PREDICTION OF THIRD PARTY DAMAGE FAILURE FREQUENCY FOR PIPELINES TRANSPORTING MIXTURES OF NATURAL GAS AND HYDROGEN Zhang, L., Adey, R.A. 2 C M BEASY Ltd, Ashust Lodge, Southampton, SO4 7AA, UK, Lzhang@beasy.com 2 C M BEASY Ltd, Ashust Lodge, Southampton, SO4 7AA, UK,.adey@beasy.com ABSTRACT As Euope is gadually moving towads a hydogen based society it has been acknowledged that adding cetain amount o hydogen, as a clean enegy caie, to the existing natual gas pipeline will help educe the CO 2 emissions which contibute to the geenhouse eect. On the othe hand, hydogen has been demonstated to be able to change the behaviou o the pipeline steel such as lowe toughness and aste cack gowth due to hydogen embittlement. Theeoe, it is necessay that the isks associated with the ailue o the pipeline caying mixtues o natual gas and hydogen be assessed. The study epoted in this pape is pat o Euopean NATURALHY poject, whose aim is to investigate the possibility o using the existing natual gas tansmission pipelines to convey natual gas/hydogen mixtues. Accoding to the EGIG database, the most common cause o ailue o the existing natual gas pipelines is thid paty damage, which mainly ees to a gouge, a dent/gouge combination o known geomety. Among thid paty damage ailues, 9% ae the esult o immediate ailue i.e. leakage o uptue o the pipeline and only % o them ae the esult o delayed ailue. While its not expected that hydogen will impact the immediate ailue it could incease the vulneability o the pipe to delayed ailue though the initiation o activation o cack like deects. This pape will pesent a methodology to pedict the pobability o inceased ailues and descibe a sotwae tool that has been developed to peom the calculations. Nomenclatue P m pimay stess P b hal cack length H dent depth D nominal pipe diamete σ nominal hoop stess t pipeline wall thickness H M bending moment Kth theshold stess intensity acto ange m atigue gowth paamete C atigue gowth paamete a cack depth 2c cack length W pipe section length K mode stess intensity acto K atio o applied elastic K to K IC L atio o applied load to yield load K toughness o mateial ρ plastic coection acto IC K P pimay stess intensity acto K S L max pemitted limit o L seconday stess intensity acto σ e eeence stess σ Y yield stength o mateial σ U ultimate tensile stength o mateial P pobability o ailue K ailue equency
. INTRODUCTION Among all ailue modes o gas pipelines thid paty damage poses the geatest theat to the saety o the pipeline netwok because it accounts o moe than 5% o the total incidents accoding to a ecent suvey []. Some o the ailues do not happen immediately ate the pipeline is damaged but occu yeas ate the stike. The aim o the cuent study is to investigate i adding hydogen to the existing natual gas pipeline netwok will incease the isk o delayed ailue. Euope is investigating the adding o Hydogen to the gas tansmission system as it is believed to be one o the most pomising enegy caies in the 2 st centuy. Howeve, the tansition om pue natual gas pipelines to pipelines caying natual gas/hydogen mixtues entails exta isk to the integity o the pipeline as hydogen has been shown to educe the toughness o the pipeline and to acceleate cack gowth. The NATURALHY poject has concluded that hydogen molecules within the pipeline pesents no hazad to the pipeline steel unless the inne thin oxide laye on the pipeline steel is distubed and a cack-like deect exists which can cause disassociation o the hydogen molecules and allow hydogen to pemeate into the pipe body, theeby educing the mateial toughness. Cack like deects may be pesent on the inne wall o the pipe as a esult o manuactuing deects, o as a esult o some kind o thid paty damage. In the case o thid paty damage a cack may be initiated on the inne wall o the pipe due to the incident. Altenatively the damage may activate an existing domant cack (i.e. a cack which is not gowing o gowing so slowly that it is not signiicant). The second case would be vey unlikely but it is theoetically possible. The ollowing sections popose a methodology to pedict the impact o hydogen on the delayed ailue pobability associated with the dent whee a cack has been initiated/activated due to thid paty damage. 2. ESTIMATING THE MAXIMUM PERMISSIBLE CRACK DEPTH 2. Set-up o the dent-cack model The dent in a pipeline is a complex stuctual poblem as the stess distibution at the oot o the dent is heavily inluenced by the depth o the dent, the included angle o the dent, the length o the dent and the location o the dent. A detailed desciption o the elastic analysis o a dent on a pipe can be ound in eeences [2, 3]. To solve this complex poblem a simpliied model has been adopted. The loading applied to the cack embedded at the bottom o the dent can be descibed by ollowing elationships [4]. m H (.8 / ) P = σ H D () M =.85σ H th (2) P M t 2 = 6 / (3) b The only dieence between a suace cack in an undamaged pipe and in a dented pipeline is that thee ae additional stesses caused by bending aound the dent but the membane stess is lowe. I the bending stess is not elaxed when the dent is constained by a ock, e-ounding is not possible. The dent depth ollows the Weibull distibution with α o.69 and β o 6.22 mm accoding to UKOPA ault database [5]. Fo the cuent study, the dent depth is chosen to be 6mm as this epesents a typical dent depth in the pipeline (see Fig. ). 2
Pobability Density.9.8.7.6.5.4.3.2... Dent depth Distibution 2. 3. 4. 5. 6. 7. 8. 9... 2. 3. 4. 5. Dent depth (mm) Pobability density Cumulative distibution..9.8.7.6.5.4.3.2. Cumulative Pobability Figue. Dent depth distibution The tansmission gas pipeline consideed is made o X7 o wall thickness 2.9 mm and diamete 9 mm. This epesents a typical moden pipeline made ate 975. The geomety o the pipe, the dent and the cack ae shown in Fig 2. The opeating pessue o the pipeline is assumed to be 6 ba. The mateial popeties o X7 pipeline in dieent natual gas/hydogen mixtues ae summaised in Table. The data epesents a typical X7 pipeline. It has to be noted that the pipelines manuactued beoe 975 might have signiicantly lowe toughness values and moe vulneable to cack popagation. In this case the igues shown in the table ae not applicable and additional expeimental esults ae equied to descibe the mateial popeties. axial intenal cack Figue 2. Geomety o the pipeline with a cack embedded in a dent 3
Table. Mateial popeties o X7 (intenal pessue = 6 ba) Gas contents Yield stength (MPa) Ultimate tensile stength (MPa) Toughness o pipelines (MPa m) % NG 482.6 565 2 5% NG+5% H2 482.6 565 5 % H2 482.6 565 The poblem o a cack embedded in the dent can be solved using a lat plate model as shown in Fig. 3. W is the pipe section which is long enough to accommodate the cack deect. The cack is axially oiented so that the hoop stess applied to the cack suace is the maximum. The pessue swing atio used hee is.35, which is a aily lage numbe o an aveage gas pipeline netwok. Howeve, this numbe is used duing the pipeline design pocess, so it is also used o atigue calculation. Hee the ailue o the pipeline ees to geneal ailue, i.e. eithe leakage o uptue o the pipeline. 2c t a W Figue 3. Geomety o the Semi-elliptical suace cack model 2.2 Calculation o the maximum pemissible cack depth Fo the cuent study BS79 level 2 Failue Assessment Diagam (FAD) is adopted. As can be seen om Fig 4 i a cack is initiated in the sae egion it does not lead to ailue immediately, but once it gows outside the saety bounday, which is depicted on the gaph, the pipeline becomes unstable and leakage/uptue can occu. In this section a seies o sensitivity tests ae peomed, o which the maximum cack depth that will not cause immediate ailue is calculated. K Bittle actue dominated egion..8 K =(L ).6.4.2 Sae egion Stat End(ail) Plastic collapse cut-o.2.4.6.8..2.4 L Figue 4. BS79 Failue Assessment Diagam (FAD) 4
Accoding to BS79 the ailue bounday is deined as ollows: { } = + o L > L max K = 2 6 o L L max K (.4 L ).3.7 exp(.65 L ) (4) The ailue citeion includes both bittle actue and plastic collapse. K measues the poximity to bittle actue and L epesents the likelihood o plastic collapse. Fo BS79 level 2A FAD, they ae given by: K K σ e L = σ Y + K P S = + KIC ρ (5) ρ is a paamete that takes plastic inteaction between pimay and seconday stess into consideation. Fo mateials, especially low stength mateials that exhibit a yield discontinuity (Lüdes plateau) L is esticted to. [2]. Othewise it is calculated though: L max σ Y + σ u = (6) 2σ Y The above equation is used o the analysis o the cuent study as the Lüdes plateau is not visible o high stength steel such as X7. The esults o the sensitivity tests ae displayed in Fig. 5. It can be seen that o moden pipelines with a elatively high toughness hydogen does not have much impact on the integity o the pipeline when the cacks have just omed in the dent. Fo these pipelines the ailue mechanism is plastic collapse athe than bittle actue. It is also discoveed that the citical depth (i.e. the cack depth which will lead to immediate ailue) declines as cack length inceases and the citical cack depth levels o as the cack length eaches about 3 mm. Howeve, o old pipelines both bittle actue and plastic collapse could be the ailue mechanism. Although peliminay sensitivity tests suggest moden pipelines with a high toughness ae less susceptible to hydogen it will be shown that once the cack stats gowing in the dent the associated isks will incease as time elapses. Replacing pat o natual gas with hydogen does make pipelines vulneable and the seveity o the damage depends on the content o hydogen. 5
6 Citical cack depth vs. cack length 5 cack depth (mm) 4 3 2 3 5 7 9 3 5 7 9 2 23 25 27 3 cack length (mm) X7 pipeline in NG, 5%H2, % H2 Figue 5. Citical cack depth vs. cack length o an X7 pipeline (6 ba, dent depth=6mm) 3. ESTIMATION OF THE DELAYED FAILURE PROBABILITY OF DENTED PIPELINES 3. Cack gowth in the dent Among all thid paty damage elated ailues, 9% o them ae immediate ailues and the est ae delayed ailues [7]. Delayed ailues ee to those that do not occu when the pipeline is damaged but occu yeas ate it was damaged. Fom the sensitivity test esults as seen in the pevious section it can be concluded that any cacks pesent in the pipeline must be smalle than those shown in Fig. 5 as any lage cacks would have led to immediate ailue. Since the cumulative pobability o ailue ove a given timeame is equied, cack popagation due to cyclic loading must be included to estimate how the gowth o cacks will impact the numbe o ailues. The Pais law with a theshold Kth is selected to calculate the cack length and depth with egad to the coesponding numbe o cycles. da o K < Kth = m dn C K o K K th (6) The actual calculation is peomed by estimating the amount o cack gowth duing a loading cycle and the detailed pocedues can be ound in authos' anothe pape [8]. Accoding to Pais equation: a = C( K ) m n+ n a a = a + a (7) whee a n coesponds to the cack depth ate n load cycles, C and m ae atigue gowth paametes. 3.2 Pocedues o estimating the ailue pobability The ailue pobability can be calculated though Monte-Calo simulation. By geneating a lage 6
numbe N o independent epetitions, the pobability o ailue can theeoe be estimated as the quotient o the ailue counts to the numbe o simulations peomed, which is given as ollows: P = N N whee N is the numbe o ailues ecoded. The whole pocess is shown in Fig. 6. The low chat is simpliied as the eal simulation involves atigue calculation, inspection and epai pogam. Relevant inomation can be ound in eeence [8]. Because cack popagation can lead to actue o leakage o the pipeline ate a cetain peiod o time, P is a unction o load cycle n. P = P( n) (8) Stat End analysis No Geneate andom vaiables? Yes Fail? Yes N=N+ No No ailue Figue 6. Flow chat o Monte-Calo simulation P denotes the cumulative pobability which monotonically inceases with load cycles. This value is only associated with the dent which has a cack at the bottom o it and the cack is vey small which does not lead to ailue at the beginning. As we do not know how many cacks ae associated with dents, we theeoe have assumed that all dents unde consideation contain a suace cack. I we assume that all dents have a depth o 6mm. The ailue equency o a pipeline aising om such deects in yea 'u' can be appoximated by: ( ) ( + ) ( ) P ( i) P i P i k u = k u (9) i= whee k is the incident ate pe km pe yea (kmy) and u o yea. Accoding to UKOPA database, the dent and gouge elated incident ate is 8.49-4 pe kmy, but the dent elated incident ate is aound.68-4 pe kmy. Equation 9 implies that the deects which do not lead to ailue in pevious yeas may contibute to the ailue equency in the utue. 7
3.3 Distibutions o deect dimensions The biggest diiculty in pedicting the ailue pobability aises om the act that inomation on the distibution o cacks imbedded in the dent is not available because cuently inspection tools ae not used to detect cacks in gas pipelines. Also thee ae many actos which inluence the pobability o ailue based on the assumed incident scenaio. Fo example i the POF o a single dent is consideed the analysis must conside the seveity o the dent, the pobability that a cack is initiated on the inside wall o the dented pipe, the pobability that thee is a pe existing cack on the inside wall o the pipe nea the dent and the equency o such a dent causing incident occuing. In addition thee is also the scenaio egading the intoduction o hydogen gas mixtues into the pipeline. Fo example has the pipeline been inspected and epaied? Was the dent pe-existing? O did it occu ate the intoduction o the hydogen gas mixtue etc. While calculation can be peomed o the dieent scenaios the esults will be vey much dependent upon the deect distibution assumed in the pipeline. The ollowing section attempts to simulate the POF o the case whee the pipeline has been peviously opeating saely with natual gas and now is opeating with a hydogen gas mixtue and the dent occus and a cack is pesent on the inside wall eithe because it is initiated o as a consequence o the dent was pe existing in the pipe at that location. Based on the sensitivity test esults shown in section 2, it is possible to estimate the distibution o the length and depth o the cack imbedded in a dent which is 6mm deep. It is poposed that the ollowing log-nomal distibutions o initial deect depth and length in dents ae used to calculate the ailue pobability. The estimation is based on the indings o the sensitivity tests which show shote cacks geneally have lage citical depth up to a cetain cack length whee the citical depth stabilises. The log-nomal cumulative cuves, which give the pobability o the occuence o a speciied size o damage o smalle, ae plotted o cack depth and cack length o a long but shallow cack in Fig. 7 and Fig. 8. Fig. 9 and Fig. ae o shot but deep cacks. Examining the cuves it can be seen om Fig. 7 and Fig. 8 that they have been designed to have a vey small numbe o deects with a depth geate than.5 mm which was the citical cack depth o a long cack in a dent in the pipeline. While thee ae still a small numbe o deects above the citical cack depth they ae not signiicant as they only lead to a small incease in the POF in the ealy yeas. Vaiable Table 2. Distibutions o cack depth and length Type o distibution Mean (mm) Standad deviation (mm) Depth (long cacks) Log-nomal.8.2 Length (long cacks) Log-nomal 5 5 Depth (shot cacks) Log-nomal 2.3 Length (shot cacks) Log-nomal 5 2 8
C ack depth distibution Pobability Density 2.5 2.5.5...2.3.4.5.6.7 Depth (mm).8.9...2.3.4.5.6 Figue 7. Cack depth distibution (long cacks)..9.8.7.6.5.4.3.2. Cumulative Pobability Pobability Density..9.8.7.6.5.4.3.2.. 3. 6. Cack length Distibution 9. 2. 5. 8. 2. 24. Length (mm) 27. 3. 33. 36. 39. 42. 45. 48. Figue 8. Cack length distibution (long cacks)..9.8.7.6.5.4.3.2. Cumulative Pobability Pobability Density.6.4.2.8.6.4.2..22.43 Log-Nomal Distibution.64.85.6.27.48.69 Depth (mm).9 2. 2.32 2.53 2.74 2.95 3.6 3.37 Figue 9. Cack depth distibution (shot cacks)..9.8.7.6.5.4.3.2. Cumulative Pobability 9
Pobability Density.3.25.2.5..5..2 2.4 Log-Nomal Distibution 3.6 4.8 6. 7.2 8.4 9.6.8 2. Length (mm) 3.2 4.4 5.6 6.8 8. 9.2 Figue. Cack length distibution (shot cacks)..9.8.7.6.5.4.3.2. Cumulative Pobability 3.4 Compaing the delayed ailue pobability o natual gas pipelines and pipelines caying natual gas/ hydogen mixtues Based on the assumed deect distibutions and the dent size the simulation esults o a single dent ae shown in Table 3 and Table 4. It is obvious that the delayed pobability o ailue o pipeline caying hydogen is much highe than that o the existing natual gas pipelines. It has to be noted that as the deect size distibutions poposed hee ae o a speciic X7 pipeline and the ailue pobabilities will change i a dieent pipeline is analysed. Howeve, it can be expected that the geneal tends indicated by these esults will emain the same o othe pipelines. Table 3. Cumulative POF values o pipelines caying dieent content o Hydogen (a long but shallow cack initiated at the bottom o a dent) Yea % NG 5% NG- 5% H 2 % H 2.54E-8.54E-8 9.92E-6.75E-8 3.7E-8 2.77E-4 2 2.32E-8 9.E-8 2.77E-3 3 2.57E-8 2.6E-7.47E-2 4 3.47E-8 4.86E-7 5.94E-2 5 4.5E-8 9.48E-7.43E- 6 5.85E-8.9E-6 2.47E- 7 7.3E-8 4.96E-6 3.67E- 8 8.7E-8.2E-5 4.78E- 9 9.58E-8 2.7E-5 5.55E-.26E-7 3.64E-5 6.22E- Table 4. Cumulative POF values o pipelines caying dieent content o Hydogen (a shot but deep cack initiated at the bottom o a dent) Yea % NG 5% NG- 5% H2 % H2.2E-8.2E-8 3.39E-5.36E-8.47E-8 2.44E-4 2.39E-8 2.48E-8.E-3
3.4E-8 6.96E-8 2.38E-3 4.45E-8.4E-7 3.7E-3 5.48E-8 2.3E-7 4.82E-3 6.76E-8 3.3E-7 5.7E-3 7 2.2E-8 7.79E-7 6.47E-3 8 2.46E-8.46E-6 7.69E-3 9 2.6E-8 2.63E-6 9.5E-3 3.58E-8 5.73E-6.5E-2 By compaing the esults in Table 3 and Table 4 it can also be discoveed that o vey shot cacks in the pipeline the associated isks ae lowe than those o long deects even i the shot cacks ae deepe. Since the shot cacks ae less dangeous than the long cacks and we do not know the eal distibution o cack sizes in the dent, the esults shown in Table 3 ae used to compute the ailue equency. Howeve, adopting this appoach will esult in consevative esults as the small cacks ae implicitly teated as long cacks and theeby having the same level o contibution to the total ailue equency as the longe cacks do. The ailue equency can be obtained by solving equation 9. Fig. shows coesponding ailue equency cuves o the same X7 pipeline. It has to be noted that it was assumed the dent depth is 6mm, but in act this is not accuate as the dent depth does ollow cetain distibution. Howeve, though this simpliication the simulation esults have aleady shown that adding hydogen will incease the delayed ailue equency but it is diicult to povide an accuate estimate without moe accuate data on the type o cacks initiated. Obviously the esults show that educing the amount o hydogen added to the natual gas pipeline system educes the isk but i inspection tools ae available which can detect cack like deects the taget isk level could be achieved though a caeully designed integity management pogamme..e+ Failue Fequency vs. Hydogen Content Failue equency (km.yea).e-.e-2.e-3.e-4.e-5.e-6.e-7.e-8.e-9 2 3 4 5 6 7 8 9 Yea % NG 5% NG - 5% H2 % H2 Figue. Failue equency o a cacked dent in pipelines caying NG and NG/H 2 mixtues 4. CONCLUSIONS A methodology has been pesented to pedict the pobability o ailue o a pipeline subjected to thid paty damage is pesented. It is based on the concept o the damage eithe initiating a cack like deect
o activating a domant cack like deect on the intenal wall o the pipeline thus leading to delayed ailue. The pocedue is based on a pobabilistic actue mechanics appoach. The ollowing conclusions have been dawn with egad to the detemination o the ailue pobability o tansmission pipelines conveying natual gas/hydogen mixtues. ) Fom the initial calculations the esults suggest that the existence o hydogen in a pipeline signiicantly inceases the delayed ailue pobability. The eason o this is that the hydogen penetates the paent mateial though the cack suace and changes the mateial popeties, which educes the ability o the pipeline to esist cack gowth. 2) An impotant element o the POF calculation is the poposed distibution o cack sizes located nea the dent. Since no data is available on such distibution, some assumptions based on a seies o sensitivity tests ae made. Futhe wok is equied to obtain data to suppot uthe pedictions. 3) Othe paametes such has pessue dop atio, numbe o cycles and the hitting ate ae eithe taken om liteatue o design guidelines. Hence, when intepeting the ailue equency esults one must ealise that the ailue equency may vay signiicantly om pipeline to pipeline. 4) In this epot, only lognomal unction is adopted to epesent the actual distibution o the deects in pipeline. Howeve, thee ae othe unctions such as Weibull and exponential distibutions, which can also be used to it the data. In addition, the cack length and depth should be examined vey caeully since the esults ae vey sensitive to these inputs. ACKNOWLEDGEMENTS The authos would like to thank the NATURALHY poject patnes o poviding data and the Euopean Union o kindly poviding the inancial suppot o this study. REFERENCES. Gas Pipeline Incidents, 6th EGIG epot 97-24, EGIG 5.R.2. 25. 2. Seng, O.L., C.Y. Wing, and G. Seet, The Elastic Analysis o a Dent on Pessuised Pipe. Intenational Jounal o Pessue Vessels and Piping, 989. 38: p. 369-383. 3. Oynyak, I.V. and Y.S. Yakovleva, Application o the cack compliance method to long axial cacks in pipes with allowance o geometical nonlineaity and shape impeections (dents). Engineeing Factue Mechanics, 28. 75: p. 452-465. 4. Bai, Y. and R. Songb, Factue assessment o dented pipes with cacks and eliability-based calibation o saety acto. Intenational Jounal o Pessue Vessels and Piping, 997. 74: p. 22-229. 5. Lyons, C., et al. A methodology o the pediction o pipeline ailue equency due to extenal inteeence. in Intenational Pipeline Coneence. 28. 6. Guide to methods o assessing the acceptability o laws in metallic stuctues (BS79). 25: Bitish Standads Institution. 7. Kiene, J., Pipeline Incidents Caused by Mechanical Damage, in Mechanical damage technical wokshop. 26. 8. Zhang, L., Adey, R, Pedicting the Pobability o Failue o Gas Pipelines Including Inspection and Repai Pocedues. in Intenational Coneence on Hydogen Saety 27. 27 2