Subscribe to our newsletter to receive the latest news and events from TWI:

Subscribe >
Skip to content

Fatigue of Deepwater Riser Welds: Crack Growth vs Endurance

   

Fatigue Performance of Sour Deepwater Riser Welds: Crack Growth vs. Endurance

 

Colum M. Holtam
TWI North America LLC,
Houston, Texas, USA

David P. Baxter
Atkins Oil & Gas,
Aberdeen, UK

Paper presented at ASME 30th International Conference on Ocean, Offshore and Arctic Engineering, OMAE 2011, Rotterdam, The Netherlands, June 19-24, 2011, Paper #49581

Abstract

Steel catenary risers (SCRs) are increasingly used in deepwater oil and gas developments. SCRs can be subject to low-stress high-cycle fatigue loading, for example from wave and tidal motion, vortex induced vibration (VIV) and operating loads, and corrosive environments (internal and external). When the production fluids are sour, higher fatigue crack growth rates (FCGRs) are expected and therefore shorter overall life compared to performance in air, as a result of the interaction between fatigue crack growth and sulphide stress cracking. Successful design of risers is critically dependent on the availability of appropriate experimental data to quantify the extent to which fatigue lives are reduced and rates of fatigue crack growth are increased. Historically there has been a discrepancy between experimental sour fatigue endurance data and fracture mechanics-based estimates of the corresponding stress-life (S-N) curves.

This paper summarises the results of recent sour FCGR tests on C-Mn pipeline steel. Tests were performed under conditions of increasing applied stress intensity factor range (ΔK), on specimens containing shallow initial flaws and at very high stress ratios (R), to obtain data close to threshold. In many cases it is material behaviour at these low values of ΔK that dominate the fatigue life (e.g. VIV loading). The FCGR data are then compared to sour fatigue endurance data, both published and from a TWI Joint Industry Project (JIP). The observed environmental reduction factor (ERF) for endurance tests is compared to that expected from the difference in fatigue crack propagation rates, to examine whether FCGR data might provide an alternative means of predicting ERFs.

This paper offers valuable insight into current best practice methods for generating sour FCGR data when qualifying girth welds for sour service, and the relationship between fatigue crack growth and fatigue endurance.

Introduction

Steel catenary risers (SCRs) are increasingly used in deepwater oil and gas developments. SCRs can be subject to low-stress high-cycle fatigue loading, for example from wave and tidal motion, vortex induced vibration (VIV) and operating loads, and corrosive environments (internal and external). When the production fluids are sour (i.e. contain water and H2S), higher fatigue crack growth rates (FCGRs) are expected and therefore shorter overall life compared to performance in air, as a result of the interaction between fatigue crack growth and sulphide stress cracking. Sour production fluids are common in oil and gas applications and therefore successful design is critically dependent on the availability of appropriate experimental data to quantify the extent to which fatigue lives are reduced and rates of fatigue crack growth are increased.

In many cases it is material behaviour at low values of applied stress intensity factor range (ΔK) that dominate the fatigue life (e.g. VIV loading).[1] Experimental crack growth rate data (da/dN-ΔK) are typically determined in simulated operating environments, and upper bound curves can be used in fracture mechanics calculations to calculate critical flaw sizes. Experimental data at low values of applied ΔK are often determined using a decreasing ΔK type test, where the crack is relatively deep by the end of the test.

At low values of ΔK (i.e. approaching threshold) crack growth rates in API 5L grade X65 C-Mn pipeline steel determined under conditions of decreasing ΔK have been shown to be substantially lower than those determined under conditions of increasing ΔK (Figure 1).[2] This is believed to be due to an influence of crack depth (attributed to bulk hydrogen charging from exposed surfaces).[2-4] The reduced influence of a sour environment at low ΔK has similarly been reported elsewhere. [5,6]

Figure 1. Sour fatigue crack growth rate data generated under conditions of increasing and decreasing ΔK, illustrating a possible crack depth effect at lower ΔK (<400Nmm-3/2/13MPam0.5)

Figure 1. Sour fatigue crack growth rate data generated under conditions of increasing and decreasing ΔK, illustrating a possible crack depth effect at lower ΔK (<400Nmm-3/2/13MPam0.5).[2] (Arrows indicate increasing crack depth in each test).


It is possible to compare the results of FCGR tests in a sour environment with fatigue endurance data based on the predicted growth of postulated flaws. An engineering critical assessment (ECA) is a fracture mechanics-based approach that is used to evaluate the significance of a flaw, based on a particular combination of material, stress and environmental conditions. An ECA can therefore provide maximum allowable flaw sizes at the manufacture and installation stage to ensure that, for example, girth weld flaws do not reach a critical size during the projected life of the component. This differs from conventional fatigue design philosophy which uses a stress-life (S-N) approach, whereby an endurance curve is generated from a series of representative tests. S-N design curves can be found in BS 7608[7] and DNV RP-C203,[8] for example, and will typically be based on the statistical mean of the experimental data minus two standard deviations of log N. The S-N approach can be used to assess the performance of a nominally defect-free weld, although joint misalignment can be allowed for. However, for welds with known defects, an ECA approach is required to demonstrate adequate fatigue life.

The assumed FCGR usually takes the form of a Paris law which relates the crack growth per cycle (da/dN) to the range of stress intensity factor (ΔK) where ΔK = Kmax - Kmin and m and C are constants (Equation 1). [9]

The assumed FCGR usually takes the form of a Paris law which relates the crack growth per cycle (da/dN) to the range of stress intensity factor (ΔK) where ΔK = Kmax - Kmin and m and C are constants (Equation 1)

It has previously been demonstrated that the assumed FCGR law has a significant influence when performing ECAs on internal surface-breaking defects in SCRs operating in a sour environment and subject to VIV fatigue loads.[1] It was shown that if the apparent diminished influence of a sour environment at low ΔK was indeed a true representation of material performance under such conditions, then significantly larger initial flaw sizes could be tolerated by closely fitting the FCGR curve to the experimental data at low ΔK. A larger allowable initial flaw size leads to fewer repairs and cut-outs, which means faster installation and significant cost savings.

Historically there has been a discrepancy between experimental sour fatigue endurance data and fracture mechanics-based estimates of the corresponding stress-life (S-N) curves. The aim of this paper is to evaluate whether the results of recent FCGR tests investigating the near-threshold (low ΔK) behaviour of C-Mn pipeline steel in a sour environment can be used to more accurately predict sour fatigue endurance behaviour.

Comparison of fatigue crack growth rate and fatigue endurance data

Figure 2 shows the results of recent FCGR tests on API 5L grade X65 C-Mn pipeline steel in air and in a sour environment,[10] plotted alongside the previous experimental data from Figure 1. Tests were performed under conditions of increasing ΔK on specimens notched in the parent material. Very high stress ratios (R) were used (up to R = 0.9) to facilitate generating data at lower ΔK (from ~100Nmm-3/2/3MPam0.5 to ~650Nmm-3/2/21MPam0.5).

Figure 2. Results of increasing ΔK tests at high stress ratio and starting at low ΔK in air and in a sour environment, plotted alongside data from a decreasing ΔK test in a sour environment

Figure 2. Results of increasing ΔK tests at high stress ratio and starting at low ΔK in air and in a sour environment, plotted alongside data from a decreasing ΔK test in a sour environment.[10] (Arrows indicate increasing crack depth in each test).


The sour environment was identical to that used for the previous tests in Figure 1, and was based on NACE TM0177 solution B.[11] This consists of 5%NaCl and 0.4%Na acetate, and is acidified to pH 3.4-3.6 using acetic acid. The basic solution was saturated with a mixture of 7% H2S in N2, to give a partial pressure of 0.007MPa (1psi) H2S, and there was a continuous passage of gas through the test solution to maintain saturation. All tests were carried out at 25°C (±3°C). The air tests were carried out at a reasonably high loading frequency (5-10Hz), since the results are not expected to be sensitive to frequency.[12] For the tests carried out in a sour environment the loading frequency was reduced to 0.1Hz (which is comparable to wave or VIV loading) to allow time for the environment to interact with the specimen.

It can be seen that there was no apparent reduction in the influence of the sour environment at low ΔK (Figure 2). The most appropriate advice, therefore, is to use an upper bound curve based on experimental crack growth rate data (da/dN-ΔK) obtained under increasing ΔK conditions in a representative sour environment. A more in-depth discussion of the test methods used and analysis of the experimental results is provided elsewhere. [10]

A Joint Industry Project (JIP) included both FCGR and strip fatigue endurance testing of API 5L grade X65 C-Mn steel welds in air and in a sour environment.[13] The results of these tests have been confidential since the conclusion of the project in 2005 but became publishable in 2010. Data from the strip fatigue tests in air and in a sour environment are plotted in Figure 3[13] alongside the Class E mean and design (i.e. mean minus two standard deviations of log N) curves from BS 7608.[7] The specimens tested in air demonstrated better fatigue strength than the Class E mean curve.[7] The sour data showed the fatigue strength to be reduced by a factor of approximately 40 on life. All tests in a sour environment were performed at a loading frequency of 0.2Hz. However, for the sour test performed at an applied local stress range of 46MPa the loading frequency was increased from 0.2 to 1.0Hz after 2,423,909 cycles.

Figure 3. Fatigue endurance data in air and in a sour environment

Figure 3. Fatigue endurance data in air and in a sour environment. [13]


There was an apparent improvement in fatigue performance of the sour test specimens at the two lowest stress ranges. It was speculated that these tests may have initially been below the threshold for crack growth, so cracking did not get underway until corrosion had notched the weld root toe further. Alternatively these low stress tests may indicate a real and disproportionate reduction in crack growth rate at low ΔK. They could also be due to experimental scatter alone.

FCGR tests were carried out under increasing ΔK conditions (from ΔK ~300Nmm-3/2/9MPam0.5 to ~1200Nmm-3/2/32MPam0.5). The sour environment used was identical to that described above and tests were performed at ambient temperature with R = 0.5 and loading frequencies of 0.1, 0.2 and 1.0Hz. Attempts to generate data at lower ΔK were reported as being largely unsuccessful (due to crack fouling by corrosion debris). At ~300Nmm-3/2/9MPam0.5 a slightly higher FCGR was observed at 0.1Hz than the other two frequencies, but this retarded until all three frequencies fell in line with each other. Regression analysis of all the sour data gave Paris law coefficients of m = 3.24 and C = 3.75x10-13 for ΔK in Nmm-3/2 and da/dN in mm/cycle (Figure 4).

Figure 4. Regression analysis of the increasing ΔK tests in a sour environment

Figure 4. Regression analysis of the increasing ΔK tests in a sour environment. [13]


One of the conclusions of the JIP was that FCGR testing was not a reliable method of predicting environmental reduction factors (ERFs) - often referred to as knockdown factors - for girth welds in a sour environment, for the range of growth rate investigated.[13] A greater influence of the sour environment was observed during endurance testing than predicted using fracture mechanics methods using the sour fatigue crack growth curve in Figure 4. It is worth noting that for the stress ranges considered (46MPa up to 200MPa) much of the early crack growth (particularly at low stress) takes place at values of ΔK less than 300Nmm-3/2/9MPam0.5, for which no FCGR data were obtained. However, in the tests presented in Figure 2, FCGR data have been successfully measured at lower values of ΔK and it is possible therefore, that FCGR testing may now correlate more closely with the fatigue endurance data.

In order to compare the relative performance of a material during fatigue crack growth and fatigue endurance testing, a Paris law curve can be used to produce a fracture mechanics based estimate of the corresponding S-N curve, within the framework of BS 7910.[14] Calculations have been performed using TWI's software CRACKWISE 4, which is fully compliant with the latest version of BS 7910.[14] A programme of endurance testing using strip fatigue specimens in a sour environment is expensive and time consuming and no such tests have been performed in the current work. Comparisons are therefore drawn with the endurance data from the previous JIP.

Input parameters

Initial flaw dimensions

An internal surface-breaking flaw was assumed to be located at a girth weld, close to the weld root toe. The weld was assumed to be a full penetration girth weld produced using mechanised processes. A typical example is shown in Figure 5. Welds are known to contain small weld toe intrusions, typically 0.15-0.4mm deep,[15,16] which provides a convenient upper and lower bound estimate of initial flaw size to be used in an assessment, assuming similar flaws were present in the strip specimens tested in the JIP. In the analyses, two different aspect ratios of initial flaw height (a) to flaw length (2c) were assumed; 0.1 and 0.3. Final failure was considered to have occurred when a surface crack propagated half way through the specimen thickness, i.e. when the flaw height reached 10.3mm based on a 14in. riser with wall thickness 20.6mm.

Figure 5. Typical pipeline girth weld produced using mechanised welding processes

Figure 5. Typical pipeline girth weld produced using mechanised welding processes


BS 7910 provides a simplified FCGR equation for C-Mn steels in air, where m = 3.0 and C = 5.21 x 10-13, to facilitate initial screening assessments that can be compared directly with calculations based on fatigue design rules for welded steels (e.g. [7]). However, these coefficients correspond to an upper bound curve. The corresponding mean values (for R ≥ 0.5) are m = 3.0 and C = 2.5 x 10-13.[17] This simplified law was used to calculate the initial flaw size that would force a calculated S-N curve to fit a Class E mean curve. For an aspect ratio of a/2c = 0.1, the calculated initial flaw size present close to the weld root toe in order to get the same life as that from a Class E mean curve was 0.12mm x 1.2mm (using the standard 2D Mk solution for an internal surface-breaking flaw[14]). This is comparable to typical weld toe intrusions[15,16] and provides a degree of confidence in the initial flaw sizes assumed in the analyses.

Stress intensity magnification factor due to presence of weld (Mk factor)

The local stress intensity magnification factor at the weld toe is characterised using the parameter Mk. Standard solutions for surface-breaking flaws are provided in Annex M of BS 7910,[14] derived from 2D (and for certain geometries 3D) finite element analyses. Mk is dependent on what is termed the attachment length, which in the case of an internal surface-breaking flaw in a full penetration pipeline girth weld is the width of the weld root protrusion. For the purposes of this work, a weld root width of 4mm was adopted. Mk is a maximum near the surface and its influence decreases as flaw depth increases. Mk is calculated automatically within CRACKWISE 4 for the selected attachment length and in the current work the standard 2D solutions for an internal surface-breaking flaw in a flat plate were used (M.3.2.2),[14] assuming a full penetration weld. While it is acknowledged that the 3D solution would provide a less conservative solution, it requires the ratio of attachment length (4mm) to wall thickness (20.6mm) to be greater than or equal to 0.5, which is rarely the case for pipeline girth welds.

Fatigue crack growth law

Figure 6 shows a two-stage Paris law based on a regression analysis of the sour FCGR data generated under increasing ΔK conditions (Figure 2) with m = 4.64 and C = 1.40 x 10-15 for Stage A and m = 2.66 and C = 8.3 x 10-11 for Stage B (for ΔK in Nmm-3/2 and da/dN in mm/cycle). The threshold value of stress intensity factor range (ΔKTH) was assumed to be 63Nmm-3/2/2MPam0.5 in line with the guidance in BS 7910 for steels in air. However, the threshold was removed from the analyses at low applied stress ranges to initiate crack growth.

Figure 6. Two-stage mean crack growth relationship in a sour environment based on the latest experimental fatigue crack growth rate data (Figure 2), plotted alongside the results of increasing ΔK tests in a sour environment from the previous JIP

Figure 6. Two-stage mean crack growth relationship in a sour environment based on the latest experimental fatigue crack growth rate data (Figure 2), plotted alongside the results of increasing ΔK tests in a sour environment from the previous JIP. [13]


The comparable sour FCGR data from the JIP are also plotted in Figure 6. It is not immediately clear why the FCGRs observed in the JIP are lower than those observed in the current work. Although the JIP tested specimens notched in the weld (as opposed to specimens notched in the parent material in the current work) recent analysis has shown that, for this material-environment combination, parent material, weld metal, heat affected zone (HAZ) and simulated HAZ microstructures exhibit broadly similar fatigue crack growth behaviour. [3]

Fatigue stresses

For each assumed initial flaw size, assessments were carried out under constant amplitude loading at applied stress ranges corresponding to the sour fatigue endurance tests in the JIP. The number of cycles to failure (defined as when the crack propagated half way through the specimen thickness) was calculated, to allow sour S-N curves predicted via ECA (i.e. fracture mechanics) to be plotted.

Results

Tables 1 and 2 show the fatigue lives predicted via ECA for each initial flaw size based on the sour FCGR curve from the JIP and the two-stage sour curve from Figure 6 respectively. Also indicated in Tables 1 and 2 are the values of ΔK at the onset of crack growth and at failure (defined as when the crack propagated through half wall). The actual observed fatigue lives from the sour endurance tests in the JIP are also shown for comparison. The corresponding calculated sour S-N curves are plotted alongside the sour experimental data in Figure 7 (JIP sour crack growth curve) and Figure 8 (two-stage sour curve).

 

Table 1 Fatigue lives predicted via engineering critical assessment based on the previous joint industry project sour crack growth curve assuming typical initial surface-breaking defects to be present at a girth weld, close to the weld root toe.

Initial flaw size (mm)Aspect ratio, a/2cStress range (MPa)Actual life - JIP data (cycles)Predicted life (cycles)ΔK at start of life (Nmm-3/2)ΔK at failure (Nmm-3/2)
0.15 x 1.5 0.1 46 3,879,512 7,440,000 50 262
60 1,533,579 3,140,000 66 334
81 212,960 1,190,000 89 464
110 76,652 441,000 120 620
146 27,198 176,000 160 810
200 9,910 63,000 219 1,013
0.4 x 4.0 0.1 46 3,879,512 4,470,000 61 265
60 1,533,579 1,890,000 79 346
81 212,960 715,000 107 468
110 76,652 265,000 145 630
146 27,198 106,000 193 844
200 9,910 38,000 264 1,097
0.15 x 0.5 0.3 46 3,879,512 9,820,000 40 263
60 1,533,579 4,150,000 52 340
81 212,960 1,570,000 71 462
110 76,652 582,000 96 618
146 27,198 232,000 127 787
200 9,910 83,500 174 1,042
0.4 x 1.33 0.3 46 3,879,512 6,000,000 48 265
60 1,533,579 2,530,000 63 336
81 212,960 959,000 85 465
110 76,652 355,000 115 616
146 27,198 142,000 153 827
200 9,910 51,000 210 1,081


Table 2 Fatigue lives predicted via engineering critical assessment based on the latest two-stage sour crack growth curve assuming typical initial surface-breaking defects to be present at a girth weld, close to the weld root toe.

Initial flaw size (mm)Aspect ratio, a/2cStress range (MPa)Actual life - JIP data (cycles)Predicted life (cycles)ΔK at start of life (Nmm-3/2)ΔK at failure (Nmm-3/2)
0.15 x 1.5 0.1 46 3,879,512 5,240,000 50 230
60 1,533,579 1,530,000 66 310
81 212,960 385,000 89 437
110 76,652 97,000 120 563
146 27,198 29,000 160 701
200 9,910 9,000 219 976
0.4 x 4.0 0.1 46 3,879,512 2,200,000 61 247
60 1,533,579 645,000 79 334
81 212,960 165,000 107 452
110 76,652 44,000 145 590
146 27,198 15,000 193 794
200 9,910 5,800 264 1,110
0.15 x 0.5 0.3 46 3,879,512 8,820,000 40 224
60 1,533,579 2,570,000 52 280
81 212,960 644,000 71 427
110 76,652 160,000 82 590
146 27,198 46,000 108 730
200 9,910 13,000 174 1,012
0.4 x 1.33 0.3 46 3,879,512 3,630,000 48 254
60 1,533,579 1,060,000 63 323
81 212,960 268,000 85 441
110 76,652 69,000 115 590
146 27,198 21,500 153 720
200 9,910 7,400 210 1,070
Figure 7. S-N curves predicted via engineering critical assessment calculations using the joint industry project sour crack growth curve, plotted alongside fatigue endurance data for specimens tested in a sour environment

Figure 7. S-N curves predicted via engineering critical assessment calculations using the joint industry project sour crack growth curve, plotted alongside fatigue endurance data for specimens tested in a sour environment. [13]


Figure 8. S-N curves predicted via engineering critical assessment calculations using the latest two-stage sour crack growth curve, plotted alongside fatigue endurance data for specimens tested in a sour environment.

Figure 8. S-N curves predicted via engineering critical assessment calculations using the latest two-stage sour crack growth curve, plotted alongside fatigue endurance data for specimens tested in a sour environment. [13]


Figure 7 confirms the finding from the JIP that FCGR testing was not a reliable method of predicting ERFs for girth welds in a sour environment, for the range of growth rate investigated. As highlighted previously, crack growth data were not generated at sufficiently low values of ΔK to accurately model the onset of crack growth. Interestingly however, better correlation is observed at lower stress as these tests exhibited better fatigue performance. An initial flaw size of 0.4mm with a/2c = 0.1 provides the closest agreement with the experimental data, but the ECA prediction is still observed to over-estimate the fatigue lives in a sour environment, which is non-conservative.

When the analysis was repeated using the two-stage sour fatigue crack growth curve from Figure 6 there was a dramatic improvement in the agreement between the ECA predictions and the experimental data (Figure 8). Overall there was excellent correlation between the sour S-N curves predicted via ECA and the experimental data. The observed ERF (for endurance tests) is comparable to that expected from the difference in fatigue crack propagation rates. It would seem that the sour environment has a similar effect on total fatigue life as it does on crack propagation alone. Examining Figures 7 and 8, the predicted sour S-N curves are not significantly influenced by the choice of initial flaw aspect ratio.

The fracture mechanics-based predictions presented in Figure 8 suggest that FCGR data may provide an alternative means of predicting ERFs. Further review and comparison of FCGR and fatigue endurance behaviour in a sour environment is undoubtedly required, to confirm the present indication that the two are directly proportional. However, there are relatively few published S-N data for steels in a sour environment, particularly at low stress. [18-21]

Buitrago et al[21] investigated the influence of a sour environment on fatigue endurance in both the low and high cycle fatigue regimes, including tests on API 5L grade X65 C-Mn steel welds in the same sour environment described above. Tests were also performed at low stress ranges. The results of the pertinent strip fatigue endurance tests in air and in a sour environment are reproduced in Figure 9. Loading frequency was 1Hz compared to 0.2Hz in the JIP.

Figure 9. Fatigue endurance data for API 5L grade X65 C-Mn steel weld specimens tested in air and in a sour environment

Figure 9. Fatigue endurance data for API 5L grade X65 C-Mn steel weld specimens tested in air and in a sour environment. [21]


The observed performance in air was significantly better than that observed in the JIP and had a shallower curve, which is consistent with what one might expect from high quality girth welds. Buitrago et al tested girth welds produced in the 1G position by a mechanised GMAW process after a STT root pass. The welds tested in the JIP were produced in the 2G position with GTAW root and GMAW fill and cap. However, the resulting sour S-N curve was significantly steeper. It was concluded that the use of a constant slope (i.e. one-stage) S-N curve represents a conservative sour design assumption.

The above published data provides another useful test case for the ECA predictions using the two-stage sour fatigue crack growth curve. Table 3 presents the results of this analysis in a similar fashion to Tables 1 and 2, this time alongside the actual observed fatigue lives from the sour endurance tests reported by Buitrago et al.[21] The corresponding sour S-N curves are plotted alongside the sour experimental data in Figure 10.


Table 3 Fatigue lives predicted via engineering critical assessment based on the latest two-stage sour crack growth curve assuming typical initial surface-breaking defects to be present at a girth weld, close to the weld root toe.

Initial flaw size (mm)Aspect ratio, a/2cStress range (MPa)Actual life - Buitrago et al data (cycles)Predicted life (cycles)ΔK at start of life (Nmm-3/2)ΔK at failure (Nmm-3/2)
0.15 x 1.5 0.1 15 23,500,000 - - -
20 7,785,439 250,000,000 22 102
30 2,587,941 38,145,000 33 167
40 2,204,733 & 1,117,474 10,040,000 44 223
50 907,739 3,560,000 55 254
70 372,359 752,000 77 380
100 230,921 148,000 110 504
161 71,610 20,000 176 868
0.4 x 4.0 0.1 15 23,500,000 - - -
20 7,785,439 105,000,000 26 110
30 2,587,941 16,000,000 40 165
40 2,204,733 & 1,117,474 4,210,000 53 218
50 907,739 1,490,000 66 249
70 372,359 318,000 92 380
100 230,921 65,500 132 537
161 71,610 10,900 213 910
0.15 x 0.5 0.3 15 23,500,000 - - -
20 7,785,439 421,000,000 17 105
30 2,587,941 64,100,000 26 147
40 2,204,733 & 1,117,474 16,880,000 35 205
50 907,739 5,990,000 44 243
70 372,359 1,260,000 61 343
100 230,921 246,000 87 537
161 71,610 30,800 140 902
0.4 x 1.33 0.3 15 23,500,000 - - -
20 7,785,439 173,000,000 21 107
30 2,587,941 26,390,000 31 170
40 2,204,733 & 1,117,474 6,940,000 42 216
50 907,739 2,460,000 52 250
70 372,359 522,000 73 390
100 230,921 104,000 105 513
161 71,610 15,000 169 811
Figure 10. S-N curves predicted via engineering critical assessment calculations using the latest two-stage sour crack growth curve, plotted alongside fatigue endurance data for specimens tested in a sour environment

Figure 10. S-N curves predicted via engineering critical assessment calculations using the latest two-stage sour crack growth curve, plotted alongside fatigue endurance data for specimens tested in a sour environment. [21]


The S-N curves predicted via ECA are much shallower than the published test data. Consequently, fatigue performance is under-estimated at high stress and over-estimated at low stress. Table 4 presents predicted fatigue lives based on Stage B only from Figure 6. This represents a conservative upper bound sour crack growth curve. The corresponding sour S-N curves are plotted alongside the sour experimental data in Figure 11. The slope of the predicted S-N curves now more closely matches that of the experimental data and fatigue lives are consistently under-estimated.

Figure 11. S-N curves predicted via engineering critical assessment calculations using an upper bound sour fatigue crack growth curve, plotted alongside fatigue endurance data for specimens tested in a sour environment

Figure 11. S-N curves predicted via engineering critical assessment calculations using an upper bound sour fatigue crack growth curve, plotted alongside fatigue endurance data for specimens tested in a sour environment [21]


Table 4 Fatigue lives predicted via engineering critical assessment based on an upper bound sour crack growth curve assuming typical initial surface breaking defects to be present at a girth weld, close to the weld root toe.

Initial flaw size (mm)Aspect ratio, a/2cStress range (MPa)Actual life - Buitrago et al data (cycles)Predicted life (cycles)ΔK at start of life (Nmm-3/2)ΔK at failure (Nmm-3/2)
0.15 x 1.5 0.1 15 23,500,000 8,500,000 16 84
20 7,785,439 3,960,000 22 113
30 2,587,941 1,345,000 33 168
40 2,204,733 & 1,117,474 626,000 44 224
50 907,739 346,000 55 282
70 372,359 141,000 77 387
100 230,921 54,000 110 517
161 71,610 15,400 176 898
0.4 x 4.0 0.1 15 23,500,000 5,800,000 20 83
20 7,785,439 2,710,000 26 112
30 2,587,941 920,000 40 168
40 2,204,733 & 1,117,474 429,000 53 225
50 907,739 237,000 66 282
70 372,359 97,000 92 398
100 230,921 37,500 132 564
161 71,610 10,500 213 882
0.15 x 0.5 0.3 15 23,500,000 10,300,000 13 83
20 7,785,439 4,800,000 17 112
30 2,587,941 1,630,000 26 166
40 2,204,733 & 1,117,474 760,000 35 225
50 907,739 420,000 44 282
70 372,359 171,000 61 384
100 230,921 66,000 74 535
161 71,610 18,700 140 900
0.4 x 1.33 0.3 15 23,500,000 7,280,000 16 85
20 7,785,439 3,380,000 21 112
30 2,587,941 1,150,000 31 168
40 2,204,733 & 1,117,474 535,000 42 224
50 907,739 296,000 52 283
70 372,359 120,000 73 379
100 230,921 46,500 105 543
161 71,610 13,000 169 840

Discussion

S-N curves have been predicted via fracture mechanics using a two-stage sour FCGR curve, developed from increasing ΔK tests starting at very low ΔK. There was excellent correlation between the predicted S-N curves and the results of sour strip fatigue endurance tests from a previous TWI JIP. This suggests that FCGR and fatigue endurance behaviour in a sour environment are directly proportional, and FCGR data may provide an alternative means of predicting ERFs.

However, the same predicted S-N curves did not correlate as well with endurance data published by Buitrago et al.[21] This is perhaps not surprising considering the significantly better air performance demonstrated by the welds tested by Buitrago et al. It was shown that the initial flaw sizes assumed in the above analyses were comparable to the size of flaw required to achieve the same life as a Class E mean curve, based on a simplified mean crack growth curve for steels in air.[17] However, the welds tested by Buitrago et al exceeded Class E mean performance, therefore smaller initial flaw sizes would need to be assumed in order to force a calculated S-N curve to fit the experimental air curve. Also, referring to the actual fatigue lives in a sour environment reported in Tables 3 and 4, the values of ΔK at the start of life were significantly lower than 100Nmm-3/2/3MPam0.5, the lowest value of ΔK for which actual FCGR data has been measured (Figure 2). Additional test data are therefore required at lower values of ΔK (i.e. <100Nmm-3/2/3MPam0.5). Possible frequency effects have also not been explored in this paper. Adopting an upper bound sour crack growth curve (i.e. Stage B from the two-stage curve) improved the correlation between the predicted S-N curves and the experimental data and ensured a conservative prediction of fatigue life.

Further investigation is required to establish whether the observed ERF (for endurance tests) is comparable to that expected from the difference in fatigue crack propagation rates. However, there are a number of lucrative benefits if FCGR data can provide an alternative means of predicting ERFs. For example, sour corrosion FCGR tests are, in general, considerably cheaper and quicker than fatigue endurance tests. If fracture mechanics calculations based on experimental FCGR data can be shown to provide accurate estimates of fatigue endurance data in sour environments then significant cost and time savings might be achieved during the initial stages of deepwater development projects. Furthermore, industry is increasingly interested in generating corrosion fatigue data at realistic service (i.e. low) stress ranges. Sour fatigue endurance tests under constant amplitude loading are often unfeasibly long and expensive due to the low frequency required in corrosion fatigue testing. FCGR testing at low ΔK may provide an alternative means of investigating fatigue behaviour in this regime.

There is much debate regarding the concept of ERFs (or knockdown factors). It is possible that two different welding procedures may exhibit a different fatigue endurance in air, but a similar performance when tested in a sour environment. The better performing weld (in air at least) is therefore assigned a greater fatigue life reduction factor, and a somewhat more stringent sour design curve. In other instances, fatigue performance in air may significantly exceed that required. The determined fatigue life reduction factor, between strip tests in air and in a sour environment, can then be very large. Applying this reduction factor to the design curve results in a very stringent sour design curve, and penalises the use of a girth welding procedure that results in good (in air) fatigue performance. The only way to eliminate the uncertainty with deriving ERFs from small scale strip fatigue specimens alone is to conduct full scale fatigue tests in a sour environment. To date such tests have not been feasible but a JIP aimed at conducting such tests commenced in 2010.

Conclusions and recommendations

Sour corrosion fatigue behaviour of C-Mn pipeline steels is a complex research area influenced by numerous environmental and mechanical variables. This paper provides encouragement that it may be possible to use fracture mechanics calculations based on experimental FCGR data to provide sufficiently accurate and conservative estimates of fatigue endurance in sour environments. Although further review and comparison is required, the potential benefits of being able to predict total fatigue life via FCGR data justify the continued investment in this area, particularly the development of FCGR test methods in sour environments.

References

  1. Holtam C M, Baxter D P, Ashcroft I A and Thomson R C, 2009: 'Influence of fatigue loading on the engineering critical assessment of steel catenary risers in sour deepwater oil and gas developments', Key Engineering Materials, Vols. 413-414, pp313-325, Trans Tech Publications, Switzerland.
  2. Holtam C M, Baxter D P, Ashcroft I A and Thomson R C, 2010a: 'Effect of crack depth on fatigue crack growth rates for a C-Mn pipeline steel in a sour environment', International Journal of Fatigue 32 (2010) pp288-296, Elsevier.
  3. Holtam C M, Baxter D P, Ashcroft I A and Thomson R C, 2010b: 'Fatigue crack growth performance of sour deepwater riser welds', Proceedings of the 2010 Deep Offshore Technology (DOT) International Conference, Houston, USA.
  4. Holtam C M, Baxter D P, Ashcroft I A and Thomson R C, 2010c: 'An investigation into fatigue crack growth test methods in a sour environment', International Journal of Offshore and Polar Engineering (ISSN 1053-5381), Vol. 20, No. 2, June 2010, pp. 103-109.
  5. Bristoll P and Roeleveld J, 1978: 'Fatigue of offshore structures: effect of seawater on crack propagation in structural steel', Proc Conf European Offshore Steels Research.
  6. Webster S E, Austen I M and Rudd E J, 1985: 'Fatigue, corrosion fatigue and stress corrosion of steels for offshore structures', ECSC Report No. EUR 9460, ECSC Steel Publications, European Commission, Brussels.
  7. BS 7608, 1993: 'Code of practice for fatigue design and assessment of steel structures', British Standards Institution, London.
  8. DNV RP-C203, 2005: 'Fatigue design of offshore steel structures', Det Norske Veritas.
  9. Paris P C and Erdogan F, 1963: 'A critical analysis of crack propagation laws', Journal of Basic Engineering (Trans. ASME), 85D(4), pp528-534.
  10. Holtam C M and Baxter D P, 2011: 'Fatigue Crack Growth Performance of Sour Deepwater Riser Welds in the Near-Threshold Regime', Offshore Technology Conference (OTC), 2-5 May 2011, Houston. Paper accepted.
  11. NACE TM0177, 2005: 'Laboratory testing of metals for resistance to sulfide stress cracking and stress corrosion cracking in H2S environments', Item No. 21212, NACE International, USA.
  12. Frost N E, Marsh K J and Pook L P, 1974: 'Metal fatigue', Oxford University Press.
  13. Slater G, Pargeter R J and Baxter D P, 2005: 'Fatigue performance of corrosion resistant steel catenary risers', final report, volume 1, TWI report no. 14134/26/05.
  14. BS 7910, 2005: 'Guide to methods for assessing the acceptability of flaws in metallic structures', British Standards Institution, London.
  15. Signes E G, Baker R G, Harrison J D and Burdekin F M, 1967: 'Factors affecting the fatigue strength of welded high strength steels, British Welding Journal, 14 (3) 108-116.
  16. Watkinson F, Bodger P H, and Harrison J D, 1971: 'The fatigue strength of welded joints in high strength steels and methods for its improvement', Proc. conference on fatigue of welded structures, Paper 7, 97-113, Brighton, 6-9 July 1970, Abington, UK: Welding Institute.
  17. King R N, Stacey A and Sharp J V, 1996: 'A review of fatigue crack growth rates for offshore steels in air and seawater environments', OMAE - Volume III, Materials Engineering, ASME.
  18. Buitrago J and Weir M S, 2002: 'Experimental fatigue evaluation of deepwater risers in mild sour service', Deep Offshore Technology Conf, New Orleans, USA.
  19. McMaster F, Bowman J, Thompson H, Zhang M and Kinyon S, 2008: 'Sour service corrosion fatigue testing of flowline and riser welds', OMAE 2008-57059, Proc of 27th Int Conf on Offshore Mechanics and Arctic Engineering, Estoril, Portugal.
  20. McMaster F, Thompson H, Zhang M, Walters D and Bowman J, 2007: 'Sour service corrosion fatigue testing of flowline welds', OMAE2007-29060, Proc of 26th Int Conf on Offshore Mechanics and Arctic Engineering, San Diego, California, USA.
  21. Buitrago J, Hudak S and Baxter D, 2008: 'High-cycle and low-cycle fatigue resistance of girth welds in sour service', OMAE2008-57545, 27th International Conference on Offshore Mechanics and Arctic Engineering, ISBN 0-7918-3827-8, ASME.

For more information please email:


contactus@twi.co.uk