Dissimilar Joint Resistance to Hydrogen-Assisted Cracking

Paper presented at Corrosion 2008, NACE, New Orleans, March 2008, paper 08095.

Abstract

Dissimilar joints between ferritic steels and nickel alloys are used in many subsea applications, typically without problems, but cracking at the interface may arise when hydrogen is present. The paper includes the results of constant load, step load and slow strain rate tests on tensile specimens, with and without notches, and slow strain rate fracture mechanics tests on notched bend specimens, all under cathodic charging in simulated seawater. Specimens were taken from an 8630 forging welded with nickel alloy 625 and 725 consumables.

Introduction

Work was initiated to address concerns raised following the failure of subsea hubs in the North Sea that employed steel to weld-deposited nickel alloy interfaces, in particular, AISI 8630 and alloy 625/alloy 725 combinations. In these reported instances, cracking occurred at the weld interface after the hubs were installed on the seabed and were cathodically protected. A literature review was carried out for information available in the public domain that might be pertinent to the understanding of the failures of dissimilar interfaces in seawater service and in the light of this, work was initiated to develop a suitable test method to reproduce the failures that had been experienced, so that the variables influencing the failure could be explored. This work was part of a group-sponsored project which was run by TWI from January 2003 to April 2006.

Literature review

Approach

Literature searches were performed using a number of databases of abstracts. The searches were limited to references related to dissimilar joints or interfaces between steels and either nickel alloys or stainless steels.

Background

Two principal areas of service of dissimilar joints have concerned investigators: cladding in high temperature hydrogen service and transition joints. The main focus of attention with regard to high temperature hydrogen service has been predominantly with stainless steel and nickel alloy claddings in pressure vessels, and, in particular, hydrogen disbonding of the interface with the steel substrate. In the case of transition joints, much of the published research has been concerned with the understanding of creep behaviour of the interface. No published research has been found to date that is specific to the operation of dissimilar joints in seawater under cathodic protection. However, some of the work concerning hydrogen disbonding of claddings in high temperature hydrogen service is considered pertinent to the current considered work. Generally, two approaches have been adopted to investigate this form of disbonding:thermal charging with hydrogen, typically at temperatures in the region of 400-500°C, which reflect service charging conditions, and electrolytic charging at room temperature. The latter method, whilst experimentally convenient,is not really applicable to the understanding of the behaviour of dissimilar joints in high temperature hydrogen service; the distribution of hydrogen generated at the interface cannot be expected to be the same as that produced by thermal charging, which is more representative of what occurs in service. Pressouyre et al ^[1] have suggested that disbonding from high temperature charging results from a supersaturation of the austenitic phase and a decrease in the diffusivity of hydrogen in austenite as the temperature falls. In this case, cracking was observed in both the martensite layer and at austenite grain boundaries in the cladding. Disbonding from cathodic charging, however, results from a supersaturation due to high external hydrogen activities and consequently the ferritephase adjacent to the interface may also be prone to cracking. Their work demonstrated that the correlation between thermal charging and electrolytic charging was good only in extreme cases, i.e. the worst disbonding was the same for each, which happened to be with type 309L stainless steel submerged-arc strip deposits. Cathodic charging produced cracking in the substrate heat affected zone (HAZ). Grain boundary or martensite cracking in the overlay was produced ononly some occasions. This is contrary to other studies, ^[2-4] which demonstrated failure modes in the cladding, by cathodic charging with hydrogen, similar to those produced by thermal charging. However, despite these differences, which are pertinent to the study of high temperature failures, the studies performed using electrolytic charging are relevant for the current work: they do represent the type of charging that may be experienced when a dissimilar joint is subjected to cathodic protection.

The crack morphology of disbonding failures has been categorised as two types, ^[5] depending on where they appear at the bond interface:

Studies of stainless steel overlays have suggested that strengthening of such boundaries, or in the case of Type II, limiting their occurrence, can improve resistance to disbonding. In this context it should also be recognised that instances of cracking in the substrate HAZ have also been reported in the specific case of electrolytic hydrogen charging. ^[1]

Reduction of the tendency to form 'coarse-grain' boundaries can be achieved at faster cooling rates which suppress planar solidification growth ^[4,5] or by increasing the bulk weld metal ferrite content which eliminates the fully austenitic zone where these features are found. ^[6] Slow cooling can also promote segregation of impurity elements, such as sulphur and phosphorus to grain boundaries. In addition, silicon has been considered to have a negative influence on disbonding resistance by promoting phosphorus segregation. ^[5]

It has been suggested ^[7] that nickel alloy claddings are more resistant to disbonding than their stainless steel counterparts because of the steep microstructural gradient across the former compared with the latter. This produces a narrower martensite and carbide precipitation region. However, the same microstructural features are observed in both alloy types at the interface, so this explanation is not very plausible. Further work is necessary to explain the different responses to disbonding of both alloy types. Possible areas of study which might explain the differences in response include the different hydrogen solubility and diffusion in nickel alloy claddings compared with stainless steel claddings. In addition, differences in thermal expansion coefficient of cladding materials, leading to differences in residual stresses in terms of both magnitude and distribution, may also play an important role.

A further area of investigation in the literature was fabrication hydrogen cracking of the dissimilar interface. This was described by Bailey ^[8] in the context of repair of low alloy steel using austenitic manual metal arc electrodes. Rowe et al ^[9] hypothesised that disbonding in high temperature hydrogen service was due to fabrication hydrogen cracking. The experimental approach adopted was to introduce hydrogen to the shielding gas of single run TIG deposits. Cracking in the cladding interface was produced on straining the sample. It may be noted that multipass welds did not crack. This was attributed to the multiple weld thermal cycles facilitating the diffusion of hydrogen out of the sample. Whilst the hypothesis may be questionable in relation to high temperature service failures, in which there is a considerable body of evidence to support a hydrogen disbonding mechanism by thermal charging of hydrogen, without the need for existing fabrication defects, it raises a valid point for the current work. It highlights that some consideration should be given to the hydrogen content of the deposited weld metal as this may influence the propensity to cracking.

Metallurgy of the steel to stainless steel and nickel alloy interfaces

A variety of effects produce important microstructural changes at the interface between stainless steel or nickel alloy weld metal and the underlying steel substrate. ^[7,10] These include the formation of unmixed zones, ^[11,12] partial mixing, causing the formation of local hard zones, and carbon migration resulting in both decarburisation of the steel HAZ and carbide precipitation within the weld metal. The majority of microstructural studies have concerned stainless steel claddings, representative of the extensive usage of such materials in pressure vessels. Gittos and Gooch described the interface microstructures between both stainless steel and nickel alloy (ERNiCr-3) claddings and 2¼Cr-1Mo steel. ^[10] The stainless steel deposits were made using both submerged-arc and manual metal arc welding techniques; the nickel alloy cladding was deposited using the submerged-arc strip welding technique. It was noted that the strip consumables produced a macroscopically much flatter interface than the rippled appearance of the manual metal arc deposits. However, on a microscopic scale, in both cases, swirls of partially mixed material were present at the interface. The work demonstrated that a compositional gradient exists across such interfaces, the limited dilution achieved at the steel interface resulting in a region of martensite, giving rise to a hard layer. This was present inboth stainless steel and nickel alloy claddings, although the width of the martensite layer in the latter was narrower, as observed by others. ^[9]

Postweld heat treatment of dissimilar joints is commonplace. Gittos and Gooch ^[10] showed that postweld heat treatment had a number of effects: martensite closest to the fusion boundary was tempered; however, hardened material was formed for some distance into the cladding. This hardening response on tempering was due to three factors: carbon migration and carbide precipitation; some regions of the diluted zone had an Ac ₁ temperature below the postweld heat treatment temperature and thus formed austenite during heat treatment which reverted to virgin martensite on cooling; and the carbon migration destabilised adjacent austenitic regions, again causing fresh martensite to form on cooling. Within the substrate, appreciable decarburisation occurred.

Effects of welding variables on hydrogen disbonding

Several investigators have reported on studies of the effect of welding variables on susceptibility to disbonding, primarily in relation to high temperature hydrogen service. ^[1-6,13] Gittos et al ^[13] demonstrated that manual metal arc cladding was more resistant than high deposition rate processes, such as submerged-arc strip cladding and electroslag cladding (Fig.1), although the latter was better than the former. The main difference between the processes that was concluded to have played a role was the geometry of the interface: the manual metal arc interface was much more irregular than that of the high deposition processes, possibly contributing to restriction of lateral crack extension. Both the work of Gittos et al ^[13] and by Sakai et al ^[5] demonstrated that Ni-Cr alloy deposits were more resistant to disbonding than austenitic stainless steel deposits (Fig.2).

Fig.1. Hydrogen disbonding test results showing the effects of different welding processes using type 309L/347 stainless steel combinations ^[13]

Fig.2. Hydrogen disbonding test results for submerged-arc strip cladding using different combinations of stainless steel and nickel alloy consumables ^[13]

It has been demonstrated that both as-welded and postweld heat treated samples exhibited disbonding. ^[13] Sensitivity to disbonding increased with increased time and temperature of postweld heat treatment. This may be understood as both may result in martensite formation, either by increased austenite reversion or destabilisation during heat treatment. It might be concluded that postweld heat treatment should be avoided. However, in most practical applications, postweld heat treatment is essential for the integrity of the substrate. The work of Gittos et al ^[13] did demonstrate that the application of a duplex heat treatment, with the second thermal cycle at a lower temperature than the first, could be beneficial ( Fig.3). This benefit was attributed to two factors: tempering of the fusion boundary martensite and that the low temperature minimised the formation of fresh martensite by either reversion or austenite destabilisation.

Fig.3. Effect of final postweld heat treatment temperature on disbonding of submerged-arc strip cladding using type 309L/347 stainless steel ^[13]

Discussion

Whilst the operation of subsea dissimilar joints does not provide a mechanism for thermal charging of the material with hydrogen, the application of cathodic protection does provide a hydrogen charging mechanism at ambient temperature. Work has demonstrated that electrolytic charging with hydrogen can produce disbonding of a similar nature to that observed in high temperature hydrogen service. ^[2-4]

Most research on hydrogen disbonding has concentrated on the effects of welding processes for stainless steel weld cladding. Only limited work has been performed on nickel-alloy claddings, which have generally performed better in disbonding tests than their stainless steel counterparts. In general, manual metal arc deposited claddings have performed better than high deposition rate processes, although the reasons for this are not fully explained in the literature. No work has systematically studied the effects of welding variables, e.g. extent of dilution, or postweld heat treatment procedure on disbonding. It would also be logical for any future study to give consideration to the influence of the hydrogen content of the deposited weld metal. ^[8,9]

Most disbonding tests are generally concerned with whether or not disbonding occurs, which makes the measurement of different parameters for ranking purposes problematic. However, constant or rising load smooth or notched tensile tests provide possible approaches. It is recognised that fracture mechanics type approaches using notched bend tests would also provide a metric with which to assess performance. However, a fracture mechanics approach presupposes the presence of a flaw or defect at the fusion line, which may be an incorrect assumption.

Conclusions

The following conclusions, pertinent to the failure of dissimilar joints in subsea applications, may be drawn:

1. The disbonding literature is primarily concerned with thermal charging, with hydrogen, of stainless steel interfaces. Limited studies have been performed on nickel alloy weld claddings. The differences in disbonding behaviour of stainless steel and nickel alloy weld claddings are not fully understood.
2. There is little published information available on fracture of dissimilar joints under cathodic protection and no data to indicate the most suitable test method.

Development of a test procedure

Materials

Materials were employed from three reclaimed North Sea hubs, which had experienced cathodic protection on the seabed and were of similar type to those which failed in service. Two samples incorporated alloy 625 (ERNiCrMo-3) weld metal deposited on an AISI 8630 (modified) steel substrate. The other material incorporated alloy 725 (ERNiCrMo-15) weld metal deposited on a 2¼Cr-1Mo steel (F22) substrate. The buttering layers were deposited by mechanised hot-wire pulsed TIG welding, with a minimum preheat temperature of 150°C, on to a square edged preparation, machined on a forged pipe section. Postweld heat treatment was carried out at 665°C. The results of chemical analyses on substrates and weld metals are shown in Table 1.

Table 1A - Chemical analyses of the hub substrate materials.

Testing Range of tests. In the course of test development, three basic designs of test specimen were employed. These comprised slow strain rate tensile, constant and rising load notched tensile and square-section bend with EDM notches. The specimen geometries are given in Fig.4. The notches in the tensile specimens were cut such that the buttering weld interface was 'slightly inclined' to the notch. This angle was measured on a metallographic section to be 3°. The application of these tests is described in the succeeding sections.

Fig.4. Test specimen geometries:

a) Slow strain rate tensile   

 b) Constant load/staircase load notched tensile   

c) Bend

Slow Strain Rate Testing

Following the literature review, it was decided that, initially, constant load exposure followed by slow strain rate testing to failure should be employed to confirm that the failed hub fracture appearance could be reproduced. A total of three slow strain rate specimens were machined from each of the material combinations. The specimen dimensions are shown in Fig.4(a). In each case, a notch was placed at the interface between the alloy steel forging and the buttering, which was located at the centre of the gauge length. To ensure that the base of the notch would intersect the weldfusion boundary, the samples were machined such that the weld interface was slightly inclined, as shown in Fig.4(a).

Duplicate specimens were tested in seawater at ambient temperature under cathodic protection at -1200mV _SCE . Two alloy 625 buttered specimens were loaded to a nominal stress of 500MPa (73ksi), calculated on the reduced cross-section at the root of the notch, and held under constant load for 15 days. One specimen was then unloaded and examined for any signs of cracking, both with the aid of a low power binocular light microscope and by sectioning. The second specimen of the pair was slow strain rate tested to failure. The strain rate selected was 1x10 ^-6 s ^-1 . Alloy 725 buttered samples were tested in a similar manner, but loaded to 400MPa (58ksi) nominal stress and held for a period of 11 days before slow strain rate testing to failure. For comparative purposes, a further specimen of each material was tested to failure in air at the same strain rate.

The results of the slow strain rate tests on notched specimens, tested in air and seawater are summarised in Table 2. In each case, samples tested in seawater failed at a lower stress than those tested in air. The alloy 625 and 725 samples fractured at nominal stresses of 794 (115) and 749 (109)MPa (ksi), respectively, in seawater under cathodic protection compared with 1174 (170) and 1109 (161)MPa (ksi) in air.

Table 2 - Slow strain rate test results

Visual inspection and metallographic sectioning of the unloaded samples from each hub exhibited no evidence of cracking after the initial holding period. The fractures in seawater in both the alloy 625 and 725 clad samples were macroscopically flat with regions of fracture identified close to the buttering fusion line, e.g. Fig.5 . The fracture faces of the samples tested in air were consistent with ductile fracture. No evidence of any flat fracture at the weld interface was observed.

Fig.5. Photomacrograph of a section taken through the alloy 625 clad 8630 hub slow strain rate test fracture face, tested in seawater at -1200mV _SCE

Notched tensile tests. Initially, a series of rising ('staircase') loading tests, with holding periods of three days, was carried out. The results were used to define loading conditions for sustained load tests. The samples were tested in seawater at ambient temperature under cathodic protection at -1100 mV _SCE . No pre-charging under cathodic protection was performed prior to loading. The samples were held at constant load for three days before raising the stress to 80% of the 'cross-weld proof stress'. The load was increased every three days until fracture occurred.

The first alloy 625 sample failed at the notch location at a nominal stress of 981MPa (142ksi). The second alloy 625 sample, which was intended to be held at constant load for seven days, failed at a lower nominal stress of 927MPa(134ksi) within two hours of the commencement of the test. In the case of the alloy 725 sample, fracture occurred at a nominal stress of 995MPa (144ksi), but in the steel threaded region of the sample. The fractures exhibited regions of flat fracture of similar appearance to those observed in the slow strain rate tests, Fig.6.

Fig.6. SEM secondary electron image of alloy 625 clad 8630 hub notched tensile sample, loaded to 927MPa, with failure within 2 hours. Tested in seawater at -1100mV _SCE . Nominal magnification is indicated by the micron marker 

The results of the sustained load notched tensile tests on alloy 625/8630 are given in Table 3 and they are presented graphically in Fig.7.

Table 3 - Results of constant load, notched tensile tests on cross-weld specimens from 8630 forging clad with alloy 625 (10mm nominal diameter gauge section). Specimens were cathodically protected at -1100mV _SCE in seawater.

Fig.7. Failure data for notched tensile tests on cross-weld specimens from 8630 forging clad with alloy 625 (10mm nominal diameter gauge section). Specimens were cathodically protected at -1100mV _SCE in seawater (arrows indicate specimen did not fail)

Slow bend tests. 12mm (0.5in) square section bend specimens were machined from the second hub sample (designated W02, alloy 625 buttering/AISI 8630 steel). After machining, the coupons were etched in 2% nital to reveal the fusion line between the alloy steel and nickel alloy weld metal buttering to facilitate marking of the sample for electro discharge machining (EDM) of a 6mm (0.24in) deep notch, in the through-thickness direction using a 0.25mm(0.01in) diameter wire.

Tests were carried out in either 3 or 4 point bending at slow testing rates corresponding to crosshead displacement rates of 10 ^-5 - 10 ^-6mms ^-1 (4 x 10 ^-7 - 4 x 10 ^-8ins ^-1) at ambient temperature. Tests were carried out both in air and seawater under cathodic protection at -1100mV _SCE. These slowly loaded bend test results are summarised in Table 4. It can be observed that maximum stress intensities sustained in air were higher (50.6, 60.8MPam ^½ (1.47, 1.76ksi in ^-½)) than those in seawater under cathodic protection (45.5-49.3MPam ^½ (1.32-1.43ksi in ^-½)).

Table 4 - Results of slow bend tests on 12mm (0.5in) square, EDM interface notched samples

A typical fracture surface, from a sample tested in seawater under cathodic protection, is shown in Fig.8. It is apparent that crack growth occurred close to the fusion line and detailed examination by scanning electron microscopy and sectioning showed that there were large areas of fracture within a few microns of the fusion line, Fig.9.

Fig.8. Fracture surfaces of slow bend test W02-06 (alloy 625 on 8630 tested in seawater under cathodic protection), after heat tinting before breaking open

Fig.9. Metallographic section cut through the fracture surface of slow bend sample W02-06 (alloy 625 on 8630 tested in seawater under cathodic protection), etched in 2% nital

Sustained load bend tests.The slowly-loaded tests were used as a guide for the selection of sustained loads for the testing of similar geometry samples (12mm (0.5in) square, 6mm (0.24in) EDM notched on the fusion line) in ambient temperature seawater under cathodic protection at 1100mV _SCE . The samples were loaded rapidly in 3 point bending using static loading jigs which allowed the vertical displacement of the sample to be monitored at one end. The results of these tests are summarised in Table 5.

Table 5 - Results of constant load bend tests on 12mm (0.5in) square, EDM interface notched samples tested in seawater at -1100mV _SCE and ambient temperature.

Two samples loaded to produce a stress intensity of 45.7MPam ½ (1.33ksi in ^-½ ) at the notch tip failed after 5.6 and 36.5h. Samples loaded to produce a stress intensity of 43.4MPam ^½ (1.26ksi in ^-½ ) at the notch tip survived test periods of 163 and 400h without failure, despite some time-dependent displacement occurring. The fracture surfaces of the sustained load samples were similar to the slow bends.

Conclusion

Although all the tests were capable of reproducing the desired fracture morphology, when carried out slowly, the tensile tests produced considerable scatter in the results, such that there was no trend between failure time and applied stress. It is believed that this was, at least in part, related to the small size of the tested area, after notching. The use of the EDM notch in the bend specimens sampled more weld beads, reliably, and monitoring CTOD provided a useful comparison between the ease of crack extension in air and in seawater under cathodic protection. On this basis, the slow bend test was identified as the most suitable for testing such interfaces.

Acknowledgements

The author would like to thank Dr A J Leonard for his assistance during his employment by TWI and Vetco Gray Inc, BP, Chevron Energy Company, Cooper Cameron (UK) Limited, FMC Technologies Inc, Shell International and Statoil ASA fortheir financial support.

References

1. Pressouyre G M, Chaillet J M and Valette G: 'Parameters affecting the hydrogen disbonding of austenitic Stainless cladded steels' Proc Conf 'Current solutions to hydrogen problems in steels', 1st International Conference, 1982, 1-5 November Washington DC, 349-355.
2. Matsuda F, Nakagawa H, Tsumata S and Yoshida Y: 'Disbonding between 2¼Cr-1Mo steel and overlaid austenitic stainless steel by means of electrolytic hydrogen charging technique' Trans JWRI 1984 13 (2) 263-272.
3. Matsuda F and Nakagawa H: 'Simulation test of disbonding between 2¼Cr-1Mo steel and overlaid austenitic stainless steel by electrolytic hydrogen charging technique'. Trans JWRI 1984 13 (1) 159-161.
4. Alexandrov O A, Steklov O I and Alexeev A V: 'Use of plasma arc welding process to combat hydrogen metallic disbonding of austenitic stainless steel claddings'. Welding Journal Research Supplement 1993 72 (11) 506s-516s.
5. Sakai T, Asami K, Katsumata M, Takada H and Tanaka O: 'Hydrogen induced disbonding of weld overlay in pressure vessels and its prevention' Proc Conf 'Current solutions to hydrogen problems in steels', 1st International Conference, 1982, 1-5 November, Washington DC, 340-348.
6. Imanaka T, Shimomura J, Nakano S and Yasuda K: 'Hydrogen attack in Cr-Mo steels and disbonding of austenitic stainless weld overlay', Kawasaki Steel Technical Report 1985 13 109-119.
7. Gooch T G: 'Corrosion of dissimilar joints ... a cracking tale', TWI Bulletin 1999 40 (5) 73-76.
8. Bailey N B: 'Hydrogen cracking and austenitic electrodes'. Metal Construction 1978 12 580-583.
9. Rowe M D, Nelson T W and Lippold J C: 'Hydrogen-induced cracking along the fusion boundary of dissimilar metal welds', Welding Journal Research Supplement 1999 78 (2) 315-375.
10. Gittos M F and Gooch T G: 'The interface below stainless steel and nickel-alloy claddings', Welding Journal Research Supplement 1992 71 (12) 461s - 472s.
11. Doody T: 'Intermetallic mixed zones in dissimilar metal welds for sour service', Welding Journal 1992, 71 (3) 55-60.
12. Karlsson L and Arcini H: 'Formation of unmixed zones in stainless steel welds', Proc Conf Innovation Stainless Steel, 11-14 October 1993, Florence, Italy. Volume 3, 273-278.
13. Gittos M F, Robinson J L and Gooch T G: 'Disbonding of austenitic stainless steel cladding following high temperature hydrogen service'. IIW Document IX-H-653-07, February 2007.