An Overview of Near-Neutral pH Stress Corrosion Cracking in Pipelines and Mitigation Strategies for Its Initiation and Growth

This paper is aimed at providing an overview of stress corrosion cracking (SCC) of pipeline steels exposed to near-neutral pH environments and corresponding integrity strategies being developed based on the understanding of the cracking mechanisms determined. Near-neutral pH SCC was first recognized as a separate form of external pipeline SCC in Canada in the 1970’s and 1980’s.^1-2  Since that time, it has been reported in many other countries, including the United States. In general, near-neutral pH cracking occurs at the free corrosion potential, −760 mV_Cu/CuSO4 to −790 mV_Cu/CuSO4 under disbonded coating where cathodic protection (CP) does not reach because of the presence of a shielding coating or some other factor. The cracks are primarily transgranular, typically with wide cracks and substantial corrosion of crack walls. The electrolyte under the disbonded coating is a dilute solution with pH in the range 5.5 to 7.5. This forms a sharp contrast with the relatively well-studied high pH SCC, in which cracks developed in the presence of a carbonate/bicarbonate environment in a pH window from 9 to 13 with cracks that are intergranular and narrow in width.³ 

Besides these features that are often mentioned in publications, near-neutral pH SCC cracks are also featured with very large ratios of crack length over crack depth, as compared with those of high pH SCC, which usually have a crack-depth profile consistent with a semicircular shape, as shown in Figure 1.⁴  This difference is primarily seen only when these cracks are in their stage of initiation and early growth before the occurrence of crack coalescence. The high ratios of crack length over crack depth can be expected in both types of SCC cracks by the time of failure because of crack coalescence, despite the fact that they undergo different mechanisms of coalescence.⁵ 

Based on the principle of fracture mechanics, the stress intensity factor (K) at the crack tip along the crack front of a crack with semicircular shape is approximately the same. This seems consistent with the fact that high pH SCC advances through repeated passivating film-formation-rupture, and film rupture caused by straining of the material at the crack tip is the limiting step.^2-3  On the other hand, near-neutral pH SCC cracks retain a semi-elliptical shape during their initiation and early stage of crack. This crack profile would lead to a much higher K at the depth tip where a higher crack growth rate can be expected if crack growth is driven by mechanical driving forces, as is the case of high pH SCC. This, however, is inconsistent with the crack profiles found in the field. As shown in Figure 2, the majority of the cracks have an aspect ratio (length/depth) larger than that of a semicircular crack. This high crack length/depth ratio should be achieved by nonmechanical driving forces, and, therefore, the crack initiation and early stage crack growth are naturally attributed to the processes of corrosion with higher rate of corrosion at the pipe surface. It should be noted that large length/depth ratio could also be attributed to crack interlinking between individual cracks with small interspacing, especially for those cracks located within tape tented areas associated with pipe long seam welds.

Developing a crack with high length/depth ratios under the process of corrosion can be achieved either because of the fast rate of dissolution at the pipe surface, the reduced dissolution rate in the depth direction, or a combination of both. The latter has been determined to be the situation for crack initiation and early stage crack growth. Under the circumstances, near-neutral pH SCC cracks are usually found to become dormant at a depth of about 1.0 mm, as shown in Figure 2.⁴  This would naturally lead to a question on crack growth mechanisms after cracks become dormant, that is, Stage II crack growth, which should be different from those found before the occurrence of crack dormancy.

Based on this brief introduction, the overview is structured to discuss the mechanisms of crack initiation and early stage growth, and crack growth after dormancy separately. This is followed by the characterization of the field operating conditions and their effects on crack growth that are very different between oil pipelines and gas pipelines. Based on the understanding of the cracking mechanisms, strategies to mitigate field crack initiation and propagation are proposed at the end.

During this stage, the conditions for corrosion have developed, such as coating damage, ground water in contact with the pipe surface, and lack of CP. Crack initiation results from localized corrosion at the pipe surface, leading to crack-like defects. This stage is usually dependent on coating conditions, soil environments, and steel metallurgy. Mechanical driving forces such as operating pressure fluctuations are less important. The rate of dissolution reduces as crack depth increases and many cracks stop growing when reaching a crack depth of ~1 mm, at which point the crack enters a state of dormancy,^6-8  as shown in Figure 2. Stage I can be controlled through effective coatings and effective CP.

It is generally believed that the initiation can be caused by many different mechanisms,^9-20  such as preferential dissolution at physical and metallurgical discontinuities such as scratches,²¹  inclusions,²²  grain boundaries, pearlitic colonies, banded structures^9,24-25  in the steel, corrosion along persistent slip bands induced by cyclic loading prior to corrosion exposure,^10,23  crack initiation at stress raisers such as corrosion pits,^11-12  and localized corrosion through mill scale-steel galvanic effects.^13-14  The initiation of the microstructurally short cracks, usually less than 100 μm, can occur under a constant stress loading. The early crack growth is generally believed to result from the presence of high tensile residual stresses at the pipe subsurface,^11,15-16,25  which adds to the applied stress. The rate of dissolution at the crack tip in crack depth direction is reduced because of a complicated process involving the gradient of CO₂ and the variation of ionic concentrations in the system.

The governing equations for crack initiation and early stage growth have been established recently based on the crack length-depth profiles determined from characterizing cracks developed during field operation.⁴  Crack depth, a, and crack length, 2c, during Stage I crack initiation and growth can be expressed as:

and

In the above equations, h represents the stable value of crack growth rate in the depth by dissolution during Stage II crack growth. The value of h can be measured and determined experimentally; r is the crack growth rate by dissolution along pipe surface, which could be regarded as a constant during crack propagation because the surface is assumed to be fully exposed to the same near-neutral environment during the process of cracking; and m is a fitting parameter which could be obtained by fitting the field crack depth and length data, as shown in Figure 3.

The determination of constants, r, h, and m, is detailed in Zhao, et al.⁴  Figure 4(a) shows three cases of Stage I crack growth with a constant h value, which was measured experimentally to be 7.69 × 10⁻¹⁰ mm/s in average when an X65 pipeline steel was tested in near-neutral pH solution.^4,26  For a given h value, three different m values were selected to fit the crack depth-length profile shown in Figure 3. Different values of m could be related to the different dissolution rates of soil environments, galvanic behavior, and materials resistance to dissolution. In particular, a crack depth-length profile with a high m value would correspond to a crack approaching a semicircular shape, which is not typical of near-neutral pH cracks. On the other hand, a crack depth-length profile with low m value is associated with cracks having large crack length/depth ratios and is very typical of near-neutral pH stress cracks.

The lower bound a-2c profile in Figure 3 has yielded a crack growth rate at the surface as high as 1.1 × 10⁻⁷ mm/s. Such a high growth rate has been found during simulation of enhanced crack initiation and growth caused by the galvanic coupling or the presence of residual stresses.^16,27  In the latter case, a high average crack growth rate was obtained with an estimated value of r of 0.8 × 10⁻⁷ mm/s. Based on this average rate and assuming an a-2c curve with m = 0.2, the value of r is estimated to be 8.3 × 10⁻⁷ mm/s. This is even higher than the crack growth rate of 1.1 × 10⁻⁷ mm/s (the blue curve in Figure 4), which was determined based on the lower bound a-2c curve in Figure 3.

The high dissolution rate at the surface, for example, at around 1.0 × 10⁻⁷ mm/s, forms a sharp contrast with the crack growth rate determined based on the dissolution rate measured by exposing a steel coupon with polished surface directly to near-neutral pH solution, which was determined to be 1.4 × 10⁻⁹ mm/s.²⁸  At the latter growth rate, it would take about 22 y for the crack to reach a depth of 1.0 mm, which is not consistent with the field observations that the life of pipeline steels with SCC is usually about 20 y to 30 y.^1-2  On the other hand, it would take less than 2 y for a crack to reach a depth of 5 mm at a growth rate of 1.0 × 10⁻⁷ mm/s. This is also unrealistic. Therefore, a high dissolution rate at the surface but a reduced dissolution crack growth rate at the crack tip is supported.

The above simple calculations have further revealed some important features of crack initiation and early stage crack growth:

• (1)

  Some galvanic processes capable of supporting very high dissolution rates must be present at the pipe surface, and it must occur within a restricted area, which must be linear in dimension in order to form a crack-like feature.

• (2)

  As passivation of the steel surface is unlikely in near-neutral pH environments,²⁹  high dissolution at the surface may also lead to planar dissolution, such as general corrosion over a wide area, or localized dissolution, such as corrosion pits. Both of them will lead to stress concentration in the area of corrosion and may facilitate the initiation of crack-like features.¹² 

• (3)

  Regardless of the crack-forming mechanisms, a reduced rate of dissolution at the crack tip must occur as the crack propagates, supported by the fact that over 95% of cracks found in crack colonies remained dormant and only less than 5% of them were able to grow out of the dormant state.^16,27 

It should be pointed out that the high rate of corrosion caused by the so-called “hydrogen facilitated dissolution”^30-31  is not justified. As stated in Lu and coworkers,^32-33  “the fact that the active dissolution is almost unaffected by the hydrogen charging and tensile stress indicates that the phenomenon of hydrogen-promoted SCC is unlikely a result of hydrogen-facilitated active dissolution.”

Whether a dormant crack can be reactivated for growth or remain dormant is a matter of Stage II crack growth. It has been determined that Stage II growth is driven by the mechanisms consistent with those for hydrogen-enhanced corrosion fatigue, which is further reviewed in the next section. A crack would remain dormant if the driving force for Stage II crack growth is below the threshold of crack growth for Stage II. Otherwise, a dormant crack would enter into Stage II growth.

The driving force for corrosion fatigue is primarily related to the pressure fluctuations during pipeline operation. For cracks in the same colony or in different colonies within a limited distance, the operating conditions can be considered to be same, and the re-activation of crack growth of some selected cracks must be a location-specific issue, which is further discussed below.

• •

  The first location-specific condition that could either reactivate a dormant crack or maintain its dormancy would be related to the nature and the magnitude of residual stresses at and near the outer surface of pipeline steel.^16,27  As the residual stress must be self-equilibrated, the high tensile residual stresses at or close to the outer surface will usually decrease toward the inner surface, and even become compressive. This would reduce the overall mechanical driving force for the growth in Stage II. In the majority of crack colonies, the overall mechanical driving forces are below the threshold for crack propagation and cracks remain in dormancy. Therefore, the nature and the magnitude of residual stresses determine largely when a dormant crack can be reactivated.

• •

  The second location-specific condition is the level of diffusible hydrogen.³⁴  It acts together with the residual stresses and the chance of a crack to remain at the dormant state will be high if the concentration of diffusible hydrogen in material surrounding the crack tip is low.

Among all of the efforts made on the effect of residual stresses, a noteworthy study by Beavers, et al.,¹⁵  on steel line pipe by hole drilling technique has shown that the mean residual stress near the SCC colonies was about twice as high as in the control areas and the difference was highly statistically significant at a 99.98% confidence level. The average residual stress for the SCC colonies was 216 MPa, with a standard deviation of 104 MPa. This gives a low bound tensile residual stress for SCC colonies of 112 MPa.

For plastically deformable materials, the residual and applied stresses can be added together directly until the yield strength is reached. For an X65 pipeline steel operated at 75% SMYS (minimum specified yield strength), the maximum tensile residual stresses can be added would be about 110 MPa (= 0.25 × 65 ksi × 7 MPa/ksi = 113 MPa). This value is surprisingly consistent with the minimum residual stresses found in SCC colonies, as indicated previously.

Figure 5(a) illustrates how the nature and the magnitude of residual stresses would affect the period of crack dormancy. The longest period of crack dormancy can be expected if the residual stress is compressive and remains high. On the other hand, a surface with high compressive stress is usually accompanied with increased tensile stresses and increased crack growth rate could be expected after cracks become active. Detailed modeling supporting the illustration in Figure 5(a) can be found in Zhao, et al.⁴ 

The surface with the highest tensile residuals stress is much more prone to pitting formation, as shown in Figure 6(b).¹⁶  However, near-neutral pH SCC cracks are found at locations with intermediate level of tensile residual stresses, as shown in Figure 6(a).¹⁷  This is because of the quick transition of high tensile residual stresses at the pipe surface to low tensile or even compressive stresses at the pipe subsurface.

Because of the reduced rate of dissolution at the crack tip and negligible contribution of dissolution to crack growth after dormancy, alternative crack growth mechanisms must be in operation. The mechanical driving forces naturally become predominant in Stage II crack growth and active growth would preferentially start at locations with higher tensile residual stresses and higher diffusible hydrogen content in the material surrounding the crack tip. Although dissolution at the crack tip contributes little to crack advance, it can make the crack tip blunt because of simultaneously dissolution occurring at the crack tip and the crack wall near the crack tip. This could be considered a beneficial effect.

The cracking of pipeline steels exposed to near-neutral pH aqueous solutions was termed as SCC when it was first discovered. With this belief, crack growth rate has also been modeled based on the mechanisms of SCC. In reality, crack propagation has never been observed under a static loading condition in laboratory testing even at high K values, except for the initiation of cracks where crack formation and early stage growth is governed by the mechanisms of corrosion as discussed previously.^9-20  Even an active crack often ceases to occur under static hold.³⁴  In contrast, crack growth is found to grow under cyclic loading above a critical condition to be defined later.^28,34  Figure 7 shows a classical example of a nonpropagation scenario under static hold at two different maximum K values.³⁴  In contrast, crack growth is observed when cyclic loading is resumed, although immediate crack growth did not occur in some environments when the mechanical driving force applied was low.

The results, like those shown in Figure 7, have revealed one of the most important prerequisites for the occurrence of cracking in pipeline steels exposed to near-neutral pH environments: the presence of cyclic loading. It has been further determined (Figure 8) that the crack growth rate under constant amplitude loading can be described by the following equation:^28,34 

where A, n (= 2), α (= 0.67), β (= 0.33), and γ (= 0.033) are all constants, α + β = 1, and h represents the stable value of crack growth rate in the depth by dissolution during Stage II crack growth. The value of h can be determined experimentally and was found to be about one order of magnitude lower than the first term in Stage II crack growth.⁴  The overall power of the frequency, f, was found to be approximately 0.1, which is a factor representing the influence of the corrosion environment on the crack growth rate. The formulation of the combined factor makes it possible to model the crack growth with all of the attributing factors included, such as crack dimension, pressure fluctuations, materials, and environments.

One dilemma may arise when the γ value in Equation (3) is further considered. Although γ is related to the influence of the corrosion environment, it is not clear how the corrosion environment may influence the value of γ, particularly considering the fact that corrosion makes little contribution to crack advance in Stage II. Further investigations have concluded that, for a given pipeline system, the presence of near-neutral pH environments and high residual stresses on the pipe surface determine whether a crack colony can be developed, while the presence of high diffusible hydrogen at specific locations within some crack colonies is thought to determine whether repeated crack growth is possible.³⁴  Hydrogen produced by corrosion at the crack tip is secondary in terms of crack growth, as compared with the amount of hydrogen generated on the pipeline surface resulting from general corrosion. The latter echoes well with the fact that the rate of dissolution at the crack tip is minimal as indicated previously.

The decisive role played by diffusible hydrogen is directly demonstrated by testing compact tension specimens with different conditions of coating coverage, as shown in Figure 9.^26,34-35  These specimens have the same crack geometry and mechanical conditions at the start of the tests. Only the specimen with a bare surface shows a sudden increase of crack growth rate after a period of incubation corresponding to the generation of hydrogen at the sample surface as a result of general corrosion and the time required for achieving a state of hydrogen equilibrium throughout the specimen.³⁴ 

Equation (3) predicts an increased crack growth rate with decreasing loading frequency. This growth rate dependence of loading frequency is experimentally proven in Figure 10 up to a loading frequency of 10⁻³ Hz or higher. However, the loading frequency during field operation can be much lower than 10⁻³ Hz. Further tests under a loading rate lower than 10⁻³ Hz demonstrate the breakdown of Equation (3). As shown in Figure 10, the crack growth rate is found to decrease with decreasing f when f is lower than 10⁻³ Hz.³⁶ 

The transition of crack growth behavior at f = 10⁻³ Hz is surprising, but has been recently found to be related to the saturation of hydrogen ahead of the crack tip at the peak stress of the loading cycle.^36-37  A theoretical model has been developed to understand this crack growth behavior transition based on hydrogen effects on the crack tip during the cyclic load condition.^36-37  It is assumed that the crack growth reaches a maximum rate when the hydrogen concentration at the crack tip reaches a certain value and the hydrogen equilibrium concentration in the plastic zone depends on the applied stresses. Therefore, the critical frequency separating the different growth behavior depends on the hydrogen diffusion into/out of the plastic zone, in response to the variation of stresses. The model suggests that this critical frequency depends on loading condition, temperature, mechanical properties of the steel, and hydrogen diffusivity at the crack tip. It is estimated that the critical frequency is of the order of 10⁻³ Hz, which has a very good agreement with the experimental results shown in Figure 10. Interestingly, similar dependence of crack growth rate on loading frequency has also been observed in an aluminum alloy charged with hydrogen. The critical frequency in the latter material was determined to be in the range from 1 Hz to 10 Hz, reflecting the much higher diffusivity of hydrogen in aluminum-based alloys.³⁸  The above theoretical analysis has further rationalized Equation (3), which can be revised to:³⁷ 

where k_B is the Boltzmann constant, T is the temperature, ν is the Poisson’s ratio, c₀ is the atomic ratio of H/Fe away from crack tip and Ω is the partial volume of hydrogen atom, and ⁠. Equation (5) has provided a clear physical meaning to constant A, which is related to the rate of hydrogen diffusion, temperature, and hydrogen concentration in the material.

During field operations, consistent and constant conditions for crack growth as those achieved during laboratory simulations are not maintained. Even under a consistent and constant mechanical and environmental condition, a growing crack may be periodically or repeatedly experience a process of active growth, dormancy, and reactivation for growth, especially when mechanical driving forces are low, such as in the beginning of Stage II crack growth. This process is illustrated in Figure 11, which was made based on the crack morphology formed by repeated process of active growth and dormancy.³⁹ 

It was found that there exist two thresholds above which continuous crack growth is observed. The lower threshold is the minimum combined factor above which continuous growth of the crack with a sharp tip is seen, while the upper threshold is the critical combined factor above which continuous growth of the crack with a blunt tip takes place. Crack dormancy during Stage II crack growth has been shown to be either mechanically induced or environmentally induced. The former is related to the room temperature creep occurring at the crack tip when the crack is loaded to a combined factor below the lower threshold (nonpropagating condition). The latter is caused by the dissolution of material from the cracked surface, which leads to an increase of crack width and the crack tip radius and, therefore, a higher upper threshold for continuous crack growth. A discontinuous crack growth mechanism was found to operate when the cracks with a blunt tip were loaded to a combined factor between the lower and the upper thresholds. It takes place by repeated dormancy-active growth cycles. Each cycle can advance the crack by a length comparable to the size of grains in the material. The active growth is initiated by forming microcracks at the weakest links, such as grain boundaries, inclusions, or phase interfaces, located in the fracture process zone ahead of the blunt tip, as a combined result of cyclic loading, stress concentration, and hydrogen segregation and trapping. The contribution to crack advance by a direct dissolution of materials at the crack tip is negligible. The transition of discontinuous crack growth to continuous crack growth would occur when microcrack initiations in the facture process zones can take place simultaneously at most locations of the crack front.

To develop appropriate strategies for integrity management of the pipelines with risk of failures caused by SCC, the mechanical loading conditions acting on the pipelines must be well characterized. Figure 12 shows a representative spectrum of pressure fluctuations. It consists of large pressure fluctuations with relatively low R ratios (minimum pressure/maximum pressure) and many smaller pressure fluctuations with high R ratios, called minor cycles or ripple loads.⁴⁰ 

Depending on the location of pipeline sections, the pressure fluctuations could be further characterized into three types based as a function of the relative pressure levels of the large loading events and the minor cycles. Examples are shown in Figure 13 for both oil pipelines and gas pipelines.^40-41 

Type I – underload pressure fluctuations: Type I pressure fluctuations are often found within 30 km downstream of a compressor station on gas pipelines. The maximum pressure of Type I fluctuations is often controlled to be at or close to the design limit, allowing fluctuations only to a level lower than the design limit. The spectrum consists of so-called underload cycles, which are large cycles with low R ratios (min pressure/max pressure) and minor/ripple load cycles with very high R ratios. Underload cycles in oil pipelines often have lower R ratios, higher number of occurrences, and a faster rate of pressure changes (Figure 13[a]) than the underload cycles in gas pipelines (Figure 13[b]). Ripple load cycles are a main feature of gas pipelines.

Type II − mean load pressure fluctuations: Type II pressure fluctuations are typically observed further down from compressor and pump stations. In the case of the Type II pressure fluctuations, the mean pressure is lower than that with Type I pressure fluctuations, and pressure spikes with a pressure level above the mean pressure but below the design limit are frequently seen. The mean pressure is still not low enough to eliminate the underload fluctuations typically seen in the case of Type I pressure fluctuations.

Type III – overload pressure fluctuations: Type III pressure fluctuations typically exist at or close to a suction site, where pressure spikes occurring above the mean pressure, also referred to as overload cycles, become predominant while the occurrence of underload cycles is minimized.

Because of the relatively high maximum pressure and large amplitudes of pressure fluctuations, Type I pressure fluctuations have been further analyzed. Typical characteristics in gas and oil pipelines are compared in Table 1. Oil pipelines are featured with higher frequency of underload cycles as compared with gas pipelines. The rate of loading is slightly different from the rate of unloading in the underload cycles. However, the range of loading/unloading frequency is very different between oil pipelines and gas pipelines. Both the unloading and loading frequencies in oil pipelines vary over a wide range, from 10⁻⁶ ~ 10⁻¹ Hz, while they are very low and over a narrower range in gas pipelines. The number of minor cycles between two adjacent underload cycles is generally higher in gas pipelines than in oil pipelines.

As discussed in the Characteristics of Pressure Fluctuations section, pipelines operate under variable pressure fluctuations. Crack growth models that have been established using experimental data obtained from constant amplitude loading experiments would yield predictions that differ significantly from the actual service life in the field, despite the fact that constant amplitude testing conditions are essential to the understanding of crack growth mechanisms, as detailed in the Stage II Crack Growth section.

Many fatigue crack growth studies available in the literature have been performed under constant amplitude loading and the fatigue crack growth rates under constant amplitude can be predicted fairly well using curve-fitting techniques. However, the problem of predicting fatigue crack growth becomes very complex when the applied load spectrum has variable amplitudes, as shown in Figures 12 and 13 for pressure fluctuations in oil and gas pipelines. This is commonly referred to as variable-amplitude or spectrum loading. It produces so-called memory effects or load-history interaction effects.

Variable-amplitude loading histories and their effects on the fatigue crack growth can vary significantly, depending on the application. It can be concluded in general that, depending on particular combination of applied loading parameters, material properties, specimen geometries, microstructure, and environmental conditions, the same variable amplitude loading sequence can produce either acceleration or retardation of fatigue crack growth.⁴²  As compared with constant amplitude loading (Case I in Figure 14), underload cycle (Case II) accelerates crack growth, while overload cycle (Case V) retards crack growth. For simple load histories containing combinations of overload and underload cycles (Cases III and IV), it has been found that if an underload immediately follows an overload, the degree of retardation resulting from overloading is reduced but not eliminated, while an underload applied prior to an overload, on the other hand, reduces the degree of crack retardation.^43-44  The main physical models proposed to explain the load-interaction effects on fatigue crack growth include: crack tip blunting,^45-47  cyclic plasticity induced residual stress around the crack tip,⁴⁸  crack tip plasticity, and plasticity-induced crack closure.^49-51 

Pipelines are operated under variable amplitude cyclic loading, as characterized in the Characteristics of Pressure Fluctuations section. As the illustration in Figure 14 shows, the Type I – underload pressure fluctuations (Case I) are the harshest in terms of crack growth, and form the greatest threat to pipeline integrity as compared with the other two types of pressure fluctuations being identified. The results of variable amplitude loading conditions conducted at two load levels showed that the crack growth accelerated on subsequent cycles following an underload.⁵²  The damage was more severe when the underloads were distributed evenly throughout the load history compared to when they were grouped together in blocks.

Figure 15 compares the effect of different scenarios of load interactions on crack growth rates, in which the underload cycles in all three cases have an R ratio of 0.5 and the minor cycles in Case I have R = 0.9.⁵³  The latter were determined to be nonpropagating without the underload but obviously lead to crack growth when combined with underload cycles (Case I) and compared with the crack growth rate measured under constant amplitude loading (Case II). In contrast, the crack growth rate was reduced when the minor cycles were replaced with a hold at the maximum load (Case III). This reflects the fact that crack growth proceeds by corrosion fatigue, and the crack growth under constant load, a situation of SCC, can be negligible.

Extensive investigations have been performed to further evaluate the crack growth behavior of underload-type of cycles with different variables/parameters, including maximum stress intensity factors,³⁶  stress intensity factor ranges for underload cycles and minor cycles,⁵⁴  R ratios for underload cycles and minor cycles, and number of minor cycles after an underload cycle.⁵³  Figure 16 is an example showing the difference in crack growth behavior between the constant amplitude loading and variable amplitude loading. From Figure 16, it can be seen that:

• (1)

  The importance of minor cycles in crack growth exists both in air and in near-neutral pH ground water solution. The enhancement in crack growth rate in air is even larger than that in the near-neutral pH environment with the presence of minor cycles. In the former case, an increase of crack growth rate by a factor as high as 20 is observed when the underload-type variable amplitude fatigue loading was applied, while it is about a factor of 5 in near-neutral pH solution.

• (2)

  Regardless of the type of loading mode, crack growth rate is determined to be higher in near-neutral pH environment than in air.

• (3)

  Crack growth in air is insensitive to the loading frequency. However, strong frequency-dependent crack growth behavior is seen in near-neutral pH environment.

It has also been experimentally determined that the crack growth threshold is much lower when loaded under a variable amplitude loading.⁵⁴  As shown in Figure 17(a), crack growth rate was comparatively measured under constant amplitude fatigue loading and under underload-minor cycle loading scheme. In both the cases, K_max and ΔK for the underload cycle and constant amplitude loading were kept the same. The minor cycle in the underload-minor cycle loading scheme has an R ratio of 0.9 and is considered nonpropagating if underload cycle is not introduced. When underload cycle is present, the critical R ratio of minor cycles above which minor cycles will not contribute directly to crack growth is as high as 0.98, as shown in Figure 17(b).⁵⁴ 

For the convenience of life prediction, the crack growth per cycle, as shown in Figure 16, was converted to crack growth per unit time and is shown in Figure 18.³⁶  The typical range of loading frequency and the number of minor cycles are also shown in Figure 18 for both the gas pipelines and oil pipelines. Several key conclusions can be drawn:

• (a)

  The constant amplitude growth behavior would overestimate the crack growth of oil pipeline in the high-frequency range but significantly underestimate the crack growth of gas pipelines typically operating in a low loading frequency regime.

• (b)

  Crack growth rate in oil pipelines is generally higher than in gas pipelines because of larger amplitude and more frequent pressure fluctuations.

• (c)

  When a prediction is made using the actual pressure fluctuation data, the constant amplitude prediction equation would significantly underestimate crack propagation in both oil and gas pipelines at lower frequency, as constant amplitude prediction models do not consider crack growth acceleration caused by load interaction.

The discrepancies between the existing crack growth models based on the results of constant amplitude corrosion fatigue and those with a consideration of load interactions resulting from variable amplitude corrosion fatigue are further summarized and illustrated in Figure 19. It demonstrates that a direct application of Paris’ law to the cracking of pipeline steels in corrosive environment is not justified. In the case of cracking of pipeline steels in near-neutral pH environments, the following three scenarios of load interactions can be identified:

• (1)

  Scenario 1: A previous cyclic loading with an R ratio different from the current cyclic loading may condition the crack tip mechanically to either increase or decrease the crack growth rate of the current cycle and/or the future cycles, which is the so-called load-history-dependent load interaction.⁶ 

• (2)

  Scenario 2: The rate of pressure fluctuations may yield different time-dependent contributions to crack growth rate, which may include the rate of corrosion, the rate of hydrogen segregation by diffusion to the crack tip,³⁴  and the degree of crack tip blunting caused by room temperature creep,^34,55-57  and hydrogen enhanced local plasticity.^58-59 

• (3)

  Scenarios 1 and 2: These scenarios can also mutually interact; for example, crack tip blunting caused by the situations described in Scenario 2 may lead to different stress states at the crack tip and therefore yield different stress-dependent load interactions reflected in Scenario 1.³⁹ 

Many investigations on near-neutral pH SCC have been performed using slow strain rate testing (SSRT).^23,60-69  It should be noted that although SSRT is fast and relevant to some extent, it could yield conclusions that are not consistent with the mechanisms of crack initiation and growth found in the field as a result of the following reasons:

• •

  SSRT cracks are very shallow, usually less than 0.5 mm. This length of crack represents only a situation of crack initiation and early crack growth during pipeline operating, under which mechanical driving forces usually play a negligible role in the field but are predominant during SSRT. A recent conclusion that hydrogen and anodic dissolution play almost equally important roles during SCC of pipeline steel in the soil solution at open circuit^63,68  is an example of the misconceptions from SSRTs. This is simply related to the fact that SSRT cracks are very shallow and crack growth by dissolution is important for shallow cracks. The importance of hydrogen during SSRT evaluation could be overestimated when compared with the cracks developed in the field because of the exaggerated loading conditions in SSRT.

• •

  The crack growth rates determined from SSRTs are usually much higher than that found in the field, and bear no connection to the crack growth rates under the variable amplitude pressure fluctuations during field operation, despite the fact that a situation of interaction between hydrogen and monotonic tensile stress is simulated. This interaction is very different from the load interactions consisting of the load-history-dependent and the time-dependent load interactions as defined earlier. As a result, the crack growth rate determined from SSRT cannot be used for the purpose of life predictions as it cannot be correlated to various situations of variable amplitude pressure fluctuations that have been identified to be very crucial to the crack growth.

Mitigating the risk of near-neutral pH crack initiation could be considered based on factors controlling the process of corrosion:

• •

  Pipe surface treatment: Probably one of the best strategies to prevent SCC is to use a white or near white grit blasted surface preparation to introduce high compressive residual stresses at the pipe surface.^70-71 

• •

  Cathodic protection: The early field work with tape coatings demonstrated that there was no SCC initiation where there was pH elevation at the pipe surface,⁷²  indicating some CP was getting to the surface. Adequate CP is not particularly effective with tape coatings; controlling CP may help in the case of asphalt coatings, where near-neutral pH SCC has also been a problem.

• •

  Coating and coating disbonding: Coating failures are inevitable. Coatings that can create long holidays and narrow gaps may be more susceptible to enhanced corrosion and crack initiation. Fusion bonded epoxy appears to offer good resistance to near-neutral pH SCC cracking.⁷³ 

• •

  pH: Environments with lower pH may shorten the process of crack initiation because of higher rate of dissolution and the time to Stage II crack growth because of higher diffusible hydrogen concentration; on the other hand, it may cause crack tip blunting because of excessive corrosion and therefore lead to a lower crack growth rate because of the reduction of stress intensity at the crack tip.⁴¹  As a result, environments with a pH in the range of 6.0 to 6.5 should be most concerned. This seems consistent with the pH susceptibility of SCC found in the field.⁷⁴ 

• •

  Materials consideration: Pipeline steels should be fabricated, installed, or serviced to avoid the conditions that could lead to galvanic corrosion, which includes but is not limited to heterogeneous distribution of residual stresses during fabrication, the introduction of localized plastic deformation from mechanical damage during installation and service, and fractured mill scale surface on pipeline steels. Although oxide scale could prevent steel surface from direct contact with the ground water in general, the fractured oxide scale may form a galvanic cell with the steel surface and lead to enhanced localized corrosion and the initiation of cracks. Pipeline steels should be fabricated to avoid galvanic corrosion related to its heterogeneity in microstructures.

• •

  Material-environment-stress interactions: Although crack initiation and early stage crack growth are primarily driven by the direct dissolution of steel in some constraint areas, the presence of tensile residual stresses and severe fatigue loading conditions may reactivate a dormant crack for growth even for cracks with a shallow depth. The material-stress interactions are further enhanced at the locations where generation of diffusible hydrogen is enhanced either as a result of corrosion of steel or the cathodic potential applied.

Based on the findings obtained from experimental investigations and modeling, it is obvious that pipelines are currently designed and operated with varied levels of risk of failures largely caused by different scenarios of pressure fluctuations. Equal levels of integrity and safety could be achieved for all pipeline sections if the following strategies could be implemented during pipeline design and operation.

If it were possible to know in advance in the design phase of a project the type of operating pressure fluctuations that the line may be exposed to, consideration may be given to the use of higher grades of pipeline steels and/or larger diameter pipelines for the purpose of lowering the cracking threat by lowering the maximum stress (K_max) and stress range (ΔK).

Pipelines should be operated in a way to avoid those pressure fluctuations that may induce large magnitude of crack growth. Details of pressure control are given next:

• •

  Overall pressure control strategies: The mitigation strategies for crack growth can be developed based on the variables that make up the combined factor shown in Equation (3), in which K_max is related to the maximum pressure involved in the operation, which is uneconomic to reduce as it controls the volume of fluid transported; however, ΔK, which is related to the pressure fluctuations during operation, can be controlled, and reducing ΔK would yield a significant benefit in terms of avoiding crack growth. The benefit can be further categorized as follows:

  • ○

    It reduces the contribution of ΔK to crack growth that is directly related to the magnitude of ΔK, which reflects both the range of pressure variation and the depth of crack. It is obvious that the pipeline is operated under load control and a control of pressure fluctuations would yield increasingly higher benefit as the crack grows.

  • ○

    It eliminates potential growth of minor cycles/ripple loading in the case of variable amplitude pressure fluctuations with underload cycles, which is typical for pipeline sections with the most frequent failures.

• •

  Process of pressurization: It is preferable to have a rate of pressurization higher than 10⁻³ Hz, which reduces the segregation of hydrogen to the crack tip during the loading portion of a pressure cycle. This could minimize the crack growth occurring during pressurization.

• •

  Process of depressurization: Crack growth is normally believed to occur during the loading portion of a fatigue cycle. However, in the case of fatigue crack growth of steel in the presence of hydrogen, additional crack growth can occur during the unloading portion of a fatigue cycle. This has been experimentally observed, as shown in Figure 20, and is believed to be caused by the mismatch between the high lattice dilation resulting from the presence of diffusible hydrogen and the reduced lattice stretching in the crack processing zone accompanied with the unloading.

  • ○

    If depressurization occurs directly following the pressurization and the rate of depressurization is the same as or lower than the rate of pressurization, the increased hydrogen segregation during pressurization cycle can be diffused out at a rate synchronized with the rate of depressurization and no crack growth would occur during the process of depressurization.

  • ○

    If the rate of depressurization is faster than the rate of pressurization, additional crack growth could take place during depressurization because of insufficient time for hydrogen effusion out of the plastic zone ahead of the crack tip.

  • ○

    Regardless the rate of pressurization, additional crack growth could take place during depressurization if there is a hold at max load at the end of pressurization and the rate of depressurization is larger than 10⁻³ Hz, which changes slightly with temperature.

• •

  Ripple loads: Ripple loads can cause crack growth following an underload cycle. Under the circumstances, only the ripple loads with an R ratio higher than 0.98 can be considered to be nonpropagating.

• •

  Static hold: Static hold cannot induce crack growth. Instead, it can reduce the crack growth rate of large cycles of pressure fluctuation. A static hold is recommended following a major underload cycle.

• •

  Overloading: Overload cycle retards the crack growth. It is highly recommended that an overload cycle be introduced periodically or at least following a major load cycle.

• •

  Hydrostatic loading: The same pressure control strategies should be adopted during hydrostatic loading to minimize crack growth induced by hydrostatic loading and to maximize the benefit of hydrostatic testing. Because hydrostatic loading is always followed by a big underload cycle, the benefit of overloading effect associated with hydrostatic testing would be limited, as illustrated in Figure 3.

Pipeline integrity management should be performed to achieve the highest level of pipeline safety equally for every pipeline section in the pipeline system.

• •

  Risk models: Risk models must be made with a consideration of the types of pressure fluctuations and their relative susceptibility to potential crack growth.

• •

  Inspection intervals: Inspection intervals must be determined with a consideration of accumulated crack growth calculated based on:

  • ○

    The pressure fluctuations recorded over the entire period of pipeline operation, which allows for determination of the crack length at the current time, particularly if results of inline inspection are not available.

  • ○

    The net crack growth from the last inspection, which should be used to schedule the next inspection, together with the consideration of achieving a certain level of probability of detection.

This paper has provided an overview of corrosion cracking mechanisms of pipeline steels exposed to near-neutral pH environments. Efforts have been made in developing predictive crack growth models that are consistent with field operating conditions. Major findings are summarized as follows:

• Crack initiation and early stage growth are featured with very high rate of dissolution at the pipe surface caused by various forms of galvanic processes and reduced crack growth in the depth direction leading to crack dormancy.

• Whether a crack remains dormant or becomes active depends on the nature and the magnitude of residual and localized stresses at the subsurface, the severity of cyclic loading, and the level of diffusible hydrogen at the crack tip.

• Pressure fluctuations combined with effects of hydrogen embrittlement are determined to be the predominant driving force for the crack growth after crack initiation.

• It has been determined from extensive experimental investigations that crack growth under Type I pressure fluctuations with frequent underload cycles can be enhanced significantly because of effects of load interactions, as compared with the constant amplitude cyclic loading.

• The load interactions during SCC of pipeline steels in near-neutral pH environments are complex, which include both the time independent load-history interactions and the time dependent load interactions related to the diffusion of hydrogen and hydrogen embrittlement in response to various scenarios of pressure fluctuations.

• Pipelines are currently designed and operated with varied levels of risk of failures largely caused by different scenarios of pressure fluctuations. If it were possible to know in advance during the design phase of a project the type of operating pressure fluctuations that the line may be exposed to, consideration may be given to the use of higher grades of pipeline steels and/or larger diameter pipelines for the purpose of lowering the cracking threat by lowering the maximum stress (K_max) and stress range (ΔK).

The authors wish to thank TransCanada Pipelines Limited, Spectra Energy Transmission, Natural Science and Engineering Research Council of Canada, Pipeline Research Council International, and U.S. Department of Transportation for financial support.