D P Baxter, S J Maddox and R J Pargeter
26th International Conference on Offshore Mechanics and Arctic Engineering, OMAE 2007, San Diego, California, 10-15 June 2007. Paper no. 29360.
The fatigue design of pipelines or risers in deepwater oil and gas developments, is often critically dependent on quantifying the extent to which aggressive service environments affect performance. Girth welds in these structures are often exposed to seawater on the external surface, and sweet or sour production fluids on the internal surface. All of these environments can lead to higher rates of fatigue crack growth and lower overall life compared to performance in air.
The seawater environment has been studied in some depth, and design codes provide advice on how steel structures are likely to behave under conditions of either free corrosion or cathodic protection. However, it is important to note that there are limits to how widely these guidelines can be applied, and for more complex environments, such as production fluids which are inevitably project specific, design guidance is rarely available.
Laboratory testing provides a means of quantifying material behaviour in a simulated service environment, and allows the impact of various environmental variables to be explored. This is important as parameters such as temperature, H2S concentration or loading frequency can have a significant effect on the extent to which performance is affected. This paper provides a review of published information and recent research data, and highlights particular areas where existing data are limited and design challenges remain.
Steel catenary or top tension risers are widely used for deep water applications and are subject to fatigue damage associated with wave or tidal motion and vortex induced vibration. Fatigue loading is particularly severe at the touchdown area where the near-vertical riser meets the pipeline or flowline. Depending on the adopted design, the latter can also be susceptible to fatigue damage from a variety of sources. For example, high pressure high temperature pipelines laid on the seabed may be susceptible to lateral buckling, where fatigue results from the thermal expansion cycle associated with shutdown-restart sequences.
Risers and pipelines are also exposed to potentially aggressive service environments on both the inside and outside surfaces. The corrosive effect of seawater on the external surface is typically controlled by the application of cathodic protection. However, this can lead to hydrogen generation on the steel surface, and degradation in fatigue performance needs to be accounted for in design. Sweet or sour produced fluids on the internal surface also have a significant effect on fatigue behaviour. The presence of salts and water, acidified by the presence of carbon dioxide or hydrogen sulphide, all contribute to the aggressiveness of the environment.
Laboratory testing to quantify the corrosion fatigue performance of welds in aggressive service environments can be performed in one of two ways. Endurance data are obtained by testing through-thickness sections of weld and plotting the results on an S-N curve. The effect of environment is then expressed in terms of a fatigue life reduction factor, by comparison with endurance data for tests carried out in air. Standard design curves, such as those given inBS7608, can then be offset by this factor. By contrast, when fracture mechanics calculations are carried out to determine critical tolerable flaw sizes, fatigue crack growth rate data are required. Standard specimens can be used to generate da/dN- ΔK data in the environment of interest, and an upper bound curve determined.
The determination of suitable endurance or fatigue crack growth rate data is of course dependent on ensuring that tests are carried out under appropriate environmental conditions. There is a need for a mechanistic understanding of the processes involved in environmental attack, and the manner in which they influence fatigue behaviour. Service environments can be complex and it is important to have an appreciation of the environmental variables that can influence the extent to which fatigue performance is degraded. This paper, which builds on  but includes more up to date information, provides a review of relevant corrosion fatigue data, and discusses the influence of variables such as pH, temperature, H 2 S concentration and applied electrochemical potential.
Mechanical variables such as stress range, stress ratio (R) or cyclic loading frequency can also have an effect on the extent to which an environment affects fatigue performance. Relevant corrosion fatigue data are again reviewed, and their relevance to particular applications discussed.
Corrosion fatigue endurance data for welded carbon steels were obtained over a period of years as part of European efforts to provide fatigue design data relevant to offshore structures operating in the North Sea. In general, these tests were carried out in either natural or artificial seawater, under conditions relevant to the North Sea. Typically this meant a temperature of around 5°C and a loading frequency of 0.15-0.5Hz. Evaluation of these data led to the widely accepted design recommendations contained in UK HSE Offshore Guidance Notes  , as adopted by DNV  and ISO  . For steels freely corroding in seawater, the fatigue life was reduced by a factor of about three, and the fatigue limit disappeared. When cathodically protected the reduction factor at high stresses was only marginally lower(approximately 2.5), but at low stresses there is more benefit as lives approach those seen in air. Typical design S-N curves are shown in Fig.1.
Fig.1. Typical design S-N curves for welded joints in seawater 
Fatigue crack growth rate data were also determined in these research projects, and under free corrosion conditions there was a three-fold increase in da/dN relative to air. At low ΔK cathodic protection was effective in restoring in-air performance, although at high ΔK the crack growth rate remained high. Cathodic overprotection (-1100mVSCE) was particularly detrimental at high ΔK due to the generation of hydrogen. These and other relevant published data were collated and analysed  to produce the recommended fatigue crack growth curves described in BS7910  . These are summarised in Fig.2.
Fig.2. Fatigue crack growth curves from BS7910 
It should be noted however that both the S-N curves described in Fig.1 and the crack growth rate curves described in Fig.2 are derived from data generated under specific test conditions. Most notably the test temperature was typically 5°C and the cyclic frequency 0.15-0.5Hz. An increase in temperature from 5°C to 20°C was seen to double the observed crack growth rate  , although any possible effect was lost in experimental scatter during comparable endurance tests. 
A reduction in test frequency has also been seen to result in an increase in fatigue crack growth rate. Vosikovsky  examined the effect of frequency on the corrosion fatigue behaviour of X65 steel in seawater, both freely corroding and cathodically protected. A schematic summary of the results is given in Fig.3. It can be seen that when cathodically protected a series of frequency dependent plateaux are observed. At 0.01Hz the crack growth rate reached approximately 6x10-3 mm/cycle before da/dN became independent of ΔK, while at 0.1Hz this plateau occurred when the crack growth rate was just under 10-3 mm/cycle. Under free corrosion conditions the shape of the crack growth curves differed, and no distinct plateaux were observed. It can be seen that the observed increase in growth rate (relative to air) is highly dependent on frequency, but also on the applied ΔK.
The occurrence of a plateau, where crack growth rate is independent of ΔK over a specified range, is a common feature of corrosion fatigue data. These plateaux can arise from diffusion limited, time dependent, environmental assisted cracking processes that cannot be sustained above a critical crack growth rate.
Fig.3. Schematic illustrating the effect of frequency on crack growth rate in seawater 
More recently, the effect of very low frequency cycling has been examined by carrying out so-called 'frequency scanning tests'. These are crack growth rate tests carried out under conditions of constant ΔK, where the cyclic frequency is varied in blocks to determine how frequency affects the fatigue crack growth rate, at a particular value of ΔK. By monitoring the crack growth rate over a relatively short crack increment, it is possible to determine da/dN for much lower frequencies than would be possible using conventional test techniques. Fig.4 shows a typical set of data for X65 parent steel tested in 3.5% NaCl at -1050mVSCE  . It can be seen that as the frequency decreases the measured crack growth rate increases until a plateau is reached, where it appears a further decrease in frequency has no further effect. At ΔK=400N/mm3/2 the plateau occurs at approximately 0.1-1Hz. At a higher value of ΔK however, the observed increase in growth rate was initially similar, but a plateau was not reached until a much lower frequency, at least as low as0.001Hz.
Fig.4. Frequency scanning test data for C-Mn steel in seawater 
Unfortunately, corrosion fatigue data to quantify the effect of frequency on endurance behaviour of welds are very scarce. However, one reference  suggests that under free corrosion conditions, lives were shorter at 0.01Hz than they were at higher frequencies, most notably at high applied stress range.
From a practical standpoint, one additional factor that may influence the observed corrosion fatigue behaviour is the formation of corrosion products or calcareous deposits. These may form on crack surfaces, and act to wedge open the crack so that it does not experience the full range of applied stress or stress intensity. In these cases a higher threshold condition may be observed, although it would be inappropriate to take advantage of this as the presence of similar conditions should not be relied upon during service.
Sweet service environments
The presence of carbon dioxide can lead to a variety of forms of localised corrosive attack. However, very few data have been published to quantify its effect on corrosion fatigue behaviour. With respect to endurance behaviour, most reports describe a significant reduction in fatigue life, and the loss of endurance limit when corrosion is involved. Mehdizadeh  reported a 41% reduction in fatigue life when parent carbon steel was tested in CO2 -containing brine. Other data are less relevant to risers or pipelines. For example, Ebara  reported that welded AISI 4330 steel suffered only a slight loss of fatigue strength at 107 cycles in CO2 gas (80°C, 90% relative humidity) compared with similar tests performed in air. However, the test frequency in this case was fairly high (13Hz), and so a minimal effect of environment might be expected.
Fatigue crack growth rate data are reported Szklarz  , where a series of tests were carried out in a simulated sweet operating environment. Tests were performed at 95°C in a pressurised solution containing 10%NaCl and 10%CaCl2 fed with CO2 at 3 bar pressure. At moderate frequencies (0.2Hz to 5Hz) crack growth rate was seen to be independent of frequency, but strongly dependent on ΔK. At low ΔK crack growth rates were similar to those seen in air. However above ΔK=500N/mm 3/2 crack growth rates were up to an order of magnitude higher than in air, as shown in Fig.5. This study also examined the influence of 500ppm inhibitor, which was found to be effective in reducing the crack growth rate only when included from the start of the test, and only at low ΔK. If inhibitor was added following a period of pre-corrosion the observed fatigue crack growth rate was the same as when no inhibitor was present.
At very low frequency (0.04Hz) crack growth rates at low ΔK were significantly higher than seen at higher frequencies, particularly when crack growth was along the weld fusion line. These data suggested that crack growth rate was independent of ΔK over the range examined. Eadie  also performed tests at this frequency in a salt solution saturated with 10%CO2. Crack growth rates were again significantly higher than in air, and higher than comparable tests carried out at higher frequency.
Fig.5. Fatigue crack growth rate data for C-Mn steel tested in sweet CO2 containing saltwater [14,15]
Sour service environments
Early work examining the effect of H2S containing environments was carried out in the 1970's by Vosikovsky. [16,17] This work examined API 5L X65 pipe (non-welded), and tests were performed in crude oil containing H2S at a stress ratio R=0.05 and cyclic frequency of 0.1Hz. The H2S concentration was varied between 20 and 5,000ppm. Watanabe et al  also tested pipeline steel (welded and non-welded) in crude oil containing H2S, this time at 400ppm, with R=0.03 and a frequency of 0.17Hz. A summary of the data from these two references is provided in Fig.6, along with the mean curve for C-Mn steels in air from BS 7910 (which is in good agreement with the baseline data). It can be seen that at high stress intensity ranges, crack growth is up to 15 times faster at 400ppm H2S, and 25 times faster at 5,000ppm H2S.
Fig.6. Fatigue crack growth rate data for C-Mn steel in crude oil containing H 2 S [16-18]
However, crack growth rates in seawater containing H2S can be somewhat higher than shown in Fig.6. Early work by Bristoll and Roeleveld  looked at crack growth rates in a non-welded plain C-Mn steel, both in seawater and seawater saturated with H2S (3000ppm). These tests were carried out at R=0.6 or higher and a frequency of 0.2Hz. In plain seawater, crack growth rates were 2-3 times higher than in air, while in H2S saturated seawater they were up to 50 times higher than in air. In both cases, the environmental enhancement was dependent on the value of applied stress intensity range, and was much lower at low values of ΔK, as also evident in the work referred to above.
Webster et al  also carried out tests in seawater saturated with H2S, this time with and without applied cathodic protection. Tests were carried out using steel conforming to BS 4360 Grade 50D (non-welded) at stress ratios of 0.05 and 0.7 and a cyclic frequency of 0.17Hz. The range of ΔK examined was slightly higher than Bristoll and Roeleveld, but there was reasonable agreement between comparable sets of data, as shown in Fig.7. At intermediate ΔK crack growth rates were typically 20 times faster than in air, but at high ΔK this factor increased to over 100.
A relatively recent study by Eadie and Szklarz  has examined the effect of various mechanical and environmental test parameters on fatigue crack propagation in a medium strength low alloy steel, tested in sour dilute brines. The partial pressure of H2S was again shown to have a noticeable effect on crack growth rate, as shown in Fig.7. It can be seen that crack growth rates were similar at low ΔK but tended to reach a plateau towards the end of the test. This was a feature of all tests carried out within this study, and was attributed to the limiting rate of hydrogen diffusion. A decrease in test frequency (from 1Hz to 0.1Hz) was seen to increase the crack growth rate plateau, and led to a diminished influence of H2S partial pressure.
Fig.7. Fatigue crack growth rate data for C-Mn steel in H 2 S containing saltwater [18-20]
This study also examined the influence of test temperature, between 30°C and 90°C. Despite the higher rate of hydrogen diffusion expected at elevated temperature, the observed crack growth rates were very similar, or slightly lower than at 30°C. This was attributed to the hindering effect of surface scaling, although the embrittling effect of hydrogen may also be lower at this temperature. The influence of stress ratio was also examined, and although increasing stress ratio had a significant effect on lowering the ΔK threshold  , crack growth rates at moderate and high ΔK were comparable over the range examined (R=0.1 to 0.6). Indeed crack growth rates at R=0.6 were slightly lower than at R=0.1 or R=0.3.
In a later study  the same authors examined crack growth rates in X70 steel in a brine saturated with a test gas containing 10%CO2 and 1%H2S. At a frequency of 0.4Hz crack growth rates were comparable to those seen in earlier work. However, at 0.04Hz the crack growth rate was significantly higher. At low ΔK crack growth rates were comparable to those seen in air, and at high ΔK became limited by a frequency dependent plateau. Interestingly, this meant that the maximum environmental effect (relative to air) was seen at intermediate ΔK, as illustrated in Fig.8.
The influence of frequency was further investigated by Buitrago et al.  where a frequency scanning technique was used to examine the corrosion fatigue behaviour of welded X80 steel in a sour brine based on NACE TM 0177 Solution B. All tests were carried out at ΔK=348N/mm3/2 . The observed crack growth rate data, summarised in Fig.8, suggest that da/dN increases steadily as the frequency is reduced, although between 0.1 and 1 Hz the crack growth rate appeared to be relatively insensitive to frequency.
Fig.8. Effect of frequency on fatigue crack growth rate in sour environment [15,23]
With respect to fatigue endurance, unfortunately there are very few published data for girth welded C-Mn steels in sour environments. The only published data relate to tests carried out by Buitrago and Weir  , where strips extracted from girth welded API 5L X80 and X65 pipe were tested in NACE TM0177 solution B at H2S partial pressures of 0.035-0.070 bara. Data for the specimens taken from 15.8mm wall thickness X65 pipe (tested at 25°C, 0.07 bara H2S, a test frequency of 0.33Hz and an applied mean stress of 150MPa) are reproduced in Fig.9. Fatigue life was reduced by a factor of 10-20 compared to both strip tests and full scale tests carried out in air.
Fig.9. Fatigue endurance data for girth welds tested in air and sour environment 
The experimental data reviewed above indicate that environmental and mechanical test conditions can have a significant effect on the observed endurance or crack growth rate. The need to ensure that design data are determined from tests carried out under conditions that realistically simulate those encountered in service cannot therefore be over-emphasised.
In seawater, the influence of applied electrochemical potential is well documented and design guides quantify the effect that this variable has on the expected life or crack growth rate. However, other variables such as cyclic loading frequency can also have a dramatic effect depending on the applied ΔK. At moderate to high ΔK the crack growth at low frequency (<0.01Hz) can be an order of magnitude greater than at higher frequency. Since design curves in standards such as DNVor BS7910 are based on data from tests carried out at frequencies representative of wave loading (0.15-0.5Hz) care should be taken when trying to determine behaviour at much lower frequencies. Both crack growth rate and endurance test data suggest that a factor of 2-3 may still be appropriate at low ΔK or stress range. However, at higher stresses the influence of frequency becomes far more significant and needs to be taken into account.
In sour environments, test frequency is again seen to have a dramatic effect on the crack growth rate, although the extent to which this is the case again depends on the applied ΔK. The concentration of H2S has also been shown to have a significant effect on the observed crack growth rate. Other factors such as temperature or stress ratio have been shown to have a smaller effect on observed behaviour.
It should also be noted, however, that most of the data referred to above relate to fatigue crack growth rate tests carried out to examine particular test variables. There are very few test data to quantify the effect of these variables on endurance behaviour, and while crack growth rate data may be useful in highlighting trends and identifying key variables, it is by no means certain that the same response will be observed in endurance tests. Endurance tests are dominated by the behaviour of relatively short cracks experiencing a low ΔK. Furthermore early crack growth is usually from a weld toe with the crack tip in a heat affected zone close to the weld root or cap. This is in contrast to crack growth rate data, which are typically determined using fracture mechanics specimens containing long cracks where the crack tip may be well advanced into the weld metal. The size of the crack, the applied ΔK and the microstructure being sampled can all affect the extent to which an environment degrades performance. Endurance tests in a simulated sour service environment should therefore be carried out to quantify the observed fatigue life reduction factor.
In some cases however, endurance testing under realistic conditions may not even be feasible. For instance, if the majority of fatigue damage is associated with very low frequency cycling, endurance tests may take a very long a time. In these circumstances it is perhaps possible to use fatigue crack growth rate data to infer the expected endurance response, i.e. use frequency scanning crack growth rate data to 'adjust' an endurance curve determined by testing at a higher frequency. Such an approach should however be validated by carrying out endurance tests at as low a frequency as possible.
Other environmental factors may also be difficult to replicate in a laboratory test. For instance in a sweet environment, the presence of an inhibitor is likely to have a significant effect, but the efficiency of inhibition may be highly dependent on the degree of solution mixing and fluid flow conditions close to the corroding surface.
Practical considerations may also limit the extent to which some variables can be explored. For instance, very few data are available for corrosion fatigue tests carried out under conditions of variable amplitude loading. Similarly, although full-scale testing of girth welded pipe has become routine, comparable corrosion fatigue testing, with either seawater on the outside or sour fluid on the inside, presents significant technical challenges.
- With or without cathodic protection, fatigue lives at high stress range are a factor of three lower than in air. However the majority of data are associated with tests carried out at around 5°C and 0.1-0.5Hz.
- When cathodically protected, fatigue lives at low stress range are comparable to those in air.
- At frequencies lower than 0.1Hz fatigue crack growth rates increase sharply, particularly at high ΔK.
In sweet CO2 containing environments
- At moderate frequencies (0.2-5Hz) crack growth rates at low ΔK are similar to those in air. However at higher ΔK (>500N/mm3/2 ) crack growth rates are up to an order of magnitude higher than in air.
- At low frequency (0.04Hz) crack growth rates are high (up to 100 times faster than in air) and independent of ΔK.
- There is limited evidence to suggest that the presence of inhibitor reduces crack growth rate. However it is only effective at low ΔK and only if present prior to any significant pre-corrosion.
In sour H2S containing environments
- In crude oil, the presence of H2S leads to a marked increase in fatigue crack growth rate, up to 35 times that in air at high ΔK. The effect diminishes with decreasing ΔK and decreasing H2S concentration.
- In seawater containing H2S, crack growth rates can be higher still, in some cases being two orders of magnitude higher than in air. Again, the effect seems to be most noticeable at moderate or high ΔK.
- Fatigue crack growth rates appear to be no higher at 90C than at 30C, indeed under some circumstances it appears that they may be somewhat lower.
- Stress ratio appears to have only a marginal effect on crack growth rate, although it does have a significant effect on the observed ΔK threshold.
- Frequency again has a significant effect on the observed fatigue crack growth rate.
- There are relatively few published sour endurance data. However, in one study fatigue lives were a factor of 10-20 lower than in air.
- Maddox S J, Pargeter R J and Woollin P: 'Corrosion fatigue of welded C-Mn steel risers for deepwater applications: a state of the art review', OMAE2005-67499, Proceedings of OMAE 2005: 24th International Conference on Offshore Mechanics and Arctic Engineering, ASME, New York, 2005.
- 'Fatigue background guidance document', HSE Report No. OTH 92 930, HMSO, London 1993.
- 'Fatigue strength analysis of offshore steel structures', RP-C203, DNV, Oslo, 2000.
- Draft revision to ISO 13819-2: '1995 Petroleum and natural gas industries - offshore structures - Part 2: Fixed steel structures', ISO/TC 67/SC 7 N222, BSI, London, 1999.
- King R: 'A review of fatigue crack growth rates in air and seawater', HSE Report OTH 511, HSE Books, London 1998.
- BS 7910:2005: 'Guide to methods for assessing the acceptability of flaws in metallic structures', BSI, London, 2005.
- Scott P M et al.: 'Rate determining processes for corrosion fatigue crack growth in ferritic steels in seawater', Corrosion Science, 23 (6), p.559-575.
- de Back J and Vaessen G H G: 'Effect of plate thickness, temperature and weld toe profile on the fatigue and corrosion fatigue behaviour of welded offshore structures: Parts I and II', ECSC Contract No. 7210-KG/601, Final Report EUR 10309 EN.
- Vosikovsky O : 'Fatigue crack growth in an X-65 line-pipe steel at low cyclic frequencies in aqueous environments', Transactions of the ASME, October 1975, p.298-304.
- Baxter D P : 'Corrosion fatigue of C-Mn steels at very low frequencies', TWI Report No. 16219.01/2007/1252.1, January 2007 (Unpublished).
- Helms R and Henke H: 'Corrosion fatigue in offshore - Influence of frequency and electrochemical conditions', Proc. 6th European Conference on Fracture (ECF6), Fracture Control of Structures, Amsterdam, June 15-20, 1986, p.1211-1226.
- Mehdizadeh P et al.: 'Corrosion Fatigue Performance of a Carbon Steel in Brine Containing Air, H2S and CO2 '. Corrosion Official Monthly Publication. National Association of Corrosion Engineers, p. 325-335.
- Ebara at al.: 'Corrosion Fatigue Strength of Steel Welded Joints in Industrial Gas Environments'. The 5th International Symposium of the Japan Welding Society, April 1990, Tokyo, p. 965-969.
- Szklarz K E: 'Aggressive CO2 corrosion and fatigue behaviour of pipeline girth welds', Paper No. 00012, Proc. Corrosion 2000, NACE International, Houston, TZ, 2000.
- Eadie R L et al.: 'Corrosion fatigue and near-neutral pH stress corrosion cracking of pipeline steel in very dilute carbonate/bicarbonate with and without the presence of hydrogen sulphide using the compliance technique', Paper No. 03527, Proc. Corrosion 2003, NACE International, Houston, TZ, 2000.
- Vosikovsky O and Rivard A: 'The effect of hydrogen sulphide in crude oil on fatigue crack growth in a pipeline steel', Corrosion, 38(1), p.19-22, January 1982.
- Vosikovsky O et al.: 'Allowable defect sizes in a sour crude oil pipeline for corrosion fatigue conditions', International Journal of Pressure Vessels and Piping, 13, p.197-226, 1983.
- Watanabe E et al.: 'Corrosion fatigue strength of ship structural steel plates and their welded joints in sour crude oil', Offshore Mechanics and Arctic Engineering Conference (OMAE 1994), ASME, 1994, Vol.III, p.151-158.
- Bristoll P and Roeleveld J: Fatigue of offshore structures : effect of seawater on crack propagation in structural steel', Proc. Conf. European Offshore Steels Research, ECSC, 1978.
- Webster S E et al.: 'Fatigue, corrosion fatigue and stress corrosion of steels for offshore structures', ECSC Report No. EUR 9460, ECSC Steel Publications, European Commission, Brussels, 1985.
- Eadie R C and Szklarz K E: 'Fatigue crack propagation and fracture in sour dilute brine', Paper No. 611, Proc. Corrosion 99, NACE International, Houston, TX, 1999.
- Eadie R C and Szklarz K E: 'Fatigue initiation and crack closure of low alloy steels in sour brine environments', Paper No. 610, Proc. Corrosion 99, NACE International, Houston, TX, 1999.
- Buitrago J et al.: 'Effect of loading frequency on fatigue performance of risers in sour environment', Paper 51641, Offshore Mechanics and Arctic Engineering Conference (OMAE 2004), Vol.III Materials Engineering, ASME, 2004.
- Buitrago J and Weir M S: 'Experimental fatigue evaluation of deep water risers in mild sour service', Deep Offshore Technology Conference, New Orleans, November 2002.