CONTROLLING FAN VIBRATION - CASE HISTORIES Donald R. Smith Seni or Pr oj ect Leader Engineering Dynamics Inc. San Antonio, Texas and J. C. Wachel Manager of Engineering Engineering Dynamics San Antonio, Texas Sponsored by ELECTRIC POWER RESEARCH INSTITUTE October 14-18, 1981 Hyat t Regency I ndi anapol i s I ndi anapol i s, I ndi ana Key Words : Vi br at i on Foundati on Nat ural Frequency Dynami c Soi l Pr oper t i es Mu1 t i - p1 ane Bal anci ng Lat er al Cr i t i c a l Speed Di sc Resonance I n1 e t Vort ex Rot at i ng St a l l CONTROLLING FAN VIBRATION - CASE HISTORIES ABSTRACT Dinald R. Smi t h and J. C. Wachel Fan vibration problems have been a serious cause of plant unreliability in large f ossi 1 -fired power plants and have resulted in operational problems, shutdowns, and reduced generation. The basic causes of most problems are dynamic resonances asso- ciated with the system. These have t o be identified before practical and effective recomnendations can be made for corrective action or design modifications. The most effective solutions can best be determined from computer models which match the measured f ?el d data. Thi s paper discusses several case h i stories and i l l ustrates methods and i nstrumenta- tion for analyzing existing fan problems. Experimental techniques are described for defining problem symptoms, and these are related t o root causes by the use of data analysis and canputer simulation techniques. Computer modeling techniques can then be used t o evolve reliable fixes or used in the design stage t o simulate interaction between rotor dynamic response and complete system response t o prevent potential dynamic problems. Secti on 1 I NTRODKTI ON Donald R. Smith y J. C. Wachel Mul t i pl e f an systems are t ypi cal l y complex and vi br at i on problems can be d i f f i c u l t t o def i ne due t o dynamic acoust i cal and mechanical i nt er act i on between fans [I]. When fans are i nst al l ed i n mul t i st or i ed bui l di ngs, t he problems are f ur t her compl i- cated due t o structure-borne vi brat i ons. Sanetimes it appears t hat "everyt hi ng shakes," i ncl udi ng t he st ruct ure, f l oor , fans, ducts, and motors. Pl ant personnel are of t en f r ust r at ed i n attempti ng t o cor r el at e t he countl ess number o f vari abl es i n t he system t o determine t he exact cause of increases i n vi br at i on l evel s. Often t he vi br at i on increases are correl at ed t o seemingly unrel at ed vari abl es such as t i me o f day, r ai nf al l , o r l oad changes on ot her uni t s. I n t he anal ysi s o f severe fan vi br at i on problems, it i s of t en found t hat t he hi gh vi br at i ons are due t o coincidences o f one or more nat ural frequencies which amp1 i f y l ow l evel exci t at i ons. A l i s t of possi bl e exci t at i on sources and nat ural frequen- ci es general l y found i n f an systens i s presented below: EXCITATION SOURCES Mechanical Defects: m i sal igrment, improper tie-down, bad grout, bent shaft, shaf t rub, warped t hr ust c o l l ar, 1 oose r i vet s on wheel, bad shri nk f i t between wheel and hub, etc. Unbalance: dust o r ash bui l dup on blades or i nsi de a i r f o i l blades, bl ade erosion, thermal di st or t i on ( shaf t bow, wheel warpage), inadequate bal anci ng a t low speeds Pul sati on: f l ow exci t at i on across obstructions, vort ex shedding, i n l e t box vortex, r ot at i ng s t al l , bl ade passage NATURAL FREQUENCIES shaft lateral cri t i cal speed torsional critical speed 0 pedestal , foundation, fl oor/support structure di sc wobble (wheel rocking) acoustical SYSTEM ANALYSlS The f i r s t step in defining and solving vibration problems i s t o determine i f the vibrations are due t o ( 1 ) high level excitations, or ( 2 ) low level excitations which are ampl ified by the coincidence of one or more natural frequencies. I f the excita- tions are high, the solution i s generally t o reduce the energy level. If coinci- dences of resonance exist, the solution i s t o modify the system t o change the nat - ural frequencies away f r m the excitation frequency. The most di ffi cul t vibration problems t o analyze and solve are those which have several resonances at the same frequency which increases the cross coupling and interaction between the resonances. Detailed fi el d t est s coup1 ed with analytical analyses are generally required t o sep- arate and identify the excitation sources and their ampl itudes and the system nat- ural frequencies. The analysis techniques and field instrumentation have been greatly improved in the l ast few years. Equipment which used t o be considered for laboratory use only can now be easily transported t o the fi el d for on s i t e dat a analysis. A typical field instrumentation setup i s shown in Figure 1-1 which includes a minicomputer, real time analyzer, trim balance analyzer, mu1 ti-channel FM recorder, oscill oscopes, signal conditioning ampl i f i er s and i ntegrators, and X-Y pl otters. In cunpl ex problems mi ni cmputers are used t o gather and process several data channels simultaneously and plot the vibration mode shapes. Mu1 ti-channel telemetry systems can be used t o collect data such as strain and temperature on the shaft and blades during on-1 ine conditions. The objective of t hi s paper i s t o i l l ust r at e techniques to identify and solve fan . vibration problems; however, these techniques are also routinely appl ied t o solve vibration problems on other types of rotating and reciprocating equipment. Sane of the mohe unique fan vibration problems analyzed by SwRI have been selected for t hi s paper t o i l l ust r at e the diagnostic procedures and solution development. Se c t i on 2 CASE HISTORIES. CASE NO. 1 Dur i ng t h e i n i t i a l s t a r t u p of t h e ID f a n s a t a l a r g e power p l a n t , t h e f a n and mot or be a r i ng hous i ng v i b r a t i o n s wer e g r e a t e r t h a n t h e s p e c i f i e d maximum a1 1 owabl e 1 eve1 o f 2 mils peak-peak a t t h e r unni ng s peed o f 900 rpm, The f o1 1 owi ng symptoms wer e obs er ved: Dur i ng t h e f i e l d ba l a nc i ng by t h e f a n ma nuf a c t ur e r , t h e f a n s were s e n s i t i v e t o s mal l bal a nc e wei ght s . The v i b r a t i o n s wer e g r e a t e r on t h e mot or t h a n t h e f a n. When t h e mot or was r un uncoupl ed, t h e mot or v i b r a t i o n s were l e s s t h a n 1 mil b u t t h e f a n out boa r d be a r i ng hous i ng v i b r a t i o n s were g r e a t e r t h a n 3 mi l s . The hor i z ont a l v i b r a t i o n s a t t h e t o p of t h e f ounda t i on wer e a l s o above t h e d e s i r e d limit. The f a n v i b r a t i o n s wer e s e n s i t i v e t o t h e i n l e t g u i d e vane s e t t i n g . The v i b r a t i o n s o f each f a n wer e a f f e c t e d by t h e v i b r a t i o n s of t h e a d j a c e n t f a n s even t hough t h e y were n o t i n s t a l l e d on a common f ounda t i on mat. Temper at ur e changes i n t h e f l u e ga s had a s i g n i f i c a n t effect on t h e v i b r a t i o n s . The v i b r a t i o n s on one o f t h e f a n s i nc r e a s e d a f t e r a heavy r a i n . n e t a i l ed i n v e s t i g a t i o n s r e ve a l e d t h a t t h e f a n s were o v e r l y s e n s i t i v e t o s mal l c ha nge s i n unbal ance, l o a d , and t e mpe r a t ur e due t o amp1 i f i c a t i o n from o p e r a t i n g n e a r t wo r es onances . The fi rst f ounda t i on n a t u r a l f r e que nc y and t h e f a n s h a f t i n s t a l l e d r e s o n a n t speed wer e bot h n e a r t h e f a n r unni ng speed. The n a t u r a l f r e q u e n c i e s c oul d n o t be a de qua t e l y s e pa r a t e d wi t hout ma j or modi f i c a t i ons ; t h e r e f o r e , t h e v i b r a t i o n 1 eve1 s were r educed by h o t ba l a nc i ng e a c h f a n. FOUNDATION NATURAL FREQUENCY Vi b r a t i o n l e v e l s measured on t h e f a n be a r i ng hous i ngs dur i ng s t a r t u p s and coast downs r e ve a l e d a na t ur a l f r e que nc y j u s t below t h e r unni ng s peed ( Fi g u r e 2- 1) . The vi br a - t i o n mode s ha pe was obt a i ne d by meas ur i ng t h e v i b r a t i o n s on t h e be a r i ng hous i ng, . b e a r i n g p e d e s t a l , and s e v e r a l l o c a t i o n s on t h e s i d e o f t h e c o n c r e t e f ounda t i on. Vi b r a t i o n d a t a wer e p l o t t e d on a s c a l e d dr awi ng ( Fi g u r e 2-2). The e n t i r e f ounda t i on Fi gur e 1-1. Typi cal Fi e l d Equipment Used i n Vi brat i on Anal ysi s and beari ngs moved t oget her as a r i g i d body and rocked about a poi nt several f eet below t he f oundat i on which i s char act er i st i c o f a f oundat i on resonance. FREQUENCY, Hz Fi gure 2-1. Spectral Anal ysi s o f Beari ng Housing Vi br at i ons Duri ng ' St ar t up FOUNDATION ROCKING MODE ( 8y) Fi gur e 2-2. Fan Foundation Vi br at i on Mode Shape On t hr ee of t he fans, t he f oundat i on nat ur al frequency was j u s t below t he runni ng speed, but on t he f our t h fan, t he resonance was sl i g h t l y above t he runni ng speed. To v er i f y t hat t he measured responses were foundation resonances, a var i abl e speed, uni di r ect i onal , mechanical shaker was attached t o t he f oundat i on and run through a speed range of O t o 1800 rpm. The shaker-exci t ed vi br at i on response ve r i f i e d t h a t t h e f i r s t foundat i on nat ur al frequency ranged f r un 12 t o 16 Hz f o r t he s e f our "iden- t i c a l " f ans. The fan wi t h t h e foundat i on resonance a t 16 Hz was t h e most s e ns i t i ve t o small changes and had t h e gr e a t e s t ampl i f i cat i on a t t h e running speed of 900 rpm ( 15 Hz). These foundat i ons were desi gned t o have t h e i r lowest foundat i on nat ur al frequency a t approxi mat el y 1500 cpm (70% above running speed). Foundation cal cul a t i ons were made i n t h e desi gn s t age usi ng a val ue of Young's Modulus ( E) of 500,000 psi f o r t h e sandst one beneat h t he foundat i ons. To eval uat e t h e di screpancy between t h e measured and cal cul at ed foundat i on nat ur al f r equenci es , t h e foundat i on-soi l dynamic system was modeled on a computer program developed by SwRI. Foundation nat ur al f r equenci es were cal cul at ed f o r a range of s oi l modul i and compared wi t h t h e shaker dat a. As shown i n Fi gure 2-3, an e f f e c t i ve s oi l modulus of 80,000 t o 120,000 psi was requi red t o match t he measured foundat i on nat ur al frequenci es. gN FOUNDATION ELASTIC SOIL MODULUS (El, PSI X lC3 Fi gur e 2-3. Comparison of Measured and Cal cul at ed Foundation Natural Frequenci es These effective E values as empirically defined are much lower than the typical values given for sandstone which can range from 500,000 t o 3,000,000 psi. The orig- inal soil analysis report revealed that the soil was highly st rat i fi ed, and the depth t o each layer varied considerably over short distances. The soil modulus beneath the foundations could also have been reduced due t o blasting and over- excavation when the foundations were constructed. This case history i l l ust rat es that the major unknown in predicting the resonant fre- quency of a fan-foundation system i s the effective soil modulus. In the design of these fans, a minimum value for the soil modulus was obtained f r m soil bore t est s, b u t even t hi s mi n i mu m value was approximately five times too large. The field data indicated that fan vibration levels were also affected by moisture content of the soil because the vibrations would change significantly aft er a rain. Moisture changes the soil modulus and can shi ft the foundation natural frequency closer to the running speed which amp1 i f i es the vibrations. Also, the damping can be reduced due t o the cohesion of the soil and separation from the concrete caused by vibration. Since the foundation dynamic design i s highly dependent upon the E value of the soi l , every effort should be made in the design stage t o obtain effective dynamic soil modulus directly below the foundation. There are several methods which can be used t o obtain more accurate dynamic soil data, as given below: 1. A 1 arge out put , 1 ow frequency shaker can be used t o excite Ray1 eigh ground waves at the foundation elevation. The soi 1 modulus can be calculated f r m t hi s data [2]. The effective soil modulus should be obtained after the s i t e has been excavated and the f i l l added. 2. The soil modulus can also be obtained by using cross hole tests. Two holes are bored t o the bottom of the foundation [ Z ] . A shear wave i s created by an impact at the bottom of the hole and measured w i t h a transducer i n the other hole which i s a known distance away. This method could a1 so be used to determine the effective modulus at the bottom of piles. 3. A shaker can also be used t o excite the natural frequencies on a small or partial ly compl eted foundation. The partial ly compl eted foundation can then be modeled on a computer and the effective soil modulus can be determined by comparing the calculated natural fre- quencies w i t h the measured frequencies. A shaker was used on a partially completed foundation block of an adjacent u n i t (Figure 2-4) t o determine the effective soil modulus. The t est s indicated that the effective soil modulus would cause the foundation natural frequency t o be near the fan running speed. Foundation modifications t o reduce the vibrations were analyzed using the computer program (e.g., increase mat size, piles, added mass, et c). Most of the foundation modi flcations i nvestf gated were i neffectf ve, were not cost effec- tive or were space limited. Other methods were then investigated t o reduce the vibrations. I S 10 I S 20 25 30 SHAKER SPEEq He Fi gure 2-4. Comparison of Measured and Ca1 cul at ed Foundation Vi brat i on Response Shaker t e s t s on one foundat i on r eveal ed t h a t t he vi br a t i ons were i ncreased by a f a c t o r of s i x when d i r t was removed from t h e s i d e of t h e foundat i on. Based upon t h i s t e s t , sand bags were t empor ar i l y pl aced agai ns t t h e s i d e of t h e foundat i on t o i ncr eas e t h e damping and l a t e r a l r e s t r a i nt . Vi br at i ons were reduced by a f a c t o r of approxi mat el y two t o one.. These t e s t s i ndi cat ed t h e s i gni f i c a nt e f f e c t of s oi l damping i n t h e oi e r a l l desi gn of foundat i ons. Foundation vi br a t i ons can a l s o be reduced i n some cases by shor t eni ng t h e hei ght of t h e foundat i on above t h e ground o r desi gni ng ' t he foundat i on i n t h e shape of a t r uncat ed pyramid i nst ead of a t a l l rec- t angul a r block a s i s usual 1y done. FAN SHAFT LATERAL RITICAL SPEED 2 The f an s h a f t l a t e r a l nat ur al frequency was measured t o be 960 cpm usi ng a va r i a bl e speed shaker at t ached t o t h e foundat i on. For t he s e t e s t s a c c e l e r me t e r s were at t ached t o t h e f an s ha f t and proxi mi t y probes were mounted a t t h e beari ngs t o sepa- r a t e t h e r o t o r response from t h e foundat i on response. The f an i n s t a l l e d resonant speeds were a l s o cal cul at ed a s shown i n t h e c r i t i c a l speed map ( Fi gur e 2-5) where t h e s h a f t nat ural f r equenci es a r e pl ot t e d ver sus t he e f f e c t i ve support s t i f f ne s s . The e f f e c t i v e support s t i f f n e s s i ncl udes t h e s t i f f n e s s e s of a l l t h e s pr i ngs from t h e r o t o r t o ground i ncl udi ng t h e oi l f i l m, bear i ng pedes t al , foundat i on, and s oi l . STIFFNESS (LWIN) Figure 2-5. ID Fan Critical Speed Map The fan manufacturer had calculated the cri t i cal speed t o be 1180 rpm; however, the calculations were based on a "rigid bearing cri t i cal ". For many years fan manufac- turers used t hi s calculation which assumes the rotor i s mounted on rigid supports because many of the supporting stiffness values were not known. As a result, there has been some confusion in the fan industry as t o the exact definition of cri t i cal speed and resonant speed. To cl ari fy the problem, the Air Movement and Control Association (AMCA) C31 adopted the following definitions: Critical S eed: A cri t i cal speed i s that speed which corresponds t o Ti --- t e natural frequency of the rotating element (impeller and shaft assembly) when mounted on rigid supports. (Note: This i s generally referred to as the rigid-bearing cri t i cal speed.). Design Resonant Speed: Design resonant speed i s that speed which corresponds t o the natural frequency of the combined spring-mass system of the rotating element, oil film, bearing housing, and bearing supports b u t excluding the foundation (foundation stiffness i s considered as i nfi ni t e). Installed Resonant Speed: Installed resonant speed i s that speed which corresponds t o the natural frequency of the combined spring- mass system of the rotating element, oil film, bearing housing, bearing supports, and includes the effect of foundation stiffness. The calculated rigid-bearing cri t i cal was 1180 rpm and the calculated instal 1 ed res- onant speed was 960 rpm. Normal ly the instal led resonant speed should be at least 20% from the running speed t o prevent excessive vibration ampl ification, In t hi s case the calculated installed resonant speed was only 7% above the running speed; therefore, amp1 i ficat i on could be expected. The rigid-beari ng critical speed i s not applicable because in the real world the rotor i s not rigidly supported. The actual stiffness of the concrete and pedestal for t hi s systen was approximately 5 x 10 6 Ib/in which i s much less t h a n rigid. A forced vibration response analysis i s normal ly required t o accurately determine the rotor installed resonant speed. The rotor should be modeled using the oil film and the combined spring-mass system of the bearing housing, bearing supports, and foundation. The bearing oil film for journal bearings i s usually represented by horizontal , vertical , and cross coupl ing stiffness and damping terms. The cross coupl i ng terms significantly affect the cal cul ated rotor response, particul arly i f the horizontal stiffness i s much less than the vertical stiffness. in these cases the rotor instal led resonant speeds cannot be accurately calculated i f the oil film i s represented as a single stiffness and damping value and the effects of the cross coupl ing terms are omitted. Therefore, i t i s important that fan users be aware of the cri t i cal speed definitions used in the fan industry and refer t o the instal led resonant speed when writing design specifications. For t hi s systen major modifications such as shortening the bearing span would have been required t o increase the installed resonant speed t o 2045 above the running speed. Since these modifications were not practical, i t was decided t o reduce the shaft vibrations by improving the fan balance. .- SYSTEM BALANCE I t was desired t o reduce the vibrations on the fan and motor bearing housings by adding balance weights only i n the fan. This means that the vibration levels on a11 four bearing housings must be considered when calculating the balance weights. To detennirie i f the vibration ampl itudes were stable and repeatable, the vibration amp1itudes were plotted versus the phase angles as shown in Figure 2-6. The varia- tion in amplitude and phase angle for each bearing created elliptical patterns due t o the interaction from adjacent fans. The best balance was obtained by reading the vibration amplitude and phase angle at the center of each el l i pt i cal pattern. On one fan a particularly strong interaction from an adjacent fan was observed and a stable vibration phase angle could not be obtained. When the adjacent fan was . tripped, the fan vibrations immediately stabilized (Figure 2-71. To further docu- ment t hi s interaction, the fan vibrations were plotted with the key phaser installed on the adjacent fan which showed that the fan vibrations were more closely related t o the adjacent fan. PHASE ANGLE, DEGREES PHASE ANGLE, DEGREES Fi gur e 2-6. Var i at i on of Motor and Fan Fi gur e 2-7. Fan Vi br at i on I nt e r a c t i on Beari ng Vi br at i on During wi t h Adj acent Fan Hot and Col d Condi t i ons Even wi t h t h e i nt e r a c t i on between t h e f a ns and t h e ampl i f i cat i on due t o t h e founda- t i o n nat ur al frequency and i n s t a l l e d resonant speed, t h e vi br a t i ons on t h e f a ns and mot ors were reduced below 2 mi l s col d usi ng a two-plane bal ance ( Fi gur e 2-6). In t h e bal anci ng procedure, a l e a s t squar es bal anci ng t echni que was used which i ncl udes t h e vi br a t i ons a t a l l f our bear i ng l oc a t i ons f o r t h e two bal ance pl anes t h a t were a va i l a bl e . HOT BALANCE When t he b o i l e r was f i r e d duri ng t h e i n i t i a l steam blows and t he t emper at ur e of t h e exhaust gases i n t he ID f an reached 250' F, t h e vi br a t i ons on t h e bear i ng housi ngs immediately i ncreased and t h e phase angl es changed ( Fi gur e 2-6). The bal ance of a l a r ge f a n normally changes when t h e f a n i s heat ed due t o thermal di s t or t i on; how- e ve r , i f t he f an i n s t a l l e d resonant speed and foundat i on nat ur al frequency a r e near t h e runni ng speed, t h e small changes i n unbal ance can r e s u l t i n s i gni f i c a nt i nc r e a s e s i n vi br a t i on amp1 i t udes. I t was t he r e f or e recommended t h a t t h e system be "hot bal anced" t o reduce t h e bear i ng housi ng vi br a t i ons t o meet t h e r equi r ed vi br at i on 1 i mi t s. Hot bal anci ng f i n s wi t h t h e pl a nt a t f u l l l oad i s a d i f f i c u l t , t i me consuming process. One of t h e major problems was t h a t t h e f a ns di d not have t ur ni ng gear s and t h e r ot or s bowed due t o t hermal s t r a t i f i c a t i o n of t h e gases i n t h e f a n housing. When t h e f a ns were r e s t a r t e d wi t h a bowed s ha f t , t h e vi br a t i ons were excessi ve and t he f a ns were imme- d i a t e l y t r i pped, Each s t a r t u p and coastdown reduced t h e bow due t o more uniform heat i ng of t h e r ot or , Sometimes a f t e r two o r t h r e e s t a r t s and coastdowns, t h e vi br a t i ons would be reduced t o l e ve l s where t h e f an coul d be operat ed. The s ha f t thermal bow problem was l a t e r sol ved by i n s t a l l i n g a temporary t ur ni ng ge a r t o r o t a t e t h e s h a f t when t h e f an was off. I t i s hi ghl y recommended t h a t turning gears be included i n the original design on all hot fans because i t i s dif- f i cul t and expensive t o r et r of i t turni ng gears. I t took approximately 10-12 hours of running time before the fan thermally stabi- lized and the vibration and phase angle quit changing. Balance data taken before t hi s time could result in inaccurate balance weights. The fan and motor vibrations were reduced by hot balancing; however, the fans remained very sensitive t o small changes. The balance weights used were much less than the allowable residual balance as specified for fans of t hi s type [4]. The bearing housing vibrations could not be maintained below the specified level of 2 mils; however, there was no bearing damage because the foundation moved with the shaft. I t was found that the bearing housing vibration levels could be increased to 5 mils peak-peak without causing excessive shaft-to-housing vibrations. The plant operated for several years wi t h these vibration I evels without bearing damage. This i 1 lustrates that a1 1 owable vibration levels must be determined for each individual piece of equipment and published vibration cri t eri a should be used as a guide or reference. CASE NO. 2 Two boilers were converted from forced draft systems t o balanced draft systems w i t h the additfon of two large ID fans on each unit. The ID fans were originally instal led with louver controls which were l at er changed t o variable i nl et vanes t o eliminate high pulsations i n the ducts. The variable i nl et vanes apparently reduced the pulsations in the fan discharge; however, the fan and motor bearing vibration increased t o unacceptable levels when the units were run above half load. The fol 1 owing symptoms were observed: 1. After a few days, the fan bearing housing vibrations would increase i n ampl itude from 2 mils t o over 5 mils and the phase angle would shi f t over 180 degrees. 2. ifot balancing the fans was not effective in reducing the vibrations for a 1 ong time period. 3. The groundborne interaction between the fans was excessive and only one fan could operate at a time. I t was determined that the fan wheel disc wobble natural frequency was slightly above the running speed which ampl ified low level excitations and resulted i n exces- sive vibrations. The problem was corrected by raising the disc wobble natural fre- quency further above the running speed by adding a stiffener plate t o the center plate. DISC WOBBLE The bearing housing vibrations and shaft vibrations were recorded during several startups and shutdowns (Figure 2-8) t o determine i f the fan was operating near a mechanical natural frequency. I t i s known that vibrations due to unbalance alone (discounting resonances) wi 11 increase as a function of the speed squared. However, the vibration at 900 rpm was significantly greater than the level due t o a pure unbalance, which indicated that the fan was operating near a natural frequency. The 1 ack of a 180 degree phase shi f t on the bearing housing vibrations indicated that t hi s resonance was probably not a shaft cri t i cal speed. Figure 2-8. Bearing Housing Vibration Response During Startup Pulsations measured in the discharge duct were synchronous and phase coherent with shaft speed. The pulsation amplitudes were approximately 0.7 psi at the fan running speed and produced a force equivalent t o a 10 oz unbalance at the outside radius of the fan. The fans were apparently sensitive t o small pulsations or aerodynamic excitations because changing the shaft dust covers had a significant effect on the fan vibration characteristics. I t i s di ffi cul t t o reduce low 1 eve1 pul sations; therefore, i t was f el t that the long range solution would be t o identify and modify the mechanical natural frequency which was ampl ifying the vibrations. To identify the resonant frequency, a large variable speed shaker was bolted t o the concrete foundation with the shaking force applied in the horizontal direction and run through a frequency range from well below the fan running speed t o 50% above. Vibrations were measured on the foundation, bearing housings, shaft, and fan wheel t o identify the resonance just above the running speed. The foundation response was found t o be highly damped and offered no significant ampl ification of the vibrations at the running speed. The major effect of the foun- dation was as a vibration transmission path between adjacent fans. The f an shaf t nat ur al frequency was measured t o be 1200 cpm wi t h t he shaft r es t i ng on t he beari ngs. Cal cul at i ons i ndi cat ed t hat t he i ns t al l ed resonant speed when t he shaf t was runni ng on t he o i l f i l m would be lowered t o 1080 cprn which was s t i l l ade- quat el y above t he runni ng speed. Vi br at i ons measured on t he fan wheel i ndi cat ed a di sc resonance a t 930 cpm ( Fi gur e 2-9). There was a cl assi c 180 degree phase s hi f t i n t he vi br at i ons as t he shaker speed i ncreased above t he resonant frequency. The vi br at i ons were measured a t sev- er al l ocat i ons on t he wheel t o determi ne t he vi br at i on mode shape. I DISC WOBBLE MODE SHAPE SHAKER SPEED, CPM Fi gur e 2-9. Disc-Wobble Mode Vi br at i on . Response t o Shaker Exci t at i on The shaker t e s t i l l u s t r a t e d t hat t he ax i al wheel vi br at i ons ar e eas i l y coupl ed i n t o hor i zont al vi br at i ons on t he bear i ng housing because t he wheel resonance was exci t ed by a hor i zont al shaking f or ce on t he foundati on. It was f e l t t h a t t he beari ng housing v i br at i on amp1 i f i c a t i o n a t runni ng speed coul d be reduced by r ai s i ng t he di sc resonant frequency. To eval uat e t he e f f e c t o f s t i f f - eni ng t he cent er pl at e, wedges were dr i ven between t he f an i mpel l er and t he i n l e t cone t o r e s t r a i n t he wheel. When t he shaker t e s t was repeated, t he response a t 930 cpm was el i mi nat ed on t he f an wheel and t he beari ng housings. An evaluation was made of a plate stiffener that could be fabricated and bolted t o the center plate in the field t o raise the disc resonant mode. Since i t i s di f f i - cult t o mathematical 1y model a complex bolted joint, a small 1/10 scale model was fabricated and tested. The model t est confirmed that the center plate stiffness could be adequately increased by bolting a reinforcement plate t o the center plate. After a stiffener plate was installed in one fan, a shaker t est was performed which showed that the wheel natural frequency increased 25%. When the fan was restarted, the bearing housing vibrations were significantly reduced as shown in Figure 2-8. The other fans were then similarly modified and have been in operation for several years and are no longer sensitive t o low level excitations even a t full load. CASE NO. 3 During the i ni t i al startup of two ID fans a t a power plant, high vibration levels were observed on the fan inlet housing which resulted in structural damage requiring extensive repairs. One of the inlet cones was broken away from the common sheet and several of the pipe braces which extend.between the common sheet and the side sheets were a1 so broken. The unit had been in operation for only 11 days during the ini- t i al firing and steam blows of the boiler. The second fan had run for only 4 days and i t t oo had suffered extensive damage similar t o the f i r s t fan. The following symptoms were observed: 1. The vibrations on the fan inlet housing were excessive and increased as a function of the i nl et damper position. 2. The vibrations on the fan bearing housing (pedestals) and foundation were low and were not affected by the i nl et housing vibrations. The problem was due t o an i nl et vortex which created a rotating st al l condition and generated high pulsations a t multiples of 20 Hz. The pulsations a t 40 Hz matched the mechanical natural frequency of the inlet cone and resulted in high vibrations and fatigue failures of the i nl et cone and the internal pipe braces. The i nl et vortex was eliminated by adding spl i t t er plates in each inlet. Accelerometers, strain gages, and pressure transducers were instal 1 ed at several cri t i cal locations on the fan housing t o identify the source of excitation. I t was found that the pulsation, vibration, and strain levels increased as a function of the damper position and the levels were excessive a t a damper position of approxi- mately 70% open. Strain levels of 350 in/in x peak-peak were measured which were higher than the allowable for t hi s material and configuration. The predominant pulsation and strain frequencies were at multiples of 20 Hz (Figure 2-10). Rotating st al l generally occurs a t multiples of 1/3 shaft speed. This fan ran at 15 Hz and the pulsations at 20 Hz were at 4/ 3 shaft speed. The st al l was due t o marginal flow or improper preswirl conditions a t the fan inlet which caused the a i r to impinge on the fan blade at a poor angle of attack. FREQUENCY, HI Fi gure 2-10. I n l e t Pul s at i on and St r a i n Spect r a Showing Mu1 t i pl e s of Rot at i ng St a l l Frequenci es Several modi f i cat i ons suggest ed by t h e manufact urer were made t o t he duct s and t ur ni ng vanes upstream of t h e f an, but no r educt i on of pul s at i on o r vi br a t i on occurred. Test i ng was t hen conducted t o det ermi ne t h e e f f e c t of di scharge' t hrot t l e cont r ol . Flow through most f a ns i s cont r ol l ed on t h e i n l e t s i de and o u t l e t dampers a r e used f o r f an i s ol a t i on. For this test, t h e o u t l e t dampers were t empor ar i l y s e t up f o r manual cont r ol . The i n l e t dampers were opened t o 45% which r e s ul t e d i n hi gh pul s at i ons and vi br at i ons. The i n l e t dampers were l e f t on a ut a na t i c cont r ol and t h e o u t l e t dampers were p a r t i a l l y cl os ed, keeping t h e fl ow r a t e const ant . When t h e o u t l e t dampers cl osed t o 5 5 X, t h e i nTet opened t o 51% and t h e vi br a t i ons and pulsa- t i o n s immediately reduced. Thi s t e s t suggest ed t h a t a f i xe d o r i f i c e i n t h e di s - charge duct coul d reduce t h e pul s a t i ons and vi br at i ons , but i t s e f f e c t s a t f u l l l oad were not known. A pr oper l y desi gned damper t o cont r ol di schar ge fl ow was impossi bl e t o obt ai n and i n s t a l l i n a r easonabl e t i me peri od; t he r e f or e , t e s t i n g was cont i nued t o obt a i n anot her s ol ut i on. I t was f e l t t h a t an i n l e t box vor t ex coul d be cr eat ed by t h e i n l e t dampers and t h e i n l e t duct conf i gur at i on C51. The i n l e t vor t ex coul d cause improper i n l e t flow con- d i t i o n s and r e s ul t i n a r ot a t i ng s t a l l condi t i on. In or de r t o des t r oy any vort ex forming t endenci es, s p l i t t e r pl a t e s were i ns t a l l ed i n each i n l e t box ( Fi gur e 2-1 1 ) d i r e c t l y oppos i t e t he i n l e t dampers. The vi br a t i ons , pul s at i ons , s t r a i n , and noi s e l e ve l s were gr e a t l y reduced. In addi t i on, t h e a i r fl ow a t comparable vane s e t t i n g s was i ncreased by a f a c t or of 1.7. SPLITTER PLATE TO BREAKUP VORTEX Figure 2-11. Inlet Flow Spl i tter Used t o Reduce Inlet Vortex Section 3 CONCLUSIONS The case hi stories discussed i l l ust r at e that fan vibration problems are usually system related and caused by mechanical or acoustical interaction between the fan, motor, foundation, and ducts. Systematic diagnostic fie1 d t est procedures and com- puter analyses of individual components are usually needed to define exact causes and optimum solutions. Several problems common t o fans were identified and guide- lines for analyzing and solving the problems are summarized below. FOUNDATIONS Foundation natural frequencies should be at least 20% away from the fan running speed because they can amp1i fy fan and motor vibrations. Accurate dynamic soil properties must be known t o properly simulate foundat ion-soil systems. The dynamic soil modul us should be experimentally measured at the s i t e of each foundation aft er the excavation has been completed and the backfill has been added. Most foundation vibration problems are due t o a rocking mode about a point below the foundation. Foundations are often designed as t al l rectangular blocks with the height greater than the width. By increasing the width of the foundation block and mat, the lateral st abi l i t y of the entire system could be improved. The outboard concrete pedestal i s typically underdesigned for the dynamic radial and axial forces from the fan. The dynamic stiffness should be canparabl e with the inboard pedestal. SHAFT LATERAL CRITICAL SPEED 1. The fan shaft "installed resonant speed" as defined by the AMCA should be used in the design or specification of a fan system. 2. The installed resonant speed should be a t least 20% removed from the running speed. 3. To accurately calculate the rotor installed resonant speed, a forced vibration response analysis should be performed modeling the bearing oil film stiffness and damping characteristics with the eight coeffi- ci ent representation. B ALANC I NE 1 . When bal anci ng, extraneous i nf1 uences such as excessive mi sal ignment, fan interaction, etc., need t o be minimized t o obtain the proper bal- ance data. 2. For sensitive systems alignment specifications may have t o be tightened. 3. Hot balancing often requires waiting for 10-12 hours for a large fan t o stabilize and q u i t shifting. The fan i s generally stabilized when the rotor stops growing i n the axial direction. Balancing w i t h data taken before thermal stabilization may result i n ineffective correction. 4. The 1 east-squares mu1t i pl ane bal anci ng method a1 lows mi nimi zi ng vibration at any number of points. This method has advantages for systems operating near resonance and can minimize the number of bal- anci ng runs which reduces the required downtime. 5. A11 hot fans should have t u r n i n g gears. WHEEL RESONANCES The disc wobble resonance should be at 1 east 20% removed from the running speed. Fans wi t h the disc resonance close t o running speed can be overly sensitive t o low level excitations such as pulsation and unbalances. The disc axial resonant vibrations can cause amp1ification of hori- zontal vibrations on the bearing housings. Disc resonances can be experimentally determined by impact t est s or shaker tests. All centrifugal fans should be checked during manufac- turi ng to determine di sc resonant frequencies. The disc wobble resonant frequency in the example was increased by stiffening the center plate. A stf ffening plate could provide a long-term fi x for existing fans with t hi s problem. ACOUSTIC EXCITATION The inlet ducts should be designed t o prevent forced inlet vortices (spi n-swi rl ) . The i nlet vortex causes improper fl ow conditions and results i n reduced fan performance and possibly rotating st al l on the fan blades. Pulsations generated by rotating st al l usually occur a t multiples of 1/3 shaft speed and, when sufficiently 1 arge, can cause structural damage. Spl i t t er plates can be installed i n the fan i nl et t o break up the i nl et vortex. REFERENCES 1. D. R. Smith and H. R. Simmons. "Unique Fan Vi br at i on Problems: Thei r Causes and Sol uti ons." Col l ege St at i on, Texas: Proceedings o f t he 9t h Turbomachinery Sym- posium. Gas Turbine Laborat ori es, Texas A M Uni ver si t y, pp. 33-43. 2. S. C. Arya, M. W. O' Nei l l , and G. Pincus. Design o f St r uct ur es and Foundations f o r Vi br at i ng Machines. Houston: Gu1 f Pub1 i shi ng Company, 1979, pp. 62-64. 3. Power Pl ant Fans Speci f i cat i on Guide1 ines. A i r Movement and Cont rol Associ ati on, Publ i cat i on 801, 1977. 4. "Balance Qual i t y o f Rot at i ng Ri gi d Bodies." Acoustfcal Soci et y o f America, ASA sm 2-1975. 5. Fan Appl i c at i on Manual, Par t 1. A i r Moving and Condi t i oni ng Associ ati on, Pub- l i c a t i o n 201, 1976.