Fatigue Measurement of Pipeline Steels for the Application of Transporting Gaseous Hydrogen¹

As a key part of the hydrogen infrastructure, a safe and efficient method for the transportation of large volumes of hydrogen must be developed. Pipelines are the safest and most cost-effective means of transporting fuels. Because the initial capital cost of hydrogen pipeline construction is high, one possibility for rapidly expanding the hydrogen delivery infrastructure is to adapt part of the existing natural-gas transportation infrastructure to accommodate hydrogen.

The ductility, fracture toughness, and fatigue resistance of linepipe steels and their welds can deteriorate with extended exposure to hydrogen, especially under pressure [1]. It has been well documented that hydrogen can greatly reduce the fracture toughness of some steels and their welds by promoting brittle failure in otherwise ductile materials, an example of hydrogen embrittlement [1,2]. Another serious issue is that the exposure to hydrogen can greatly increase the fatigue crack growth rate (FCGR) [3–11].

The work presented here was based on needs expressed by the ASME B31.12 Committee on Hydrogen Piping and Pipelines [12]. The relevant code was founded on a large volume of tensile property data, which showed the general trend of hydrogen embrittlement increasing as the strength of C-Mn steel increases [1,2]. However, pipelines operate well below ultimate tensile strength. Therefore, a code modification based on fatigue was desirable. In 2009 when this program began, there were few data sets in the literature on fatigue testing of pipeline steels in pressurized hydrogen gas [3–11]. In response, FCGR testing and FCGR modeling of four representative pipeline steels, with the primary variables of hydrogen gas pressure, in this case 5.5 MPa and 34 MPa, and cyclic loading frequency, in this case, 1 Hz, 0.1 Hz, and 0.01 Hz, was undertaken. This work focused on base metals; welds and heat-affected zones will be the focus of future research. The ASME B31.12 committee has approved a code modification based on fatigue. This paper will present details of the FCGR testing of the base metals that forms the basis of that code modification. A companion paper will present the details of the phenomenological FCGR modeling work that is the framework and substance of the code modification [13].

The materials used for this study were X52 and X70 pipeline steels. There were two X52 steels used, one that was placed into natural gas service in 1964 and removed from service at an unknown time, and the other a modern alloy from 2011 that was made specifically for hydrogen application, but was obtained before it saw service. The reasoning for including these two was to test an older steel that could be a candidate for repurposing a pipeline section for hydrogen use and to test a modern steel that is now in hydrogen service. There were two X70 steels from the early 2000 s used, where one saw natural gas service (X70A) and the other did not (X70B). The vintage X52 pipe was 914 mm (36 in) in diameter and had a wall thickness of 10.6 mm. The modern X52 pipe had a 508 mm (20 in) diameter and had a wall thickness of 10.6 mm. The X70 pipes were both 914 mm (36 in) in diameter; alloy A had a wall thickness of 18 mm, and alloy B had a wall thickness of 22 mm. The use of X70 steels was intentional, with the reasoning that perhaps future hydrogen applications may use this classification of higher-strength steel. This grade of steel is not currently used for hydrogen service because the code applies a large increase in wall thickness for steels above the specified minimum yield strength (SMYS) of 360 MPa (52 ksi). Additionally, a comparison of FCGR for steels of different SMYS was needed to determine whether a correlation existed between strength and degradation of mechanical properties that could be attributed to hydrogen applied to fatigue, as it does for monotonic tension.

Pipeline-steel grades are regulated by the American Petroleum Institute's standard 5 L [14] and designated by their yield strength (σ_y) in ksi (kilo psi, 1 ksi = 6.9 MPa). The chemical compositions of these steels have maximum limits on C, Mn, S, P, and dispersoid-forming elements such as niobium and vanadium. The variations in strength (e.g., between X52 and X70) do not result primarily from variations in alloy composition, but rather from variations in the processing of the steel. Thermo-mechanical processing allows the yield strengths of pipeline steels to be tailored through combinations of grain refinement, precipitation hardening (micro-alloying), and phase transformations [15–17].

Table 1 shows tensile properties measured in air in the hoop (transverse) direction (perpendicular to fluid flow direction in the pipe), where the yield stress is defined as the 0.2% offset from the linear elastic line. The tensile tests were conducted in accordance with ASTM E8, where at least three replicates were used [18]. Note that the vintage X52 material did not reach the SMYS of 360 MPa (52 ksi). The reason for this is unknown. The table also shows the pipe sizes.

Table 2 shows the chemical compositions of the four steels. The measurements of chemical composition were performed by a commercial laboratory that used ASTM E415 and ASTM E1019 methods [19,20]. Uncertainties of the measurement methods can be found in Sec. 15 of Ref. [19] and Secs. 9, 20, 31, 42, 54, and 65 of Ref. [20]. Carbon content is low in modern steels to improve weldability. A large difference can be seen between the vintage steel and the modern steels, where the vintage steel has more than three times the carbon content of its modern counterpart. Each X70 steel has even less carbon than the modern X52. The modern steels have low sulfur and contain micro-alloying elements such as niobium and titanium.

Comparison of quantified results included nonparametric statistical analysis (Kruskal–Wallis), and p < 0.05 was considered significant.

The microstructures of the four materials of this work are shown in Fig. 1. X52 vintage is made up of polygonal ferrite and pearlite. X52 modern, X70A, and X70B are likely made up of polygonal ferrite, acicular ferrite, and possibly other shear transformation products (e.g., bainite).

Quantified grain sizes for these materials are shown in Fig. 2. Longitudinal, transverse, and short transverse orientations were investigated, but only results for the longitudinal orientation are shown as this orientation displayed the most potential for anisotropy. For quantified grain sizes, three locations were analyzed for each material: edge, quarter, and midline. These locations cover half of the pipe thickness, and it was in this direction that possible anisotropy was observed qualitatively upon initial inspection of the microstructures. It was unnecessary to show results for the entire thickness as both halves of the thickness were observed to be the same for all four materials. For each location, six different images were quantified and the results averaged for that location. Comparisons between the locations of a material (anisotropy) or comparisons between materials were analyzed by nonparametric statistical methods (Kruskal–Wallis), and p < 0.05 was considered significant. It is important to note that comparisons within one material include all combinations of locations, but comparisons between materials were confined to the same location (e.g., X52 vintage edge to X70A edge). Figure 2 shows all statistically significant differences (*) except the differences between X52 vintage and the rest of the materials, which were found to be significantly different for all locations. There is statistically significant anisotropy observed for X52 modern, X70A, and X70B. The trend observed for all three materials is increasing grain size from the edge toward the midline, and the difference is approximately 1 μm in all three materials. It is important to point out that although there is a statistically significant difference in grain size, the practical significance of a 1 μm change in grain size is most likely negligible in terms of its effect on mechanical properties. For a comparison between materials, X52 vintage has a larger grain size than the other three materials, but there is no difference between the other three materials.

There are some additional microstructural observations that show anisotropy for X52 vintage. In Fig. 3, microstructural banding is apparent in the transverse (T) and longitudinal (L) orientations. Also observed in Fig. 3 is an inhomogeneous distribution of ferrite and pearlite in the short transverse direction. This inhomogeneity is the focus of ongoing work and will not be discussed in this paper. Sulfide stringers were also observed in X52 vintage (Fig. 4). No microstructural banding was observed for X52 modern (Fig. 5), X70A (Fig. 6, left), or X70B (Fig. 6, right).

The observed grain size anisotropy in X52 modern, X70A, and X70B (Fig. 2) most likely arises from a slightly slower cooling rate toward the midline of the pipe thickness. Sulfide stringers in the X52 vintage material (Fig. 4) are expected considering the relatively high sulfur content as seen in chemistry results.

As mentioned earlier, one aim of the study was to determine whether any of these steels showed any pressure sensitivity on the FCGR. Another aim of the study was to measure the effect of cyclic loading frequency on the FCGR. Therefore, testing was conducted at 1 Hz, 0.1 Hz, and 0.01 Hz, with the objective of generating enough data such that the effect of cyclic loading frequency could be included in the phenomenological FCGR model and subsequent testing at 0.01 Hz would be unnecessary. Certain data in the literature showed an increase in FCGR as the cyclic loading frequency decreased, because hydrogen had more time to diffuse to the crack tip [2]. Test pressures of 5.5 MPa and 34 MPa were selected. All FCGR measurements performed in hydrogen gas were conducted at a load ratio, R, of 0.5. Baseline measurements were conducted in air at 1 Hz and R = 0.1. Table 3 shows the test matrix. Only a few select conditions were run at 0.01 Hz because of time constraints. The results of all 94 of these tests are discussed herein. The table is provided to define the scope of the test program and to show the number of repeat tests, if any, for each given condition.

FCGR tests were conducted in accordance with ASTM E647-11 [23] for C(T) (compact tension) specimens with a crack mouth opening displacement (CMOD) gauge attached to the load line and W = 45 mm, where W is the specimen width. The thickness, B, of the specimens ranged from 6 mm to 19 mm. The specimens were polished to an average surface roughness less than 0.25 μm and measured with a profilometer, in accordance with ASTM G142 [24]. The specimens were machined with a chevron notch to aid in growing a precrack with a straight front. For the C(T) specimens, the crack runs parallel to the axial (longitudinal) pipe direction, and the force is exerted in the circumferential (transverse), or hoop stress, direction, designated as T-L or C-L. All C(T) specimens were fatigue precracked in air at 15 Hz, R = 0.1, with a final stress intensity factor, K, held to less than 15 MPa m^1/2 to obtain a sharp initial crack, also in accordance with ASTM E647-11 [23]. Fatigue precracks were nominally 3.2 mm long. All tests were run in load control, where the stress–intensity-factor range, ΔK, increases as the crack grows. As an example of the range of loads used in this dataset, for the thinnest specimens, the loads ranged from 2.0 kN to 4.0 kN, and for the thickest specimens, the loads ranged from 6.2 kN to 12.4 kN.

Prior to FCGR testing, the specimen geometries were used to calculate the maximum and minimum forces from Eq. (A1.3) of the ASTM E647-11 test standard [23]. We had planned on all tests beginning at 8 MPa m^0.5. However, if excessive time was spent waiting for FCGR crack incubation and the onset of growth, then loads were increased, and, correspondingly, initial ΔKs were increased. When testing a chain of specimens, the starting loads were determined from the thinnest specimens. The chain was run until all specimens were completed, or the test was stopped by the operator. If the chain of specimens contained specimens of very different thicknesses, then when thinner specimens completed, loads were increased to levels corresponding to the desired starting ΔK for the next thicker of the unfinished specimens. Prior to performing each test, the chamber was purged with research-grade helium (99.9999%) two times at a gas pressure of 14 MPa. Then, the chamber was again purged with helium at the test pressure or 14 MPa, whichever was higher. The chamber was then evacuated with a vacuum pump for at least 30 min then purged two times with research-grade hydrogen (99.9999%) at 14 MPa and again at 14 MPa or the test pressure, whichever was higher. Following the purging, the chamber was pressurized with hydrogen again to the test pressure, and the test was run. The hydrogen gas was analyzed before testing for each batch of gas purchased (typically a lot of six cylinders). Early in the program, gas was sampled after tests to verify that our purging procedure resulted in gas that had less than 3 ppm oxygen and less than 7 ppm water. Since then, we have assumed that this same procedure will result in clean gas. Sampling of each lot of incoming gas has been subsequently performed. Oxygen can lower the FCGR and water can either increase or decrease it, depending upon the crack growth regime [2,8,25]. The maximum measured oxygen content was 3 ppm and the maximum water content was 7 ppm, and the resulting FCGR tests were considered to be unaffected by that low concentration of gas impurities. This was based on results in the literature where no effects were documented with water or oxygen impurities for a level less than 10 ppm [26,27].

Two chambers were used for this testing program. The smaller chamber has approximately 1 L of internal volume and can accommodate a single tensile or fatigue specimen for testing in gas pressures up to 140 MPa. The second chamber, described more fully later, is larger and can accommodate up to 10 C(T) specimens and test in gas pressures up to 34 MPa. Each chamber was designed specifically for testing in hydrogen gas. Each chamber has a pull-rod on one end that enables the application of load and displacement to a specimen or specimens. These chambers each use three U-cup seals that allow sliding of the pull-rod. The chambers actually contain two separate pressurized regions, such that the load inside the main chamber does not exert load externally on the pull-rod. The second pressurized region balances the forces of the gas pressure on the pull-rod.

In order to generate this large amount of data in a reasonable time, an apparatus that could simultaneously test ten specimens was designed and built [28,29], although some tests were run as single specimens in a separate, single specimen chamber. The chain of ten specimens is shown both schematically and photographically in Figs. 7(a) and 7(b). Clevises of one design attached one specimen to another, and those of a different design attached the top specimen to the pull-rod and attached the bottom specimen to the internal load cell. All clevises had flattened holes such that the pins could freely rotate. The design of the fixture permitted the entire set of specimens to continue fatiguing, even after some had achieved the desired crack length. This was accomplished by way of the links, shown in gray in Fig. 7, that spanned the holes of each specimen and which had elongated holes such that load was transferred through them to the next specimen after the crack length reached 0.75W. This saved considerable time and resources when testing under pressurized hydrogen gas, because there was no need to vent and repressurize when a given sample reached the maximum crack length. All specimens within the chain were exposed to the identical test conditions, eliminating possible variability in gas concentrations or impurities between tests. Polytetrafluoroethylene spacers of varying thickness were placed between specimens and clevises to provide alignment of the chain. Aluminum spacers, shown in Fig. 7(c), were used on top and below the inner clevises to keep the chain erect so that the chamber could be placed over it and the pull-rod attached blindly. Each set of 10 ran 24 h a day and 7 days a week, for a typical duration of over 6 weeks for 1 Hz tests and over 10 weeks for 0.1 Hz tests. Each specimen had a CMOD gage from which data were used to calculate the crack length from compliance. The load was monitored and controlled by the use of an internal load cell attached at the bottom of the set of specimens. The use of an internal load cell is necessary during fatigue testing in pressurized gas because of the forces on the actuating rod from the friction of the seals. This force is inconsistent, but generally increases as the pressure of the test vessel increases. For instance, the maximum load on the external load cell can be as much as 50% higher than that of the internal load cell, for testing at 34 MPa hydrogen gas pressure. The load cell drifted slightly during exposure to hydrogen gas. However, the drift over a month resulted in a change in load ratio, R, of less than 6% over the course of these tests. Pressure was maintained to within 3% of the target pressure throughout the duration of each experiment in both chambers.

Following FCGR testing in the chamber, each specimen was immersed in liquid nitrogen for a few minutes and rapidly cracked open to reveal the precrack length for optical measurement, and so that fractography could be performed on fracture surfaces. From the measurement of the precrack length, an effective modulus was calculated for each specimen. In accordance with ASTM E647-11, a test was invalid if the effective modulus differed from typical (210 GPa) by more than 10% (21 GPa). Data that fell outside of this range are not included here. The data were analyzed by way of the procedure given in Appendix XI of ASTM E647-11 [23].

In austenitic stainless steels, precharging time is a concern. However, precharging works very differently for ferritic steels such as pipeline steels, because the diffusivities of ferritic steels are a couple of orders of magnitude higher and hydrogen solubility is a couple of orders of magnitude lower than in austenitic steels [11]. Therefore, precharging time was not considered to be an issue for this testing program. Specimens of widely varying thickness, from 6 mm up to 19 mm, were tested. While one might think that thicker specimens may behave differently because they soaked in hydrogen much longer before cracking commenced at a measurable rate than for thin specimens, that was not the case. Figure 8 shows FCGR results for two specimens, one of which ran for 4 days, and the other for 28 days. The FCGR data overlap, demonstrating that specimen thickness and test duration do not affect the FCGR results.

As a baseline, FCGR tests were conducted in air on all four steels. Two sets of ten specimens were run, one of X52 steels and the other of X70 steels, which included many repeat specimens for evaluation of the expected variability of the test method. Experimental spread and variability will be used here in place of experimental uncertainty for fatigue crack growth data, as the determination of experimental uncertainty is less representative than variability. The ASTM E647 standard test method for the measurement of fatigue crack growth rates provides no guidance on uncertainty. That test method does give values for variability within a given laboratory (±27% on average, with a range from ±13% to ±50%) and between laboratories (±32%), where a “highly homogeneous 10 Ni steel” was tested [30]. Pipeline steels would not be classified as highly homogeneous, and the ASTM method for calculating variability was not performed on data obtained in hydrogen. For guidance on the expected variability of these data sets, experimental variability of the four steels discussed here and tested in air had variabilities from ±9% to ±44% of da/dN and is rigorously calculated elsewhere [29].

These tests in air were run without the test chamber in place. Figure 9 shows the results of those tests. The tests were run with a loading frequency of 1 Hz, R = 0.1, and a sinusoidal waveform. Subsequent tests in hydrogen gas were conducted with a triangular waveform. From a theoretical viewpoint, there should be no difference between triangular and sinusoidal waveforms, because there is no dwell time at the endpoints of load for either. For details on effects of waveforms for fatigue loading, see Wei and Simmons [31]. Previous results from other authors show that FCGR in air is insensitive to loading frequency [8,32]. We would expect that strength would have little effect in air also, because there is no corrosion component for fatigue in this case. The FCGR of all four materials is remarkably similar in the range of ΔK from 10 MPa m^0.5 to 25 MPa m^0.5. The Paris-law constants for these data in air are given in a companion paper on modeling of the FCGR results [13]. Paris' law for crack growth is a mathematical construct where a straight-line fit is made to data from a plot where both axes are logarithmic.

Figure 10 shows FCGR data for X52 steels in 5.5 MPa (800 psi) hydrogen gas at a cyclic loading rate of 1 Hz. At very low ΔKs, the hydrogen and air data converge. Above ΔK ≈ 10 MPa m^0.5, FCGRs in hydrogen are higher than those in air. We separate the FCGR data in hydrogen into three regions. The first region is at or below ΔK ≈ 10 MPa m^0.5, where FCGRs are similar to that in air, and therefore, the corrosion effect of hydrogen is small. The second region shows an increase in the slope of the FCGR curves in hydrogen, which we attribute to an increase in hydrogen diffusion to the crack tip from an increase in driving force from stress. We believe that the crack extension per cycle is less than or equal to the size of the hydrogen-affected region ahead of the crack tip, which is the size of the fatigue process zone. The crack extension per cycle is equal to the process zone size at the “knee” in the curve [33]. The third region is above the “knee,” which can be found at values of ΔK between 12 MPa m^0.5 and 17 MPa m^0.5. In this region, the slope of the curve is approximately that seen in air. We hypothesize that this region is dominated by the increase in ΔK, where the diffused hydrogen at the crack tip is essentially saturated. In this region, the crack extension per cycle is greater than the size of the hydrogen-affected region due to the stress field at the crack tip. At this low pressure of 5.5 MPa, the FCGR of the vintage X52 steel is below that of the modern steel.

Similar characteristic behavior is seen for both X70 materials at a cyclic loading rate of 1 Hz, R = 0.5, and hydrogen gas pressure of 5.5 MPa (800 psi). Figure 11 shows FCGR results for the two X70 materials at these conditions. The general character of the FCGR data is the same as that for X52 under the same conditions. The FCGR data for both X70 materials at 5.5 MPa show some overlap. A dataset from Sandia National Laboratories, performed in hydrogen gas pressurized to 21 MPa at R = 0.5 and 1 Hz, on an X52 steel of 1990 s vintage, is shown for comparison. The shape of the FCGR data is similar to the National Institute of Standards and Technology tests, except the Sandia National Laboratories test, run at higher pressure, shows a higher FCGR up to 15 MPa m^0.5.

In all four steels that were tested, the FCGR tends to be higher as pressure increases. Figure 12 shows FCGR data for the X52 steels at a hydrogen gas pressure of 34 MPa (5000 psi), as well as 5.5 MPa. The data at 34 MPa gas pressure lie above the data at 5.5 MPa for both materials. The data for the X52 vintage material show a larger separation between the two pressures, indicating that the material is more sensitive to hydrogen pressure than the other three steels.

Figure 13 shows FCGR data for both X70 steels at hydrogen gas pressures of 5.5 MPa and 34 MPa. The X70A steel appears to have slightly higher pressure sensitivity, as compared with X70B, but the 34 MPa data from both X70 steels appear to be within the experimental spread of the data. At values of ΔK above 13 MPa m^0.5, the FCGR data for both X70 steels at both hydrogen gas pressures are essentially the same, as they have the same experimental spread.

An important comparison is whether the X70 steels have higher FCGR than the X52 steels. Figure 14 shows this comparison at a hydrogen gas pressure of 5.5 MPa. The width of the lines of data represents the approximate spread of the data for each steel, and the lines represent the Paris Law relationships of each steel from the previous work [35]. FCGR data from tests of pipeline steels in pressurized hydrogen gas often show two slopes and a knee. Therefore, two Paris law fits are done, one for data below the knee and one above. The FCGRs of both X70 steels show some overlap over much of the range of ΔK. The modern X52 steel shows overlap with the two X70 steels over much of the range of ΔK.

Figure 15 shows a similar comparison of FCGRs of all four steels at a hydrogen gas pressure of 34 MPa, cyclic loading frequency of 1 Hz, and R = 0.5. Once again, the line widths are approximately the spread of the data, and the lines themselves are two-part Paris Law fits to the data, where the fitting parameters can be found in a previous paper [35]. Since we have fewer data sets at a hydrogen gas pressure of 34 MPa compared to that at 5.5 MPa, and some steels have only two valid data sets, and therefore not much spread in the data, we have used the same spread of the data here at 34 MPa as was used for Fig. 14 showing representative lines of FCGRs for 5.5 MPa hydrogen gas. There is significant overlap in the FCGRs of all four steels at this higher hydrogen gas pressure of 34 MPa. The X70A steel has the highest FCGR from 11 MPa m^0.5 to 14 MPa m^0.5, but otherwise there is little to differentiate between these steels at this high pressure of hydrogen gas.

One of the motivations of this work was to determine the extent to which these pipeline steels are sensitive to cyclic loading frequency when tested for FCGR in hydrogen. If the functionality of cyclic loading could be determined with confidence, and that functionality had low sensitivity to microstructure, testing in the future at lower frequencies (below 1 Hz) in pressurized hydrogen gas could be eliminated, given that testing at low frequencies takes a very long time and is prohibitively expensive.

Tests were run on the four steels used in this study in pressurized hydrogen gas at three decades of cyclic loading frequency, 1 Hz, 0.1 Hz, and 0.01 Hz, all at R = 0.5. There were no successful tests on X70B at 0.01 Hz, but there was enough data at enough conditions to show that the sensitivity of pipeline steels to loading frequency is low and can be adequately determined analytically and included in the phenomenological FCGR model [33,36]. If the argument for increased FCGR is based upon diffusion of hydrogen to the crack tip, then a decrease in loading frequency might be expected to have a similar effect to an increase in gas pressure. That is, in the case of decreasing loading frequency, the effect on diffusion is to provide more time per cycle, whereas an increase in pressure increases the driving force for diffusion, but both effects increase diffusion of hydrogen to the crack tip. Figure 16 shows FCGR data for all three decades of loading frequency for the two X52 steels, although no FCGR data were completed at 0.01 Hz for the X52 vintage steel. FCGR increases for the X52 modern steel at the lowest loading frequency. Because the variation of FCGR for both X52 steels from 1 Hz to 0.1 Hz is similar, in that there is little difference between the FCGRs at those two loading frequencies for either steel, we assume that the increase in FCGR for the vintage X52 steel will be similar to that of the modern X52 steel at 0.01 Hz. The lines shown in Fig. 17 are visual fits to the combined data for those test conditions, in order to better differentiate between different loading frequencies.

The effect of loading frequency on X52 vintage and X70A steels at a hydrogen gas pressure of 34 MPa is shown in Fig. 17. Sensitivity to loading frequency is similar or even lower than that seen at 5.5 MPa. We propose that this high gas pressure essentially saturates the X52 vintage material, because decreases in loading frequency show no difference in FCGRs. The X70A material shows a slight increase in FCGR as loading frequency is decreased. The increases in FCGR as loading frequency is decreased are, however, within the experimental spread of the data.

Another way to evaluate the effect of cyclic loading frequency is to determine the FCGRs of the four steels for different loading frequencies at a particular stress intensity factor range, ΔK. Figure 18 shows bar charts of the crack growth rates of all four steels at both hydrogen gas pressures for all cyclic loading rates for which we have FCGR results at a ΔK of 14 MPa m^0.5. The plots show that the X52 vintage steel has the smallest sensitivity to loading frequency, and the modern X52 and X70A have the largest. The effect, seen over two or three decades of frequency, is small for all four of these materials. This effect of cyclic loading frequency can be handled by a power law factor in the phenomenological FCGR model, where the power is on the order of −0.1, such that continued testing in pressurized hydrogen gas at low frequencies, below 1 Hz, is unnecessary. Figure 18 also shows that the effect of increasing pressure from 5.5 MPa to 34 MPa is larger than the effect of decreasing cyclic loading frequency by two or three decades.

where dA/dN_T is the crack length per load cycle and the constants are given in Table 4.

The first version of the ASME B31.12 code based materials performance factors on SMYS for pipeline steels because FCGR data for materials testing in pressurized hydrogen gas were not available, and because monotonic tensile tests showed that higher yield strength strongly correlated with an increased potential for hydrogen embrittlement [1,2]. The result was that pipeline steels with a SMYS greater than 360 MPa (52 ksi) had performance factors of less than one [12]. Figure 20 shows FCGR results from pipeline steels with a measured yield strength ranging from 325 MPa up to 800 MPa. The data show no definitive correlation between SMYS and FCGR. Therefore, higher-strength pipeline steels, such as X70, can be used with the same safety and reliability as X52 for hydrogen transport applications, which will allow for cost savings. The cost savings can be realized because the difference between X52 and X70 steels comes predominantly from processing, rather than chemistry, and the increased strength of X70 yields thinner pipe wall thicknesses and less tonnage of steel [15–17,37].

All four of the pipeline steels tested for FCGR showed some pressure sensitivity, and the vintage X52 material showed more sensitivity to pressure than the other three steels. Sensitivity of FCGR to pressure appears to have a maximum at the highest slope of the da/dN versus ΔK plots for hydrogen data (below the knee).

The effect of loading frequency on FCGR can be accommodated by the use of an analytical term where loading frequency is raised to an exponent. The loading frequency raised to a power of −0.1 applies for X70A, and lesser exponents apply for the other three steels. Other materials or even other steel microstructures may have greater sensitivities to cyclic loading frequency. Those materials or microstructures would require testing to evaluate sensitivity to cyclic loading frequency.

Pipeline steels are specified based on minimum yield strength and historically steels with higher yield strengths have shown greater losses of ductility due to hydrogen embrittlement. Analysis of six steels, including three X52 s, two X70 s, and one X100, shows the expected decrease in ductility with increasing yield strength, but no corresponding change in FCGR.

Both the pressure and cyclic loading results fit into a framework of hydrogen diffusivity. Microstructural variations in these materials, even simply based on grain size, likely lead to differences in diffusion of hydrogen to the crack tip. Increasing hydrogen gas pressure increases the driving force for diffusion, and decreasing cyclic loading frequency gives more time for hydrogen to diffuse. Therefore, in order to design new materials that are resistant to hydrogen damage, lowering of hydrogen diffusivity may be key.

A portion of this research was conducted while Robert Amaro was employed as a National Research Council post-doctoral candidate. Early work on the test facility and single-test fatigue measurements were performed by Y.S. Levy and N.E. Nanninga.

• •

  U.S. Department of Transportation (DOT) (Grant No. DTPH56-09-T-000005).

• •

  Material Measurement Laboratory (Hydrogen Pipeline S).

• •

  Pipeline and Hazardous Materials Safety Administration (Grant No. DTPH56-13-X-000013).

CMOD =

crack mouth opening displacement

FCGR =

fatigue crack growth rate

SMYS =

specified minimum yield strength