Вы находитесь на странице: 1из 6
Shaft CenterLINES Some realities of field balancing by Robert C Fisenmaun, PE, President Machinery Diagnostics, ne Minder, Nevadr speed, some rotors can only be balanced inher casings inthe fick, using every technigue, trick, and pioce of informa tion available to the speci, Most turbomachinery rotors ate sie ‘cessflly balanced in slow speed shop balancing machines. This aporoach pro vides gond access to all correction planes, and the option of porforming multiple rans to achieve a satisfactory ‘lance. I is generally understood that -unaing st low speeds with simple bear our carte 900 2 i EEE ings or eollers does not duplicate the tojatiooal dynamics of the feld installa tion, Orher rotors are shop balanced on high speed machines installed in vasa, chambers. These machines. provide an improved simulation of the installed rotor behavior. due to the higher speeds andthe use of bearings that resemble the normal running bearings. In these bigh speed chambers, the influence of blade sand wheel aerodynamics is significantly redced by operating wit Some machines, soch as Isege steam turbines, often requite afield trim bal- duet the influence of higher order modes oF the limited sensitivity of the Tow speed hakancing techniques. ar0up of ‘There is ako a sm machines. which, except for comp balan ‘These machines have segmented rotors, ‘that are assembled concurrently withthe liaphragms and/or casing. In these ‘machines, the inal ror assembly isnot achieved umil mos of the stationary ele ‘ments are bolted into place An example ofthis type of rotor i shown in Figure 1 Machine configuration This rotor consists of an overhung hot as expander wheel, a pair oF midspan compressor wheels, and three stages of overhung steam turbine wheels. A series of axial through bolts ae used co onnest the expander stub shaft through the com pressor wheels and into the turbine stub shat. This typeof rotor assembly is sim ilarto many gas turbine voors, However im this machin, the rotor ust be built concurrently with the inner casing Specifically, the horizonslly split iner nal bundle is assembled withthe it u-aluminur compressor wheels, stab shafts, plus bearings and seals, End cas: ings are wtachel, the expander whee i bolted into position. snd te tree turbine Sages are attached with snosber se of ‘hough bolts “The eight rotor segments are joined with Curvic® couplings. which are iden ified as #1 through #7 on Fipure 1. Even with properly ground and tight fing 12 orbit Figure 1 zn i J compressor.seam turbine rater and vibration transducer arrangement. June 1997 ‘couplings. there is potential for relative movement of rotor elements. Although exch of the rotor segmen's is component balanced, any minor shift between ele ‘pens will produce a synebronous unbel tance force, Since this unit operates at 18.500 pm afew grams of unbalance oF sami oc two of eccentricity will result in excessive shaft vibration and a strong potential for machine damage. Furthermore, the distribution of oper= ing temperatures noted on Figure 1 reveals the complexity of the thermal effets hat rms he tolerated by this unit. “The expander wheel inlet temperature is 677°C (1,250°F), and compressor Gis: charge temperature is in excess of 221°C (430, The steam turbine oper- tes with @ 371°C [7O0%F) inlet temper- ‘ature and a final exhaust temperature of 71°C (160°F) By any definition, this must be eon dered dificult cotcr to dea with ‘More than one person hs referred to this rotor as a “squirt,” meaning & rotor ‘whieh responds erratically. On the posi- tive side, this machine is a compact design that yields a high thermal fie ciency. Therefore, when the unit is prop ly assembled, and balanced i very cost efTetive to operate. eld balancing this 413 kg (910 1b) rotor is achieved by adding or removing ‘weights from a series of teaded holes in four balance planes (Figure 1). Balance Plane #1 coniains 20 axial holes ate expander exhaust. Plane #2 and #3, cach have 20 radial holes. They are Tocated in two rows on the expander and turbine stub shafts, respectively. The outboard end of the turbine Jed Stage ‘wheel contsin balance Plane #4 with 30 axial holes. Shaft vibration measurements are ‘made with Bently Nevada XY proximity probes (Figure 1). These radial probes fare connected to a Bently Nevada 3300 ‘Monitoring. System. which in turn is connected t a Transient Data Manager? (TDM2) Communications Processor (CP). The CP sends static and dynamic signals from the monitors to a personal computer running TDM2 Software ‘This instrumentation is used for mon June 1997 itoring, tending, and machinery diag nostics, The 3300 Monitors can auto ‘matically activate alarms when vibration setpoints are exceeded, and the eomput- er system has additional Alert and Danger indications. Tis instrumentation also provides the steady state and tae sient vector data necessary for field trim halonsing. this complex rot. Balancing calculations During shop balancing, itis easy 10 make a eight change; iis inexpensive to make a run, and tere is litle physical risk to the mechinery or the operator. In the case of field balancing, iis often dif- Feu to ehange weighs, and it is genor- ally expensive make a fll speed run. Furthermore, if an incorrect weight is used, the results may be hazardous tothe ‘machinery, ard the health of the opera- Buch field balance shot should be a meaningful move, and should contribute to the overall database descring the behavior of the muchine. Bslancing behavior is based on the fundamental relationship between the shaft response tnd the applied force expressed by’ Force Farce Pm Brant Dyna Sis 0 ‘This genera expression has many specific applications in rotor dynamics. Within the balancing discipline, the re- sponse is the weasined shaft oF casing motion and the fore isthe magnitede of the unbalance force. The restraint is the dlynamie stiiness. Another way to view this term isco consider it a the sensitiv ity ofthe machin to rotor unbalance. ‘The variables in Fg.1 are. vector ‘quantities. That i, each parameter cat= ries both a magnitude and a diretin. the vibration of a rotor is described by the A vector, with ampliude vais of mi pp. and the unbalance magnitude is defined by the U vector with units of rams instead of force, then the sensitiv- ity $ veewor carries the ameipaed enai- neering units of grammil pp. Equation | nay now he rewritten follows U ard @ ‘The vibration vector A is measured ‘iret, andthe problem resolves to one ‘of determining the unbalance based upon an unkown Sensitivity vector Accord. ing to most sources, this problem was ‘originally solved by ELL. Theare in his 1934 ASME paper. He basically provi. ced a way (0 practically determine the sensitivity vector. That procedure involves the addition of known calibra tion weight co the rotor and determining the response vector C duc to that calibra tion weight. Ifthe calibration weight ‘vector is defined by W, and the resultant vibration response vector ofthe rotor is defined as the AC, Equation 2 may be expanded as follows: This last expression may be solved for the sensitivity $ vector: % The § vector may now be combined ‘vith he nil vibration response vector in Eg, 2, and the unbalance vector com: poted. The concept has been around for ‘over 60 years, and has been suecessfully applied on many machinery rors ‘The previous diseusson covered only ‘one measurement plane. The same logie may be applied to two oF more measure- ment and correction planes. With respect to the curtent rotor under investigation, there are only two measurement planes, and four separate correction planes. This Js different from most machines where the number of measurement snd comtee= tion planes are equal ‘Thus, the general form of e@ 3 is: y, Sm Gr « One ‘where mis the measurement plane and p js the weight conection plane. Orbit 13 ‘The bakincing technique of “ad a veeight, make a run, toke dara, and do vector calelations” generally produces acceptable results on many machines However, this technique may fail on a complex rotor. A review of the balance history of the machine may help ientify ‘he corret technique. Historical data review ‘A double overhung rotor with an appreciable midspan mass (such as Figure 1) has the potential for multiple resonances with both forward and reverse modes, In order to better under ‘and the behavior of tis machine, var tus historical data sets were reviowee. Fortunately, TDM2 data was available This included steady state orbits, plus transient speed Bode and polar plots. rom this data it was clear that reverse cnbits peated around 7,000 and 17,000, rpm. In ation, tabular ists of sempled vectors were also documented, It was ‘noted that the slow roll vectors were ‘ot repetitive, and this was a cause for Previous field balancing activites on this machine were generally success when a two step correction was used The first step consisted of a balance at the process hold point of 14,500 rpm, tusing the outboard Planes #1 and M4. ‘This was followed by a trim at 18,500 pm on the inboard Planes #2 and #3 located next to the compressor wheels Te was evident that ifthe rotor wasn’t adequately balanced at 14,500 spm, it probably woulda’t run a 18,500 rpm. For comparative purposes. the Sensitivity vectors from three differen ata runs were computed based upon steady state data at 14,500 rpm. These S veciors were computed in. accontance ‘with Eq 4. The results are summarized in Table 1. The_ two shaded vectors in this table are oF questionable sceurzey, ‘due tothe Fact thatthe differential vibra tion vector C was less than 0.1 mil pp. ‘This small differential vibration is indicative of low sensitivity 10 the applied weight. ‘The S vectors in Table 1 have some similarities, but the variations ate signif icant, For example, the magnitude of S varies from 22.610 764 gramimil, and 14 Orbit 76° angular difference is noted. On Sy ‘he amplitudes change from 24.2 14.9. frum/mil, but the angles reveal a 59° Spread. AC this point, a preliminary conchision might be reached that this rotor is nonlinear and eannot be feld balanced. Clearly, additional informa- tion vas’ nesded define the rotor behavior, Process observations Further examination revealed thet vibration severity changed in accordance ‘with the machinery operational state. Peak vibration amplitides occur at a rotor critical (balance resonance) that spears between 7,600 and 8,100 rpm, This resonance displays the following characterises versus operating. conic + Cold statup 16 14,500 rpm Peak response of 2.0 0 5.0 mil pp + Warm coastdown from 14,500 rpm Peak response of 4.0 t0 5.0 mil pp + Hot shutdown from 18,500 xpm Peak response of 60 8.0 mil pp ‘These amplitude variations, cots bined with changes in the rotor Syn- chrosous Amplification Factor through he resonance, suggest a change in the damping term (Quadrature Dynamic Sititfiness) oF dynamie stiffness Furthermore, it was determined that the different operating conditions were directly associated with variations in shaft slow roll vectors, which were taken at 1,000 pm. In al eases, the runout vee tors for each cold startup. warm coast- down or hot emergency shuxdown were consistent within each condition. When the rotor was shut down from a hot state, the slow roll vectors always reuaned 10 the same initial stating vectors as the rotor cooled. Therefore, the apparent changes in slow roll were resolved. I was also determined thatthe previ ous vibration response vectors used for brlancing were acquired at process hold point of 14,500 apm. Under this condition, the muchine speed was held constant, but rotor and casing tempera tures were changing a the process sta: lized prior to continued startup of the plant. Next, the complete transic response picture naedad to be added to this process dat, Transient data review “There iss adage that states, “If your only tool is «hammer, thew all ef your problems begin to resemble nails.” This Is very appropriate for this panicular rotor. [Fhe “put a weight in and take a weight our” balance technique is the only tool available, the results are often isappointing. his is evident from the spread in sensitivity vectors shown in Table 1. Attempting to balance a machine when these values keep chang ing is dificult at best. Many runs ste required to alain @ barely acceptable balance state. tis quite clear that additonal tools must be employed to properly field bal- ance this machine. Certaily, the plant process observations discussed in the Tow | pmasant | Omesee | oma sers gonna’ | ign soy | (janis) 5 | mene | wee | mamane [ra meeue | Maen aceon | wren | meer 3, | eae me | tases | sormmne ta | eter | sete | 1058 100 ee wrens | size mi 3, | mee or | ose | wae Staton etre ih eirson lhe sated let Semitivty vectors base om steady state dat at 14,500 ep, dune 1997 previous section are significam. How- correetions at this intermediate point changes between 450 and 18,508 pm, ‘ever, this information must be supple: were based upon the transient vectors Other vector changes ate visible as the mented by an examination of the vari- acquired by the TDM2 System as the machine approcetes the normal ope able speed (transient vibration response rotor passed through 14,000 rpm, ing speed of 18,500 spr. Some of these lata (Bods and polar plots), ples an The Bode plot in Figure 2 also changes are due to the influence of a ‘undersanding ofthe rotor critical (bal- gyi a variety af amplitude and phase backward mode around) 17.000 rpm. lance resonance) speeds, and mode shapes ‘The vibration response from the installed proximity probes was extracted from the TDM2 database; atypical stat: ‘up Bode plot is shown in Figure 2. This data displays the Y axis probes from ‘both measurement planes. Both pots are corrected for slow rll rout, (Note: the ‘Bode plots stam at 4,000 rpm for bet resolution.) Thus, the resultant data representative of the true dynamic "motion ofthe shat at each ofthe two at cal measurement panes, ‘The major resonance ecurs at approximately 7,800 rpm. Note thatthe process hold point at 14500 rpm die plays substantial amplitude and phase ‘excursions, Tis is logically du to the heating of the rotor and casing, plus Variations in “settle out” ofthe operating system (Le, pressures, temperatures, 10 flow rates, and molecular weights). Although this process stabilization isa z ‘necessary part of the startup, the var ‘eco 6008 G09 ToEco 1260) 14099 100 4000 ations in vibration vectors megate the atta smn om) validity of this information for use as repetitive balance response data, Figure2 'At speeds above 14.500 rpm, there Raneut-compensaed Bode plot of Y-axis probes during (pica machine train startup. apuomeettnt ie) ‘pose sre additional vector changes, and a desirable far spot (plateau) inthe ampli- the canes dsm ape Twas | Banani] bam Sue ] Dae serss ‘The only essentially flat area that pro- Kcreavint © doh | (gram © dos) | (orurie @ cos) scone yor vee ace aaron waakae Inthe 12000 pm coz @ ze | 4240 ‘To test the validity of this conclusion, Su ini wet 1000p wee tse mr | mse ee | mse Te Seid TDN owen a Paseser | ere sar} sor sor Scion wi he ned weg were 3, | eens | sea | Atrtantane Ose comput the ses vets so baise ier | mares | vase re intel wi Ba Te alo there computtons ve pes on moan | Tae ne Tle. Ths vector ae det coe Sop ree ur | ane ioe | oes 6 16 pare the sect oewsed ne I Hoever scleral conte, oh anne nena aon Sot Succes anan the er da ciency Seat Sho strsopnorn the ute town mus Tans 2 therefnesfatre alae Sexy vaso bavd or sn st 1490-9 June 1997 Orbit 15 Figure Runoutcompensats polar plots of Vans This highor speed data is dificult to Fully snprend inthe Bode pot. I becomes more definitive when this same data is plotted ina polar format (Figure 3). Both ¥ axis probes show the primary rovor resonance at 7,800 pm, The point ‘of major interests that, at ful sped, the turbineend shall is moving towards the 9 o'clock direction, and the expander: ‘end shaft is heading towards 4 o'clock. ‘This behavior indicates the presence of « Figure Cateuated damped mode shapes at maka rotor resonance nominally 7.30 rp. 16 Orbit 3 ines during typical machine startup pivotal mode occurring at a frequency above 16,000 ep, Tn many cases, this type of response ‘would nat he unusual, However, for this the historical files had no indication fof a resomanee around the acrmal rin- ning speed. Dac to the response of ‘midpan balance weights (Planes #2 and 4) at 18.500 rpm, it was clear that the vibriondata was correct, an the histor jal undamnped mode shapes were awed. ‘Analytical model ‘As was previously stated, there are ‘only to lateral measorement planes along the entire 2 metre (79 inc lenguh ‘of this rotor. Since there were no othor feasible locations for shaft probes, si tional measurement options were elim nated. Therefore. the only viable approach resided with a meaningflsna- Ivtical model ofthis rotor system, A 65-station damped mode was con strocted. This computer model included bearing sifiness end damping that var- Jed with speed, plus Mlexible bearing p> pons. The caleulations produced eigen- values (natural frequencies) plus eigen- vectors (mode shapes) ‘The firs four mores are stiff shaft pivotal and translational shapes. These ‘modes di not appear in the vibration ata due to the high dynam stiffness for each mode. A backward_ mode was detected at 7.080