion joints shouldbe used if there is any doubt in the load transmitted to thesuction and discharge flanges. If the system vent isdesigned to exhaust to a hotwell, the pipe should besubmerged to a maximum of 12 in. If the discharge

exhausts to atmosphere, the sound pressure level shouldbe checked for meeting OSHA standards, paragraph 1910.95and Table G-12 and/or the local standards.A thermostatic type condensate trap should be avoided sincethey have a tendency to cause a surge or loss of steampressure when they initially open. This could cause theejector to become unstable.

OperationStart-upThe ejector motive line should be disconnected as near aspossible to the motive inlet and the lines blown clear. This isextremely important on new installations where weld slagand chips may be present and scale particles could exist.These particles could easily plug the motive nozzle throats. Ifa strainer, separator, and/or trap is present they should beinspected and cleaned after the lines are blown clear. Thevapor outlet of the aftercondenser and condensate outletsshould be open and free of obstructions and the coolingmedium should be flowing to the condenser(s).All suction and discharge isolating valves, if present, shouldbe opened. If the unit has dual elements with condenserspresent, ensure the condenser is designed for bothelements operating. If the condenser has been designed forone element operating, the suction and discharge valvesshould be opened to only one element (the other elementbeing isolated).The motive valve to the last ejector stage (‘Z’ stage) shouldthen be fully opened. For optimum performance during anevacuation cycle the motive valves should always be openedstarting with the ‘Z’ stage and proceeding to the ‘Y’, ‘X’, etc.stages. If a pressure gauge is present near the motive inlet,the reading should be taken to ensure the operatingpressure is at or slightly above that for which the unit isdesigned. The motive pressure gauge should be protectedwith a pigtail to insure protection of the internal working partsof the gauge. The design operating pressure is stamped onthe ejector nameplate.

ShutdownThere are two procedures to be considered when shuttingdown: method A is appropriate if it is desired to maintain thevacuum upstream of the first stage ejector (an isolating valvehas to be present at suction) rather than allow pressure torise to atmospheric pressure, in which case the valvesshould be closed in the following order:• Close 1st stage suction valve.• Close 1st stage motive inlet valve.• Close 2nd stage suction valve.• Close 1st stage discharge valve.• Close second stage motive inlet valve.• Close 2nd stage discharge valve (if present).If there are more than two stages, then the second stagemotive inlet valve should be closed on all ejectors before thesecond stage discharge valve is closed. If the systemcontains an isolating valve at the first stage suction only, theprocedure would be to close this valve and then either shutoff the motive to all ejectors at once or shut them off by stagesstarting at the first stage. When all the motive valves havebeen shut off, the cooling medium may be turned off. If theunit is going to be shut down for a short period of time toservice the ejectors or for some other reason, it is notnecessary to shut off the cooling medium. Energy savings

should be considered when making this decision. If the unit isgoing to be down and freezing of the cooling medium ispossible, then measures must be taken to prevent freezing orthe unit drained as much as possible to prevent damage.Allowing a small amount of coolant to continuously flow willusually prevent freezing.Method B is employed if it is not required to maintain a vacuumupstream of the first stage ejector and the valves should beclosed in the following order:• Close motive valve to all ejectors or close the motive

valve(s) to each individual stage starting at first stage andcontinue on to second, etc.

• The cooling medium may be turned off as explained in thepreceding paragraphs.

Switching ejector elementsShould it become necessary or desirable to shift from one twostage element to another while the unit is in operation, thenthe procedure is as follows:• The standby Z stage ejector discharge valve (if provided)

should be opened.• The Z stage motive valve should then be opened.• The Z stage suction valve should then be opened. When

this has been accomplished, this standby Z stage ejectorbegins to take suction from the intercondenser along withthe other Z stage element.

• The Y stage discharge valve on the standby element shouldthen be opened.

• This is to be followed by opening the Y stage motive valve.• The Y stage suction valve should then be opened. At this

point both two-stage elements are in parallel operation. Theprocedure then continues as normal. The operatingelement can now be secured by closing the valves asfollows:

• Close 1st stage suction valve.• Close 1st stage motive valve.• Close 2nd stage suction valve.• Close 1st stage discharge valve.• Close 2nd stage motive valve.

• Close 2nd stage discharge valve (ifprovided).

Again the sequence then continues asnormal.

Operating surveyThe goal here is to introduce a systematicway to troubleshoot a crude vacuum system.The first task is to review design data andthen go out into the field and take data. Thisleads to the most important part of vacuumsystem troubleshooting: how and what datashould be taken.Figure 11 shows the appropriate test pointsfor a three stage crude vacuum system. Thefollowing test points are mandatory for propersystem troubleshooting:• Suction and discharge pressure on each

ejector.• Motive steam pressure at each ejector.• Cooling water inlet and outlet pressures for

all condensers.• Cooling water inlet and outlet temperatures

for all condensers.It is essential that all of these readings areaccurate. The most common cause ofmisdiagnosing vacuum system problems isinaccurate or inconsistent measurements.

For this reason, certain guidelines must be followed.Accurate suction and discharge pressures at eachejector are the most important and most difficultreadings to take.All ejector suction and discharge pressures, except forthe last stage discharge pressure, will be in the rangefrom I - 400 mm HgA. Measuring pressure in this rangerequires a high accuracy absolute pressure gauge.Wallace & Tiernan absolute pressure gauges arecommonly used. This gauge should not be permanentlymounted to the system. It should be kept in a lab until itis needed. All absolute pressure measurement devicesare delicate and prone to being knocked out ofcalibration by process vapors and liquids. A commoncompound pressure gauge with a range of 30 in.HgV/0/30 psig is often used by refinery personnel to takethese measurements. This type of gauge is simply notaccurate enough to yield useful vacuum measurements.The motive steam pressure and cooling water inlet andoutlet pressures should be measured with a properlyranged and calibrated pressure gauge. The coolingwater temperatures should be taken with a bi-metallicthermometer using thermowells. All of the vacuum,motive, steam, cooling water pressure and temperaturemeasurements should be taken with one instrument.For instance, the steam pressure measurement shouldbe taken at the first stage ejector. The same gaugeshould then be physically moved to the second stageejector and then to the third stage ejector. Thiseliminates any possible difference in gauges caused bywear, over pressurization, shock, etc. Quite often, smallball valves are permanently added to the equipment tofacilitate this type of testing.

Table 2 is a compilation of design andtest data taken for the three stagecrude system shown in Figure 11. Thecolumn marked ‘Design’ shows thedesign values for all the test points.The design suction, discharge andmotive pressures, P1-9, are all takenfrom the system performance curveshown in Figure 12. The ejectordischarge pressures are calculatedfrom the curve assuming a maximumpressure drop of approximately 5%across each condenser. The designvalues for condenser inlet and outletcooling water temperature and coolingwater pressure drop, ∆p, are obtainedfrom the manufacturer’s condenserdata sheets. As shown, there are nodesign values given for the coolingwater inlet and outlet pressures. Fordesign and troubleshooting the only

Measurement data can then be compared to the designdata. This is done using the system performance curveand data sheets. It is often very helpful to be able tocompare new data to baseline data taken when the systemwas operating correctly

important number is the pressure loss across thecondenser, not the actual pressure.

Case studies 1 to 4 represent examples of different types

of common performance problems. In each case, adifferent problem was found with the equipment. Aftereach case has been dicussed, there will be an additionalsection on how mechanical failures can also contribute tothe symptons shown.

Case study 1:fouled condenserThe most common performance problem with steamejector systems is lower than design steam pressure. Forthis reason, motive steam pressure is always the first data

steam pressure is always the first data that should beexamined. In this case, the motive steam pressures ateach ejector, P7-9, are all above design and should notpose any performance problems. Next, the ejector suctionand discharge pressures are examined, starting with thethird stage ejector. The process begins with the last stagebecause if that is not working, then the other stages will notwork either.Here, the third stage discharge pressure, P6, and thirdstage suction pressure, P5, are both below design. Thus,the third stage ejector is operating correctly and its loadmust be within design limits. Since the third stage ejectorload is within design limits, the second intercondensermust be working properly. Next, the second stage ejectordischarge pressure, P4, is examined. It is also belowdesign, indicating an acceptable shellside ∆P of 3.5%.Remember, pressure drop across a vacuum condensershould be less than 5% of its operating pressure.Moving to the second stage ejector suction, P3, thesystem’s problems begin to show up. P3 is 13 mm Hghigher than design. It is not possible for the first stageejector to compress its load to 96 mm Hg Abs, 13 mm Hggreater than the 83 mm Hg Abs design, and still maintain asuction pressure of 20 mm Hg Abs. The higher thandesign first stage discharge pressure is causing the firststage ejector to break operation. The logical cause of thehigh second stage ejector suction pressure is a fouled firstintercondenser. To confirm this, the cooling water data isexamined.The cooling water pressure drop on all three condensersis normal, indicating cooling water flow rate isapproximately at design. The cooling water temperaturerise is low across the first intercondenser and high acrossthe second intercondenser. The low temperature changeon the first intercondenser indicates that the cooling wateris not absorbing as much heat as it should and therefore,must be fouled. As previously discussed, a fouledcondenser allows greater vapor carry over to thedownstream ejector. This accounts for the high secondstage ejector suction pressure and high secondintercondenser cooling water temperature rise.Case study 2:excessive noncondensible loadingFollowing the same thought process as case study 1,motive steam pressure is not a problem. The third stageejector discharge pressure is also under design. It isnoted that the third stage ejector suction pressure is higherthan design, measured at 305 mm Hg Abs versus adesign of 277 mm Hg Abs. This appears to affect first andsecond stage ejector performance.Possible causes of an elevated suction pressure arecooling water flow rate below design, cooling water inlet oroutlet temperature greater than design, condenser foulingor higher than design loading to the ejector. Reviewingcooling water data suggests no abnormalities, i.e.pressure drop across each condenser seems acceptableand cooling water temperatures are below design values.With cooling water pressure drop and temperature rise ateach condenser close to design values, fouling may beruled out. The remaining possible cause is an increasedload to the ejector.Common performance problems arise whennoncondensible gas loading exceeds the design value.

Higher non-condensible loading results in increasedloading to downstream ejectors. This is due to a highermass flow rate of noncondensibles plus their associatedvapors of saturation.The elevated pressure at the third stage ejector suctioncauses the second stage to break operation. Again, this isbecause the second stage ejector is unable to compressits load to a pressure greater than 292 mm Hg Abs.Therefore, there is an increase in the suction pressure ofthe second stage as it breaks operation. This, in turn,forces the first stage to break operation and the suctionpressure to the system increases from 20 mm Hg Abs to62 mm Hg Abs.Case study 3:excessive condensible loadingThis case is characterized by a modest loss in lower toppressure. Once again, the steam pressure to each ejectoris satisfactorily above design. The third stage ejectorsuction and discharge pressures are below design. Thesecond stage ejector suction and discharge pressures arealso below normal, as is the first stage ejector dischargepressure. The only pressure that is abnormal is the firststage ejector suction pressure.The cooling water data indicates all three condensers havehigher than design cooling water pressure drops andlower than design temperature rises. This indicates that:the high cooling water pressure drop is an indication ofeither fouling or high cooling water flow rate. The low ∆Tindicates that either the condensers are fouled or that thereis a high cooling water flow rate. The previous analysis ofthe suction pressures of the second and third stageejectors show no signs of fouling, i.e. elevated suctionpressures. The conclusion must be that there is a higherthan design cooling water flow rate to the condensers.Higher cooling water flowrate does not affect ejectorsystem performance. The elevated first stage suctionpressure and tower top pressure must be the result of ahigh condensible load causing the ejector to run out furtherout on its curve.Case study 4:low motive steam pressureUsing the same method as previous case studiesprovides a quick answer to this performance problem. Thesteam pressure on the second stage ejector is belowdesign. As discussed earlier, this will cause the secondstage to break operation. When this second stage ejectorbreaks operation, its suction pressure rises above themaximum discharge pressure of the first stage ejector.This results in broken operation for the first stage ejectorand increased tower top pressure. This situation willcorrect itself if the second stage ejector steam pressure isincreased.Mechanical problemsNow that examples of how process conditions, fouling andutilities will affect system performance have been seen, itneeds to be understood what affect mechanical problemswill have on a system. A common mechanical problem is aloose steam nozzle. When a steam nozzle becomes looseit begins to leak steam across the threads. The leakingsteam then becomes load to the ejector. If the loose nozzleoccurs in the first stage ejector the affect will be anoverloaded first stage ejector. If the leak occurs in the

second or third stage ejector, the data will look similar to afouled condenser.Inspection of ejector internals should be done periodically.Proper cross-sectional area and smooth internal parts areimportant. The ejector manufacturer will provide thediameter of the motive nozzles and diffuser throats. Ifinternal surfaces show signs of erosion or corrosion, or ifthe two key diameters have increased by more than 4%, itmay be necessary to replace the ejector. Product build upwithin an ejector similarly affects performance in anadverse way.

Condenser condensate drain legs function as gravitydrains. The height of the drain leg must be sufficient toovercome the elevation of liquid maintained within thedrain leg due to the pressure differential betweencondensate receiver and the condenser. If the leg is tooshort, the condenser will flood. If the drain leg becomesplugged, the condenser may flood. A flooded condenserperforms poorly and broken ejector operation is a commonresult.

ConclusionEjector systems support vacuum tower operation. Properoperation of an ejector system is important; without it, thevacuum tower performance is not optimal. When towerpressure increases above design operating pressure,flash zone pressure increases proportionally. Theconsequence of higher flash zone pressure is reducedvacuum gas oil yields and increased vacuum resid. Whencharge rates to the tower are less than design, the ejectorsystem will pull the tower to a lower pressure. Lowerpressure in the tower may adversely affect tower hydraulicsand cause flooding. This will affect vacuum gas oil quality.With annual performance evaluations of ejector systems,improved product quality, increased unit throughput orreductions in operating costs can often be realised.