SOLVING THE VIBRATION PROBLEM OF A VERTICAL MULTISTAGE CRYOGENIC PUMP by Ching M. Chang Engineering Associate and Fred W. Braun Manager Union Cart jide Corporation Linde Division “Tonawanda, New York. Ching M. Chang isan Engineering A socate tn the Research and Developmen Department of the Linde Division, Union Carbide Corporatio, in Tonawanda, New York. One of his specialty ares i turbo. sachinery research and development “He recived a PhD. degree Enginger- ing from Technological Universiy Aachen, Aachen, Wet Germany and a MB.A. de- (gree from State University of New York at Brafflo, in Buffaic, NewYork DrChang is alo an Adjunct Full Profesor of Enginering at SUNY.Bugfalo He isa member of ASME and a vegiered profes Sonal engineer inthe State of New York Fred W. Braun is Manager of Mechani- al Equipment Engineering nthe Gas Pro ducts Engineering Department ofthe Linde Division, Uniom Carbide Corporation, Tonawanda, New York. Heis responsible for specifying and troubleshooting mechanical ‘equipment used by the Division. ‘Mr. Brasen received his B.S. degree from ‘he State University of New York at Buffalo, im Buffalo, New York. He is 2 member of ASME and a registered professional eng ner in the State of New York ABSTRACT Recent experience in solving the shaft breakage and seal rub problem of vertial, eight-stage liquid nitrogen pam sta plant site i documented, Several inal Held tents Conducted to diagnose the vibration problem are described {n'deul The pumps modelled asa dua level rotor bearing System, to account for the vibration characterises of the Sippor sructare and pump easing bundle. A subsynchro- nous mode assocated with the eantlever bending of the pump casing bundle was corvetiy predicted by this dual level nalydeal model” This subsynehronoss mode would remain undetecable ifthe pump were modelled asa single level system with rigid pomp cating bundle and inilexible support structure, Aso included in te analysis are pertinent representations for throtle bushings. Wear rings, bearing 85 bushings, and statorimpeller forces, based on empirical ‘models selected from the open literature. The pump model ‘was calibrated with vibration data of the support structure and the pump shaft. A comprehensive parametric study produced four cost-effective solutions, one of which was chosen for implementation, Subsequent field tests confirm- ed the reduction of shaft vibrations predicted by the analysis. ‘The pump continues to operate satisfactorily. ‘The combined experimental and analytical approach out- lined should be of interest to engineers responsible for solving the vibration problems of vertical multistage pumps. INTRODUCTION combined experimental and analytical approach which led to the successful resoltion of the severe vibration problem of a vertical mulssage cryogenic pump a plant Ste is described, The pump design and a number of intl Fld. tests conducted to assess the pump’ vibration problem are pre- sented as background in his cae history! The subsequent fection is devoted tothe description of an analyis in which the pump i modeted as a dual level rotor bearing system, ‘The models for wear rings and statorimpellr forces are dlscusted. The iterate on fluid impeller interactions is briefly reviewed. The analytical model i calibrated against selected vibration measurements ‘Alarge number of design fctors have been vatied to develop a set of couveffective design modiieations. The implementation of one ofthese design solutions produced reduction of shaft vibration, which ssuicent toresolve the porphin oem, Suaequet fd tx valid the magnitude ofthe predicted vibration reduction, Conclusions are presented, Inluding assessment of the general applicability ofthe outlined approach to solve wibra- tn problems of vertical multistage purmp, BACKGROUND “The subject pump is used to pump liquid nitrogen at 88.6 K (~300°F) under typical conditions. The pump operates at a flow rate of 27.18 liters (430 gpm) and a discharge pressure of 8.90 Mpa (565 psia). The unit is an eight-stage Vertical turbine design with a direct mounted 200 hp, 3600 ‘cpm motor The impellers have a specific speed of 1450. A. buffered labyrinth gas seal is employed to prevent liquid from leaking out the pump shaft. The pump is constructed ‘of appropriate stainless steel, Monel and bronze alloys suitable for cryogenic temperatures.86 PROCEEDINGS OF THE FOURTH INTERNATIONAL PUMP SYMPOSIUM ‘The pump and the electric motor is depicted in Figure 1 ‘The pump assembly is attached to a reinforced insulation box by 20.045 m (1.75in) thick mounting plate, The motor is installed onto the plate by a motor pedestal 0.533 m (21.0 in) in length. The motor pedestal is open on two sides to permit —— ‘access toa rigid coupling that connects the motor and pump. shaft. The coupling consists of the motor-hal, spacer, and the pump-half parts, The labyrinth gas seals secured on the top side of the mounting plate directly beneath the motor 1" SUCTION VENT Figure 1. Vertical Multistage Cryogenic Pump. a MANHOLE ° Fv purce Gas tne - BLIND FLANGE 3/4" DRAIN 38" os PLATE + PURGE 5 cas 8 j Cine SUCTION ve : i exe . J ecorsctiance van ee ROLLED SiSPLATE “ye” sis 3/4” DRAIN PATE CONCRETE FOUNDATION Figure 2. Cryogenic Pump Cold Box Design Figure 3. Installation of the Pump,SOLVING THE VIBRATION PROBLEM OF A VERTICAL MULTISTAGE CRYOGENIC PUMP a Figure 5. Motor Pedestal coupling. The eryogenic pump cold boxisshown in Figure 2 ‘The pump assembly ands installations depieed i Figures Band 4. A view of the motor pedestals presented In Figure 5 The upper motor bearing is a combined radial and thrust antfricion bearing, whereas, the lower motor bearing isa Fadia rolling element all bearing. Suffnes coefficients for these motor bearings were available from the vendor “The pump is locaed entirely below the mounting plate, A column pipe serves s the purnpinlet manifold, with the first Stage vengar the bottom most poston. The discharge of the Firat tage is directed by intectal passages to the suction of the second stage. The liquid js pumped pwards in this manner, The section view of the_punnp-motor assembly is presented in Figure 6. A throtde bushing separates the Dump discharge from the pump suction, and i exposed reste difference of 8.748 Mpa (543 pia) there between Graphite bronze bearing bushings are placed between stages and a the inlet of the rst stage, All bearing bushings are lubricated by liquid nitrogen tinder a stage differential pressure, Nitrogen leakage between stages is controled by Small deatances at the impeller and casing wear rings. The detailed arrangement of the pump stages is depicted in Figure 7. The largest unsupported span between the lower motor bearing andthe throttle bushing is 1-39 m (52 in) The span f — MOTOR-HALF SPACER — a ‘COUPLING PUMP-HALF ‘COUPLING casyaintH —/ GAS SEAL * 1 Je Scoome THROTTLE PLATE BUSHING . SUCTION YS piscHarce PUMP ——SHAFT coLuMN ——~ PIPE, Figure 6. Pump Cross-Section, ‘between the throttle bushing and the stage number 8 bear ing bushing is 0876 m (14.8 in). ‘The impellersare spaced at. 4921 (7.5625 in) intervals They are secured to the pump shaft by keys, snap ings and fasteners in the axial direction. The process of installing impellers at room temperature is shown in Figure 8. Under ‘exyogenic operating conditions, the impeller shrinks onto the shaft. The pump assembly cannot be mass balanced at room temperature, due to the long and very flexible sha, along with an impeller-to-shaft clearance fit of 0.089 mm (3 {0 4 mils). The individual impellers are shop balanced to below 36.00 gm-mm (0.05 ozin), The shaft diameter is 0.0865 m (3.436 in). Initially, the pump was insalled and commissioned with- out much difficulty. A significant calendar time elapsed, before the pump was restarted in 1984. After approximately 25 hours of operation, the pump shaft failed at the upper (eighth stage) snap ring positon. The failure surface was badly smeared. While no positive identification of the failure ‘mode could be made, i was hypothesized thatthe failure was likely due to bending or torsional fatigue rather than over load. No damage was made to the bearings or throttle bushing; however, the labyrinth gas seal showed some signs of contac.88 PROCEEDINGS OF THE FOURTH INTERNATIONAL PUMP SYMPOSIUM l Ley eer IMPELLER ue WEAR Hee RING. RQ \ LEY HE casine Rs ‘WEAR RING coLUMN we RC Pasar BEARING BUSHING ~ M ; IMPELLER BOTTOM BEARING BUSHING Figure 7, Enlarged View of Pump Stages. “The unit was rebuilt using a new shaft with new beating bushings of known clearance, and a new labyrinth gas teal The unit was returned to service. On initial startup, the pump shaft wiped the labyrinth gat seal. Velocity” and Eceleraion uaneducers were placed on the motor pedestal and mounting plate nan atempc to understand the pump’ Sibrational medes and frequencies, No instrumentation was placed on the pump end, since twas inaccessible within the Insulation box: Resits indicated that she mounting plate was neatly stationary, while the motors vibraion orbitwas highly tliptcal The major axis ofthe elipcal path coincided with the weak axis of the motor pedestal. The major vibration Frequency coincided with motor speed Figure 8, Installation of Pump Impellers “The motor pedestal was then reworked to make the stiff: ness more or fess uniform in all directions, The motor ws Febalanced and the reworked unit was reinstalled ‘Additonal non-contating edly current proximity probes were installed to monitor the rave motion of te shaft with respect to motor pedestal. The probes were located {1.0127 m (0.5 in) above te labyrinth as sea and placed 80 dlegrees to each other, They were orieted to coincide with the major and minor axes ofthe motor ort A keyphasor wa also inwalled to monitor the phate angle “A calrated hammer was fabricated that could apply a known force of random frequency content to the pump Support structure: The motor wat impact tested numerous times at different positions andthe vibrational response was captured for further evaluation. No frequency response was found that eould have been excited bythe motor running speed or mule of i ‘Upon restart, normal operating conditions were achieved was found that the vibration amplicude of the motor was reduced significantly and the orb was now more neatly Geile. However, the shaft vibration remained excessively High near the abyrinth gas seal and the shaft contacted the seal atone point. As shown in Figur 9, the shaft vibration was in excets of 0.198 mm (7.60 mis) peaktorpeak “The unit was restarted, brought up to full speed normal operation condition and shut down for several met All dita were recorded on magnetic tapes for subsequent data Feduction and evaluation, Jt became rather apparent that there was a serious rotordynamie problem within the pump assembly. Iwas decided to perform a rotordynamic analyse and, atthe same time, disassemble he pump for inspection "The motor and impellers were rebalanced to tighter tolerance. Alleritel dimensions that could affect alignment were reworked to ensure tat the assembly would be prop- rly aligned when reasembled. The rigid coupling wat also feworked to-a tighter tolerance both on algnment and balance. Reassembly awaited results of the rotordynamic ’A maximum peak-to-peak shaft vibration of 0.076 mm (6.0 mil) in the labyrinuh gs seal location was selected as feaonable goal to be achieved by modilying the pump design. Pump mdifeation propomls are to be developed tased on ths rotordynamic analy. Preference i given to TIME A 10/09/84 LINDE eu f AN ©2008 A 10.0 V AMS 261 PROX 00 MSEC 100.0 oe MER 10.0 v RMS ‘BEFORE SHUTDOWN eu ° t PROX-Y| 00 MSEC 700.0 X83 MSEC Y (A) Y (g) 7.64 EU Figure 9. Vibration Signals Indicating Seal Rub.SOLVING THE VIBRATION PROBLEM OF A VERTICAL MULTISTAGE GRYOGENIC PUMP 89 design solutions which require the least amount of changes to the pump assembly, while providing an adequate reduc- tion of the shaft vibration. ANALYSIS ‘The vertical pump is characterized by a very long flexible shaft with a large number of rotating stages, @ pun impeller bundle whichis ree-hanging inside of another free hanging column pipe, and a heavy motor which sits on top, driving the pump shalt. This design cals for arotordynamic analysis which considers the vibrations of the motor pedestal above ground and the pump casing bundle below ground, Preliminary undamped critical speed analyses showed that a subsynchronous mode could not be detected by the analysis, ifthe pump were modelled asa single level system, assuming both the support structure and pump casing bundle Being maintained rigidly in space. The resulting mode shape computation confirmed that this subsynchro- nous mode is due to the cantilever bending of the pump casing bundle. A. comprehensive stability analysis showed further that modelling the pump as a single level systems resulted in overestimating the pimps stability margin. ‘The pump is thus modelled as a dual level rotor bearing system, Pertinent models for throttle bushings, wear rings, bearing bushings and stator impeller forces are incor” porated into the analysis. Since the primary concern was the excessive shaft vibration at the labyrinth gas seal location, the ‘emphasis ofthe analysis was placed on studying the pumps forced response characteristics. Although the method of modelling a turbomachine as a dual level system is not news its application to a vertical, multistage pump has probably not been attempted. Gunter, etal [1], performed stability analyses fora vertical 124stage high pressure injection pump of double case con strudion. Various bearing designs were studied to improve the pumps stability margin during startup, when there were insufficient axial pressure drops across the seals and wear rings to stabilize the system. The pump was modelled as a single level rotor bearing system, without taking into account the support structure vibration, An adjustment was made, however, tothe lower motor bearing stiffness to account fora flexible bearing retainer cartidge. The pump’ impeller bundle is pivoted in a spigot fit connected to the outside casing. Operating conditions were shown to strongly affect the sufness and damping characteristics of seals, wear rings and throttle bushings. No forced response analyses were carried out, Gopalakrishnan et.al. [2,3] also stressed the importance of modelling seals, wear rings and statonimpeller forces properly in defining the crtial speeds of pumps. A differ- enation is made between dry and wet critical speeds, which are related to the absence and presence, respectively of the so-called Lomakin effect in wear rings, seals andlor bush- ings Bossman [4], published the ces results ofa vertical water pump and concluded that: + Hydraulic effects in pumps appear to be several orders, ‘of magnitude higher dian those due to imbalance and alignment, + Enlargement of clearance causes the pump to become less stable-and to exhibit higher vibrations. + Only when located near the impellers can a proximity probe register the subsynchronous vibration of the pump. His results supported the finding that it is important to properly model the wear rings, seals and bushings under pertinent operating conditions Flows in wear rings and bushings under axial pressure gradients have been the subject of many investigations inthe literature, for example, Black [5], Childs [6], and Allaire, et al.[7], and Kim and Childs 8]. Important test data on stator- impelier forces were published by Jerry, etal. [9, 10], and Chamieh, et al- 1111 More recently, Gopalakrishnan [12], reviewed the data for wear rings, bushings and stator-impeller forces available in the literature. For a horizontal multistage feed pump, he found that the effect of wear rings on the pumps stability and critical speeds is much more significant than that die statorimpeller forces. The wear ring effects tend to sup- press critical speeds and postpone instability, whereas those Of statorimpeller tend to lower the critical speeds as well as the instability onset speed. No results were presented for vertical pumps. Many experimental methods were described in open literature to resolve vibration problems of specific pumps, such as Starr [13], Jones [14], Makay [15], Meyer [16], Sosa, et al. (17), Marenco, tal. [18], Goldman {19}, Mondy, et al {20}, and Mondy (21), and Smith, et al. (22). Most of these methods, while successfully applied to the pumps studied, are not generally applicable to pumps of different designs, particularly the vertcal pump described herein, ANALYTICAL MODEL Unlike the pump studied by Gunter etal. {1}, wherein its casing bundles pivoted ata spigot fit of ts outer cating, the Casing bundle ofthis pump i unsupported atts lower end. Both the easing bundle andthe cyldrical column pipe are freelanging, 8 the entire motor pump unit is Ngidly Supported oniyatthe mounting plate ecation (igures and 2)" Becruse of thie specfe pump support design, it was elected tomodel the pump motor Suppor structure seve 1 rotor and the rotaing assembly a evel rotor ina dual level rotor bearing system model “The imbalance of individual coupling pieces has been measured in the shop. ‘The resuling imbalance of the assembled coupling, on the other hand, is difficult to es tate accurately. In this analysis, itis being treated as an Unknown parameter which hasbeen systematically varied to Callrate the analytical model, The analytical model s re garded as propery calrated, if the field measured vbra- tons atthe folowing three tations sre matched simultane ‘ously by the models predic the absolute support struc ture vibrations a the motortop and motorbatom stations, and the relative vibration ofthe shaft with expect motor pedestal atthe shaft probe location, 12°70 mi (05 in) above the labyrinth gas seal sation “The Following additional measurements have been incor- porated into the analytical mode: + Mass and support siffnes of the motor pedestal + Mass and imbalance of impellers and motor rotor. * Diametral earances in impeller wear rng, casing wear ring and bearing Musing ofeach pump stage measured at room temperatures, adjusted for expected shrinkage atthe Cryogenic operation temperatures The shrinkage adjst- tment was made based on reals of sample ests conducted in the laboratory. “A 4Station mode was established forthe motor pump unit, including 38 bearings consisting of two motor bearings, one trol bashing, ase of impeller wear ring casing wear fing, and bearing bushing for each of the eght impeller Sages and the Bottom beating bushing “Fre following models for smooth wear rings (also bush- ings) and statorimpeller forces are adopted from {13} fore PROCEEDINGS OF THE FOURTH INTERNATIONAL PUMP SYMPOSIUM use in this study. The smooth wear ring coefficients are Tunctions of the pressure drop across the rings: Kee=Ky =m 0? oy) Cu=Oy=2€0 ® Kyo —Kyneot ° Gy ® where AL. m= 2p r20*8) apg, a RL ge +E) 6 pup lite 3H aL pe 2¢ Ose og! © with &=GyC*= Damping coefficienveritical damping coefficient = Fluid density = Speed of rotation = Shaft Diameter = Length of Wear Rings .P = Pressure drop ‘= Gravitational Constant ‘= Friction Factor & =Entrance loss coefficeent H =Radial clearance ra prose ‘The short circuit factor 6 is defined as ° uF Equations 7-10 set — (GUIDE). Kees — my @®=Kyy o Ko=m! of —Ky. ° my= Kaa 1 p ra” bo (9) my! =KYyy a1 p 12? by aa) where Ka, KYyy=Non-dimensionalstifnes coefficients, dependent on sttorimpeler geometry and flow factor. pene Iunpele ip radius. “Tip width at impeller ex n be Fora given geometry, this model indicates thatthe stiffness coefficients for the statorimpeller effects vary with the fuid density and speed only: Based on test data published by Chamich et al. [11], the following values are selected for use in our model: Krya2o Kr =09 a The pertinent stiffness and damping coefficients for wear rings, throttle and bearing bushings and statorimpeller forces for the present pump are summarized in Table 1 Table 1. Summary of Stiffness and Damping Coeficiente ono orig ectticont, ‘The statorimpeller forces although assumed to act atthe impellers centroid, are modelled by distributing the perti- nent stiffness coefficients to the impeller casing wear rings of the same stage, equally Sic ine jor eine fing wn ibeaaal moda ist properly predict the unbalance responses ofthe pump. shaft especially at the labyrinth gas seal (station number 13) ‘where seal rub occurred; itis very essential that the model be Calibrated using avallbie test data It is alo important to assess the sensitivity of the model to some ofthe less certain, design factors, such a8 the actual clearances of wear rings and bearing bushings. The decision was made to use the RESP2V3 program deveoped by the University of Virginia ROMAG Industrial Consortium [21]. For use in this study, the original version of this dual level response analysis program was extended to handle up to 80 bearings RESULTS: ‘The baseline vibration characteristics of the pump, as predicied by the calibrated analytical model, are shown in Figure 19. Both the absolute vibration of the support struc- ture and the relative vibration of the pump shaft are approximatsly illustrated by straight line segments. TheSOLVING THE VIBRATION PROBLEM OF A VERTICAL MULTISTAGE CRYOGENIC PUMP 91 SERIES BUSHING DESIGN RADIAL CLEARANCE: 0.115 Mm THROTTLE BUSHING DESIGN RADIAL CLEARANCE: 0.156 MM BASELINE VIBRATION OF PUMP SHAFT RELATIVE TO CASING/STRUCTURE BASELINE ABSOLUTE VIBRATION OF SUPPORT STRUCTURE. “ - 261082 ©6103 461022 461012 4 UNBALANCE RESPONSE (MM, O:P) LEGEND 1 MoToR Tor 24 STH BUSHING 7 MOTOR BOTTOM 2g STH BUSHING 12 SHAFT PROBE B aTH BUSHING 33 MOUNTING PLATE 36 3RD BUSHING 18 THROTTLE @USHING 40 2ND BUSHING te STH BUSHING 44 IST BUSHING 207TH BUSHING, 4% BOTTOM BUSHING Figure 10. Base-Line Vibrations ofthe Pump. absolute vibrations of the support structure at the motor top {station number 1) and motor-bottom (station number 7) locations and the relative vibration of the pump shaft with respect t0 the support structure at the shaft probe (station number 12) location are matched sith test data, ‘A sensitivity analysis indicated that che shaft vibration at the labyrinth gas seal location (sation number 13) changes insignificant, if the following factors are varied in the ranges shown: + Wear ring stiffness (= 20 percent) + StatorImpeller forces (+ 100 percent) + Series Bearing stiffness (+ 50 percent) ‘Thus, observations have been that for modelling the unbalance response of the above-grotind sections (stations 1 t0 13) of vertical pumps ofthe type presented, very accurate representations of wear rings, bearing bushings and stator impeller forces do not appear to be requived Further studies indicate thac the motorbearing stiffness and impeller imbalance affect the shaft vibrations at station number 13 also only slightly, A comprehensive parametric study followed which iden- ‘tied the following practical design modifications as effec: tiveln reducing the shaft vibrations a he labyrinth gas seal (tation number 18) and throte bushing (sation number 15) locaton + Reduce the coupling imbalance. + Shorten the supportfre shaft pan txtween the lower rotor bearing and the throtde bushing by relocating the present throtule bushing, or + ddding another throttle bushing in between thet + Increate the shrote bushing stiffness. + Reduce the bottom beating bushing sites. PROPOSED SOLUTIONS Four specific solutions were developed to employ acombi- nation of the aforementioned design modifications in order attain the desired goal of vibration reduction at the labyrinth ges location (station number 13): Proposed Solution 1 (Figure 11). Relocate the throutle bush- ing toa new station, number 14, which about 0.211 m (8.3 jn) above its current location (station number 15), and trim balance the coupling to reduce its imbalance by 40 percent. This solution is expected to reduce the shaft vibration by BASELINE GROLIER, —Repuceo yigRaTion oF pum QEERATION. PERRET SHAFT RELATIVE TO QF RVPPOR! vidaTion CASING/STAUCTURE Bue To PROPOSED SOLUTION NO." STATION NO. fewt2 461032 461022 461012 4 UNBALANCE RESPONSE (MM, 0-P) LEGEND 1 MOTOR TOP 24 GTH BUSHING 7 MOTOR BOTTOM 28 STH BUSHING 12 SHarrPRoge 32 4TH BUSHING 33 MOUNTING PLATE 36 SRD BUSHING Ye NEWLOCATION FOR 42ND BUSHING THAOTTLESUSHING 4g. 1ST BUSHING 18 GOLOLOCATION FOR 4 BOTTOM BUSHING THROTTLE BUSHING 16 ary BUSHING 20° FTW BUSHING Figure 11. Proposed Solution 12 PROCEEDINGS OF THE FOURTH INTERNATIONAL PUMP SYMPOSIUM BASELINE ute QRSOMEE, —nepuceo VIBRATION OF pump GIBRATION REDUCED PAL rve tO See PORT . viBRATION CASING/STAUCTURE Bue TS pROPOSED S8utION No.2 STATION NO, 46we? 46032 4610272 461012 4 UNBALANCE RESPONSE (MM, 0) LegeND 1 moron ror 24 oTH sushinG 7 MOTOR BOTTOM i stnausning 12 SuaerrAoge BATH ousHINa 13 MOUNTINGPLATE 38 SnD eUSHING 1 THROTTLEBUSHING — 40 2ND BUSHING 16 grHeusHiNG a4 israussine 20 TM RUSHING &% sorrow susHing Figure 12. Proposed Solution 2, 56.3 percent, from 0.1930 mim (7.6 mil) to 0.0868 mm (3.4 mils) peak-to-peak, at station number 13. Propased Solution 2 (Figure 12). Increase the stiffness ofthe throttle bushing 10 times, reduce the coupling imbalance by 40 percent and decrease the bottom bearing stiffness by 90 percent, This solution is expected to reduce the shaft vibra- tion by 65.8 percent, to 0.066 mm (2.6 mils) peak-to-peak, at station number 13. Proposed Solution 3 (Figwe 13). Add a second throttle bushing of equal stiffness to station number 14, and trim balance the coupling to reduce its imbalance by 40 percent. This solution is expected to reduce shaft vibration by 57.9 percent, 00.0813 8.2 mil peak-to-peak at sation number Proposed Solution 4 (Figure 14). Reduce the motor pedestal stiffness to 1/100 of its present value, raise the throule bushing stiffness 10 times, reduce the coupling imbalance by 40 percent, and reduce the stiffness of the bottom bearing. bushing stiffness by 90 percent. This Solution is expected to reduce shaft vibration by 74.5 percent, 6 0.0493 mm (1:94 mis) peak-to-peak, at station umber 13, All of these proposed solutions require that the coupling be trim balanced. Of the parameters studied, coupling imbalance isthe single most important factor affecting the shaft vibration at station number 1. Au increase of throctle bushing stiffness must be accompanied by a reduction of the bottom bearing bushing stiffness in order to prevent a buildup of excessive vibration at the lower pump end. The proposed Solution 2 has an advantage over Sofution 1, because it is the simpler solution co implement. Both Solutions } ane 3 require thata thrate bushing be installed ac the station number 14. Solution 3 produces a slightly larger expected benefit over Solution 1 plementation of Solution 4 requires that a new motor pedestal be constructed rendering it somewhat impractical, although it would produce the largest reduction of shaft vibration among the solutions proposed. CHOICE OF SOLUTION Each of the proposed solutions has advantages and disad- vantages. The easiness of implementation is of practical importance. In addition, tis also preferable to change one thing ata time to enhance the likelihood of achieving success ina timely manner. Also considered was the possiblity that the throttle bushing stiffness would be reduced over time, due 10 wear. The proposed Solution 1 was thus adopted for field implementation. FIELD TESTS ‘The shaft vibration atthe probe location (station number 13), before and after the implementation of Solution 1, is illustrated in Figures 15 and 16, respectively. The vibration BASELINE AGS0LUTE REDUCED VIBRATION OF PUMP SISRATION. —PSRAPT” SHAPT RELATIVE TO QFSUPPORT viBwATION CASING/STAUCYURE DUE TO PROPOSED SOLUTION NO. STATION NO, 7 2 3 1a 5 16 2» foes 460032 461022 461012 4 UNBALANCE RESPONSE (MM, 0-?) LEGEND woTOR TOP MotoR BorTom SHAFT PROBE MOUNTING PLATE THROTTLE BUSHING THAOTTLE BUSHING STH BUSHING 7TH BUSHING. Figure 13. Proposed Solution 3. ™ 2 B 36 a a 6TH BUSHING. STH BUSHING ATH BUSHING SRO BUSHING. 2ND BUSHING 1ST BUSHING BOTTOM BUSHING The im-SOLVING THE VIBRATION PROBLEM OF A VERTICAL MULTISTAGE CRYOGENIC PUMP 98 data are compared in Table 2. Since the phase angle data were not recorded, the measured vibrations could net be properly compensated. Assuming that the vibrational sig. nals are all in phase, the running speed shaft vibration is shown w have been reduced by 70.7 percent to 54.9 percent. The vibration level of the pump remained steady altera field BASELINE GRIER, repuceD yrBRATION OF puMP GIRTON. PERRET Sart RELATIVETO SPRUEFORE 4 VIBRATION CASiNG/STHUCTURE PROPOSED SOLUTION, No.4 ‘STATION NO. 2614s 460032 461022 461012 4 UNBALANCE RESPONSE (MM, 0-P) LEGEND 1 MOTOR TOP 24 STH BUSHING MOTOR EOTTON 26 STH BUSHING 12 SHAFT PROBE 2 atTHousnine {3 MOUNTING PLATE 36 SRO sUSHING. 18 THROTTLE SUSHING 42ND BUSHING 16 erWeusHING 44 IST BUSHING 20 FH aUSHING 43 BOTTOM BUSHING Figure 14. Proposed Solution 4 qistete@leja(e(2lela) wie CEPT ome ‘L Pt “ “TTL DPR [| eee we | | Cy hin ewe in Figure 15. Pump Vibration Before Implementation of Solution 1 me COPPELL sate a3 “TOTP aeSce So! Rae Figure 16. Pump Vibratin After Implementation of Solution 1 Table 2. Comparison of Vibration Data Before and After Des Medication _ rrObe rAGse PROG PROBE PROBE YROBE TOME vipeanron 0,204 0.199 6.0922 9.126 0.117 0,206 ‘operation of 2500 hr between the two measurement dates indicated in Table 2 SUMMARY AND CONCLUSIONS, ‘A problem-solving experiece related tothe resolution of wibration problem of a verical mulistage cryogenic pun har been cuitined. A combined experimental a anal approach was aken to ates the nature ofthe problem and to define practical design modifications to resorethe prob lem. The pump was analyiclly modelled as dual evel ‘oon beating hen to propery ake ino account the ‘bration characteristics of the pump suppoft structure Care bas been taken to model al parts of the pump, ncding wear rings, bearing bushings, throtle bushing, totiom bearing bushing coupling, motor pedestal, and the motor Tes data were used to calibrate the analytical mode, which yielded several practical solutions t0 the pumps ‘ibration problem, Feld data confirmed the Usefulness of the implemented solstion. ‘The vibration problem of the rump Was successfully resolved. The following conclusions fay bedtawns * * The modelling of this vertical pump aa dual level roto bearing system isintuistely valid and technialy conece Modelling the pump as single level system shown to cause an important sobsjnchronous mode to remain undetected Sind the system stability matgin overevssated. The dual«4 PROCEEDINGS OF THE FOURTH INTERNATIONAL PUMP SYMPOSIUM level model used herein is propeny calibrated by matching prediction for the baseline case with test data of support Erniture vibrations and relative shaft vibration taken at three different stations. Proper modelling of the motor pump unto account forthe vibrations ofthe pump support fracture and pump casing bundle sas important asthe use ofcorrect models forts components andor accurate numer- ical valvet forthe pumps cital dimensions. + The empirical models for wear rings, bearing bushings, and statonimpeller forees used inthis sdy are obtained from current literature, Experience reported herein has been that the synchronous responte predicted forthe above ground stations of tis pump with fong flexible shat does ot vary sensitively with the numerical values of he dynamic Coefficents of these components in the ranges studied * A combined analytical and experimental approach isan efiecive ay stenacal sxe bation pres ot ‘ertcal mulstage pumps of the type studied. Many empin Ga methods of acute deeibed im the erature ae machine specifi, not generally applicable to other pum On the other hand, the method of solution prevented, cspecaly the dual level modeling technique, should be Spplicabe to all vertical pumps REFERENCES 1, Gunter, E.J., Lindsey, §. E, and Muschick, R.P, “Stabil ty and ‘Time Transient Analysis of Oconee 24-Stage High Pressure Injection Pump,” Duke Power Company Report, Charlotte, North Carolina (June 1985), 2. Gopalakrishnan, S., and Usui, Y,, “Critical Speed of Centrifugal Pumps,” The Shock and Vibration Digest, 16, (4), pp. 3-10 (April 1984), 3. Gopalakrishnan, S., Fehlau, R. and Lorett, J. Calculate Critical Speed in Centrifugal Puinps, Gas Journal, pp 113-119 (December 7, 1981} 4, Bosmans, RE, “Hydraulic Induced Instability on a Vertical Service Water Pump: Case History,” Proceed- ings of Symposium on Instability in Rotating Machinery, Bently Nevada Dynamics Research Corporation, Carson, City, Nevada (June 10-14, 1985). 5. Black, H. F, “Effects of Fluid Filled Clearance Spaces on Centrifugal Pump and Submerged Motor Vibrations,” Proceedings ofthe Eighth Turbomachinery Symposive, Tarbo- machinery Laboratories, Department of Mechanical En- gineering, Texas A&M University, College Station, “Texas (1979), 6. Childs, D. W, "Finite Length Solutions For Rotordy- namic Coefficients of Turbulent Annular Seals,” Jour- nal of Lubrication ‘Technology, Trans. ASME, 103, pp. 437 -445 (July 1983), 7. Allaire, P-E., Gunter, EJ» Lee, C. P and Barrett, LE, “The Dynamic Analysis of the Space Shuttle Main Engine High Pressure Fuel Turbopump,” Final Report, Part IIL, Load Capacity and Hybrid Coefficients For ‘Turbulent Interstage Seals, NASA Marshall Space Flight Center (September 1976) 8. Kim, C,,and Childs, D. W, “Analysis for Rotordynamic Coefficients of Helically.Grooved Turbulent Annular Seals,” ASME Journal of Tribology, 109, pp. 136-143, (anuary 1987). 9. Jery,B., Acosta, A. J., Brennen, C. E., and Caughey, T: K, “Hydrodynamic Impeller Stiffness, Damping and Inertia,” Proceedings of the Third Workshop on Rotor 10. 12. 18, 4 1. 18. 19, 20, 2 2. 28, dynamic Instability Problems in High-Performance ‘Turbomachinery, College Station, Texas, NASA Confer- cence Publication No, 2538, pp. 187-160 (1984). Jery,B., Brennen, C. E., Caughey, T. K., and Acosta, A. J.,"Forces on Centrifugal Pump Impellers,” Proceedings of the Second International Pump Symposiam, Texas ARNE Universi, College Staion, Texas pp. 21.29 (Apri Chamieh, D.S., Acosta, A. J., Brennen, C. E,, Caughey, TK, and Frang, R., "Experimental Measurements of Hydrodynamic Stiffness Matrices for a Centrifugal Pump Impeller,” Proceedings of Second Workshop on Rotordynamic instability Problems in High Perform- ance Turbomachinery, Texas A&M University, College Station, Texas, NASA Conference Publication No, 2250, pp. $82-398 (1982), Gopalakrishnan, S., "Pump Instability Phenomena Gen- erated by Fluid Fores,” Proceedings of Symposiam on Instability in Rotating Machinery, Bently Nevada Dy- namics Research Corporation, Carson City, Nevada (une 10-14, 1985), Start, D.E, "Trouble-Shooting Vertical Pumps Uniizing Vibration Techniques,” Proceedings of Machinery Vs bration Monitoring and Analysis Seminar, Vibration Institute, Clarendon Hills, lines, pp. 131-133 (1983). Jones, R..M., “Vertical Reactor Coolant Pump In- stabilities” Proceedings of Symposium on Instability in Rotating Machinery, Bently Nevada Rotordynamics Re- search Corporation, Carson City, Nevada (1985). Makay, E,, "How Close are Your Feed Pumps o Instabil ity Caused Disaster?” Power, pp. 69-71 (December 1980), Meyer, R.J.,"Solve Vertical Pump Vibration Problems," Hydrocarbon Proceeding, 56, (8), pp. 145-149 (August 1977) Sosa, F and Baker, E. W, “Reduced Casing Vibrations on Large Vertical Circulating Water Pumps,” ASME Paper, No. 77-FE-28 (1977) Marenco, S. G.,et al, "Elastic Solution for Vibration ‘Trouble on Vertical Pumps,” Energ. Ellewte, $9, Suppl t No. 10 (1982), paper in Conference on Rotordynamic Problem in Power Plants, Rome, lal, pp. 429437 1382) Goldman, S., “Solving Pump Problem Using Vibration Spectrum Analysis," Noise Vibration Control World: wide, 12, (8), pp. 326-381 (November 1981), Mondy, 8 E,and Sud}, “Pump Modifications Solve Complex Vibration Probleins,” Power, pp. 41-13 (Feb- ruary 1985) Mondy, R. E., “Absorber Stops Elusive Multi-Stage Pump Vibration,” Power, pp. 51 ~ 52 (March 1985). Smith, D. R.,and Woodward, G. M., “Vibration Analysis Of Vertical Pumps," Proceedings ofthe Fifteenth Turbomach- inery Symposium, Turbomachinety Laboratory, Depart- ‘ment of Mechanical Engineering, Texas A&M Universi- 1y, College Station, Texas, pp. 61-68 (1986). Li, D, F and Gunter, F. J, “Unbalance Response Analy ‘of Dual-Rotor Systems: A Manual for Use with the Computer Program RESP2V3,” Report MAE 78/150, Department of Mechanical and Aerospace Engineering, University of Virginia, Charlotesville, Virginia (1978)SOLVING THE VIBRATION FROBLEM OF A VERTICAL MULTISTAGE CRYOGENIC PUMP ACKNOWLEDGEMENTS The authors acknowledge the helpful discussions with Drs.PE. Allaire and L. E. Barrett, and Mr Joe Knight of the University of Virginia ROMAC Industrial Consortium, con- cerning the modeling of bushings and the numerical accu racy of the extended RESP2V3 computer code. 95