PIPING VIBRATION ANALYSIS by J. ©, Wachel President Scott J. Morton Project Engineer and Kenneth E. Senior Project Atkins Engineer Engineering Dynamics, Incorporated San Antonio, Texas J.C. “Buddy” Wachel is Prosdent of Engineering Dynamics, Incorporated, an independent consulting firm. He hasocer 2B years of experience and has published over 30 technical papers. He has solved problems worldwide in piping vibrations, lateral and torsional critical speeds, rotor instabilities, and crankshaft follures. In addition to the practical field experience, he supervises the computer design audits of machinery and piping to ensure that te systems will have acceptable vibrations ond stresses. ‘Mr. Wachel hasa BSME and MSME degrees from the Univer- sity of Texas. He isan ASME Fellow and a member of ASM, The Vibration Insitute, and isa Registered Professional Engineer in the State of Texas Scott J. Morton is a Project Engineer swith Engineering Dynamics, Incorpo- rated. He has experience in’ computer ‘modelling in finite elements, rotor ‘dynamics and acoustics. He is also in- tolved in the diagnostic ila testing of rotating and reciprocating. equipment and in the evaluation of structural and piping etbration problems. Mr. Morton holds Bachelors and Mas ters degrees in Maritime Systems and ‘Ocean Engineering from Texas A&M University and isa Reals tered Professional Engineer in the State of Texas us Kenneth E. Atkins is a Senior Project * Engineer with Engineering Dynamics, Incorporated, and has experience in per forming lateral and torsional eriical speed analyses, rotor stability analyses ‘and the evaluation of structural vibration problems using finite element methods. He has been actively involved in field troubleshooting of aide variety of rotor- dynamics, structural, and piping vibra tion probiems. Mr. Atkins received a B.S. degree in Engineering Science {from Trinity University in 1978. He 1s a member of ASME and isa Registered Professional Engineer in the State of Texas, ABSTRACT Excessive piping vibrations are a major cause of machinery downtime, leaks, fatigue failures, high noise, fires, and explo- sions in refineries and petrochemical plants. Excessive vibration levels usually occur when a mechanical natural frequency ofthe piping system is excited by some pulsation or mechanical Source. The vibration mode shapes usually involve lateral vibra- ‘ons and/or shell wall radial vibrations. Simplified methods are presented for analyzing lateral and shell wall piping vibrations and jodsing, their severity. The methods are thought to be conservative and are intended to be used as screening eriteria to determine if more sophisticated analyses, such as compiiter stress modelling or strain gage test ing are necessary. Frequency factors for calculating the mechan: ical natural frequencies for the classical piping configurations (uniform straight beams) and various piping bend configurations are presented. Factors are presented to compensate the natural frequency calulations for concentrated and ditbsted weight ‘The relationships between piping vibration displacement, ve locity and stress are presented and criteria for judging the ever: ity of piping vibration in terms of the endursoae stress Kit are shown. The mechanisms thet ean excite piping vibrations willbe discussed, as well as methods for controlling their severity INTRODUCTION Piping vibration failures have been one ofthe major causes of downtime, fres and explosions in industrial plants over the past 30 years, For example, one piping failure at a petrochemical120 PROCEEDINGS OF THE NINETEENTH TURBOMACHINERY SYMPOSIUM plant in 1974 caused over $14,000,000 in property damage [1], ‘due to an explosion. In nuclear pressurized water reactor power plants, over 80 cases of eracks ot leaks occurred in the piping systems of charging pumps over a two-year period (2). There- fore tis vitally important thet piping vibration amplitudes ina system be evaluated to determine if the levels are acceptable. I the vibrations levels are judged to be excessive, the piping coofiguration, support structure, span length, or material may hhave to be modified to make the system acceptable. Alternately, if these factors cannot be changed, the excitation mechanisms ‘must be altered or eliminated, In order to make practical modifi- cations tothe piping to solve vibration problems, iis necessary to understand all the principles involved in the determining the natural frequencies and the excitation sources that cause the problems, “Methods are presented for calculating the natural frequen- cin The lini atten blo pe ad nae stress is shown In addition the excitation souraes that cause exces- Sive vibrations and methods for minimizing theie harmful effects ae diseussd. ration problem areas of typical piping systems include the excitation ofthe following: Piping span natural Frequencies Piping shell walleircumferentil and aval natural frequencies + Piping appurtenances (vent and drain lines, gage, and test connections) ‘Valves and valve components _ erctingcompresoelinderand manifold bot nat. ural frequencies "The principles involved in understanding the behavior of pip- {ng vibrations ofthe components listed above are covered CALCULATION OF PIPING NATURAL FREQUENCIES [3-8] “To ensure that piping systems are fee from excessive vibra- tions, i is necessary that the individual piping spans not be mechanically resonant to system excitation frequencies gener ated by compressors, pumps, flow excitation meckanisms, et ‘To accomplish this, the frequencies ofthe excitation fores and the mechanieal natural frequencies of the piping must be calcu lated. With experience, simplified design procedures ean be used to evaluate the piping system with a minimum of detailed computer analyses. For complex systems, stress analysis com- ter programs should be used to evahste piping system reliably Straight Piping Spans Actual piping span natural frequencies deviate from the theo- es exist in typical plant piping bave boundary conditions that dif from ideal values. Nevertheless, ideal beam theory gives avalu- able starting point for understanding piping vibration behavior, ‘The natural frequency of any piping span can be calelated if the frequency factor, the span length, the diameter, wall thcke ness and the weight per length are kaown. Fora straight uni- form piping span, the natural frequency can be calculated using the following relationship @ where: f, = Span natural freyueney, He § = Gravitation constent, 386 in/sec* E = Modulus of elasticity, psi 1 = Moment of inertia, int 1 = Span length, in, A. = Frequeney factor, dimensionless = Weight per unit length of beam (including fluid and insulation), Ibs/in b= Density, Ibvtin® A. = Pipe cross-sectional area, in? By substtting in material propotios fr steel, E = 90 10° Tbfin®, p = 0.285 Ibvin?, and g = 386 in/sec*, Equation 1 can be simplified to: f, = 2290, +t @ where: I= radius of gyration, inches L= length of span, ft Note that this equation does not include the weight of the ‘uid and the insulation. The frequency factors (\) for caleulating the first two natural frequencies for ideal straight piping spans are given in terms ofthe overall span length in Figure 1 oo TSE Peo FA __ [tov an 1 FE [rmtrow fae ras asa) small | epee ou {4s sn | muy sa¥e8| 20: am ale =| Gh (RIL I Tro | | raat ou|a4 nal am[a tl ot |ane mals asl sna} ai 1 = Forme | P= ey | 5= Kav SOP |S = KV SCF tows | yi | ye | eee 0 p03 B=) Figure 1. Frequency Factors and Stress Factors for Uniform Steel Pipe Configurations. Piping Bends ‘The natural frequencies of selected pipe configurations with piping elbows (L-bends, U-bends, Z-bends, and three dimen-‘TUTORIAL ON PIPING VIBRATION ANALYSIS 121 sional bends) were analyzed. using a finite element program (ANSYS) to generate frequency factors fer the first two modes. Inthis analysis, acutved beam (elbow) element was usedso that more accurate frequency factors for the piping configurations could be established. In doing so, the frequency factors may be slightly different from other published data for square comer ‘beams or beams without cylindrical cross sections. The fie ‘queney factors were generated for a range ofaspect ratios tode~ velop suffiient information so that the natural frequency of pip- ing spans could be approximated regardless ofthe eonllguration ‘The accuracy of the analysis was verified by comparison of the frequency factors with the theoretical values at the limits of the aspect ratios. The frequency factors asa funetion of the aspect ratios ofthe leg lengths are given in Figures 2, 3, 4,5, 6,7, 8 rasqvenc excron, 3 ya vee tener RATIO Figure 2, Frequency Factors for Uniform L- Bend Piping Config- rations anquenoy meroR, 4 s yA tec ueverm nacIO Figure3. Frequency Factors for Uniform U-Bend Piping Config trations for First Outof-Plane Mode. REQUENCY FACTOR, 4 i ee sya ec texera Ramo Figure 4. Frequency Factors for Uniform tuations for First In-Plane Mode tend Piping Config- PREQUENCY FACTOR, + Tae ere yyatee emer RADIO Figure 5. Frequency Factors for Uniform Z-Bond Piping Config- turations for First Out-of Plane Mode. Effect of Concentrated Masses Applying energy methods (9, it ean be shown that the frst natural frequency ofa beam with aconcentrated lad canbe cal- culated by: @ Pipe span mtual frequency with concentrated weight, Hewe PROCEEDINGS OF THE NINETEENTH TURBOMACHINERY SYMPOSIUM {= Pipe span natural ftequency without concentrated weight, Hz Concentrated weight, Ibs Weight of beam span, Ibs B= Weight correction factor, dimensionless Weight correction fctors tobe usedin calculating the natural fiequencies of ideal pipingspansforweightsatthemaximumde- —& flection locations are given in Figure 8. two weights ae lo. cated in one spn, the fllowing equationsean be used toealeu- g arc uexorm RATIO : Figure 8, Frequency Factors for Uniform 3D-Bend Piping Con- 2 figurations for Second Mode. i Es Inte the effet of the second weight. The frequency for one weight Pi yA bec uewere Rario Figure 6. Frequency Factors for Uniform Z-Bend Piping Config- trations for First In-Plane Mode. ‘uequency #4Cr0R,» ya ue0 LENGTH RATIO Figure 7. Frequency Factors for Uniform 3D-Bend Piping Con- {figurations for First Mode If the second weight in the span is considered by itself, the equation is: i-— C) Vive ‘The frequeney for the span with both weights can be obtained from the following equation: f+ 6 Correlation of Caleulated and Measured Natural Frequencies ‘Theoretical beamn natural frequeney calculations ean be com rected to make them agree more closely with measured field data (7]. The correction factors, given in Table 1, are based on the non-ideal end conditions typically encountered in actual pip- ing installations. ‘The procedures for caleulating the natural frequency ofpiping spans can be used to select clamp spacings which ensure thatthe piping spans will be resonant above some selected frequency. ‘The use of these correction factors will normally give answers ‘that are within 15 percent of measured values. For the majority ofpiping vibration problems, this accuracy should be suficient.TUTORIAL ON PIPING VIBRATION ANALYSIS 133 otha Sith “gre ee ee sm Bog aa rowma ye » Figure 9. Weight Correction Factors for Uniform Piping Config rations Table 1. Natural Frequency Correction Factors For Piping End Conditions. Ping Conunine ae ay Ed Contin Sypeena Crane chceleee ne Sige Wald Fontfad aga Sight Speed stperied Fen-Fie a Sieiet West Put ne 05-18 Sinidt——Seppene Pe poem gaa Peat anatad o Siig Pye Pac Fae e wate wo Esp net Shell Wall Vibrations High frequency piping shell wall vibrations can be caused by excitation of eircumferential radia) mode natural frequencies (6), The nodal patterns are illustrated in Figure 10 fr simply supported eylinder showing he combination ofthe lateral beau vibration modes and circumferential modes. A number of theories are used to calellate the natural frequencies and the stresces due to shell wal vibration. According #0 Blevins (6), the Flugge and Sanders shell theories are generally felt to be the most accurate. igure 10. Axil «nd Circumferential Nodal Patterns for Simply Supported Cylinder Without Axial Constraint ‘The curvature ofthe shell ouples the Mexural and extensional vibrations and considerably complicates the analysis af shell batons, The shel theories describe the motion of the shell im {eras of an eighth-order ciferential equation. Because of the complet ofthe shell equations and their solutions, few closed- form solutions are available for the natural Frequenctes and mode shapes of shells, Blevins [6] gives the illowing relationship for calculating the natural frequencies for eylindicl shells of tfinite lent ® where: f= Shell wall natural frequency, Hz A, = Frequency factor, dimensionless J = Mean radius of pipe wall, inches issn’ ratio lass density of pipe material, Ib-secin* ipe wall thickness, inches Mode number, 1, 2, Amold and Warbuston [9] investigated the effects of the end conditions on the natural frequencies of shells and compared rmeasuted test results with their calculations, Fung, Sechler and1 PROCEEDINGS OF THE NINETEENTH TURBOMACHINERY SYMPOSIU Kaplan [10] included the effects of internal pressure on the nat- tural frequencies of shells. Fora steel pipe with Poisson's rato of 0.8, the equation ean be simplified eo FE ssH9) ® “The frequency factors fonction of the thickness divided by the mean aus, therefore, the natural requeney varies ineasly with the shell wall thickness and is inversely proportional tthe Square ofthe radius VIBRATION-INDUCED STRESS LEVELS In order to determine if piping vibration amplitudes are ae- ceptable, the resultant dynamic stresses caused by the vibra- tions must be compared tothe allowable endurance stress lim. ‘Toaccomplish this, the maximum stress in a piping span vibrat- {gat resonance must be expressed as a function ofthe dynamic deflection or velocity measured atthe maximum vibration point within the span ‘There have been attempts to develop criteria for acceptable piping vibration levels as afunetion of frequency. Probably he most widely used are the vibration amplitude vs frequency charts [4, 7] that were developed, based on experience inthe petrochemical industry. These amplitude versus frequency ‘hats are used asa sereening criteria inthe evaluation of piping systems experiencing high vibration levels, "The material presented inthis scetion gives the relationship between vibration and stress in typieal piping configurations and presents definitive methods for evaluating piping system re- Tiablty using the actual vibration-induced stresses Vibration Displacement Amplitude Vs Frequency Criteria “The vibration versus frequeney riteria chart given in Figure areas bo ured ovat csluation of th soe fap bration problem. These curves are based on experience and Ihave been used in the petrochemical industry for over 25 years with good suecess [4,7] These eriteria are very conservative for Tong flexible piping spans, such as those used in centrifugal VmeAmtox Fanquancy, 2 Figure 1. Allowable Piping Vibration Loves Vs Prequency. equipment plant piping They are nt applicable to shell wall stron, ‘The author’ experiences have shown that, whenever piping ‘shraton amplities st the measured equences se peter than the danger line, piping flres area yp ocumence. ‘When vibration level were below the design ine, very ow a= Us live occurred. Therefore hese vibration versus equency exter cn serve at» good Hattog point in evalating piping “ibvtions to sereen those sytetsthet need further analyse. Stress as a Function of the Vibration Displacement (Deflection) Amplitude etter method to evaluate the severity of piping vibration deflection amplitades ts to compare the maximum resonant vibration-induced dynamic stresses to an allowable endurance limit stess. There is general agreement that the low eyele fatigue curves for earbon steel given in the ASME USAS B31.7~ 1969 can be used to obtain an acceptable endurance limit stress {i}, ANSVASME Code OM3-1957 [12} uses this stress versus exclescto-fulure curve as a basis for specifying eriteria for evaluating the vibration-induced stresses in nuclear powerplant piping for preoperational and startup testing. API Standard 618 {13} uses the same deta to specify the allowable dynainie stress Jevel fr steel pipes as a design requirement. ‘OMSis the frst code that has attempted to establish amethod for evaluating piping vibration-induced stresses based on mea- sured. resonant’ vibration amplitudes or velocities. The methodology used in the code involves a vhree-step process in determining the acceptability of piping vibsations. The ist step is ategoried as Vibration Montoring Coup 3(VMG3}) nd ‘volves a visual or perception wallovn ofthe piping to deter tine if the vibration are acceptable, based on the experience Of the analyst with the typeof piping system being examined I, in the judgement ofthe anslst, the vibrations are nt obvi sly safe, the pipingis judged o bein the Vibration Monitoring Group 2 WMC2). In VMG2, the acceptability of the piping vi trons edged by seen exiting he rir induced stresses by measuring the vibrations and calculating etoosnduced” stressor by implied. method, These tnethods are based on modellingthe vibration portion of he pig- via a simple beam analogy and determining the vibration limits in terms of the displacement or velocity. (Note that this paper gives the information required to make these caleula tions.) ‘The third category is Vibration Monitoring Gm 1 (WMG) and ivolves a rigorous qualietion method requiting thatthe vibrational sizesses be determined with a high degree “of nuracy. VMC1 qualification may involve detailed correlation, between analysis and experimental esults or instrurgentation of the piping with asuficient number ofstrain gages to determine the magnitude ofthe highest stresses. In VMI, computer mod- els ofthe system are developed, in conjunction with the meas- ‘ured vibration amplitudes, to predict the maximum vibrtion- Induced stresses “The same methodology is used in evaluating piping vibrations ‘throughout the petrochemical industry; however, the methodol- ‘ogy Is not specially detailed in an applicable code. Olson [2] hhas compared the acceptable vibrations determined by using the ANSVASME Gode OM 3 {12}, which i based on stess, to ‘he amplitude versus frequency eutves presented in Figure i “The vibration-induced stress in piping span vib nance has been shown to be related tothe maxi amplitude (2eflection) in the span 4, 5,6, 7]. The relationship is given in the equation below ao)TUTORIAL ON PIPING VIBRATION ANALYSIS 2 where, S = Dynamic stress, psi Ky = Deflection stress factor y= Maximom vibration amplitude (deflection) mea- sured between nodes (normally at supports), mils D = Outside pipe diameter, inches L = Span length, fe SCF = Stress concentration factor “The deflection stress factor afanetion ofthe boundary con- ditions and the vibration mode shape at resonance. Blevins [6] {Gres the normalized vibratory mode shaper at esooance or the classical beams. This data cant be used to generate the deflection Stross ctr fo these modes sing the methods presented by ‘Wachel [4]. The deflection stress factors for the first two modes cf the ideal chsal beams aod the piping configurations with tlbow! are als given n Figure |. These actorsare uted toca. inte thestresal the piping pan natural requeney and the stress ts tobe eomectd i the plpe fe vibrating at 8 ilrent Be- quency. Mode correction factors are given in a later section. Tor the piping configuration sith elbows, the ses defo tion fctors were aleaated withthe fate clement program ANSYS. The accuy ofthe eesatons was ered by com parison withthe fers obtained for the caseal beams. ‘The lta the deflection stress tors are given in Fires 2,1, Tab, 16,7; and 1 forthe various piping conSgurtions wth trend for the out-of plane and the plane modes, The sess ed inthe clelatons es the esto resin stress sine Staton pigs the maine Tonal sreses “Theallowae vibration amplitde canbe ck based on theendurance init, OMS species 0000 psizere to peas the alowable endurance mit or carbon tel nd spe thatthe toinimum safety factor 13 which equate 0 00 pol zero 10 le AFTOIS states that th stresses sal De ss than 2500 ps Pal to peak, or 13000 pal zero to peak, and is normally sed Sh vey oro Te allow ation ome given by: | i a I z = at g Z | 2 oe 1 4 zg |_| : e eee al or ak Pee aan | [ sre nme atten |] a bya rag swat RArI9 Figure 12. Deflection Stress Factors for Uniform Steel L-Bend Configurations. DEFLECTION STRESS PACTOR, Ke ByADEG LENGTH RATIO. Figure 13. Deflection Stress Factors for Uniform Steel U-Bend ‘Configurations for First Out-of-Plane Mode. 8, LB v sone (x0) on a teil eo Scr = Suomen ter Sr > sano ta KG = Desens tr where: s, ‘ya tee LexorH RATIO Figure M. Deflection Stress Factors for Uniform Steel U-Bend Configurations for First In-Plane Made, Ifthe API 618 allowable of 26000 psi peak to peak is used as the endurauce limit combined with stress concentration factor of 4.33, a safety factor of 2, and a stress defletion actor of 3000126 PROCEEDINGS OF THE NINETEENTH TURBOMACHINERY SYMPOSIUM bya Lea LeNrH sao Figure 15. Deflection Stress Factors for Uniform Steel Z-Bend Configurations for First Outof Plane Mode. T i =e DEFLECTION STRESS FACTOR, Ke i ByAtga uaneru waTIO Figure 16. Deflection Stress Factors for Uniform Stel Z-Bend Configurations for First In-Plane Bode {applicable for a fixed-ixed pipe), the allowable vibration in peak to peak mils ean be ealculated, Equation (11) becomes: Le ; 4 “Rate of Thumb 12) ‘This can be used (conservatively) as a screening criteria for straight runs of piping or for piping with bends, based on the deflection stress fators given n Figures ad 1218. The span length isthe length between megsuted vibration nodes which [ =a (oe Ee | PTriil : ANTI - TOES i feat _ “| Ste T ee Figue if. Deflection Stress Factors for Uniform Steel 3D-Bend Configurations for First Mode DEFLECTION sTHESS FACTOR, Ka i sya Leo nora RAT Figure 18. Deflection Stress Factors for Uniform Steel 3D-Bend Configurations for Second Mode. are normally tthe supports, Ths criterias overly conservative Forcantilever beams. Stress asa Punction of the Vibration Velocity {Ina piping span vibrating at resonance, itis also possible to relate the maximum stres to the measured velocity [4]. Inorder todevelop a closed-form solution of the dynamic stress asa fane- tion ofthe velocity, the radius of gyration has to be expressed as ‘afunetion ofthe outside diameter of the pipe. A comparison of the radius of gyration for different sizes of pipe versus the simplified equation of 0.34 D, where D, the outside pipe diameter shows that, fora significant range of pipe sizes, this‘TUTORIAL ON PIPING VIBRATION ANALYSIS wt simplified equation is within afew percent for pipe schedules from 10 to 160 (4]. By making the substitution of 0.34 D, for the radi of eration, the relationship of the maximum pipe veloc ity in the span to stress ean be developed. The result show that the stress in an ideal beam equal ta constant, Ky mulled by the masimu velocity measured nthe piping span “The velocity stes factors forthe ist two modes are gin in Figure I forthe classi! types of straight spans as defined by the end conditions For thepipingconfigurations with piping elbows, the velocity stress factors were algo caleulated tn the analysis which de- ‘eloped the frequeney factors andthe deflection stes factors “The velocity stress factors are ven in Figures 19-25 for the frst two modes forthe various aspect ratios ofthe leg lengths c t i I rT H cat i Set ‘veLociTY STRESS PACTOR, IK, Dya ues Lenore RAcIO Figure 19. Velocity Strees Factors for Uniform Steel L-Bend Con- Agurations. “The actual maximum span stress is equal tothe velocity stress factor times the maximum measured velocity times the stress ‘concentration factor. This equation for the stress is: S]K,xVXSCF a3) where: S = Dynamic stress, psi K, = Velocity stress factor SOF = Stress concentration factor V = Maxime velocity in pipe span, in/see ‘Theallowable velocity isalsoa function ofthe enduranceliinit and is given in the Equation (14). To account far system un- knowns, itis necessary to include a safety factor, usually 2 for fatigue analysis * SEXSPSGF an LYRLOcITY STRESS TACTOR,K 57a tes eewerx nario Figure 20. Velocity Stress Factors for Uniform Steel U-Bend Configurations for First Out-of-Plane Mode, | Figure 2 Velocity Strees Factors for Uniform Steel U-Bend Con- Figurations for First In-Plane Mode. where: V,__ = Allowable vibration velocity in pipe span, in/see S, = Allowable endurance limit stress, psi K, = Velocity stress fitor SCF = Stress concentration factor SF = Safety factor In calculating allowable vibration, st customary to use the zero to peak stress allowable, since velocity is always expressed 2 zero to pet, Based on an allowable endurance iit of 13000 psi zero t0 peak, a matimum velocity stress constant of 318, a128 PROCEEDINGS OF THE NINETEENTH TURBOMACHINERY SYMPOSIUM \vevocrry staass FACTOR, K, aya Leo Levon mato Figure 22. Velocity Stress Factors for Uniform Steel Z-Bend Configurations for First Out-of Plane Mode. the first natural frequency when calevlating the actual dynainic stresses ofa piping span, For evaluating the maximum vibration-induced stresses, the equation becomes: Sq=(8) (SOF) Ky Ka Ky on where: Sq = Maximum dynamic stress, psi S = Dynamic stress caleulated at the natural frequency, psi ay vevocrry stness cron, a 9 Nie eee i” mai : < J ae Figure 23. Velocity Stress Factors for Uniform Steel Z-Bend Configurations far First In-Plane Mode. ‘stress concentration factor of 5, and a safety factor of 2, the ale lowable velocity is equal to: exe 2 #4 infsee zero to peak 08) (Other Considerations for Vibration-Induced Streses Te is necessary to consider other factors, suchas concentrated or distributed weights, and responses at frequencies other than ya. Wa RATIO Figure 24. Velocity Stress Factors for Uniform Steel 3D-Bend Configurations for First Mode 3 : F z yan Nora RAO Figure 25. Velocity Stress Factors for Uniform Steel 3D-Bend Configurations for Second Mode‘TUTORIAL ON FIPING VIBRATION ANALYSIS. 129 Ky = Stress correction factor for concentrated weights Ky = Stressoorecton factor for moderfrequency considerations Ky = Stresscortecton factor for pipe contents and insulation Stress Weight Correction Factors (K;) ‘When a concentrated weight is located in a pipe span vibrat- ingatis first atual fequency, the strestscleuated using the norm beam equations canbe in error, therfore, coretion Eetors must be applied to the sires caltltions. The sect steight correction factor given in Figure 26 can be ued to com Tents the dynamic street ealculatonr for the elec of concen- Hated weights on piping spans with the lst! boundary con- Aiton. Two curves are presented, one which gives the fadors tiicionaa coker eonecing bose shes ed mand the otherfr correcting the clolated steses on messured ciation eeloetty, To obtain these tr, concen- trated weights were placed at the vibration antinode locations for ‘Re canblever, sinpl-tportd;Bred-ppored, fe Bred, L- Bend, UsBend, Zends and 3D-Bends piping confgwations ‘The cover are composite curves which shold be conservative for piping spans vibrating at their first natural frequency. This factor increases the stress; therefore, the allowable vibration is reduced tthenever concentrated weight present sT1ess WEIGT CORRECTION FACTOR, Ki ‘mo OF CONCENTRATED WEIGHT TO SPAN WEIGHT Figure 26. Stress Weight Correction Factors (ky). For these curves, iFthe concentrated weight isnot exactly at the antinode, the weight correction factor can he approximated by linear interpolation Mode Correction Factor (Ks) I the piping span is not vitating at its fst lateral bending, mode, mode correction actors whieh depend upon whether the ‘pan i vibrating above ordelor the first natural frequency must be applied. ‘If the piping spar is vibrating below its lowest lateral bean vibration mode, the mode shape wil be similar in shape tothe Statie deflection mode shape and will gradually change to the vibration fist mode shape asthe frequeney approaches the fist natural frequeney. Data analyses performed to determine the ‘mode correetion factors are summarized in Figure 27 [4] for the deflection stress method, These data show thatthe stresses eal- culated using the vibration deflection for the piping span vibrat= lg at it first mode will be within a few percent of the correct stress for most piping configurations. DEFLECTION MODE CORRDCTION FACTOR esquecr nario Figure 27. Off-Resonant Mode Stress Correction Factors for De Jflection-Stress Calculations (Ke). If the piping pan is vibrating at requencies higher than the fies natural frequency the calnlated testes based onthe de- flection made shape atthe frst natural requency an be fn err. ‘The mode correction factor needs to be developed for a particu- lar configuration this method sta bewsed for requencies that ‘ae greater than approximately 50 percent above the first natural frequency. the stresses are calculated using the velocity stress method, the mode correction factor from Figure 28, given as a function ofthe frequency ratio forthe classical beams, ean be used [4] Tt ean be seen that the mode comection factor is inversely pro- portional tothe frequency; the lower the frequeney, the higher the mode correction facto. This means thatthe velocity stress caleulations should not be used for frequencies below the first natural frequeney unless the exact mode correction factor is de~ veloped. This is recognized in OMS which specifies that the stresses should be multiplied by the ratio ofthe natural fre~ quency to the exciting frequency. ‘When the excitation frequency is higher than the fist natural frequency, Figure 28 shows thatthe stresses calculated at the first natural frequency would be conservative Ifthe piping span is vibrating at its second natural frequency, the deflection and/or velocity stess factor given forthe second mode should be used to ealeuate the stresses. The natural fe~ Aaveney forthe span length used should match the measured fe- quency or the resulting calculations could be in error. In com- plex piping systems, this an bes problem, since adjacent spans tan cause severe off-resonance vibrations. Stress Correction Factor for Weight af Pipe Contents and Insulation (Ka) ‘The primary effect ofthe increased weight of piping contents ee since the insulation and contents do not add appreciable sti:130 PROCEEDINGS OF THE NINETEENTH TURBOMACHINERY SYMPOSIUM ‘veLoorry noo conECHON FACTOR requency mAs10 Figure 28. Off-Resonant Mode Stress Correction Factors for Velociey-Stress Calculations (Ks. ness. In the original derivation of the natural frequency, the linear density was in the éenaminator of the square root fune- tion. Therefore, the natural frequency i reduced by the rato of the square root ofthe original linear weight pex unit length to the new overall weight per unit length. For the stess calculations using the dellection stress factors, the correction factor is 1. For the stress calculations using the velocity stress factors, the correction factor will be equal to: | as) ‘Weight per unt length of contents Weight per unit length of insulation ip = Weight per unit length of pipe A screening velocity value can be obtained by assuming val- ues of the correction factors that are maximum. OMS arrived at 4 sereening eriteria of 0.5 in/see; however, this number has proven to be very conservative in many piping systems (14) OMS uses a factor of Ky of 1.5 and a K, of8, The concentrated weight corteetion factor K, of 8 that is used was bused on a con- ‘eentrated weight to span weight of20 to 1. In practice, the ria {more likely tobe less than to 1. Fora maximum concentrated weight to span weight of3, K, would be approximately 27, and the screening eiteria would be LS in/see. If vibration measure- ‘ments indicate that the screening eviteiais exceeded, the actual stress factors and the correction factors should be applied for the span and the acceptably f the vibrations based on these numbers. ‘Shell Wall Vibration-Induced Stresses ‘When the high fequency piping shell wall wibrations (axial and cireuraferential mode shapes) ae exited, the equations re~ Inting vibration displacement, velocity and aceleration tothe piping stress based on the lateral beam vibration between sup- ports do not apply. Therefore, the equations relating vibration Emplitdes to stress for these modes must be determined in ‘order assess the reliability of the vibrations ‘The stresses in the shell wall are given by the following iso- twopie stress strain relationships: Le lect ve) (9) E Tor 6) 20) where 5, = Lateral bending stress, psi Sp = Circumferential stress, pst y= Poisson's rntio=0.3 for steel 6, = Strain in aval direction ‘© = Strain in circumferential direction Mikasinovio [15] presented an expression relating vibration velocity measured on the cylindrical shell wall to dynamic strain: ir a where: Vibration velocity i/sec, ze10 to peak © = Velocity of sound in metal TEypl= 202,254 inlse Tor steel] «= Dynamic rain, ivin ‘The assumptions made in the derivation above are that the “bration measurements are peak measurements sd several te- Sonant mods are involved, such that the peak vibration veuty is approximately the same around the citeumference and slong the axial length ofthe piping between the constraints. Its not Ienown bow practical this assumption fs, sine the vibration con ditions ina given pipe length ae a function ofthe piping confi ‘uation, the wall thickness, and the internal diving forces. Mikasinovie tested eifferent pipe sizes, wal thicknesses and end conditions with saisietry results Using this formula, would be posible torelate the vibration velocity to the fatigue endurance limit. In ANSVASME OMS, the allowable endurance limit stress 8 10000 ps, and the minimum safety factor is 1.3, which makes the allowable sess {eq 0 7690 psi zero to peak. For the shell wall vibrations, the rmasimum stress concentration fctr inthe hest-alfected zone ofa weld would be. This means thatthe allowable stiess could bbe as low as 1538 ps zero to peak. Ifwe divide the stress by the elastic morules of 30,000,000 psi, the allowable strain is ob- tained at 1.3 microstran tina © 10-*) This values consistent withthe experience ofthe authors 16) ‘Using this value forthe acceptable strain, the allowable veloc sty is equal to yg BURR (1.3% 109) = 2) V,=17 inches per second zero to peak (23)‘TUTORIAL ON PIPING VIBRATION ANALYSIS 131 If the stress concentration factor is less than the maximum, the allowable vibration velocity would be higher by the ratio of the actual stress concentration factor to 5. For a butt weld, the stress concentration factor is approximately 2; therefore, the al lowable velocity would be 4.1 ips ‘The vibrational velocity ofthe shell wali also related tothe sound pressure level (C weighting); however, no closed-form so- lution exists. Field experience with strain gages installed on pip- ing with high frequency, broad band vibrations has shown that ‘the sound pressure level (SPL) measured approximately I lnch away from the pipe wall is proportional to the dynamie stain. Although the relationship hetween dynamic strain and SPL. amplitude is not exact, the overall Tevels as presented below have been used to estimate the severity of shel wall vibrations and asa sereening mnethod tohelp determine where strain gages should be installed on a piping system to determine the safety factor Piping SPLand Strain Criteria When the SPL. is measured with the sound pressure mete using C weighting approximately 1.0 in from the vibrating pipe wall, the following eriteria have been found to be applicable 190 dB is equivalent to spproximately 100 microsteain 196 dB is equivalent to approximately 200 microstrain In aidition tothe erteria outlined above, it has been shown by field experience that allowable strain levels can be specified [16]. These allowable strain levels (6) are given below: € < 100 microstrain Safe om) 100 microstain <€ < 200 microstrain Marginal (25) 2200 mierastrain Excessive 6) ‘These strain limits are based on measurements that are lo- cated away from the high intensified stress locations, such as the Deatsaffected zone. Typically the strain gages are installed about one-half inch away from the weld, This strain limit criteria Is equivalent to an allowable stress of 3000 psi peak to peak. VIBRATION EXCITATION SOURCES Piping vibrations are most often excited by pulsation forces in- side the piping or, secondarily, by mechanical excitation fro ‘machinery unbalanced forees and moments at one and two times the running speed. Potential excitation Sources are included in the following list and are also summarized in Table 2 + Mechanical energy ftom machinery unbalanced forces and moments + Polsations generated by reciprocating compressors and pumps + Pulsations generated by centrifugal compressors and pumps + Pulsations generated by flow through or across objects + Pulsations gene ted by pressure drop at restrictions + Pulsations generated by cavitation and flashing «+ Pulsations generated by waterhammer and surge Pulsation Generating Mechanisms Reciprocating Compressors and Pumps ‘The intermittent flow ofa fluid through compressor or pump cylinder valves generates uid pulsations which are related toa number of parameters, including operating pressures and tem- peratures, horsepower, capacity, pressure ratio, clearance vol- umes, phasing between cylinders, fluid thermodynamic proper ties, and eylinder and valve design, Pulsations are generated at discrete frequency components corresponding to the multiples ‘of operating speed, ‘The pulsation amplitudes depend on the magnitude of the pulsation generated and the reflected amplitudes of the fre- aweney components ws they intrct with the scout reso- ‘ances in the system. Pulsation amplitudes can be predicted by modelling the acoustic characteristics of the piping, the pulsations generated by the compressor or pump and the interaction ofthe two. Digi- tal [5] and analog simulation techniques [7] have been de- veloped to model the piping and the pulsation generating chatacteristis of compressor and pump systems. The analog technique, which was developed in the 1950s, solves the diller eatial equations by building electrical models ofthe piping and the compressors and pumps, In the digital technique, the difer- ential equations of the acoustic phenomena are solved directly with complex matrix algorithms using modern high speed computers, Centrifugal Compressors and Pumps Pulsation amplitudes generated by centufugal machines gen- erally occur atone times running speed and blade passing fe {quency and thelr multiples. They are «function ofthe radial vik brations, the radial impeller clearance, seal and wear rings clearances, the symmetry ofthe impeller, diffuser and case, and the volute characteristics. As operating conditions deviate hom the design or best efficiency point a variety of secondary flow patterns [17] may produce additional pressure fuctuations Significant low frequency pulsations ean also be produced as a result of dynamic interaction of the acoustical response of the piping, the head-flow curve ofthe unit, the dynamie flow damp- ing, and the lation ofthe unitin the piping geometry (7, 18] Flow Through or Across Objects Flow through aestrction o past an obstruction or restriction {nthe piping may produce turbulence or fowinduced pula tions (19, 20]. ‘These flow generated pulsations (commonly called Strouhal excitation) produce noise and wibration at fre- ‘quencies which ae related to the flow velocity and geometry of the sbstruction ‘The acoustical modes of piping system and the location of the turulent eatin have strong intense on the fe quency and amplitude ofthe vortex shedding, The frequencies fenerated by the turbulent energy are centered wound a re- ‘quency which ean be determined by the following equation SAY. D £2 en where: {, = Stouhal vortex equeney, Hz S,, = Strovhal number, dimensionless (0.2 to 0.5) V_ = Flow velocity in the pipe, fvsec D = Charecterstc dimension ofthe obstruction, f132 PROCEEDINGS OF THE NINETEENTH TURBOMACHINERY SYMPOSIUM Table 2. Piping Vibration Excitation Sources. s-acttanteat InDuCKO ig tev Peaeney Js = ane pan rh 2. PULEKTION INDUCED ‘htepoctng ‘compro Repeating Pape © Grtctag ‘A Flow Through Proare ‘eon Ohtesne 8. Flew Pa ibe Mode Asoc Bay = Meighan 4 1uguiD (oR MIXED PHASE) FLOW EXCITED, ‘Pow RatwineeDeete Random Virtnntaw J-= 0-201 ‘Ge Sas For Ln (era Sie) Cavan and igh cnc Beeh,_— Broad Band +. pressune sunag/ Tvbeavuie water ae For flow over tubes, Dis the tube diameter, and forexcitation by flow pasta branch pipe, Disthe diameter of the branch pipe. Pressure Drop Through Restrictions Pressure regulators, flow control valves, relief valves, and pressure letdown valves produce pulsations (noise) associated ‘with turbulence and flow separation, and the relatively broad band frequency spectrum is charaeterstcally centered around a frequency corresponding to a Strouhal number of approx mately 0.2, Cavitation and Flashing Flashing and eavitation can occur in the low pressore ego cf liquid system pressure contol valves when the pressure drags teow the vapor revere, When ailation coco gasbubole is formed and moves with the low. As the presse ineeaies, the pressure res above the vapor pressure, the gas buble cal lapses, and a high amplitade shock pulse results nthe Mud "To avoid flashing akera restriction, suficent back pressore should be provided by taking pressure dropt several locations. Alternately te restiction could belacated near an open end 0 thatthe Bashing energy can dsipate nto larger volume Hydraulic Waterhanmer and Surze Starting and stopping pumps with the attendant fast opening, and closingof valves isa major cause of severe transient pressure ciation rogues yp) ‘Type Problems Inrtramentton Lee chai ade Aste " png Sem Figo Frm, Ensim Bune ape Sean aor Acie Ping Fotgu Fare Coops Vane Moth Acti oe i) ann Rong Suiacs. ‘Donmre figs Cava Comes Vlain Mads Ygu ale S anh eis ‘cenit Piping Sinaaata Doe te Gc le in ‘api Pomp Strays surges in piping systems. Increasing the closure time of valves can reduce the severity ofthe surge pressure. Methods areavail able to evaluate the severity of waterhammer ina particular pip- ing configuration for various closure rates (2) Centrifugal compressors and pumps can sometimes surge when they are operating at alow flow, high-head condition. The flow-vereus-head curve can actually cause backflow to occur and significant pulsations can be generated which area function of the piping acoustical natural irequencies and the overall impe- dance characteristics [18] ‘Coupling Mechanisms For vibrations to occur, there must be an energy generating source plus coupling mechanism to convert the pressure forces into shaking forces. Therefore, in evaluating the piping vibration characteristics of an installation, its essential to understand the ‘coupling mechanisms which cause shaking forces to occur in the Piping system. Pressure pulsations couple to produce shaking foreest piping bends, closed ends of vessels and headers, discontinuities oF changes in the piping diameters and at restrictions, such as orifices, valves, and reducers. In a continuous straight pipe of ‘constant diameter, pulsations will nat produce a significant vib- ration excitation fore.TUTORIAL ON PIPING VIBRATION ANALYSIS 133 EVALUATION OF THE SEVERITY OF PIPING VIBRATION ‘When a vibration problem occur, itis necessary to evaluate its severity and determine the most effective way toalleviate the problesn. The first step is to make an initial survey or walkdown ‘ofthe piping system to determine piping spans with high vibra- tion levels. During the walkdown ofthe piping, sis necessary to look for common symptoms of piping vibrations problems. ‘These include fatigue cracks in the piping, leaks a flanges, bre- ken or loote pipe clamps or hangets, cracked concrete piers, rubbed weight supports (bright metal, damaged. pressure gages, nose related tothe pipe hitting its restraint, or high shell wall vibrations. ‘The second step is to make vibration measurements to evaluate specific piping spans that are thought to have excessive vibration amplitudes. The acceptability is judged by performing the calculations necessary to obtain the dynamic stresses using the simplified techniques presented inthis paper. Since the re~ lationships between vibration and stress were developed for te- sonant piping spans, the frequeney factors presented ean be used to veriy that the span is a resonance. Ifthe system vibration characteristics are complex and itis de~ sired to ensure the safety ofthe piping, it may be necessary to develop a computer model ofthe piping. This model could be forced to have the measured vibrations and the resultant stres- ses calculated. The acceptability of the vibrations could be Judged by comparison ofthe calculated stresses to the material tendurance limit. Strain gages ean be installed atthe suspected high stress locations to measure the dynamic strains which ean be compared to the criteria presented herein. SOLUTIONS TO PIPING VIBRATION PROBLEMS Solutions to most piping vibration problems involve reducing, the excitation forces, eliminating the coupling mechanisms, oF eliminating the mechanical o: pulsation resonances. The most ‘effective solutions are those that eliminate the resonances since the amplification factors for mechanical resonances are typically 10 ~ 30, Amplification factors for pulsation resonances ean be ashigh as 50, although the range of 16 — 30 is more typical Modifications to Solve Mechanical Resonences Sinco the span natural frequeney is an inverse funetion ofthe square of the span length, the most efective way to solve a ‘mechanical resonance is to add pipe restraints such as piers, ‘supports or clamps to shorten the Vibrating span. Many times, temporary bracing with hydraulic jacks, wooden beams and ‘werlges can be used to confirm that support ata particular loca- tion will reduce the vibrations ‘Some ofthe general guidelines which can be used in selecting aiiations to detune the mechanical resonances ae ouined + Pipe supports and clamps should be installed on one side of cach bend, at all heavy weights, and atall piping discontinuities. + The support and clamp stiffness should be adequate to re- strain the shaking forees inthe piping to the desired amplitudes and should be greater than twice the basic span sifiness in order toeffectively enforce a node at the support location. + Vents, drains, bypass, and instrument piping should be braced tothe main pipe t eliminate relative vibrations between, the small-bore piping and the main pipe. + Restraints, supports, or gussets should not be directly welded to the pressure vessels or the piping unless they are sl jected to the appropriate heat treatment. Its more desirable to ‘add saddle-type clamp around the pipe and weld the braces to the clamp, * Pipe guides with clearance are used as thermal expansion control devices and are generally ineffective in controling pip ing vibrations * To resist vibration, the piping clamps should have contact withthe pipe over 180 degrees of the circumference. Rubber ot gasket-type material can be used between the clamp and the pipe to improve the contact. + The piping span natural frequeney should not be coincident withthe excitation frequencies. + In piping thathas high shell wall vibrations, eduction ofthe vibrations and the noise can be accomplished by adding constainedayer damping, if proper desian procedures are + In systems with pressure reducing valves, the wall thiek- ness of the piping should be one-half inch or greater if there is possibilty of sonic low downstream ofthe valve [20]. Fullsad- le reinforcement tees or welding tees should be used ‘downstream of sonically choked valves or where there isa poss: bility of sonic flow occurring st the branch pipe intersection Solutions to Pulsation Resonances ‘When a pulsation resonance is found, acoustic changes to the ‘piping system can be the most effective way to detune or reduce the amplitudes of the pulsations. Probably the most effective clement that can be conveniently used in existing ystems is an orifice plate, which is an acoustical resistance element, and is most elective when located ata pressure pulsaion node. Gener ally, without additional information, an orfice plate with @ diameter ratio of approximately 0.5 will give suflicent pressure drop acoustical resistance) to evaluate whether such an ecousti- cal modification will bean effective solution orifice plates are ineffective or impractical due to the pres sure drop, it may be necessary to install pulsation filters to re- duce the amplitudes of the pulsations. These could be volute bottles, Helmholtz-type filters, gasbladder type aceumulators, ete, Itimay be necessary to acoustically model the piping system using digital or analog techniques to determine the level of changes that wll be required to detune the system and solve the problem. A combination of mechanical and pulsation changes ay be needed to reduce the severity of the problem to the point where the vibrations are acceptable. CONCLUSIONS Some ofthe basie principles necessary for understanding the behavior of piping vibrations and determining the acceptable vibration levels in piping spans have been presented. The infor nation presented can be used by engineers to perform the follwing: + Calculation ofthe first and second mechanical natural fre- {quencies of uniform piping spans, including piping bends with various aspect ratio. * Calculation of the maximum vibrationinduced stresses based on the masimum measured vibration deflection within a ‘uniform piping spa Calculation ofthe maximum vibration-induced stresses based fn the maximum measured vibration velocity within a uniform piping span.1 PROCEEDINGS OF THE NINETEENTH TURBOMACHINERY SYMPOSIUM * Calculation ofthe maximum acceptable vibration deflection within a uniform piping span, based on the endurance stress lint. + Caleulation of the maximum acceptable vibration velocity ‘within a uniform piping span, based on the endurance stress Timit. + Calculation of the effects of concentrated and distributed (contents and insulation) weights on the natural frequencies and the vibration-induced stresses. * Calculation ofthe effects of offresonant stresses + Determination ofthe acceptable sereening criteria fora pip- ‘ng span to eliminate those spans with adequate safety factors so that detailed vibration and stress analyses will nt have tobe per formed on all piping spans, + Determination ofthe possible excitation sources that could be causing the piping vibrations. + Determination of the possible mechanisins that could be ‘coupling the pulsation or mechanical energy into shaking forces. + Determination ofthe possible mechanical solutions to solve the vibration problem, * Determination of the possible acoustical solutions to solve the pulsation problem which may be eausing the vibrations REFERENCES 1. Garrison, W. G., "Major Fires and Explosions Analyzed for 80xear Period,” Hydrocarbon Processing (September 1988) 2. Olson, D. E., “Piping Vibration Experience In Power Plants,” Pressure Vessel and Piping Technology, A Decade of Progress, Book No, H00330, ASME (1985). 3, Wachel, J. Cand Bates, C. L., “Techniques for Controlling Piping Vibrations and Fallure,” ASME. paper 76-PETIS 0976) 4. Wachel, J. C., "Piping Vibration and Stress,” Proceedings ofthe Machinery Vibration Monitoringand Analysis, Vibra tion Institute (April 1981). 5. Wachel, J.C, Atkins, K. E., et al, “Vibrations in Recip- rocating Machinery and Piping,” Enganeering Dynamles Incorporated Report 85-305 (1985). 6. Blevins, R. D., “Formulas for Natural Frequency and Mode Shape,” Var Nostrand Reinhold Company, (1979), 7. Sparks, ©, B., etal, “Controlling the Effects of Pulsations and Fluid Transients,” Report No. 160 (revised), SGA- 10. a. 4, 6. n, al PCRC Open Seminar, (November 1981) Harris, C. M., and Crede, C. ., Shock and Vibration Handbook, 1, Baste Theory and Measurements, New York, New Yorke McGraw-Hill (1961), Amold, R.N., and Warburton, G. B., The Flexural Vibra tions of Thin Cylinders,” Journal Proceedings Institute of ‘Mechanical Engineers, 167, (1953), Pung, Y.C., Sechler, E. E., and Kaplan, A.,"On the Vibtax tion of Thin Cylindrical Shells Under Internal Pressure,” J. Aeronautical Sciences, 2¢ (1951) USAS BSL.T - 1969 “Nuclear Power Piping,” ASME Code, New York, ANSUASME Standard OM:3, "Preoperational and Initial Startup Vibration Testing of Nuclear Power Plant Piping Systems,” ASME, New York, New York (1987) API Standard 618, “Reciprocating Compressors for General Refinery Services,” American Petroleum Institute, ‘Washington, D.C. (February 1986) Tulk, J. D., "Correlations Between Dynamic Stress and Vibration Velocity in Piping Systems,” Pipeline Dynasies and Valves— 1989, ASME PVP, 180, ASME. Pressure Ves- sels and Piping Conference, pp. 19-25 (1988). Mikasinovic, M., “Vibration Acceptance Criteria of Cicu- lar Cylinder Sheils,” Pipeline Dynamies and Valves 1989, ASME PVE, 180, ASME Pressure Vessels and Piping Con: ference, pp. 47-52 (1989). Wachel, J. C., “Field Investigation of Piping System for Vibration—Induced Stress and Failures," Pressure Vessel and Piping Conference, ASME Volume No. H00219 June, 196 Makay, £., “Centrifugal Pump Hydraulic Instability,” CS- MAS Research Project 1266-18, Electric Power Research In- stitute, Palo Alto, California (1980), Sparks, C. R., "On the Transient Interaction of Centrifugal Compressors and Their Piping Systems,” Journal of En- gineering for Power, 105, pp. 891-901 (October 1983) Blevins, R. A., “Flow-Induced Vibration,” New York: Van ‘Nostrand Reinhold (1977 Carucci, V.A., end Mueller, R.T., “Acoustically Induced Piping Vibration in High Capacity Pressure Reducting Sys- tems,” ASME Paper 82- WA/PVP-8, Pressure Vessel and Piping Conference (1982) fer, R., "Prevention of Power Plant Water Hammer ASME Paper 87-PVP-19, Pressure Vessel and Piping Cor erence (1987) Southwest Research Institute