Transgranular Stress Corrosion Cracking of High-Pressure ... Fracture/Final_student... · Although intergranular stress corrosion cracking (IGSCC) of high-pressure gas pipelines has

CORROSION ENGINEERING

394 CORROSION–MAY 1994

✫ Submitted for publication August 1993.* Herschel Bldg., University of Newcastle upon Tyne, Newcastle upon

Tyne, NE1 7RU, England.** Amoco Corp., Amoco Research Center, Naperville, IL, 60566.

*** TransCanada Pipeline Co. Ltd., Calgary, Alta., Canada.

Transgranular Stress Corrosion Crackingof High-Pressure Pipelines in Contactwith Solutions of Near Neutral pH✫

R.N. Parkins,* W.K. Blanchard Jr.,** and B.S. Delanty***

INTRODUCTION

The majority of instances of stress corrosion cracking(SCC) in high-pressure gas pipelines has beenassociated with the propagation of intergranular cracks(IGSCC), but several instances of transgranularcracking (TGSCC) have been documented. TGSCChas involved service and hydrotest failures, andTGSCC frequently has been found in the region ofdents in pipelines.

In those respects, TGSCC and IGSCC areidentical, but the difference in crack morphologysuggests that the mechanisms of growth, and thereforethe conditions under which the two forms occur, aredifferent. Since the steels involved and the operatingpressures do not differ essentially for pipelinesdisplaying the two different modes, the implication isthat the different mechanisms result from variations inenvironmental conditions. IGSCC is associated withrelatively concentrated carbonate-bicarbonatesolutions with pH values in the region of 9.5. TGSCCoccurs in the presence of relatively dilute solutions withpH values ≈ 6.5. These differences in pH sometimeshave prompted investigators to refer to the two formsas “high-pH” and “low-pH” cracking, although “low” ishardly applicable to solutions of pH ≈ 6.5

Other differences between the two modes ofcracking relate to differences in environmentalconditions in the broadest sense.1 Thus, IGSCCdisplays a temperature sensitivity not shown byTGSCC in the field and in laboratory tests. Similarly,

0010-9312/94/000091/$5.00+$0.50/0© 1994, NACE International

ABSTRACT

Although intergranular stress corrosion cracking (IGSCC) ofhigh-pressure gas pipelines has been known for more than20 years, a transgranular form (TGSCC) was detected morerecently. Instances of TGSCC have been associated withdilute solutions with pH values in the region of 6.5 because ofthe presence of carbon dioxide (CO2). Such pH valuesindicate relatively little, if any, cathodic current reaches thepipe surface, since hydroxyl ions would be generated and pHwould increase to values in the region of 10 if current didreach the pipe surface. Slow strain rate testing (SSRT) ofpipeline steel specimens in dilute solutions of pH in the regionof 6.5 suggested dissolution and hydrogen (H) ingress intothe steel are involved in the crack growth mechanism. Theinitiation of TGSCC in specimens subjected to cyclic loadingand maximum stresses approximating those of an operatingline was facilitated by pitting. The geometry of the pits allowedthe localized generation of solutions of lower pH than that ofthe bulk solution outside the pits, thereby facilitatingdissolution and H discharge.

KEY WORDS: carbon dioxide, dilute solutions, hydrogeningress, intergranular stress corrosion cracking, pH, pipelines,pitting, slow strain rate testing, steel, stress corrosioncracking, transgranular stress corrosion cracking


395CORROSION–Vol. 50, No. 5

longitudinal weld and usually are located on only oneside of the weld. The weld reinforcement probablycauses the tape to bridge the immediately adjacentpipe. The void on one side of the weld is filled withadhesive, while the other side remains empty until theingress of ground water occurs. A similar effect canoccur where successive wraps of tape overlap andallow the ingress of liquid to produce a narrow strip ofshort cracks. A major crack may develop at the weldtoe, where stress concentration or structuralmodification as a result of welding may facilitate crackgrowth. However, low-pH cracking is not invariablyassociated with welds, joints having an asphalt coatingcracking elsewhere.

TG cracks at pipeline surfaces frequentlycoalesce. This issue has been studied in detail inrelation to IGSCC.3 Crack coalesence is manifest mostreadily at the outer surfaces of pipes or laboratoryspecimens. The nearest tips of adjacent cracks passone another before turning and growing toward oneanother to achieve coalescence. It is to be expectedthat the chances of a pair of cracks of given lengthinteracting in this way increases as their transversedistance of separation decreases. Measurements ofthe transverse separations of adjacent cracks and theirlengths have been made for several TGSCC coloniesin pipelines. Data are shown in Figure 2 as a plot of thetransverse separation against the mean length of theadjacent cracks. Those pairs of adjacent cracks thatshowed clear evidence of coalescence are separatedfrom those pairs that did not by a line that defines thecondition for coalescence as:

y < 0.14 (2a) (1)

effects of potential on the two forms of cracking aredifferent, as are the associated reactions and theirproducts. For example, TGSCC often is associatedwith the formation of relatively large amounts of whiteiron carbonate between the coating and pipe surface.In IGSCC, however, the very small amount of ironcarbonate that sometimes is present is incorporated inthe thin magnetite films that invariably form. Thesefilms are strongly adherent to the crack sides andeffectively prevent any lateral dissolution on the sides.As a result, IG cracks are narrow or fine. Conversely,the TG crack sides suffer significant lateral dissolution(Figure 1), with appreciable amounts of looselyadherent corrosion products forming in the crackenclaves. The original crack faces are destroyed, andthe TG paths can be established with certainty onlynear the crack tip, where there has been relatively littletime for dissolution of the crack sides. In contrast, thethin magnetite films that protect the IG crack sidesallow the crack morphology to be retained over thewhole crack length.

These differences between the two modes ofcracking indicate the reactions involved in their growthare markedly different, reflecting the differences in thesolutions found between pipe and coating in the vicinityof the two types of cracks. Reactions involved withIGSCC have been studied extensively, and themechanism is understood reasonably well.2 However,the conditions of TGSCC have not been defined aswell.

TGSCC in pipelines is characterized by very highdensities of cracks in localized regions. In IGSCC, orhigh-pH cracking, cracks are spaced more widely.Colonies of TG cracks often are formed near a

FIGURE 1. TGSCC in a section from an operating pipeline, showing initiation at a pit and increasing amountsof lateral corrosion on the crack sides as the crack mouth is approached.



where y is the transverse distance of separation and2a is the mean length of the adjacent cracks. Thesame expression as Equation (1) has been found tohold for IGSCC in pipelines and for other steelsexposed to solutions other than those to whichpipelines are subject, indicating crack coalesence isrelated to mechanics and is not dependent upon thematerial or environment involved.

TGSCC colonies vary in relation to size of thecolony and the maximum crack length within a colony.Together with the maximum estimated crack depthwithin a colony and the number of colonies per unitarea, these parameters have been used to calculate aSCC severity factor for tape-coated sections ofTransCanada Pipelines Co. Ltd. (TCPL) lines.4 ThatSCC severity factor has been used in conjunction withinformation relating to the environmental conditionsalong a line to identify sections where the risk ofcracking appears high. This approach also has beenused to identify variability in the severity of cracking indifferent sections when variability showed little relationto factors that influence high-pH cracking (i.e.,temperature). Thus, TGSCC has been detected as faras 67 km downstream of a compressor station.5 Inhigh-pH cracking, the incidence of cracking has beenshown to decrease markedly with increasing distancefrom a compressor station so that 90% of crackingoccurred within 20 km downstream of a station.6

FIELD DATA

Other organizations probably have experiencedTGSCC in pipelines, but TCPL has carried outextensive field investigations relating to this phenome-non.4-5 Service and hydrotest failures 15 years after

installation prompted a series of investigations toobtain data on the prevalence of cracking andassociated environmental circumstances. The X-65grade pipe studied had field-applied coatings of asphaltor polyethylene tape. The incidence of cracking wasappreciably greater with the polyethylene tape. About70% of 251 excavated samples of tape-wrapped pipe,usually ~ 50 m length, revealed SCC, with an averageof 11 crack colonies in each excavated sample inwhich cracks were discovered. By contrast, only ~ 14%of 189 excavated asphalt-coated sections showedcracks, with an average of 6.3 colonies of cracking inexcavated samples found to contain cracks. Theexcavation locations were chosen as those mostvulnerable to cracking.

At each site, the soil type and drainage patternwere determined. Location of each site was recordedwith reference to its position on a slope or otherfeatures of the landscape. Data indicated cracking wassignificantly more prevalent, by a factor of ~ 7.5, inclays and silts (lacustrine soils) than in sands andgravels (glaciofluvial materials). A similar propensity forclays to enhance the chances of IGSCC was observedwith a pipeline in Australia.7 In both instances, thetendency for such regions to drain poorly resulted inretention of the cracking environment and probablycontributed to the enhanced cracking. Similarly, thegreater density and cohesive characteristics oflacustrine soils likely enhanced the chances of coatingdisbondment due to soil stresses.4

Data from analysis of samples of liquid collectedfrom under coatings revealed the incidence of crackingcorrelated with pH, with cracking most marked whensolutions of pH ≈ 6.5 were found in the vicinity (Figure3). The pH values may have been lower than themeasured values in some cases because the liquidwas effervescent in the hypodermic syringe when liquidsamples were taken from between the coating andpipe. This suggested the presence of an evolving gas,most likely carbon dioxide (CO2), the egress of whichfrom the solution would have increased pH.

Figure 3 also shows low pH environments werenot invariably associated with cracking. Of the liquidswith pH values ≤ 7.5, only ~ 25% were associated withcracking.

Apart from pH, the TCPL solutions were fairlydilute ground waters. Such solutions prepared in thelaboratory have a pH in the vicinity of 8, but theaddition of CO2, which often is formed by decayingorganic matter in soils, readily reduces pH to 6.5 orother values depending upon the partial pressure ofthe CO2 in equilibrium with the solution. The retentionof such low-pH liquids in the vicinity of pipe is possibleonly if negligible cathodic current reaches the pipesurface, since hydroxyl ions would form and pH wouldincrease if sufficient current reached the surface.8

FIGURE 2. Spatial and dimensional characteristics of adjacentTG cracks produced in service, according to whether theyshowed evidence of coalescence.



LABORATORY TESTS WITH DILUTESOLUTIONS CONTAINING CO2

Experimental MethodsWhen TCPL first detected TGSCC, the effects of

factors such as environment composition, potential,and temperature were not known. It also was notknown whether TG cracks similar to those in operatingpipelines could be reproduced in laboratoryenvironments similar to those in the field. Slow strainrate testing (SSRT) was well suited to rapid surveys ofthe type needed initially. SSRT was conducted toencompass a range of environmental factors onsamples of the X-65 steel involved in one of the jointsin which TCPL experienced cracking.

The solutions used in SSRT were based uponanalyses of the liquids found in the field (Table 1). Itwas probable that appreciable amounts of free CO2were present in the liquids because of the relativelylarge amounts of iron carbonate between coating andpipe in some regions, the most likely source of whichwas by reaction between solvated iron and CO2.Consequently, SSRT was conducted in the solutionswith and without CO2. Tests were conducted at the freecorrosion potential as well as at various controlledpotentials and at temperatures from 5°C to 45°C.Below-ambient temperatures were achieved bypumping a cooled liquid from a refrigerated tankthrough plastic tubing wrapped around the outside ofthe cell. Above-ambient temperatures involved heatingthe cell with an external electrical resistance winding.Temperatures were controlled to within 1°C of thevalues quoted.

SSRT specimens were cut from the pipe in thelongitudinal and transverse directions, although thestress corrosion responses showed no markeddifferences with change in orientation. Tests wereconducted using cylindrical, waisted specimens12.5 mm gauge length and 0.25 mm in diameter.Specimens were finished with 5/0 emery paper anddegreased in acetone before testing. The specimenswere contained in glass cells that were closed withrubber stoppers through which the ends of thespecimens protruded for gripping. The cells containedfacilities for bubbling gases through the solution and aprobe to an external saturated calomel electrode(SCE). All potentials were quoted with respect to theSCE.

A platinum counterelectrode within the cell wasused for controlled potential tests, and a bufferamplifier was used for all potential measurementsbecause of the relatively low conductivity of the dilutesolutions. The strain rates applied were ≈ 2 x 10–6 s–1Results were considered in terms of the reduction inarea (RA%) to fracture, although all specimens were

examined by metallography after completion of thetests.

SSRT is a severe form of testing in that, asperformed most frequently, specimens are takenmonotonically to total failure. High stresses and strainsusually are involved. Consequently, after SSRT wasused to outline the environmental conditions likely tolead to cracking in the low-pH solutions, cyclic loadingtests were performed over a range of stressingconditions down to those representative of anoperating line. High maximum stresses and low Rvalues (ratios of minimum to maximum stress) wereinvolved in some of the tests because one objectivewas to determine whether relatively large cracks withmorphologies identical to those in service cracks couldbe produced in reasonably short test times. SSRTproduced rather short cracks because of the shortexposure times involved. Moreover, in view of thescatter in field and SSRT data, it was anticipated thatsimilar scatter would occur in cyclic loading tests, withscatter increasing as the maximum stress decreased.By conducting cyclic loading tests over various rangesof maximum stresses and R values, viewing the data

TABLE 1Solution Compositions (g/L) Representativeof Solutions Found in Field Investigations

Substance NS1 NS2 NS3 NS4

KCl 0.149 0.142 0.037 0.122NaHCO3 0.504 1.031 0.559 0.483CaCl2·2H2O 0.159 0.073 0.008 0.181MgSO4·7H2O 0.106 0.254 0.089 0.131

Designation

FIGURE 3. Numbers of liquid samples taken from belowasphalt and tape coatings with pH values shown with theincidence of SCC in those locations on TCPL lines.5



as a whole, and using extrapolation from the moreadverse stressing conditions, it was likely that datafor more realistic stressing conditions would beenhanced.

In the light of the service data and data obtainedfrom SSRT, all of the cyclic loading tests wereconducted at room temperature (RT) in the solutiondesignated NS4 (Table 1). A mixture of CO2 andnitrogen (N2) was bubbled through the solution toachieve pH ≈ 6.5. Test cells were essentially the sameas those used for SSRT. The majority of tests wasconducted at open-circuit potential (OCP), which variedfrom ~ –0.60 VSCE to –0.72 VSCE but was mostly in therange from –0.68 VSCE to –0.72 VSCE. A small number oftests were carried out at the controlled potential of–0.7 VSCE, but results were indistinguishable from thoseperformed under OCP. Tensile specimens 12.5 mmlong, 5 mm wide, and 2.5 mm thick gauge length werecut in the longitudinal direction from pipe. The outersurface of the pipe was left intact on one surface ofeach test specimen, and the other surfaces wereground and polished. Surface condition was importantin that cracks rarely generated at polished surfaces.The original pipe surface, with the coating removed bywater blasting, proved much more susceptible to crackinitiation. Tests were conducted at 7.4 x 10–2 Hz and at3.2 x 10–4 Hz to determine whether frequency, asopposed to test time, influenced cracking. Results didnot differ significantly. Specimens were subjected tocyclic loading in machines that applied weights by

means of a lever system, with the cyclic componentderived from the movement of a cam operating througha spring of appropriate characteristics. A load cell inthe system allowed the maximum and minimum loadsto be determined. Test times ranged from ~ 1 week to3 months. At the conclusion of the tests, surfaces ofthe specimens were examined by scanning electronmicroscopy (SEM) and by optical microscopy (OM)after appropriate preparation. OM was used todetermine the depth of the deepest crack. The averagecrack velocity was calculated by dividing the depth bythe test time.

SSRT RESULTS

Effects of SolutionComposition and Potential

Results from tests involving solutions NS1 throughNS4 (Table 1) to which no CO2 was added showed nomajor differences, although some interesting effectswere observed, especially in terms of RA% as afunction of potential.

Figure 4 shows plots of RA% against potential forthe NS1, NS3, and NS4 solutions. The NS2 solutionwas used in only a few tests but with no significantdifferences. Although there was considerable scatter inthe data, each solution displayed a similar trend in thatmarked reductions in RA% were apparent at potentialsin the vicinity of –0.7 VSCE, at least in several tests,followed by an increase in ductility to fracture and a

FIGURE 4. Reductions in area from SSRT on X-65 steel at various potentials in three of the solutions inTable 1 and without CO2 added. Horizontal bars refer to tests at the free corrosion potential, with the potentialthat persisted through most of the test indicated by the symbol at one end of the line.



CO2 and N2 was bubbled to produce pH ≈ 6.5. Results(Figure 6) allowed comparison with trends observed forsolutions without (pH ≈ 8.2) and saturated with CO2.Results for tests in solutions with pH of ~ 6.4 weresimilar to those for solutions of pH ≈ 8.2 in that therewas a significant tendency for the ductility to recover atintermediate potentials, although the data as a wholetended toward higher potentials in tests using thesolution of lower pH.

Scatter in RA% values in Figures 4 through 6might have detracted from assessing the effects ofaddition of CO2 to the solutions. A more appropriateway to consider the data was by using Weibull plots ofRA% over restricted potential ranges. Figure 7 shows

further decrease as potential was loweredprogressively. The intermediate potentials, at whichlow ductilities were observed, were a little above thepitting potential (Epit), as measured in potentiodynamicpolarization experiments, with RA% increasing fromthe lowest values as the potential was raised above Epitand dissolution became more widespread. Atpotentials just below Epit, the steel behaved in a ductilemanner, with RA% returning to values approaching70% at potentials of ~ –0.75 VSCE. However, furtherreduction in potential decreased ductility, with RA% inthe region of 35% at –0.9 VSCE. This further reductionwas most probably a result of the ingress of hydrogen(H) into the steel.

Data for the longitudinal and transversespecimens in Figure 4 did not appear essentiallydifferent, although there were some indications ofdifferences in the tendency for cracks to develop.Relatively few longitudinal specimens displayed cracksother than in the necked region, even when markedreductions in ductility were observed. However,transverse specimens developed cracks over thewhole of the gauge length, although only at potentialsof ≤ –0.8 VSCE with the NS1 and NS4 solutions, but at≤ –0.65 VSCE in the NS3 solution. The significance ofthese differences in relation to the different solutionswas not known, and they were surprising in that NS3was the most dilute of the solutions. Nevertheless, thetransverse specimens tended to develop cracks morereadily than the longitudinal specimens despite nomarked influence of such apparent in the RA% data.

Effects of CO2 AdditionsThe effects of saturating the solutions with CO2

were dramatic. RA% values obtained during SSRTwere reduced markedly even at potentials (≈ –0.8 VSCEand –0.5 VSCE) where Figure 4 indicated increasedductilities.Cracks also initiated over most of the gaugelength of the specimens in all solutions and at mostpotentials, and average crack velocities were generallyhigher than in solutions to which no CO2 was added.Figure 5 shows a plot of results for tests at RT in thevarious solutions saturated with CO2 and involvingspecimens cut longitudinally and transversely frompipe material. While there again was considerablescatter, the low RA% values shown over a wide rangeof potentials indicated influence of the environment,since several SSRT in air all gave RA% values in therange of 70 to 75. As in tests involving solutionswithout CO2 additions, there did not appear to be anysignificant difference in the behaviors of specimens cutin different orientations.

Solutions in Figure 5 had pH values ≈ 5.8, whichwas lower than in the field data (Figure 3) for the mostprevalent cracking conditions. Consequently, SSRTwas conducted in solutions through which a mixture of

FIGURE 5. RA% from SSRT on X-65 steel in the solutions inTable 1 but saturated with CO2 throughout tests conducted atvarious potentials. Horizontal bars refer to tests at the freecorrosion potential.

FIGURE 6. RA% to fracture in SSRT at various potentials onX-65 steel in dilute solutions to which different amounts of CO2were added to produce different pH values.



Weibull plots9 for tests in the solutions with and withoutCO2 saturation and for potentials in ranges including –0.65 VSCE to –0.7 VSCE, which approximated OCP. Theeffect of CO2 in promoting lower ductilities at a givenprobability was clear for all ranges, even though aboveand below OCP, values were likely to be of much lesssignificance to low-pH cracking in the field. Compari-son of the plots in Figure 7 showed ductility wasreduced significantly by the addition of CO2 to thesolutions and by decreasing the potential. Both trendswere significant indicators of the cracking mechanism.

Effects of TemperatureThe SSRT data above related to tests at RT, but

pipelines can operate above and below typical RT.Consequently, SSRT was conducted at temperaturesfrom 5°C to 45°C. All of the specimens in this phase ofthe work were cut in the transverse direction. NS1,NS3, and NS4 solutions were involved, and differentpH values were achieved by bubbling CO2 at differentpartial pressures through the solutions.

Figure 8 shows results from tests at varioustemperatures conducted in solutions of pH ≈ 6.4.Figure 9 shows the equivalent plot for solutions ofpH ≈ 5.8. The lack of systematic trends in the data withtemperature concurred with the field findings in whichthe incidence of cracking was unrelated to distancefrom compressor stations and, hence, to temperature.

FractographyIrrespective of the exposure conditions, where

relatively low ductilities were achieved in SSRT, thefracture surface invariably displayed areas of quasi-cleavage (Figure 10) such as is associated often with

hydrogen-induced cracking. A further indication of theentry of H into the steel was the manifestation ofsecondary cracks that were not connected to the outersurface of the specimen and usually were nucleated onbands of pearlitic material (Figure 11). Nearer to thefracture surface, these cracks yawned and linked,sometimes to cracks at the outer surface, but they alsowere detectable in regions where lesser deformationoccurred and where the yawning apparent in Figure 11was less marked. In specimens that fractured with highRA% values, the fracture surface invariably displayeddimples indicative of fracture propagation by microvoidcoalescence, patches of which also were apparent onfracture surfaces that displayed quasi-cleavage,especially in the case of intermediate RA% values.Cracks invariably followed TG paths, whether viewedby SEM of the fracture surface or by OM of polishedand etched sections. Cracks apparent at the outersurface of the specimens contained corrosionproducts, except when the test potential was in theregion below –0.8 VSCE.

CYCLIC LOADING TEST RESULTS

Crack VelocitiesMost of the cyclic loading tests involved maximum

stresses of 345 MPa (50 ksi), 414 MPa (60 ksi), or483 MPa (70 ksi), with a few at 552 MPa (80 ksi) orabove. The crack velocity data is plotted in Weibullform in Figure 12. As with some plots of Figure 7, theraw data did not always fall about a straight line asrequired by the Weibull function. This usually wasbecause there was an incubation period before anydetectable change in the measured parameter,

FIGURE 7. Weibull plots of RA% data from SSRT in dilute solutions, with and without saturation by CO2, atpotentials in the ranges shown.



causing the line to be convex toward its lower end. Inthe present work, the sharply undulating nature of theoriginal surface of the pipe material led to the arbitrarydefinition of a crack as a feature that was sharp at itstip and had a depth of 4 to 5 times the width, with aminimum depth of 0.005 mm that was measuredreadily at a magnification of 600x. The effect ofintroducing those lower boundary conditions resulted insome curvature of the line in Weibull plots. This wascorrected for in the usual way by introducing the so-called location parameter (shown on each plot inFigure 12). Very few data were obtained for stressesof 552 MPa or above. Those obtained are shown inFigure 12(d). The regression line for the data wasdominated by the two points at the extremes of thecrack velocity range.

Clearly, in view of the few data for maximumstresses of 552 MPa or above, the result shown inFigure 12(d) was the least reliable. Even so, themaximum stress increased as the slope of the Weibullline decreased, implying that the crack velocity for agiven probability increased with increasing stress,which was expected. Figure 13, derived from Figure12, shows the 70%, 50%, 30%, 10%, and 5% chancesfor the average crack velocity to reach the levelsshown for the various maximum stresses. Thisapproach reflected the scatter, or element of chance,involved with this system, a fact apparent in laboratorydata and in service (Figure 3). As Figure 13 indicates,two tests proceeded to total failure, but in two testsinvolving maximum stresses of 345 MPa and R valuesof 0.7 and 0.8 and in one test with a maximum stressof 620 MPa and a R value of 0.67, no cracks were

detected in exposure periods lasting 1,199 h, 1,655 h,and 634 h, respectively. Moreover, the position of theabscissa in Figure 13 corresponded to the maximumallowable operating stress for a line fabricated in X-65steel. The extrapolated lines showed crack velocitymight vary by over 1 order of magnitude for thatmaximum stress, depending upon chance.

Multiple crack initiation in cyclic loading testsusually is indicative of corrosion playing an active rolein cracking. Nevertheless, the topography of theoriginal surface of the pipe was such that queries wereraised as to whether multiple cracks could initiateunder cyclic loading conditions without corrosion.Consequently, a cyclic loading test was conducted in

FIGURE 9. RA% from SSRT on X-65 steel in dilute solutions,through which CO2 was bubbled to achieve pH ≈ 5.8, conductedat various potentials and temperatures.

FIGURE 10. SEM micrograph of part of the fracture surface ofSSRT specimen tested at –0.5 VSCE in NS1 solution saturatedwith CO2.

FIGURE 8. RA% from SSRT on X-65 steel in dilute solutions,through which a CO2-N2 mixture was bubbled to achievepH ≈ 6.4, conducted at various potentials and temperatures.



loaded specimens following total failure. Figure 17shows part of one of those fracture surfaces.Disposition of the crack arrest marks indicated at leasttwo separately nucleated cracks coalesced in theearlier stages of growth. At higher magnification, thefeatures of the fracture surface in Figure 17 showedvery similar characteristics to those in Figure 10 fromSSRT. Although derived from data from a differentspecimen than Figure 17, Figure 18 shows the stresscorrosion fracture surface near a crack tip exposed bybreaking the specimen in liquid N after completion ofthe cyclic loading test (compare to Figure 10).Corrosion products near the crack tip did not obscurethe fracture surface, but at short distances from the tip,the fracture face was obscured by corrosion product,some of which was so loosely adherent that it wasremoved simply by washing. This left a blackenedfracture face. The black corrosion product, mostlikely magnetite, had to be removed electrochemicallybecause of its adherence. This usually revealedsignificant corrosion of the fracture surface, whichincreased as the mouth of the crack was approached.

OM was performed mostly to determine maximumcrack depths. It also confirmed SEM observations ofessentially TG cracks (Figure 19). Figure 19 alsoshows that opposite faces of the crack sufferedcorrosion subsequent to crack extension, since theopposing parts did not match except very near thetip.

DISCUSSION

Polarization curves for pipeline steels in the high-pH environment at a concentration that promotesIGSCC show typical active-to-passive transitions in thepotential range within which cracking occurs. Suchbehavior was not shown by equivalent curvesdetermined for the solutions in Table 1, whether or notCO2 was present. That difference implied a differencein the mechanism of cracking for the high- and low-pHsolutions. For the high-pH solutions, crack growth wasby a dissolution-controlled process, which could bemodeled to give predicted velocities in good agreementwith experimental measurements.2 The presence ofrelatively large amounts of corrosion products in low-pH cracks showed dissolution processes occurred, butpolarization curves indicated the highest anodic currentdensities at relevant potentials were unlikely to exceed10–3 A/cm2, which corresponded to a crack growth rateof ~ 4 x 10–9 mm/s (writing Faraday’s second law as apenetration rate). Such a growth rate was appreciablybelow the highest value determined in SSRTexperiments, where velocities in excess of 10–6 mm/swere frequent, or in cyclic loading tests experimentsinvolving higher maximum stresses (Figures12[c] and[d]). Consequently, a simple dissolution model was

dry air (created by containing the sample in a cell filledwith magnesium perchlorate, which removes moisturefrom air). The test was continued for 35 weeks (1.5 x106 cycles) over the stress range 517 MPa to 241 MPa(R = 0.47). Examination of the specimen surface bySEM and by OM after sectioning revealed no evidenceof crack initiation, which underlined the important roleof the CO2-containing environment in cracking in thissystem.

FractographySEM revealed cracks of various length at the

surface corresponding to the original outer surface ofthe pipe. Cracks tended to lie within narrow bands,compared to the length over which the cracks in agroup were apparent. Figure 14 shows an examplewith no cracks visible other than those within the band.Cracks in Figure 15 showed a similar tendency, butalso indicated more readily the bending of cracksbefore coalescence. Measurements were made of thetransverse separation and average lengths of adjacentcracks to determine whether the coalescenceconditions for small cracks were similar to those forlarge cracks (Figure 2). The data for small cracksobserved in the cyclically loaded specimens are shownin Figure 16. The line that separates the coalescedfrom the non-coalesced cracks indicated coalescenceoccurred if:

y < 0.156 (2a) (2)

which was sufficiently close to the conditions inEquation (1) for large cracks to suggest that the samerule applied, irrespective of crack size. This also hasbeen found for IGSCC of line pipe steel.3

Evidence for the coalescence of cracks also wasapparent on the fracture surfaces of two cyclically

FIGURE 11. SEM micrograph of etched section through aSSRT specimen tested in NS3 solution saturated with CO2, ata potential of –0.7 VSCE, showing the association of internalcracks with bands of pearlitic material.



insufficient to explain the crack growth in low-pHsolutions, although it was clear that dissolution wasinvolved, as indicated by attack on the crack sides andby corrosion products in the crack enclave.

Pitting also may have played an important role,especially in the initiation of cracks, as exemplified byFigure 1. Moreover, the cyclic loading tests showedcracks were initiated with relative ease from the sidesof specimens that displayed the original surface of thepipe compared to polished surfaces. While theundulations on the original surface provided somedegree of stress concentration, that was insufficient toexplain the different behaviors of the two types ofsurface. If so, the polished surfaces should haveproduced cracks at sufficiently high applied stresses,which they did not.

Conversely, when the potential was controlled alittle above Epit during SSRT, pitting, reduced ductility,and cracks were observed sometimes. In addition, theundulations on the original pipe surfaces wouldconstitute pre-existing pits within which the changes insolution composition that accompanied pittingreactions would be facilitated. Thus, chlorides andsulfates in the low-pH solutions could result in localacidification within pits. In turn, that acidification couldfacilitate some of the reactions involved in the crackgrowth process.

Perhaps the most significant pointer to animportant component of the mechanism of crackgrowth in the-low pH solutions was from SSRT, whichshowed marked reductions in the ductility to fracture atrelevant potentials, especially in the presence of CO2.

FIGURE 12. Weibull plots of the crack velocity data from cyclic loading tests on X-65 steel involving various Rvalues and maximum stresses in the ranges: (a) < 414 MPa, but mostly at 345 MPa; (b) 414 MPa to 449 MPa,but mostly at 414 MPa; (c) 483 MPa to 518 MPa, averaging 490 MPa; and (d) > 552 MPa, averaging 577 MPa.



FIGURE 13. % chance (appended to each line) of the averagecrack velocity achieving the values shown for various maximumstresses (data from Figure 12). The two data points shownrelated only to the laboratory tests specimens that failed totallyand suggest there was a 7% to 8% chance, at very highmaximum stresses, of achieving the velocities necessary topromote failure in times of the order used in such tests.

FIGURE 14. Tendency for cracks to form in colonies that arenarrow compared to their length, at the surface correspondingto the original outer surface of the pipe, in a cyclically loadedspecimen exposed to NS4 solution at pH ≈ 6.5.

in Table 1 with or without CO2 in terms of the extent ofcracking over a range of potentials and pH values. Thesolid line represents the equilibrium potentials belowwhich H discharge was possible. Most of the datapoints for tests in which cracking was observed beyondthe necked region fell below that line, just as themajority of tests involving cracks solely in the neck ortests involving no cracks lie above the H line. Figure 20also shows Epit for the dilute solutions with (pH ≈ 6 to 7)or without CO2 (pH ≈ 8). Above those potentials, ifpitting occurred and the potential was not so high as topromote more general corrosion, pH values of thesolutions within the pits would be below those of thebulk solutions, and H discharge facilitated. Thatprobably accounted for those data points relating toextensively cracked specimens that were above theH line in Figure 20 (i.e., those points should have beensomewhere to the left of the positions shown, becausethe actual pH within a pit or crack would be less thanthe bulk solution pH used in plotting the data). Theimportance of CO2 in cracking, displayed in Figures 6and 7, probably was related largely to its influence inlowering pH and facilitating the discharge of H, asreflected in Figure 20 by the CO2-containing solutionswith pH below ~ 7.

Other data supported a H-related crackingmechanism for solutions with or without CO2, such asthe progressive shifts in the lines on Figure 7 as thepotential was lowered. The internal cracking apparentin Figure 11 also would be difficult to explain withoutinvoking a role for H. The quasi-cleavage shown inFigures 10 and 18 is typical of the fractographyassociated with ferritic steels exposed to H.

Field and laboratory data showed considerablescatter, with cracking not invariably occurring for

This reduced ductility could not have been simply aconsequence of the formation of cracks in thespecimens, since no cracks were detected in somespecimens showing a marked reduction in the ductilityto fracture or the cracks were so few and small thatthey were unlikely to adversely influence ductility.Rather, it was most likely that a large proportion of thereduction in ductility was a result of H ingress into thesteel. Figure 20 shows SSRT results for the solutions

FIGURE 15. Small cracks at the surface corresponding to theoriginal outer surface of the pipe, produced by cyclic loading ofa specimen exposed to NS4 solution, pH ≈ 6.5, at OCP, andshowing bending of the cracks prior to coalescence.



apparently identical exposure conditions even incarefully controlled laboratory tests. There were anumber of possible reasons for such scatter, beginningwith the pitting process suggested as a forerunner ofcracking. The stochastic nature of pitting in pipelineshas received some attention from the viewpoint ofpitting as a cause of failure.10 There was no reason todoubt that the scatter inherent in pitting as a cause offailure was also manifest when pitting was a forerunnerof cracking. When no significant cathodic currentreached the pipe surface, OCP would have been in theregion of the pitting potential, but that did not mean pitswould have nucleated inevitably. Even with theundulations of the original surface of the pipefacilitating pit nucleation, pits did not appear to forminvariably, although they formed markedly morefrequently than at polished surfaces. However, pitsappeared in some specimens subjected to cyclicloading tests from which cracks did not grow. Thus, thechance occurrence of pitting was added to the chanceof crack nucleation from pits, compounding theelement of scatter.

Other features of the data collected from the cyclicloading tests appeared relevant to this issue. In high-pH cracking, tests similar to those reported in thepresent work showed a statistically significant trend forcracks to continue to nucleate with increasing testtime.3 That trend plays an important role in crackgrowth because of its relevance to crack coalescence.While evidence for crack coalesence in low-pHcracking was readily apparent from examination of fieldand laboratory samples, the cyclic loading testsproduced no evidence of a systematic trend for crack

numbers to increase with increasing test time. In somespecimens, high crack densities were observed(approaching 10 cracks per mm of section length). Inother specimens, the density was markedly less forappreciably longer test times. While there was nocorrelation between crack density and test time, therewas a very significant correlation, despite considerablescatter, between crack density and crack velocity. Thiscorrelation was not surprising in view of the evidencefor crack coalescence, with increasing crack densityenhancing the chances of coalescence and, thus,increasing crack velocity. Because of the stressdependence and scatter in the crack velocity data, aswell as the various test times involved, the trendbetween crack density and velocity was most readilyapparent simply from averages of those quantities fordifferent stress levels (Table 2).

The lack of a consistent trend for crack densitywith time may have been related to the method used todetermine the number of cracks, coupled with theirtendency to occur in bands (Figures 14 and 15).Obviously, a metallographic section through such aband was likely to reveal only one or two cracks. Thenumber of cracks in a band could have increased withtime, but such a change would not have been revealedin a microsection. SEM would have been the mosteffective method to count cracks in a band, but thequestion of why cracks form in bands was of perhapsgreater significance, especially from the viewpoint ofeventual modeling of low-pH cracking. The distributionof IG, or high-pH, cracks approximates to random-ness,11 but that did not appear to be likely with TGcracks in view of their tendency to form in bands. This

FIGURE 16. Spatial and dimensional characteristics of cracksproduced in cyclic loading laboratory tests in low-pH NS4solution, according to whether they showed evidence ofcoalescence.

FIGURE 17. Part of the fracture surface of a large crackproduced in X-65 steel by cyclic loading while immersed inNS4 solution, pH ≈ 6.5, at OCP. The edge shown correspondedto the outer original surface of the pipe and showed evidenceof multiple crack initiation and coalescence.



cracks. That appeared to conform with banding of themicrocracks in Figures 14 and 15 and with the fracturesurfaces produced by the low-pH solution. The fracturesurfaces did not appear to show the markedlyundulating crack fronts usually seen with high-pHcracking and that might have been a consequence ofthe closer spacings of the TG cracks. However, it alsomight have indicated that, while coalescence occurredin the early stages of the growth of TG cracks, the laterstages of growth involved much less coalescence andthat the later stage was dominated by the growth of anindividual crack. That appears to be the case for thespecimen in Figure 17, where evidence of thecoalescence of a few cracks near the original surfaceof the pipe is apparent, but where the subsequentgrowth showed no evidence of further coalescence,even on those parts of the fracture surface not shown.That specimen was subjected to a very high maximumstress (552 MPa) and a low R value (0.44), and so itmight have been expected that single crack growthwould dominate the later stages of crack extension.However, careful examination of service crackssuggested many small, almost coplanar, crackscoalesced to produce the widest cracks and that thelength of such cracks increased markedly bycoalescence with further, almost coplanar cracks. Itwas conceivable that coalescence resulted in a longsurface crack because of banding until reaching a pointwhere the crack’s subsequent growth, especially indepth, was as a single crack.

Apart from the differences in mechanism, thelargest difference from a modeling point of view was inrelation to the later stages of growth, with thecoalescence that plays a major role in high-pHcracking being replaced by growth of individual cracksin low-pH cracking, assuming the above speculationwas valid. This proposition could be tested using thecrack velocity data from Figure 12 to determine thepredicted failure times for an operating line, includingthe element of chance incorporated in the data of thatfigure. It was assumed for the purposes of thecalculations that a crack would need to reach a depthof 6 mm for failure to occur. Results are shown inFigure 21 as a plot of the time to failure in yearsagainst the chance of failure for various maximumstresses, although only the curve for 345 MPa (50 ksi)approached reality. That curve indicated times of thecorrect order for relatively low chances of failure, risingto about a 20% chance of failure in 15 y (i.e., the timeto the first indications of cracking in the TCPL line).However, the estimates were based upon laboratorydata from experiments in which the environmentalconditions for cracking persisted and those conditionsmay not hold true in the field, in which case the curvefor the maximum stress of 345 MPa would be movedto the right by some unknown amount.

tendency suggested sites for the nucleation of newcracks were related to the locations of existing cracks.That relation may have been a consequence of Haugmenting the tendency for shear localization orinstability, as indicated by the influence of H on themixed-mode (I/III ) toughness of a low-alloy steel.12

The relation between new cracks and existingcracks may be relevant to cracking on operatingpipelines. While the distribution of those cracks often isrelated to the weld reinforcement causing the tape tobridge the immediately adjacent pipe, the length andstraightness of some of the cracks suggested thatmuch of the coalescence was of essentially coplanar

FIGURE 18. Fracture surface characteristics near the crack tipof a cyclically loaded specimen exposed to NS4 solution, pH≈ 6.5. The specimen was broken after immersion in liquid N,which created the cleavage shown to the lower left, theenvironment-sensitive fracture extending to the right from aline joining the 11 o’clock to 5 o’clock positions.

FIGURE 19. SEM micrograph of an etched section through acyclically loaded specimen showing the essentially TG pathfollowed by a secondary crack.



A further point in the context of Figure 21concerned laboratory data and reproducibility. Themaximum time used in the cyclic loading laboratorytests on which Figure 21 is based is shown on thefigure. This time indicated that failure was likely only atvery high stresses in such times, and even then with arather low probability. That statement remained valideven when allowing for the thinner sections oflaboratory test specimens, which would cause thecurves to be moved to shorter times by a factor of 2.4.Of course, useful data may be obtained from laboratorytests interrupted before the point of failure, but thequestion of reproducibility remains unresolved and thenecessity for replicate tests is clear.

CONCLUSIONS

❖ Data collected in the field showed TGSCC of high-pressure gas transmission pipelines was associatedwith dilute solutions with pH ≈ 6.5. This pH region wasnot consistent with significant amounts of cathodiccurrent reaching the pipe surface.❖ SSRT was used to rapidly survey environment-related factors. Results showed CO2 played animportant role in the TGSCC of a grade X-65 steel.Marked reductions were seen in the ductility tofracture, even in the absence of secondary cracking, atpotentials in the region of OCP.❖ SSRT showed no tendency for cracking to dependupon temperature, in the range from 5°C to 45°C, inconformity with field observations.❖ Cyclic loading tests over wide ranges of maximumstress and R values in solutions of pH ≈ 6.5 producedcracking from surfaces corresponding to the originalouter surface of pipe material. Those cracks had thesame features observed in samples from operatinglines.❖ Results suggested the mechanism of TGSCC bydilute CO2-containing solutions of pH ≈ 6.5 involveddissolution and H ingress into the steel.❖ Cracks probably initiated at pits, where a localizedenvironment was generated that had a low enough pHto allow the discharge of H at the operative potential inthe pit.❖ Continuing anodic dissolution in the crack enclave isa necessary corollary to the component of crackgrowth resulting from H ingress, especially in theabsence of appropriate applied cathodic currentreaching the pipe surface. The anodic reaction iscomplimentary to the generation of H by the cathodicreaction.❖ Some evidence from laboratory and field samplesindicated the cracks formed in bands, suggestingnucleation of new cracks occurred a little beyond thetips of existing cracks.

TABLE 2Average Crack Densities

and Crack Velocities from Cyclic Loading Tests(A)

345 414 483 552

No. of cracks/mm 1.20 1.86 1.96 3.93

Crack velocity10–9 mm/s 7.5 12.1 66.2 185.5

(A) Various R values and test times not recognized in determining theaverages also were involved.

Maximum Stress (MPa)

FIGURE 20. Extent of cracking in SSRT specimens in dilutesolutions at various potentials and bulk solution pH values, thelatter achieved by using different partial pressures of CO2. Alsoshown is the equilibrium line below which H can be dischargedfrom water, together with Epit of the X-65 steel in dilute solutionsof pH ≈ 6 to 7 or pH ≈ 8.

FIGURE 21. Chance-dependent times required to reach acrack 6 mm deep in a pipeline for various maximum stressescalculated from the Weibull plots in Figure 12.



❖ There was clear evidence from field and laboratorysamples of cracks coalescing. The later stages ofgrowth may have been dominated especially by thedepthwise extension of what became a single longcrack as the result of much coalescing in the earlierstages of growth.❖ Estimates of the lifetime of a pipe from depthwisecrack velocities determined in laboratory tests andinvolving realistic maximum stresses gave predictedvalues of the correct order, but the laboratory data alsoindicated the chances of such a failure were not high.❖ Field and laboratory data showed considerablescatter, with evidence of cracking from a particular setof exposure conditions not being reproduced whenthose conditions were replicated. It probably isinevitable, therefore, that appropriate statisticalmethods will need to be involved in any approach tomodeling or to the collection of laboratory data relatingto TGSCC of high-pressure gas transmission pipelines.

ACKNOWLEDGMENTS

The authors acknowledge the assistance ofrepresentatives of TransCanada Pipeline Co. Ltd., whocollected most the field data. Most of the laboratory

data was obtained at the University of Newcastle uponTyne for TCPL.

REFERENCES

1. R.N. Parkins, “Environment Sensitive Cracking (Low-pH StressCorrosion Cracking) of High-Pressure Pipelines,” NG-18 Report No. 191(Arlington, VA: American Gas Association, 1990).

2. R.N. Parkins, Corrosion 43 (1987): p. 130.3. R.N. Parkins, P. M. Singh, Corrosion 46 (1990): p. 485.4. B.S. Delanty, J.E. Marr, “Stress Corrosion Cracking Severity Rating

Model,” CANMET Int. Conf. Pipeline Reliability, Calgary, Canada,1992(Ottawa, Canada: CANMET, 1992).

5. TransCanada Pipelines, “Report on 1987 Pipe Integrity Program,SCC Research Program and Planned 1988 SCC Research Program,”1988.

6. R.R. Fessler, “Stress-Corrosion Cracking Temperature Effects,” 6thSymp. Line Pipe Research, Catalog No. L30175, R-l (Arlington, VA:American Gas Association, 1979).

7. T.R. Baker, R.N. Parkins, G.G. Rochfort, “Investigations Relating toStress Corrosion Cracking on the Pipeline Authority’s Moomba-to-Sydney Pipeline,” Proc. 7th Symp. Line Pipe Research, Catalog No.L51495, 27-1 (Arlington, VA: American Gas Association, 1986).

8. R.N. Parkins, “Stress Corrosion Cracking in Pipelines,” NACEInternational Canadian Region Western Conference, Calgary, Alta.,Canada, Feb. 1994 (Houston, TX: NACE, 1994), p. 423-455.

9. W. Weibull, “Fatigue Testing and Results” (Elmsford, NY: PergamonPress, 1961).

10. G.G. Eldredge, Corrosion 13 (1957): p. 51t.11. R.N. Parkins, B.N. Leis, T. K. Christman, “Spatial Densities of Stress

Corrosion Cracks in LinePipe Steels,” NG-18 Report No. 195, 1992).12. J.A. Gordon, J.P. Hirth, A.M. Kumar, N.E. Moody Jr., “Effects of

Hydrogen on the Mixed Mode [I/III] Toughness of a High-Purity RotorSteel,” Metall. T-A 23A (1992): p. 1,013.

Documents

Transgranular Stress Corrosion Cracking of High-Pressure ... Fracture/Final_student... · Although intergranular stress corrosion cracking (IGSCC) of high-pressure gas pipelines has