THE NEW FIFTH EDITION OF API 618 FOR RECIPROCATING COMPRESSORS WHICH PULSATION AND VIBRATION CONTROL PHILOSOPHY SHOULD YOU USE?

Transcription

1 THE NEW FIFTH EDITION OF API 618 FOR RECIPROCATING COMPRESSORS WHICH PULSATION AND VIBRATION CONTROL PHILOSOPHY SHOULD YOU USE? by James D. Tison Senio Staff Enginee and Kenneth E. Atkins Senio Staff Enginee Engineeing Dynamics Incopoated San Antonio, Texas James D. Tison is a Senio Staff Enginee at Engineeing Dynamics Incopoated, in San Antonio, Texas. He has been extensively involved in field measuements and compute modeling of otating and ecipocating equipment fo ove 24 yeas. M. Tison was with Southwest Reseach Institute fom 1977 to 1982, and cofounded Engineeing Dynamics Incopoated in Fo the last 19 yeas, his wok has been pimaily elated to development and impovement of softwae fo the simulation of pulsation in compesso and pump piping systems, as well as oveall esponsibility fo acoustic and mechanical design studies of ecipocating compesso and pump piping systems. M. Tison holds B.S. and M.S. degees (Mechanical Engineeing, 1975, 1977) fom the Univesity of Floida. He is a membe of ASME and is a membe of the API 618 and API 674 pulsation and vibation contol sub-task foces. Kenneth E. Atkins is a Senio Staff Enginee with Engineeing Dynamics Incopoated, in San Antonio, Texas. He has extensive expeience in the design and toubleshooting of a vaiety of mechanical systems involving ecipocating machiney, stuctual, and piping vibation poblems. M. Atkins was a Reseach Enginee with Southwest Reseach Institute (1978 to 1981) and a Machiney Enginee with Exxon Chemical Ameicas (1981 to 1982). In 1982, he cofounded Engineeing Dynamics Incopoated. He has authoed seveal technical papes in the aeas of ecipocating machiney, piping, and stuctual dynamics. He has lectued fequently at the Texas A&M Tubomachiney and Pump Symposia with both tutoials and shot couses. M. Atkins eceived a B.S. degee (Engineeing Science, 1978) fom Tinity Univesity. He is a membe of ASME and a egisteed Pofessional Enginee in the State of Texas. He is also a sub-task foce membe fo pulsation and vibation contol fo API Standads 618 and 674. ABSTRACT The poposed Fifth Edition of API 618 ( Recipocating Compessos fo Petoleum, Chemical, and Gas Industy Sevices ) incopoates significant changes in the section concening pulsation and vibation contol. Thee ae still to be thee design appoaches, but the equiements to pefom cetain analyses that wee pesented as optional in the Fouth Edition will now be dependent on pessue pulsation and foce levels detemined fom the acoustical simulation. The confusion concening when piping foced mechanical esponse calculations should be pefomed, which oiginated in the Fouth Edition, has been eliminated; foced esponse calculations ae not equied to satisfy API 618 Fifth Edition when pulsation levels ae contolled popely. A sepaate aticle on pulsation and vibation contol is being developed by the API 618 sub-task foce on pulsation and vibation contol as an appendix (annex) to API 618. This text will be a stand alone RP (Recommended Pactices) document in the API system, which would then be efeenced by API 618 as well as othe API standads (e.g., API 674 fo Positive Displacement Pumps) fo which pulsation and vibation contol ae an issue. This document, to be issued in 22, will discuss the diffeent design philosophies inheent to the new edition of the standad. The pupose of this tutoial is to povide the use with a woking knowledge of good engineeing pactices fo pulsation and vibation contol of ecipocating machiney in elatively high mole weight gases (e.g., natual gas), as well as an indepth undestanding of the poposed changes in API 618 and the diffeing design philosophies. Seveal case histoies ae used to illustate why obust pulsation contol is impotant fo ecipocating compesso piping systems. The authos ae membes of the API 618 sub-task foce on pulsation and vibation contol and each has ove 2 yeas of expeience in this field. INTRODUCTION In the 195s and 6s, design techniques wee developed using analog simulation tools fo the contol of pulsation in compesso piping systems. Acoustical designs utilizing eactive pulsation contol (acoustic filteing), in combination with esistive elements (oifice plates) whee necessay, became vey successful in contolling pulsation levels tansmitted to piping, piping shaking foce, and bottle unbalanced foce. Ove the last 2 yeas, digital techniques have pogessed significantly as the speed and capacity of computes have developed, and today, digital techniques fo acoustic simulation ae in geate oveall use woldwide than analog methods. Howeve, in ecent yeas thee has also been a tend in some industy segments away fom utilization of effective pulsation contol techniques and towad moe eliance on mechanical techniques to contol vibation. Thee ae seveal easons fo this distubing tend. Fist, the basic pulsation contol technology has histoically been popietay to cetain oganizations. Many uses of acoustical simulation softwae do not undestand eactive pulsation contol and/o thei softwae does not pemit them to be cost competitive 183

2 Piston Velocity 184 PROCEEDINGS OF THE 3TH TURBOMACHINERY SYMPOSIUM in designing eactive filte systems; esistive designs equie significantly less engineeing effot and technical expetise. Anothe eason fo this tend is the polifeation of finite element based stuctual dynamics softwae fo piping. Vitually evey pipe stess analysis package on the maket today has some dynamic capabilities. Mechanical natual fequencies and foced vibation levels of complex piping systems, once modeled, can be calculated faily easily; howeve, it is the lack of undestanding of the limitations on the accuacy of these calculations that leads to seious poblems and in some cases disastous consequences. As will be shown heein, even if the stuctual dynamics calculations wee extemely accuate, thee is no justification fo the isk involved by designing systems with inadequate pulsation contol. Howeve, the new Fifth Edition of the API 618 (21) standad will continue to include language concening detailed mechanical esponse and natual fequency calculations, implying that these calculations ae sufficiently accuate to be useful in the design stage. While such calculations can be pefomed to any degee of accuacy in theoy, pactical consideations put limits on the accuacy that is actually achievable. It is the goal of this tutoial to illustate this point, and to pesent well-established design techniques that can educe the dependence on expensive and poblematic foced esponse analysis fo the qualification of piping system designs. SOURCES OF VIBRATION IN RECIPROCATING COMPRESSORS Excitation Mechanism Recipocating compessos geneate flow modulations that in tun geneate pessue pulsations. The flow modulations come about as a esult of intemittent flow though the suction and dischage valves, as well as geomety effects due to the (finite) length of the connecting od. Figue 1 shows a schematic of a compesso cylinde. The suction flow (Q S ) entes the cylinde, and the dischage flow (Q D ) exits the cylinde. The velocity of the piston, shown in Figue 2, is appoximately sinusoidal in shape. The deviation of the actual piston motion fom the sinusoidal shape is due to the finite length of the connecting od. As the atio of the connecting od length to the cank adius (L/R) is inceased, the shape becomes moe closely sinusoidal. The pessue pulsation geneated by the compesso is popotional to the flow (Q S o Q D ) modulation. Since the flow is based on the poduct of the piston velocity and the piston swept aea, the shape of the dischage flow at the piston face is of the same shape as the piston velocity cuve (Q = Aea Velocity). Since the suction and dischage valves of each cylinde end (e.g., the head end) of a compesso ae neve open simultaneously, the suction and dischage piping systems ae isolated acoustically. Theefoe, we can look at the flow excitation of eithe the suction o dischage independently fo the pupose of undestanding the pulsation excitation mechanism. w Cank End Q s Head End Figue 2. Piston Velocity fo Slide Cank Mechanism. Figues 3-6 show the effect of the valve action on flow though the dischage valves of a compesso. Figue 3 shows the dischage valve flow vesus time fo the head end of a cylinde. Duing compession, the suction and dischage valves ae closed. When the pessue in the cylinde eaches the dischage back pessue, the dischage valve opens, and the flow vesus time wave though the valve has the shape of a potion of the piston velocity cuve shown in Figue 2. As the cylinde eaches top dead cente (TDC), the dischage valves close, and the flow etuns to zeo. Flow Amplitude Figue 3. Single Acting Compesso Cylinde (L/R =, Ideal Valves). A fequency analysis of the flow wave of Figue 3 is shown in Figue 4. Due to the epetitive action of the compesso cylinde, excitation is geneated only at discete fequencies, which ae multiples of the unning speed. These fequencies ae commonly efeed to as hamonics. The highest amplitude occus at 1 unning speed, with the levels geneally deceasing at highe hamonics. Flow Amplitude Actual Piston Motion Sine Wave 1 V max V min Cank Angle (Degees) Flow vs. Time Degees of Rotation Hamonic Figue 4. Flow Fequency Spectum fo Single Acting Cylinde. Q d Figue 1. Recipocating Compesso Slide Cank Mechanism. Fo a pefect double acting cylinde (symmetical head end and cank end flows, L/R = ) the flow vesus time contains two identical flow slugs 18 degees apat in time. Theefoe, the odd hamonics (in this idealized case) cancel, so that the nonzeo

3 THE NEW FIFTH EDITION OF API 618 FOR RECIPROCATING COMPRESSORS WHICH PULSATION AND VIBRATION CONTROL PHILOSOPHY SHOULD YOU USE? 185 cylinde flow excitation occus at even hamonics of unning speed (2, 4, ). Actual cylindes have piston ods, diffeences in head end/cank end cleaance volumes and finite length connecting ods, so that the two flow slugs geneated each evolution ae not identical (Figue 5). Theefoe, even in double acting opeation, the cylinde will, in geneal, poduce flow excitation at all hamonics of unning speed as shown in Figue 6. These flow hamonics act as excitations to the piping acoustics, and the acoustic esonances of the piping will amplify pulsation at paticula fequencies. Flow Amplitude Figue 5. Unsymmetical Double Acting Compesso Cylinde (L/R = 5, Ideal Valves). Flow Amplitude Degees of Rotation Flow vs. Time H.E. C.E Hamonic Figue 6. Flow Fequency Spectum fo Double Acting Cylinde. Mechanical Excitation Mechanisms In addition to acoustical excitation, anothe souce of excitation in ecipocating compesso systems is mechanical excitation due to ecipocating inetial foces of the compesso itself, and cylinde stetch caused by intenal pessue eaction foces acting on the cylindes and fame. These foces ae typically stongest at 1 and 2 unning speed, and ae pimaily a concen only in the immediate vicinity of the compesso. A RISKY APPROACH TUNING OF MECHANICAL NATURAL FREQUENCIES BETWEEN SIGNIFICANT EXCITATION HARMONICS Real Wold Inaccuacies of Mechanical Natual Fequency Calculations To avoid potential vibation poblems in piping systems, the single most impotant concept is to avoid coincidence of mechanical natual fequencies with significant pulsation o mechanical excitation fequencies. Howeve, field expeience shows that the accuacy of pedicted mechanical natual fequencies in piping systems is suspect even unde the best of cicumstances. Eo magins of ±2 pecent ae obtainable only in situations whee accuate bounday conditions ae known, and extensive, detailed modeling of both the piping system and the suppoting stuctue is pefomed. Realistically, many mechanical natual fequencies cannot be calculated within a magin of 2 pecent o even 5 pecent. Inspection of eal wold chemical, gas tansmission, and gas gatheing stations eveals that in many cases, pipe suppots have become loose o do not even touch the piping at some locations, which negates modeling effots. Othe items that influence the accuacy of these models ae: Uncetainty of stiffness (six degees of feedom) of clamps/hold downs. Uncetainty of stiffness of clamp/hold down suppoting stuctue. Difficulty in accuately pedicting coefficients of fiction. Nonlinea effects (e.g., gaps closing due to themal gowth). Uncetainty in attached weights (valves, actuatos, etc.). Uncetainty of as-built piping layout and dimensions. Difficulty and complexity of modeling ack suppot stuctue. Uncetainties in soil stiffness effects on concete pies. Settling of suppots esulting in loss of piping contact. A piping/stuctual suppot system is not a polished machine pat fo which finite element models ae easily defined and analyzed. Futhemoe, many vibation elated poblems ae not associated with the main pocess piping itself, but with othe attached components, examples of which ae listed below: Valve actuatos Tubing Conduit and cable tays in ack systems Inspection openings and instument connections (themocouples, pessue tansduces) Flow measuement instumentation Scubbe level contol instumentation Small banch connections (fo instumentation connections, vents, and dains) Instument panels mounted on compesso decks It is impotant to emembe that even when the main pocess piping has low vibation, the main line can act as a base excitation to attached mechanically esonant banches. Often, small banches attached to the main pocess piping ae not consideed in mechanical natual fequency esponse modeling. Figue 7 shows conceptually how a banch, if esonant to the fequency of the vibation of the main line, can cause high vibation of the banch itself. Theefoe, maintaining vey low foce levels in the piping is impotant. Figue 8 shows a valve actuato that vibated so seveely that the suppot backet failed; the vibation of the main piping was less than 2 mils peak-to-peak, while the actuato itself had vibation in excess of 5 mils peak-to-peak. Main Pocess Piping Mechanically Resonant 5 Hz 1 mils p-p Base Excitation = 2 mils 5 Hz Figue 7. Example of Vibation of Attached Banch Line Due to Base Excitation by Main Pocess Line.

4 186 PROCEEDINGS OF THE 3TH TURBOMACHINERY SYMPOSIUM Tansmitted to Piping Fo 9 RPM Compesso Amplitude Hz Hamonic 1x 2x 3x 4x 5x 6x 7x 8x 9x 1x Figue 9. Chaacteistics without Acoustic Filteing. Amplitude Tansmitted to Piping and Mechanical Natual Fequencies Minimum Mechanical Natual Fequency 36 Hz Calculated Piping Mechanical Natual Fequencies Figue 8. Valve Actuato Vibation 5 mils peak-to-peak at 24 Hz, 2 mils peak-to-peak on Main Line. A detailed knowledge of the mechanical natual fequencies and esponse chaacteistics of the above components is geneally not available in the design stage. Unfotunately, failues of small banch connections attached to main piping, as well as othe items listed above, epesent a lage pecentage of vibation elated poblems and actual failues that occu in ecipocating compesso piping systems. The Effect of Inaccuacies in Mechanical Natual Fequency Calculations Table 1 shows how eos in effective stuctual stiffness (due to pipe suppot system and stuctual stiffness of the piping itself) affect the accuacy of mechanical natual fequency (MNF) calculations based on the elation: MNF k. Table 1. Effect of Uncetainty of Effective Stuctual Stiffness of Piping System on Actual Mechanical Natual Fequencies. Hz Hamonic x 2x 3x 4x 5x 6x 7x 8x 9x 1x Figue 1. Chaacteistics without Acoustic Filteing: Calculated Mechanical Natual Fequencies Supeimposed. Table 2 shows the effect of uncetainty on the location of mechanical natual fequencies elative to hamonics of unning speed fo a 9 pm compesso. The fequencies between 1 and 2, 2 and 3, 3 and 4, and 4 and 5 unning speed shown in the table epesent MNFs that ae tuned between hamonics. Assuming a bette than typical ±2 pecent mechanical natual fequency calculation accuacy, the actual ange of each natual fequency is also shown in the table. At all fequencies above 2 unning speed, the actual possible ange of mechanical natual fequencies between hamonics is too lage; the actual ange of the mechanical natual fequencies exceeds the fequency gap between hamonics. Theefoe, in the design stage, it is impossible to tune any calculated mechanical fequencies above 2 unning speed fequency away fom pulsation excitation fequencies. Even the ange between 1 and 2 unning speed is within 1 pecent of 2 unning speed. Figue 11 shows this concept gaphically. Table 2. Range of Actual Mechanical Natual Fequencies Based on ±2 Pecent Uncetainty (9 RPM Compesso). As an illustation of the difficulty of pedicting mechanical esonance fequencies in a piping system, conside a 9 pm fixed speed compesso. The fundamental (1 unning speed) fequency is 15 Hz. The fequencies of the fist 1 hamonics ae shown as bas on a gaph in Figue 9. Piping systems have numeous mechanical natual fequencies; in fact, they have an infinite numbe of natual fequencies. Figue 1 shows, conceptually, the locations of mechanical natual fequencies of a piping system supeimposed on the pulsation spectum. Geneally, most designes agee that the minimum mechanical natual fequencies of the piping should be maintained at least 2 pecent above the second hamonic of unning speed when pactical to avoid significant acoustical and mechanical excitation. The lowest piping mechanical natual fequency in Figue 1 is shown above 36 Hz (2 pecent above 3 Hz), assuming this can be accomplished. Anothe impotant concept to emembe is that even if the mechanical natual fequencies could be calculated within an accuacy of say, ±1 pecent, significant amplification still occus when the focing fequency (i.e., the pulsation fequency) is 1 pecent above o below the mechanical natual fequency. As shown in Figue 12, the effective amplification facto at fequencies 1 pecent away fom a paticula mechanical natual fequency is 5:1 fo all values of citical damping atio less than 5

5 THE NEW FIFTH EDITION OF API 618 FOR RECIPROCATING COMPRESSORS WHICH PULSATION AND VIBRATION CONTROL PHILOSOPHY SHOULD YOU USE? 187 pecent. Theefoe, the excitation fequency need not be exactly coincident with the mechanical natual fequency to cause excessive vibation; a magin of 1 pecent is not necessaily sufficient even when the exact natual fequency is known. Amplitude Hz Hamonic 36 Hz Tansmitted to Piping and Mechanical Natual Fequencies Range of Actual Mechanical Natual Fequencies ( ± 2%) x 2x 3x 4x 5x 6x 7x 8x 9x 1x Figue 11. Chaacteistics without Acoustic Filteing: Actual Range of Mechanical Natual Fequencies Supeimposed. 2. ς = Q = A BETTER APPROACH TO VIBRATION CONTROL PULSATION AND FORCE CONTROL THROUGH REACTIVE ACOUSTICAL FILTERING Fotunately, many of the inheent difficulties of mechanical vibation and natual fequency pediction may be ovecome though obust acoustical design. Wheeas the mechanical natual fequencies of piping can be difficult to pedict within ±2 pecent o even ±5 pecent, acoustical natual fequencies and theefoe eactive filte fequencies can be calculated elatively accuately (within ±5 pecent). Futhemoe, the technique of acoustic filteing can be used effectively and confidently to contol pulsation in elatively high mole weight, elatively low speed of sound systems (less than 2 ft/s) in the design stage. (In low mole weight gas systems, whee eactive filtes ae impactical, pulsation contol can be accomplished though the use of esistive o pessue dop elements.) Acoustic filteing involves the use of two volumes joined by a elatively small diamete pipe, which is known as a volume-chokevolume filte. Figues 13 and 14 show vaious foms of the volume-choke-volume filte. Such devices have the pulsation esponse chaacteistics shown in Figue 15. At fequencies above its chaacteistic esonance (f H ), tansmitted pulsation levels dop off apidly. Equation (1) is used to calculate the filte fequency, f H, of an ideal filte with no piping attached: ω n ς = Q = ω ω n Figue 12. Effect of Sepaation Magin fom Mechanical Natual Fequency on Amplification Facto. ω n 5:1 Amplification Fo 1% Magin whee: f = c = Speed of sound (ft/s) A c = Aea of choke tube (ft 2 ) L c = Length of choke tube (ft) L c =L c.6d (ft) d c = Choke diamete (ft) V 1 = Volume of pimay bottle (ft 3 ) V 2 = Volume of seconday bottle (ft 3 ) f H c = 2π A c L V V c L c, A c, d c 1 2 (1) Inaccuacies in Mechanical Foced Response Calculations The new (Fifth Edition) API 618 (21) standad allows that in the event the pulsation levels and foce levels acting on uns of pipe exceed the amplitude guidelines, and sepaation magins ae not met, foced esponse calculations may be pefomed to qualify the system. If the pedicted vibation levels ae below the vibation allowable guideline, then the pulsation and foce levels ae consideed acceptable. The accuacy of mechanical esponse amplitude (vibation and stess) calculations is influenced by the same items that affect the calculation of natual fequencies. Pactically, because of the uncetainty of the actual mechanical natual fequencies, mechanical esonance must be assumed. If the condition of esonance is not assumed, the pedicted esponse levels will almost always be low. The Q value (amplification facto), which may vay fom 1 to 1, must also be assumed. This makes esonant amplitudes extemely difficult to pedict. In fact, since the esonant amplitudes computed ae defined by the assumed damping, the pedicted esults ae abitay. Mechanical esponse calculations ae best left to situations in which field data ae available to adjust the model to give the pope natual fequencies and damping; accuate simulation of foced piping mechanical esponse at the design stage is geneally impactical. V 1 V 2 L c, A c, d c V 1 V 2 Figue 13. Nonsymmetical Volume-Choke-Volume Filte Staight Choke Tubes. Design Using Acoustic Filteing in Conjunction with Good Mechanical Suppot Pactices Figue 16 shows how the pulsation contol though use of such a filte contols vibation, eliminating the concen of uncetainty of piping mechanical natual fequency calculations. and esulting foce levels ae contolled to insignificant levels above some cutoff fequency (usually below 1 unning speed). Piping mechanical fequencies ae placed well above this cutoff fequency.

6 188 PROCEEDINGS OF THE 3TH TURBOMACHINERY SYMPOSIUM V 1 L c, A c, d c V 2 Figue 14. Nonsymmetical Volume-Choke-Volume Filte Elbows in Choke Tube. 1. f H Compaison of Contol Devices in Dischage Piping Systems Figues 17, 18, 19, 2, and 21 compae vaious pulsation suppession techniques in an infinite length dischage line (noneflective bounday condition) of a compesso opeating ove a speed ange of 7 to 1 pm. The assumption of a noneflective bounday eliminates acoustical esonances of the piping itself, and is a convenient method fo compaison of the effectiveness of pulsation contol devices. The pulsation contol teatments ae: None (no suge volume, Figue 17). A simple suge volume with a volume equal to 5 pecent of the volume calculated using the API 618 Design Appoach 1 sizing fomula ( 1 /2 API suge volume, Figue 18). A simple suge volume with a volume equal to 1 pecent of the volume calculated using the API 618 sizing fomula (1 API suge volume, Figue 19). Volume-choke device (Figue 2). Volume-choke-volume filte with f H < 1 unning speed (Figue 21). Case No. Contol 1 None 5 Dischage Line PSI P-P L = psi p-p 2x 4x 1x 3x 5x 6x Figue 17. Compaison of Contol Devices Case 1. Case No. 2 Contol 1/2 x API Suge Volume (4'-" x 1.75" I.D.) 5 Dischage Line Figue 15. Typical Fequency Response to Flow Excitation of Volume-Choke-Volume Filte System (No Passbands). Tansmitted to Piping psi p-p 2x Amplitude ~1 Hz Filte Cutoff Fequency Mechanical Natual Fequencies L = 6x 1x 3x 4x 5x Figue 18. Compaison of Contol Devices Case 2. Hz Hamonic x 2x 3x 4x 5x 6x 7x 8x 9x 1x Figue 16. Chaacteistics with Acoustic Filteing. With this design concept, the mechanical natual fequencies of the piping ae well above the lowe significant hamonics of pulsation, emoving the concen ove the exact location of the vaious mechanical natual fequencies. Within pactical limits, even 1 unning speed pulsation levels can be contolled to any desied level. Compaison of the pulsation amplitudes fo the five cases shows the significant eduction in pulsation levels obtained by the use of acoustic filteing. (Note that no esonances occu in the piping because of the assumed infinite length line bounday conditions; theefoe, these cases can only be used fo elative compaison of pulsation amplitudes.) Mechanical Analogies and Intepetation of Suge Volumes and Filtes Suge Volumes At fequencies below the length esonances of the bottles themselves, volume bottles act pedominantly as acoustic

7 THE NEW FIFTH EDITION OF API 618 FOR RECIPROCATING COMPRESSORS WHICH PULSATION AND VIBRATION CONTROL PHILOSOPHY SHOULD YOU USE? 189 Case No. Contol 3 1 x API Suge Volume (4'-" x 15.25" I.D.) 5 Dischage Line Volume Sping psi p-p Sping m m m m Figue 19. Compaison of Contol Devices Case 3. Case No. 4 L = Contol 1 x API Suge Volume with Choke (4'-" x 15.25" I.D. Bottle) + 4'-" x 2.9" I.D. Choke 5 (.34% P) 2x 1x 3x 4x psi p-p Dischage Line Figue 22. Mechanical Analogy of Suge Volume. fequencies below the open-open esonant fequencies of the choke tube length, and the closed-closed esonant fequencies of the bottle lengths. The mechanical analogy of such a filte is a high flexibility (volume) in seies with a lage mass (choke) and anothe high flexibility (volume) as shown in Figue 23. At fequencies above the esonant fequencies of the mass sping system, the piston motion is isolated due to the momentum chaacteistics of the choke tube fluid. The acoustic filte has chaacteistics analogous to those of L-C filtes used in electical systems. K B Volume Choke Volume k k k L = 2x 1x 3x 4x Figue 2. Compaison of Contol Devices Case 4. Case No. Contol Dischage Line 5 Volume-Choke-Volume I[ ( f < 1x) Vol. Each Bottle > 1 x API Suge Volume (5'-" x 15.25" I.D.) + (5'-" x 15.25" I.D.) 1'-" x 2.9" I.D. Choke 5 (.42% P) L = 1x 2x Figue 21. Compaison of Contol Devices Case 5. compliance. Acoustical compliance is analogous to mechanical flexibility as shown in Figue 22. The pipe beyond the suge volume contains the gas that has mass and elastic popeties. This fluid is set into a vibatoy state by the motion imposed upon it by the piston. If a highly flexible (low stiffness) element is placed between the piston face and the pipe fluid, the piston motion is essentially isolated fom the fluid in the piping, and less vibation (and theefoe less pessue vaiation) of the fluid occus. This is simila to the concept of vibation isolation commonly used in machiney. Filtes Volume-choke-volume filtes have, in addition to two compliance components (two volumes), a choke tube that acts as an acoustical inetance to esist changes in velocity of the fluid contained in the choke tube. As fo the single suge volume, these lumped compliance and inetance popeties ae valid at psi p-p Sping K B1 Mass M = 1 KB1 + K 2 p M Figue 23. Mechanical Analogy of Volume-Choke-Volume Filte. Design Pocedue fo Heavy Gases Using Reactive Contol (Acoustic Velocity < 2 ft/s) f H m m m The use of eactive filteing in conjunction with contol of mechanical natual fequencies esults in a safe magin between significant pulsation induced foces and mechanical natual fequencies. The pocedue fo designing eactive filtes is: Detemine volume-choke-volume filte design to filte all hamonics of unning speed. Geneally, the filte fequency is set at 5 to 8 pecent of 1 unning speed fo heavy gases, o between 1 and 2 unning speed fo lighte gases. Pefom pulsation simulation to detemine pulsation levels and acceptability of filte design. Detemine maximum fequency (f p ) of significant pulsation and foce in piping. Detemine minimum allowable mechanical natual fequency (f m ) based on (f p ). Set f m 1.5 f p. Locate vibation estaints nea all concentated masses (e.g., valves). Use pipe suppot span tables (Table 3) to detemine additional suppot locations based on f m. Detemine minimum stiffness (k) of each suppot: k 2 lateal span stiffness = 2 48 EI ( = suppot span). 3 K B2 B2 k k

8 19 PROCEEDINGS OF THE 3TH TURBOMACHINERY SYMPOSIUM Table 3. Pipe Suppot Span Spacing Table. Use of this acoustic filteing concept in conjunction with contol of minimum piping mechanical natual fequencies ensues that esonance is avoided. CHANGES FOR API 618 FIFTH EDITION Residual Nonesonant Foce Evaluation A significant change cuently being made to API 618 (21) is with egad to pulsation induced unbalanced foces acting on piping uns. Although this is not a new concept, past vesions of API 618 specified limits only fo pulsation levels. The new standad will addess allowable foce levels fo nonesonant conditions. Figue 24 shows how acoustically induced foces ae calculated fo a potion of the piping system (assuming centifugal effects of dynamic gas/fluid flow at elbow ae small). The staight un of pipe between elbows is consideed to be a igid body, and the foce acting along the un is the sum of the foce acting at the elbows at each end as defined in Equations (2) and (3): ΣF = F + F π Σ F P id π = P id A B 4 4 whee P A,P B ae vectos epesenting amplitude and phase of pulsation at points A and B at a paticula fequency. These foce calculations ae easily made based on the known piping geomety and calculated pulsation levels at the elbows. FA FA = PA π id Figue 24. Dynamic Foce on Piping Run. A FTotal While calculating the foces acting on the piping is faily staightfowad, detemining an acceptable foce level is quite difficult. It is impotant to ealize that the shaking foce guideline fo the new API 618 (21) standad is based on specific nonesonant configuations. In geneal, much highe foce levels may be toleated at fequencies below the lowest mechanical natual fequencies; howeve, foce levels should be contolled to vey low levels at fequencies nea o above the lowest piping mechanical natual fequency. A foce evaluation citeia poposed and used by the authos is given in Table 4 fo the allowable foces acting on each piping A B 2 2 FB = FB FA = PB π id (2) (3) FB estaint in a staight un of pipe between elbows (o othe foce coupling points). The foce pe estaint is calculated as the foce in the piping segment divided by the numbe of estaints that esist axial vibation of the piping un. Note that the impotant concept of these citeia is that foces at highe fequencies should be contolled to much lowe levels, since the accuacy of mechanical natual fequency calculations is such that the only easonable engineeing assumption that can be made is that esonance can potentially occu at the highe fequencies. Table 4. Allowable Foce Guideline. Cae should be taken to apply foce citeia with caution. Vey low foce levels in main piping may cause vey low vibation levels in the main line; howeve, if banch piping, apputenances, instumentation lines, etc., ae esonant at the same fequency, vey high vibation of the attached elements can occu. Definition of Mechanical Response Calculation Tigges API 618, Thid Edition (1986), descibed in geneal tems the concepts of esonance avoidance and the use of filteing techniques. The Fouth Edition (1995) added desciptions of vaious pocedues (Appendix M) in an attempt to claify specific pocedues to meet the equiements of Design Appoaches 2 and 3. Pocedue M.7 of the cuent Fouth Edition is as follows: A piping system dynamic stess analysis calculates the mechanical system esponses and associated mode shapes. The significant pedicted pulsation foces ae imposed on the piping to the extent necessay in ode to calculate the expected vibation and stess amplitudes at the citical points in the system. These stesses ae compaed to the levels identified in With this shot paagaph, Design Appoach 3 analyses wee foced with poviding a level of calculation that is subject to a geat deal of intepetation. What ae the citical points in the piping system? What pulsation foces ae significant? Based on the authos involvement with the API 618 sub-task foce, it appeas that in Euope this paagaph has been taken liteally. Finite element models of piping systems ae outinely made, and piping foced esponse calculated. This appaently fits in well with thei nomal design pocedue since: The bottles ae designed pio to knowing the piping details. Acoustic filteing is usually not used. This mechanical analysis is used to design and justify moe complex suppots and anchoing systems to contol vibation. Table 5 shows the pocedues specified by Appendix M of the Fouth Edition fo each of the thee design appoaches. Table 6 descibes each of these pocedues. The flow chat shown in Figue 25 descibes the wok pocess used to satisfy the thee design appoaches fo the new Fifth Edition of API 618 (21), and vaious analysis pocedues

9 THE NEW FIFTH EDITION OF API 618 FOR RECIPROCATING COMPRESSORS WHICH PULSATION AND VIBRATION CONTROL PHILOSOPHY SHOULD YOU USE? 191 Table 5. Summay of Analysis Pocedues of Fouth Edition fo Design Appoaches 1, 2, and 3. Peliminay PSD Sizing (M.1) Design Appoach 1 Piping Layout Final? No Pefom Pe-Study pe Table 6. Desciption of Analysis Pocedues fo Fouth Edition fo Design Appoaches 1, 2, and 3. Yes Pefom Acoustic Simulation pe and evise PSD's o pipe as equied. (M.2 & M.3) Design Appoach 2 Revise Design Citeia and and pessue dop citeia Met? Yes Complete Maximum Piping Span Table (M.4) Design Appoach 2 Analysis is Complete No OR coesponding to the Fouth Edition ae noted. Design Appoach 1 (DA-1) is not changed fom the pevious editions and involves no simulation of the system. The pulsation suppession devices (PSD) ae sized based on the fomula included in the specification o by vendo popietay methods. If a simulation is to be done, thee is a new optional pe-study that can be pefomed. This comes fom a common pactice in Euope, also efeed to as a dampe check. This simulation is conducted pio to the finalization of the piping layout and sizes the PSDs using noneflective piping. The intent is to allow pocuement of the PSD vessels ealie in the poject timeline, which is desiable fom a commecial standpoint. Howeve, fom a technical standpoint, this is not the best appoach since the attached piping affects the optimum PSD design. Once the piping layout is finalized, a complete acoustic simulation is caied out. If Design Appoach 2 (DA-2) is specified, the pulsation and pessue dop citeia must be met and a table of maximum clamp spacing is developed based on avoidance of esonance with any significant pulsation enegy. No evaluation of shaking foces o compesso manifold system esponse is equied. Design Appoach 3 (DA-3) includes evaluation of the mechanical esponse chaacteistics of the compesso manifold system and piping, just as fo the pevious editions. The key changes include the intoduction of shaking foce citeia and specific steps that can be taken to satisfy DA-3. As shown in the flow chat, if the pulsation and pessue dop allowables ae met, and thee ae no mechanical natual fequencies of the manifold system o piping that ae coincident with significant pulsation enegy, and the nonesonant shaking foces ae acceptable, then the system is acceptable and the analysis is complete. This is the technical appoach that has seved the industy well fo decades. If, howeve, the above-mentioned citeia ae not met, specific additional steps can be taken to justify (with calculations) that the system may be acceptable even though it violates the DA-3 citeia. The DA-3 citeia again ae low pulsation, low shaking foces, and nonesonant mechanical systems. If these citeia ae violated, the additional steps include calculation of vibation levels fo compaison to a newly included allowable vibation cuve and ultimately calculation of cyclic stesses if the vibation citeia ae not met. A close examination of these steps eveals that one should not like to be in the situation of having to esot to these analyses. Revise Design Revise Design Step 3a Mechanical Analysis of Compesso System and Piping System pe (M.4, M.5) Sepaation Magin Citeia Met? OR OR Step 3b1 Pefom foced mechanical esponse analysis on Compesso Mechanical Model pe (M.6) Compesso System and Piping Vibation Citeia Met? OR Cyclic Stess Citeia Met? OR No No No Step 3b2 Pefom foced mechanical esponse analysis on Piping System pe (M.7) Design Appoach 3, Continued Figue 25. Flow Chat fo API 618 Fifth Edition. No Yes Yes Yes Shaking Foce Citeia Met? Design Appoach 3 Note: M.8, M.9, M.1, M.11 Optional Yes Design Appoach 3 Analysis is Complete Design Appoach 3 Analysis is Complete Design Appoach 3 Analysis is Complete

10 192 PROCEEDINGS OF THE 3TH TURBOMACHINERY SYMPOSIUM Fist, to be in the position of having to calculate vibation levels equies that the shaking foces ae eithe too high o thee is a mechanical natual fequency of the manifold system o piping that violates the sepaation magin. If the shaking foces ae too high, the vibation levels should also calculate to be too high, since the shaking foce citeia wee deived fom the allowable vibation cuve. If the sepaation magin is violated, then fo pactical puposes the chances of the system being esonant ae high, which is obviously an undesiable situation. Finally, if the point is eached whee the cyclic stesses must be computed, this means that the foces ae high, the vibation is high, and the system is likely esonant. It follows then that the cyclic stesses, if computed popely, will likely be significant. Even if the calculations show the stesses to be acceptable, given the limitations on the accuacy of these calculations, a system equiing steps 3b1 o 3b2 to satisfy DA-3 is necessaily a high isk system. In addition, the stess calculations typically apply only to the main pocess piping, not the numeous banch connections, instumentation, etc., that ae esponsible fo the majoity of failues in industial piping. The flow chat (Figue 25) illustates that Design Appoach 3 can be satisfied by any of the design steps, 3a, 3b1, o 3b2. This is an impovement ove the Fouth Edition in that Step 3a clealy satisfies the design citeia as oiginally intended in the Thid Edition. Pogession to Steps 3b1 and 3b2 ae only equied if the designe has failed to contol pulsation levels popely and/o the system does not meet the sepaation magin guidelines (i.e., the potential fo esonance exists). This is not to say that stuctual analysis tools ae of no pactical use. They can be used with good success in the design stage to avoid esonance if the limitations ae undestood, and they ae extemely useful in designing coections when used in conjunction with field measuements. Table 7 shows analysis pocedues that will be optional in the new Fifth Edition. Table 7. Optional Pocedues (New Fifth Edition). Figue 26. Failue of 1 Inch Suction Piping. Speed (RPM) 6X 1 1X 2X 3X 4X 5X 1 DIV X X 75 Mechanical Resonance in Suction Piping Figue 27. Mechanical Resonance of Suction Piping. Example Compesso Field Test Data Speed Run RPM Unit Designed pe API 618 4th Edition -D/A 3 Using Foced Response Analysis to Justify Design 1) S Pipe Vetical (Watefall) 2 MILS PP/Div 2) S Pipe Vetical (RPM Slice) 5 MILS PP/Div Piping System Themal Flexibility 1 1X 2X 3X 4X 5X 6X Example Compesso 1 DIV Field Test Data CASE HISTORIES ILLUSTRATING LIMITS OF MECHANICAL MODELING Failue of 1 Inch Suction Pipe This fist example illustates the false sense of secuity ceated by a design that (with good intention) met the API 618 Fouth Edition (1995) equiements. An acoustical simulation was pefomed; howeve, pulsation filtes wee not designed and a poo mechanical layout was used. The basic design philosophy of avoiding esonance was not followed. Instead, a stuctual analysis of the system was conducted, including foced esponse calculations to justify the design even though high pulsation and potential esonances existed within the speed ange. The calculations pedicted low vibation and stess levels. The authos wee consulted to assist the use company afte a failue occued in the main suction piping. The layout, showing long unsuppoted spans is illustated in Figue 26. The failue occued in the 1 inch suction piping at the location indicated. Field tests showed that the piping mechanical natual fequency (Figue 27) was excited by pulsation enegy (Figue 28) within the opeating speed ange. Measued vibation levels wee well above the allowables. The measued vibation mode shape showed that the failue location was as would be expected (Figue 29). Speed (RPM) Acoustical Resonance in Suction Piping Figue 28. Acoustical Resonance in Suction Piping. Failue of Suction Bottle Nozzles Even if pulsation levels ae contolled to acceptable levels, excessive vibation can occu if mechanical natual fequencies exist nea significant mechanical excitation fequencies. Figue 3 shows a suction bottle system in which a volume-choke-volume filte was constucted in a single vessel. Fom a pulsation contol standpoint, this design was effective. Howeve, it esulted in a lage 7X 8X Speed Run RPM Unit Designed pe API 618 4th Edition -D/A 3 Using Foced Response Analysis to Justify Design 1) Sct Yoke 2 psi PP/Div 2) Sct Yoke (RPM Slice) 1 psi PP/Div

11 THE NEW FIFTH EDITION OF API 618 FOR RECIPROCATING COMPRESSORS WHICH PULSATION AND VIBRATION CONTROL PHILOSOPHY SHOULD YOU USE? Hz Phase = 126 Deg 1 DIV Compesso Manifold System Impact Response Data FW (#3 Cylinde End) All 3 TLA-6 Units Down Othe Engines Running New 42" Suction Filte Impact and Measue on Both Ends of Suction Bottle in Stetch Diection 15 Hz Phase = 126 Deg Note: 15 Hz=Suction Bottle Cantileve Mode Rise (#1 Cylinde) End Figue 31. Impact Response Measuements of Suction Bottle. 33 3X 5X 6X 7X 8X 9X 1 DIV Vibation Measuement 32 Suction Bottle Rise End Z Diection Figue 29. Measued Vibation Mode Shape. diamete, elatively heavy suction bottle, which is difficult to dynamically suppot. It was pactically impossible to aise the calculated mechanical natual fequency of the suction bottle cantileve mode above the ange of expected excitation fequencies, so the designe specified longe suction nozzles to place this mode in between the thid and fouth hamonics. Once again, thee was too much confidence placed in the ability to calculate the mechanical esponse with such pecision. The pedicted fequency was 17.4 Hz, while the measued fequency was 15 Hz, Figue 31. This was within the ange of excitation by the thid hamonic fo this vaiable speed machine (25 to 3 pm). Excessive vibation levels occued (Figue 32), and a suction bottle nozzle failue esulted. Figue Inch OD Filte in Single Bottle. Gas Tansmission Station Piping System Figue 33 shows a sketch of a six-compesso dischage system fo a gas tansmission and stoage facility. Six units, with fou diffeent compesso types, wee opeating in a common dual heade system fo stoage o withdawal. Compessos wee added to the system ove a 25 yea peiod as demand inceased. Unit 1 was the only compesso initially. It utilized volume-choke-volume filtes and opeated without poblems. Unit 2 was added a yea o Speed (RPM) mils (p-p) at 15 Hz Figue 32. Measued Vibation on 42 Inch Suction Bottle. 1) Suction Bottle (Watefall) 2) Suction Bottle (RPM Slice) 2 MIL PP/Div so late and the seconday volume was eliminated to save on costs. Vibation levels emained acceptable. A few yeas late, the thid unit was added and the logic was that since Unit 2 was acceptable without an acoustical filte, the addition of Unit 3 without filteing should also be acceptable. At this point, mino poblems began to occu with pipe clamps and instument tubing. A fouth highe speed unit was installed that did not inteact significantly with the othe thee units. Unit 5 was installed afte a few moe yeas, again without a filte. Chonic poblems developed with sevee vibation occuing at cetain opeating conditions. The sixth unit was added including a volume-choke-volume filte, but the vibation poblems pesisted. Pipe clamps in the dischage heade epeatedly failed and pneumatic actuatos expeienced excessive vibation and failues. Field testing showed that seveal system acoustic esonances wee excited by the unfilteed units. This pulsation enegy caused excessive vibation of the heades and also coincided with mechanical natual fequencies of seveal valve actuatos (Figue 34). The ultimate solution was to install volume-choke-volume filtes fo each unit. TYPICAL PACKAGED HIGH SPEED DESIGN Figues 35, 36, 37, and 38 compae layout equiements fo simple suge volumes and eactive filte designs. These examples ae fo a single stage, fou-thow compesso typical of many packaged high-speed units. The eactive acoustical design (Figues 36 and 38) utilizes intenal choke tubes and extended length dischage bottles. The scubbes ae utilized as seconday suction volumes. These designs geneally do not equie significant

12 194 PROCEEDINGS OF THE 3TH TURBOMACHINERY SYMPOSIUM Figue 33. Gas Tansmission Dischage Piping System Model Six Compessos in Paallel. Figue 36. Skid Mounted Compesso Layout with Reactive Acoustical Filte Designs. 1 DIV Test 4 Conditions Vaying Speed of Unit 5 Only Unit 5 Pumping Heade B All Units on Same Suction Ps = 48 psig Unit 1 Down, A Dsch Unit 2 Down, Both Blkd Unit 3 31 pm, A Dsch Unit 4 Down, A Dsch Unit 5 Vaying, B Dsch Unit 5, B Lead Line - 5 psi/div Unit 6 3 pm, A Dsch Pd (A Heade) = 631 psig Pd (B Heade) = 698 psig Tacking 4th Ode Speed: RPM B-1 Heade Nea Unit 9 Lead Line - 5 psi/div 34 mils at 18.5 Hz 2 mils at 2 Hz Actuato Vibation - 2 mils/div Figue 34. Measued and Vibation in Dischage System. Figue 37. Plan View of Layout without Reactive Acoustical Filtes. additional skid aea o fabication and mateial costs, while offeing supeio pulsation and bottle foce contol as compaed to simple suge volumes and oifice plates. Note that two scubbes ae used to avoid the type of poo layout that esults when a single scubbe is used (Figues 39 and 4). This is impotant since the choke tube connecting the scubbe and suction bottle is subject to significant mechanical and acoustical excitation. Figue 38. Plan View of Layout with Reactive Acoustical Filtes. Figue 35. Skid Mounted Compesso Layout without Reactive Acoustical Filtes. Figues 41 and 42 show special dual choke tube designs that ae equied to eliminate cetain acoustical modes on compessos with wide speed anges (e.g., 2:1 tun down). CONCLUSIONS The intent of API 618 at the Thid Edition was that Design Appoach 3 meant effective pulsation contol. This usually equied that eactive filteing be used in elatively high mole weight systems. Fom the use s pespective, a Design Appoach 3 system was a safe and eliable system. This standad seved the industy well fo many yeas. The Fouth Edition attempted to define the steps equied to qualify a piping system, in the event that the allowable pulsation levels wee exceeded. This ceated confusion and led to systems being designed with less emphasis on pulsation contol, and

13 THE NEW FIFTH EDITION OF API 618 FOR RECIPROCATING COMPRESSORS WHICH PULSATION AND VIBRATION CONTROL PHILOSOPHY SHOULD YOU USE? 195 Figue 39. Poo Suction Layout fo Skid Unit Resulting fom Use of Single Scubbe. Figue 41. Example of Dual Choke Tube Reactive Filte Design Using Scubbe Volume Suction. Figue 4. Altenate View of Poo Suction Piping Layout Resulting fom Use of Single Scubbe. justified with mechanical esponse calculations of questionable validity. This is not to say that all systems designed this way ae unsafe o uneliable. Such systems can wok acceptably well, but it is moe often the esult of good fotune athe than accuate esponse calculations. Since the Fouth Edition was published, numeous uses have been in the situation of having puchased a Design Appoach 3 System, yet ending up with unacceptable and potentially catastophic esults. The Fifth Edition will claify the confusion that esulted fom the addition of the language concening mechanical foced esponse calculations in the Fouth Edition. The use will now be able to detemine if the system meets Design Appoach 3 by the use of the technically sound pulsation and shaking foce contol philosophy (Step 3a), o though the use of the highe isk philosophies based on mechanical foced esponse calculations of steps 3b1 o 3b2. The authos expeience, as pesented in this pape, shows that obust pulsation contol though the use of eactive acoustical filtes is equied to achieve safe and eliable piping systems of ecipocating compessos. REFERENCES API 618, 1986, Recipocating Compessos fo Geneal Refiney Sevices, Thid Edition, Ameican Petoleum Institute, Washington, D.C. Figue 42. Example of Dual Choke Tube Reactive Filte Design with Seconday Bottle Dischage. API 618, 1995, Recipocating Compessos fo Petoleum, Chemical, and Gas Industy Sevices, Fouth Edition, Ameican Petoleum Institute, Washington, D.C. API 618, 21, Recipocating Compessos, Daft of Fifth Edition, Sub-Task Foce Committee Coespondence. BIBLIOGRAPHY Atkins, K. E., Tison, J. D., et al., 1999, Vibations in Recipocating Machiney and Piping Systems, Engineeing Dynamics Incopoated Semina Manual, San Antonio, Texas.

14 196 PROCEEDINGS OF THE 3TH TURBOMACHINERY SYMPOSIUM