aa Feature Report Part1 Evaluate Heat-Exchanger Tube-Rupture Scenarios Using Dynamic Simulation Applying dynamic simulation models to tube-rupture scenarios can help ensure more accurate sizing and hazard assessments Soumitro Nagpal Fluor Daniel India Put. Lid IN BRIEF TUBE-AUPTURE FLOW -STIMA PELEF OE SELECTION FIGURE 1. Heat exchangers in many CP applications can experince tue rupture and these scenarios mus be evaluated and “dees inorder to ensure continued sale operations ibe rupture in heat exchangers is an ‘extremely serious issue in the chemi- cal process industries (CP; Figure 1), Heat-exchanger tube-rupture sce- narios can be evaluated by various reans, including dynamic and steady-state process simulation. Performance of reliable dynamic simulations of tube-rupture scenarios re- ‘uires suitable experience, expertise and significantly higher effort than steady-state evaluations. This article discusses a number Of heat-exchanger tube-rupture scenarios, including: the requirements for scenario eval- uation; tube-rupture flow estimation; selec- tion and sizing of relief devices and associ- ated piping; and guidelines for identification CHEMICAL of situations where a dynamic evaluation is appropriate, Ih addition, several dynamic-simulation case-study summaries are presented, illus- trating the utlity of this method in the proper selection and sizing of relief systems. Spe- cial emphasis is given to liquid-filed sys- tems, which see very rapid pressure-rise rates upon tube rupture. These cases show that dynamic evaluation can lead to better reliet-device selection and sizing, as well as improved materials selection for various con- ditions, including some that do not require dynamic evaluation under American Petro- leum Institute (API; Washington, D.C.; www, api.org) guidelines. ENGINEERING WWW.CHEMENGONLINE.COM FEBRUARY 2015,HP Fi Natural gas! LP ud Propane ‘evaporator case Design pressure bang 89/17 wen) Operating pressure berg 09/1 wen) Operaing tempera. °C 1610-20 ture HIP) 128 Blockod:inpressue- basis 0.16 sorale Setpressue, Fy bang 17 ole even Pov Pak pressure, agg bang 184 Pan! 22% Prakriptuefow — kals 37 Prakrelatfow kas 48 STU Td tural gas! Natural gas/MydrotreaterNatura-gas Ethylene /—_—Ettylene/ Propane —-LPsteam[q_ effluent) condensate Liquid guid ‘condenser Stripper Liquid propane propylene [7] mathanol [7 bottoms (6) case 2 cases case 4 cases case ase? wir 9/55 183/128 aa/7 aig iz oie na/astsa/19 4/72 woisi19 101/96 2wi0/18 — awa0/147 §—27S19310/ 20/18 -261011/25 111015/64 23010 200, 092 5 ™ or a 6 7 55 2a w 162 8 Psy Pov Psvia) ——upure dk PS Pov 182 6 a7 105 205 16 % % ani 03% 26% 26% 36 wa 57 134 7 18 125 a9 1052 Noteoported Not reprted Note: Cases 4 through 7 refer to exchanges that ae uid onthe LP side Evaluation criteria Process-hazard analyses for shell and-tube heat exchangers call for evaluation of tube-rupture scenarios if the difference in maximum allow- able working pressure (MAWP) be- tween the low-pressure (LP) and high-pressure (HP) sides is greater than that covered by the 10/13 (or 2/3 rule, as may be applicable). This is based on ASME Boller and Pres- sure Vessel Code Section Vill, Div.1 [7], which requires a system to be hydro-tested at 120% of its MAWP. This code requirement eliminates the need to evaluate a tube-rupture scenario, as the LP side is protected ifits design pressure is no less than 10/18 of the HP side's design pres- sure. API-521 6th Ed. [2] notes that “Pressure relief for tude rupture is, Not required where the low-pressure exchanger side (including upstream and downstream systems) does not exceed the criteria noted above. The tube-rupture scenario can be mit gated by increasing the design pres- sure of the low-pressure exchanger side (including upstream and down- stream systems), and/or assuring that an open flow path can pass the tube-upture flow without exceed- ing the stipulated pressure, and/or providing pressure relief." This option is often used to elii- nate a tube-rupture scenario evalua- tion, However, for an exchanger with a large pressure difference between the HP and LP sides, uprating the LLP side design pressure can be ex- Pensive, and a tube-rupture scenario evaluation is necessary. The required activities in a tube-rupture sce- rrario evaluation are detailed in the following sections, ‘Tube-rupture flow estimation ‘AP520, Part 1 [3] provides the nec- essary guidelines and equations for single- and two-phase flow estima- tion. For two-phase flow, API-521 [2] recommends the following: “A two- phase flow method should be used in determining the flowrate through the failure for lashing liquids or two- phase fluids. The flow models de- veloped by DIERS (the Design Insti- tute for Emergency Relief Systems) and others can be adapted for this purpose. In cases where the fluid flashes at the low-pressure side of the heat exchanger, two-phase flow methods based on the homogenous equiibrium model (HEM), such as those proposed by DIERS, may be Used for the flow through the tube to the break.” Darby's work in this area [also presents a good summary of these steady-state sizing methods, includ- ing the homogenous direct-integra- tion (HDI) method, Relief-device selection In. systems where the shellside- tubeside pressure difference is high, CHEMICAL ENGINEERNG WwWW.CHEMENGONLINE.COM FEBRUARY 2015 the pressure rise on the LP side can be very rapid. In those cases, relief valves provided to protect the LP side from overpressure may not re- spond quickly enough to protect the system. Such a scenario may call for the installation of a rupture disk. Many companies have developed their own practices for identifying scenarios that require protection with rupture disks. Some companies pre- fer that the rupture disk is mounted directly on the exchanger — not on the inlet or outlet piping — because the disk might not react fast enough with the intervening pipe. This is not always practical to implement, since other factors must be considered, such as layout constraints or free draining requirements for two-phase Or liquid-relief ines. Thus, it is rec- ‘ommended that rupture disks be in- stalled as close as possible to, if not directly on, the heat exchanger. Liquid-flled systems, especially those filed with an incompressible liquid, such as water on the LP side, can experience a very rapid high” pressure spike upon tube rupture. Even a rupture disk may not be able to respond. Multiple disks at diferent locations on the exchanger itself may be required. Depending on the toxic- ity, flammability and reactivity of the fluids, consideration of tube-rupture- relief contingency may be advisable, evenif the 10/13 ruieis followed. The acoustic pressure wave that occurs a |ry RECOMMENDATIONS ON DYNAMIC MODELING FROM API-521 [7] £6 A. one-dimensional dynamic model can be used where the approach is to simulate the pressure profle and pressure transionts {developed in the exchanger fom the time ofthe rupture. These methods generaly include the dynamic model ofthe tube- pure reli scenario and the response time ofthe relief device, the accuracy of which i erica in caiculating the accuracy ‘of pressures generated, The opening tme forthe device used should be verified by tha manufacturer and should also be compatibie with the requirements ofthe system. ‘This ype of anaysisis recommended, i addition othe steady-state approach, where theresa wide difrence in design pressure between the ‘wo exchanger sides for example, 7,000 kPa -1,000 ps} or mor]. especialy where the lw pressure side is quid-land the High-pressure sido Contains gas oa fui that flashes across the rupture. Modeling has shown that under these circumstances, transient conditions can produce overpressure above he test pressure, even when protectedbyaPRD (pressue-rlief device), nthase cases, adctionalprotectonmeasures should bbe considered,” in the case of incompressible-liquid- filed systems results in a waterham- mer-type effect that cannot be simu- lated with the dynamic simulation tools commonly used for analysis of these scenarios. Sometimes @ rup- ture pin is also considered, allowing faster open times than rupture disks. ‘Typical opening times for relief de- vices are 50-100 milliseconds (ms) for spring-loaded pressure safety valves (PSVs) and around 1-10 ms for graphite rupture disks [5] The capacity of a relief device han- dling high pressure-rise rates can be estimated using Equation (1), tak- ing into account the pressure rele device's (PRD) opening capacity (00), set pressure (SP), allowable overpressure (AOP) and relet-valve ‘opening time (7). OC = SP x AOP/T a For example, for a PSV with a set pressure of 5.5 bars, with 10% over- pressure allowed, and an opening time of 100 ms (0.1 s), OC = 5.5 bars 0.1/0.1 s = 5.5 bare/s, The blocked-in pressure-rise rate (where no credit is taken for any flow out of the exchanger) caleu- lated in dynamic runs can be com- pared with the PRD opening capac- ity to determine the suitability of the selected type of relief device. If the PRD's opening capacity is lower than the blocked-in_pressure-rise rate, the device may not be able to ‘open sufficiently in time to capture the pressure rise. Relief-device sizing The steady-state calculation pro- cedures described previously are applicable to tube-rupture flow est mation. However, sizing of the rele! device itsef and its inlet and outiet lines requires relief-iow estimation. ‘The API-521 [2] guideline states that for steady-state relief device siz- ing, “PRD size should be based on the gas and/or liquid flow passing through the rupture The volumetric flowrate of the LP fluid equivalent to the volumetric flow- rate of the HP fluid passing through the rupture location should be used when the fluid properties at the re- lief location are significantly different from those at the rupture location, or when the relief device is located at some distance from the exchanger. Alternately, API-521's recommenda- tions with regard to dynamic model ing are shown in the box above. in summary, API-521 recommends that for exchangers with differential pressures ofless than 70 bars, reliet- device sizing should use the tube- rupture flow estimate. For exchang- ers with differential pressures greater than 70 bars, use a dynamic simula- tion of the system to estimate relief flows and relief-device sizing Olther situations where a dynamic evaluation is sometimes considered are the following; ‘+ Exchangers with iquid-flled shel ‘Sizing of relief piping, because steady-state methods can lead to oversized relief piping, especialy for short, sharp-peak rele flows, a8 pipe holdup effects are not considered ‘Long shells with baffles on a tight pitch ‘For existing units where the over- pressure protection was mounted (on the relief piping ‘* For existing units with high differ ential pressures where a relief valve was used Examples Several tube-rupture dynamic-sim- Ulation studies are presented in the following sections. Cases 1 through 5 are recent tube-rupture studies that utiize commercial dynamic-sim- Ulation software. In these cases, the heat exchanger’s LP side is spit into several volume segments to facilitate the observation of the physical phe- nomena occurring during the rupture event. The details of this approach are discussed in Ref. 6. Cases 6 and 7, as reported by Ennis and others [7], are also discussed here for com- parison with these resutts for liquid- filed tube-rupture cases, Table 1 gives a summary of the cases. Case 1: Natural-gaspropanechiller. Natural gas is cooled in a chiller to reduce the hydrocarbon dewpoint, where liquid propane evaporates at 1 barg and -24°C on the shellside. ‘The natural gas is on the tubeside at 68 barg pressure and tempera- tures around ~16 to ~20°C. The ex- changer in this example is specified as type AKL based on Tubular Ex- changer Manufacturers Association (TEMA; Tarrytown, N.Y.; woww.tema, org) standards. TEMA exchanger type codes, such as AKL, provide a shorthand for basic designs and manufacturing methods of heat ex- changers. For more information on TEMA standards and codes, see Specifying Shell-and-Tube Heat Ex- changers, Chem. Eng., May 2013, pp. 47-53. Dynamic evaluation of a single-tube, ful-bore rupture in this exchanger ‘indicated a slow pres- sute-rise rate of 0.16 bars/s in the shell, as the pressure rises from 1 barg operating pressure to 17 barg relief pressure in 100 s. A spring- operated safety valve is adequate to protect the LP side of the exchanger for this scenario. Figure 2 shows the exchanger sheliside Segmentation considered for the simulations. Figure 3 shows the temperature, pressure and flow transients for the run. In this run, the rupture wes considered at the ex- treme left of the exchanger in seg- ment V1. The gas cools as it expands, CHEMICALENGINEERING WWW.CHEMENGONLINE.COM FEBRUARY 2015,negnte toate : 7 wl) ow | | " wl ow ws victece tect ' Tae FIGURE 2. Tis schematic shows an exchanger shelside segmented into several volumes to factate the observation ofthe piysical nomena occurring during the rupture een. n this case P as flows from et right the propane gud ivelin the shel i 1800 mm and te tubes are not shown fon the LP side upon flowing across the tube rupture, resuiting in its tem- perature dropping to -87°C immedi ately after rupture, and subsequently increasing as the shelside pressure rises. While the fluid temperature in the shellside is estimated to drop to ‘a minimum of ~48°C near the rupture location, direct impingement of the rupture fluids on the intact adjacent tubes, shell and tubeshest is pos- sible. Conservative design practices led to a selection of the minimum design metal temperature (MDMT) for the exchanger based on the minimum rupture fluid temperature. These calculations did not consider the thermal inertia of the metal mass in the system, inclusion of which can allow a higher MDMT selection. Case 2: Natural-gas-cooled refrig- erant condenser. A propane-reftig- erant-based natural-gas dewpointing plant utilizes the cooled, dewpointed gas to condense refrigerant dis- Charging from compressors. Reftig- erant at 5-8 barg condenses on the sheliside, whereas HP natural gas at 67 barg flows on the tubeside of the exchanger, which is TEMA type BEM. Dynamic evaluation of a single-tube, fullbore rupture in this exchanger indicated a relatively slow pressure-rise rate of 0.92 bars/s in the vapor-filad shellside, as the pressure rises from 8 barg operating pressure to 17 barg relief pressure in 98s Figure 4), ‘A spring-operated safety valve is, adequate to protect the LP side of the exchanger. The dynamic simula- tion indicated that a PSV size based on the tube-rupture flow is inade- uate to limit the pressure ise to be within the required 10% overpres- sure. The PSV size was increased, and while this leads to an increased peak relief above the tube-rupture ‘low, the peak pressure rise can now be contained to the prescribed limit. Case 3: Natural-gas heating with LP steam. This study's work [9] focused on HP natural gas on the tubeside of an exchanger heated with condensing LP steam on the shelside in a TEMA-type BEM ex- eebes be 100 » 18 16 up tmpeatu,¢ iu eportanperse uk apr tanger = VF ui tempers, {8 bk iu empertc, c ‘Noone fw hos Fal | Vives esi, bag ves ese ag a ar Tam = FIGURE 2. case 1's tube-rupture event occurs 18s nto the run, while eto starsat 118 nthe ‘un, Note thatthe temperature rises in Segment WE due to pure compression without exposure to cold rupture tuids CHEMICAL ENGINEERNG WwWW.CHEMENGONLINE.COM FEBRUARY 2015 changer. The natural gas is at 113, arg pressure and the temperature ranges from 3.2 to 80°C. The LP steam is at 3.5 barg and 147°C. The set pressure on the relief device pro- tecting the LP side is 5.5 barg. Dynamic evaluation of a. single- tube, full-bore rupture in this ex- changer indicated a shellside pres- sure-rise rate of 5 bars/s, increasing from 3.5 barg operating pressure to 5.5 barg relief pressure in 0.4 s. Upon relief valve opening, the pressure-rise rate drops to 0.27 bars/s. it was de- termined that this pressure-rise rate is acceptable for the installation of a spring-loaded relief valve. The study also indicated that the PSV set pres- sure should be set lower than the exchanger design pressure, as the pressure near the tubesheet can be higher than at the PSV inlet due to inlet line losses. Case 4: Hydrotreater effluent stripper-bottoms exchanger. A hy- drotreater effluent stream is cooled against the stripper bottoms stream of the unit in a shell-and-tube heat exchanger. The exchanger com- prises two shells in series, with the HP hydrotreater effluent’ on. the tubeside at 158 berg and tempera- tures of 275-310°C. The stripper bottoms pressure on the shellside is 17-19 barg and the temperature is 280-290°C. The sheliside design pressure was originally selected as 128 barg, based on the 10/13 rule. Dynamic evaluations for single and muttiple tube-rupture scenarios were executed [6]. On tube rupture, a very rapid pressure rise in the liquid-filed shell was estimated at 74 bars/s, with pressure rising from 18.9 barg operating pressure to 23 barg relief- valve set pressure in just 55 ms. Peak pressure rise in the shell pro- tected with one, two or three spring- loaded relief valves was estimated. The dynamic runs indicated that the peak pressure rise can be limited to 87.7 barg, 50.1 barg and 81.4 barg using three, two, and one PSV, re- spectively. Thus, a significant margin for shell design-pressure reduction is available below the originally se- lected design pressure of 128 barg. The study showed that despite the rapid pressure-rise rate, PSVs with a set pressure selected well below the design pressure to capture the pressure rise can, in fact, be used to lower the LP design pressure a |ry 52 a 1800 00 000 ‘00 00 00 200 oo b ape tos Feet tts J sat pessue.bay times FIGURE 4. For Case 2, his transient ekagram shows the exchanger LP shale pressure rise on tube rup- ‘ur, ruptrefws and rele flows, fora PSV set pressure of 17 barg Case 5: Liquid propane cooled by natural-gas condensate. A p10- pane-refrigerant-based natural-gas dewpointing plant uses natural-gas condensate 10 sub-cool liquid reftig- erant. Liquid propane at around 5-8 barg pressure is on the tubeside, and HP natural-gas condensate at 43 barg flows on the shelside of the exchanger, which is TEMA type AEL. The condensate state is two- phase, vapor-iquid in and out of the ‘exchanger. Dynamic evaluation of a single-tube, ful-bore rupture in this ‘exchanger indicated a fast pressure- rise rate of 65 bars/sin the liquid-filed tubeside. The pressure rises from an operating pressure of 7.2 barg to 17 1.00 barg relief pressure in 0.16 s. ‘Arupture disk was selected as the relief device, Dynamic simulation of this scenario was carried out, and shows that the peak relief flow on disk rupture in liquid-filed systems can be an order of magnitude higher than the associated tube-rupture flow driving the event, as shown in Figure 5. The dynamic model was Used for sizing the rupture disk’s inlet and outlet lines based on the tran- sient relief flows, with the criteria that the LP side must stay below the al- lowable design overpressure, Accurate modeling of this system involving two-phase choked flow in pipes is a challenge, even with the 1a Nass nbs Faye toe bos resre e9 ° ue or a Timms FIGURE 5. Case 5's rupture occurs 0.05 into the run, flowed by rapid pressure ris o rupture dis burst ‘rossre of 17 bry a 0.2, nial rll Mow canbe sen Tobe sgniiarly higher than te rupture Now advanced dynamic simulations pack- ages available commercially. A com- mercial simulation model, which uses a momentum conservation equation based on Darcy's law, was used in this study to model the rupture disk’s inlet and outlet lines. The exchanger tubeside was modeled using three volume segments, one each for the channels, and one to account for the volume of the intact tubes. Fig- ture 6 shows the simulation process flow diagram (PFD) used for the transient analysis Cases 6 and 7: Liquid-filled ex- changers inan ethylene plant. Enris [6] presented dynamic simuiation re- sutts for two liquid-filed exchangers, the first with liquid propylene, and the second with methanol on the LP side. Table 1 provides further syste de- tails. For both exchangers, the HP-LP pressure differential is >70 bars. The study used models that solve one- dimensional partial ditferential equa- tions for the conservation of mass and momentum for the liquic-flled riser to the PSV attached to the exchangers. This method is @ more rigorous @p- proach for two-phase pipe flow than available in some dynamic-simulation software packages. ‘The blocked-in pressure rise cal culated was 28 bars/s for the pro- pylene-fled system, and 64 bars/s for the methanol-filed system. PSVs ‘were proposed as the relief device for both cases. Note that as in Case 4, the PSV set pressure was significantly below the LP design pressure, allow- ing pressure to rise to above 10% of the PSV set pressure. Closing thoughts The folowing are some conclusions, that can be drawn based on the in- formation presented in this article Dynamic simulations can allow for more accurate PSV sizing for tube- rupture scenarios than steady- state methods # Based on dynamic simulation studies, PSVs can be used to protect’ liquid-fled exchangers with high pressure diferentials of over 70 bars ‘* Dynamic simulation should be con- sidered for evaluating tube-rupture scenarios of iquid-filed systems for releF-device selection and sizing, even when exchanger differentia pressures are less than 70 bars, # the exchanger LP side is not pro- CHEMICALENGINEERING WWW.CHEMENGONLINE.COM FEBRUARY 2015,pre + owe ] ese i vikoo esto! ‘esa Eo rape noo fe oviove 2 sone ame tat TA in slr dec ARETE wie e100 Retigeane ———PRETO2 fp 1 ia 1 et tiger ev. oe gis ky aa wee | Venere rao To a [ reser 7230 oom my ET asec tox [et [HUSTED [Eashn —TE2E igh] tected by the 10/13 rule ‘© Line sizing for the rupture disk’s inlet and outlet should be based (on the transient relief flows, with the criteria that the LP-side’ pres- sure rise is limited to the allowable design overpressure ‘© Dynamic simulations allow evalu- ations of suitable exchanger mini- mum metal design temperatures ‘© Dynamic evaluation of the heat- exchanger tube-rupture scenario is often simplified by looking at the exchanger in isolation from the larger system of which it is a part, since the assumption of instanta- neous closure of inlet and outlet isolation valves and emergency shutdown valves (ESDVS) on tube rupture will give the fastest, and thus most conservative, pressure- rise estimate. Hazard analysis, should, however, consider the im- pact of actual ESDV closure rates (on the system beyond the ESDVs, especially for liquic-filed systems, which see very rapid pressure” rise rates For further guidance, Figure 7 presents a decision chart for evaluat- ing tube-rupture scenarios based on API-521 guidelines. Figure 7 and the cases discussed above show that dynamic evaluation can lead to bet- ter reief-devioe selection and sizing, and materials selection for various eee isthe prtd Se nt RE ES nteaeur suet tons EE Pe nageuraen: oP se ‘of HP design pressure? a eas Is desgn es Ines eee ananece —aEy~ ane fnce® pd vase 1. Eauaton may ibe equre feat ts 2 Asta Case orn cours where be #4? presare fons <7 as, ani senator may tent fhe elie Tho esau rca lt tonap Se 5 apn capac Fen #3 rupees se, rt an cal in sing bed say sie lo alkene 2. ett by Cate 2. an ition an law far er aca RD ng FIGURE 7. Based on AP-521 guidlines, this decision chart for evaiating tube-rupture scenarios can hap ‘nginersdearmine i steady-state or dypamic evaluation is most suitable CHEMICAL ENGINEERNG WwWW.CHEMENGONLINE.COM FEBRUARY 2015 FIGURE 6. For Case 5, the heat exchanger'sLP ‘def sygmente into three volume elements: V1 (oletchanne, v2 (bes) and V3 (ott ctanne) rs wer wre ete tas yor i esov2 | Ss Paige out a te Tero Venger rosr —[e a ae fears is fon — OST [hah conditions, including some that do Not require dynamic evaluation under the API guidelines. / Edited by Mary Page Bailey References 1. ASME Boia nd rescue Vesa Code, Sacto Bi 1 20g) U6- 200, 401 Sanda 821, Presse ledng and Dap. suring Syst" i aon, Janay 2014 401 Slandard 520, ‘Siang, Saletan staan of Presse lag Deces i atoais Pat] — Suing and Slot, 8h Editon Doenber 2008 ty, Su Salty Ret aves Ay Cantons, (han, Sop 2005, pp 42-50, 5. Ofsure Tecinolny ‘Rap 2000023, “Ts ing and ana of alt dco opening ues." Propard. by’ PS @Ppeine Smuaion and. tog) Lid Yor to Hath and Satay Excite, UK 9.65. Urano, and Loni, namic Smuaton of Tube Fuge High Prose Heat Exctarges ‘fa jtoacing Unt tobe pubis a yaca ban Processing, Fans, C.J, Babes, KK, and Pa, C, “Danle del awa oxéuago ube ntire dicta 4 hgh ress aching qu nolo grass Hiqu-tld ste, Joual ol Loss Preeragn te Poca insti pp. 111-121, Vl 242011 8. Ha, Hl, “Dynami Sinvston Aras for he Tube Rap of a Natal Gas Hoar LNG Pant” Gah Caradan Chemical Eno. ing Confrence, agama als, Ona, Carat, (etter 19-22, 2014 ‘Soumitro Nagpal sue po ass pet with For Oni hada PA Lid, Bung 88, 3d fl OL je Cy cua Nayana 122002, inde Ea spentanegpanurcan) wi 21 years of Wak exgerence process covlpmat sin, en ‘Goong, and mandgamet. He fas wide-ranging egorance in roca odolng and seul an is 2 Fs tan Feder for eacutng rani siuiaon roots oe tian gas and pe ins. Sou a gad ale of i inst of Techy and Since Pla {BE} in chomcal engoang, and recabed 2 n., from to Unversy a Uh, capt