Stainless Steel Cladding Disbonding after Hydrogen Service

IIW document Commission IX-2234-07. February 2007.

Summary

The influence of welding process, consumable type and postweld heat treatment (PWHT) on the phenomenon of disbonding was investigated. Small weld clad test blocks were thermally charged with hydrogen in an autoclave. A range of austenitic stainless steel consumables and a Ni-Cr alloy were used to make the claddings on 2.25Cr-1 Mo parent metal by submerged-arc strip and wire, electroslag, plasma hot wire, and manual metal arc processes. The effects of simple and multiple postweld heat treatments at 565-690°C were determined and local hydrogen contents in clad materials were determined both by measurement and also finite element modelling.

Examination of the interface microstructures and disbonding cracks showed that the cracks were located in the compositional/microstructural transition zone within about 100µm of the fusion boundary. In this region, martensite was present after PWHT, giving rise to high hardness and susceptibility to hydrogen embrittlement.

Some disbonding sensitivity was found with all consumable types, but the stainless steel overlays showed appreciably higher susceptibility than the Ni-Cr cladding. Disbonding resistance was higher with manual metal arc (MMA) welding than with the various high deposition rate methods. The microstructure of the bulk first-layer cladding exerted an influence on susceptibility with both high ferrite and high martensite being beneficial. Minimising postweld heat treatment time and temperature reduced disbonding, as did duplex postweld heat treatments with the second stage temperature at or below 650°C.

Computed and measured hydrogen levels indicated that a peak in hydrogen concentration formed near the interface on cooling.

Introduction

Vessels for elevated temperature hydrogen service are typically fabricated from 200mm thick 2.25Cr-1Mo low alloy steel which is weld clad with austenitic stainless steel. Service temperature may be up to about 450°C with a hydrogen partial pressure of up to some 175bar. A number of instances have been reported where disbonding of the cladding has occurred (i.e. local separation of the cladding from the substrate) after some time in service. The disbonding is believed to occur during shutdown, and can arise in vessels such as hydrodesulphurisers, hydrocrackers, heat exchangers, and vessels in coal conversion plant.

Although there have been no serious problems associated with disbonded vessels in service, disbonding has been viewed with some concern because, while the probability of service failures and resulting unscheduled shutdowns appears to be low, the possibility cannot be completely ruled out. A number of studies into the phenomenon have been carried out (e.g. 1-7), and it is generally agreed that the problem stems from a hydrogen embrittlement mechanism, hydrogen diffusing through the cladding (and vessel wall) in service and causing cracking on cool down to normal ambient temperature. However, it had not been possible to define, unambiguously, either the mechanism of cracking or the roleplayed by material, welding and postweld heat treatment variables in determining its occurrence. In view of this situation, a project was initiated to generate appropriate data with the aim of defining measures enabling disbonding to be avoided.

Objectives

• To evaluate the influence of welding process and procedural variables on interface disbonding.
• To determine the optimum combination of consumables and PWHT conditions for resistance to interface disbonding.
• To elucidate the mechanism of disbonding.

Experimental approach

Weld clad specimens were made using processes and procedures representative of industrial practice. A range of consumable types was used with postweld heat treatment carried out with varying time and temperature conditions. The samples were charged with hydrogen, in an autoclave at elevated temperature, and the extent of disbonding was assessed following cool down.

Metallurgical examination was carried out on the microstructure of the interface region and on the morphology of cracking developed. Modelling of hydrogen diffusion was employed to examine the local hydrogen levels reached at different locations within both the samples and pressure vessel walls.

Experimental programme

Materials

Parent 2.25Cr-1Mo steel was obtained with a composition representative of normal production. The analysis is given in Table 1. The bulk of the work involved austenitic stainless steel claddings as indicated in Table 2. It will be noted that one and two layer claddings were investigated, the consumable types being selected to produce a range of deposit compositions and microstructures. A nickel-chromium consumable was used for comparative purposes.

Table 1 Chemical analyses of first layers of deposited claddings and parent steel

SAW = Submerged-arc strip welding
   (1) low current density conditions
   (2) high current density conditions
MMA = Manual metal arc welding
ESW = Electroslag welding
HS = High speed submerged arc-strip welding
PHW = Plasma hot wire welding
SASA= Submerged-arc series arc welding
- = not analysed

Table 2 Claddings tested

Welding processes

Much of the work involved the submerged-arc strip cladding method in view of the widespread industrial use of this technique. Variants were employed, including low and high current densities and a high speed version of the technique. In addition, test blocks were produced using the following processes:

1. manual metal arc cladding
2. plasma hot wire cladding
3. electroslag cladding

A summary of welding conditions used is given in Table 3.

Table 3 Welding conditions

Postweld heat treatment (PWHT)

The following aspects of heat treatment were investigated:

1. heat treatment temperature
2. heat treatment time
3. multiple heat treatment with a high finishing temperature
4. multiple heat treatment with a low finishing temperature

A standardised PWHT of 30 hours at 690°C was employed for all cladding types. In addition, samples were heat treated at temperatures between 650 and 690°C for times of 1-30 hours. Data generated in the course of the investigation suggested that multiple PWHT could be of benefit in reducing the incidence of disbonding, and studies were therefore carried out with two stage tempering at temperatures of 690°C for the first stage and565-640°C for the second stage.

Hydrogen charging and cracking assessment

Other work had shown that disbonding could be induced by isothermal charging of clad samples in high pressure hydrogen at temperatures between 400 and 500°C followed by fairly rapid cooling to room temperature. ^[1-6] Following such work, samples with dimensions indicated in Fig.1 were produced from the postweld heat treated clad blocks. {It is noted that this work was carried out prior to the introduction of the cylindrical test samples specified in ASTM G146-01 ^[8] } These samples were charged at different hydrogen pressures, up to 190bar at a standard temperature of 425°C for a time of 3 days, the time being selected on the basis of available hydrogen diffusion data, and intended to promote close to the saturation equilibrium hydrogen content in the test materials. Following charging, the samples were cooled to room temperature, normally at a 'standard' rate of ~135°C/h. Selected samples were cooled at different rates for comparative purposes.

Fig.1. Test block geometry, dimensions in mm. Clad on one surface only as shown above

The extent and location of interface disbonding was assessed initially by ultrasonic testing. Metallographic sectioning was used to confirm the ultrasonic assessment of crack development.

Metallurgical studies

Detailed metallographic examination of the clad/parent metal interface was carried out for the different samples using light and, where necessary, electron microscopy. Microanalysis was performed using an electron probe micro analyser and an energy-dispersive X-ray analyser attached to a scanning electron microscope.

Hydrogen diffusion

Hydrogen concentration profiles, which evolved during the autoclave charging of the test blocks and during shutdown of a reactor vessel with a 200mm thick wall clad with 8mm of austenitic stainless steel, were calculated using finite element analysis to model diffusion. This information was derived as a function of temperature and time after initiation of cooling using 1 and 2 dimensional solutions to the diffusion equations for the vessel wall and test blocks, respectively. In order to model the effect of the relative changes in hydrogen solubility between the parent metal and the weld cladding during cooling, super saturations, rather than concentrations, were employed, as described by Sakai et al. ^[3] In addition, direct measurement was undertaken of hydrogen levels in the charged blocks by cooling them rapidly to subzero temperatures and machining samples from the different locations using a cold cutting technique, so that hydrogen loss was minimised. Hydrogen contents were determined using a vacuum hot extraction method.

Results

The results of the individual disbonding tests are plotted in Fig.2 in terms of area % disbonded against hydrogen charging pressure in the autoclave. The charging temperature was the same in all runs, namely 425°C. In Table 4, the disbonding susceptibility is characterised in terms of threshold pressure to initiate disbonding and the area % disbonded after charging at 150bar.

Fig.2. Results of disbonding tests:

a) All results for 309L/347 submerged-arc strip cladding using low current density welding conditions

b) The effects of different welding processes using the 309L/347 combination of consumables    

 c) Results for submerged-arc strip cladding using different combinations of consumables

d) Effects of single and double layer claddings for submerged-arc strip welding

e) Effect of heat treatment on 309L/347 submerged-arc strip cladding (high current density)

f) Effect of heat treatment on 309Nb series arc cladding 

g) Effect of heat treatment on 309Nb plasma hot wire cladding

h) Effect of postweld heat treatment on 309L/347 submerged-arc strip cladding (low current density)

Table 4 Criteria for disbonding susceptibility for all claddings tested

Welding process

Manual metal arc welding was the most resistant process. The high deposition rate processes tended to be grouped together but, for the particular process variants tested in this project, electroslag had the lowest susceptibility and conventional submerged-arc strip and wire processes the highest sensitivity to disbonding.

Consumable type

The Ni-Cr alloy cladding showed the highest disbonding resistance of all the claddings tested. The conventional stainless steel strip consumables, with and without niobium, had a similar susceptibility, while the 309LMo consumable had the highest susceptibility and the lower alloy content 308L first-layer produced a more resistant deposit. For the high deposition rate processes, there were a trends of increasing disbonding resistance with increasing ferrite in the bulk as-deposited micro structures of the first-layer claddings, Fig.3, and increasing cladding thickness, Fig.4.

Fig.3. Effect of as-deposited ferrite content (IIW Ref. Atlas) in the first-layer claddings (equal to or greater than 7mm in thickness) on disbonding susceptibility

Fig.4. Effect of cladding thickness on disbonding susceptibility for 309L, 309Nb and 347 first-layer consumables

Postweld heat treatment

The results in Table 5 and Fig.2e show that decreasing the PWHT temperature or time decreased the amount of disbonding and the data in Fig.5 show that the amount of disbonding was approximately proportional to the log of the PWHT time. The results for the single and multiple heat treatments (finishing at 690°C) applied to the submerged-arc series arc cladding fall in the same scatterband, Fig.2f and, although the multiple heat treatment applied to the plasma hot wire cladding did cause less disbonding, Fig.2g, this is probably attributable to the lower total time at temperature for the multiple PWHT. Using a lower PWHT temperature in the range 565-640°C, for the second stage of a duplex heat treatment, decreased disbonding susceptibility. Times of 5 and 16 hours were used for this second stage of heat treatment, but the results of both the single and duplex heat treatments were consistent with the final heat treatment temperature being the controlling variable, whether in single or duplex heat treatments, Fig.6.

Table 5 Effect of postweld heat treatment temperature on disbonding of 309L/347 submerged-arc strip cladding deposited using low current density welding conditions

Fig.5. Effect of time of postweld heat treatment at 690°C on area % disbonded for 309L/347 submerged-arc strip (high current density) deposits tested using the standard procedure after charging at 425°C and 80bar hydrogen pressure

Fig.6. Effect of final postweld heat treatment temperature on the area % disbonded of submerged-arc strip 309L/347 deposited under low current density conditions and testing using the standard procedure after charging at 425°C and 150bar hydrogen pressure

Metallography

The cracking observed in the test blocks was similar to that observed on a sample of service disbonding from a reactor vessel, Fig.7. For the most part, it was located in the cladding within 100µm of and approximately parallel to the fusion boundary, although very occasional excursions into the heat affected zone were observed, Fig.8.

Fig.7. Disbonding cracks from a reactor vessel wall clad by submerged-arc strip welding using 309 consumable, etched electrolytically and in 2% nital

Fig.8. Disbonding cracking from an autoclave test block, prepared from 309L manual metal cladding, etched electrolytically and in 2% nital

The microstructure in the region where cracking was located, Table 6, was complex and, following PWHT, four microstructural zones could be identified ( Fig.9). viz:

1. At the interface there was a region of tempered and carburised martensite.
2. A narrow band of virgin martensite was present on the edge of the tempered and carburised martensite zone.
3. The next region moving towards the cladding was a light-etching zone in which grain boundaries were heavily delineated by carbide precipitates.
4. From the position where weld metal delta ferrite was observed, there was a further region heavily precipitated with carbide penetrating approximately 100µm into the cladding.

Table 6 Measurements of various cladding characteristics

SAW = Submerged-arc strip welding
   (1) low current density conditions
   (2) high current density conditions
MMA = Manual metal arc welding
ESW = Electroslag welding
HS = High speed submerged arc-strip welding
PHW = Plasma hot wire welding
SASA= Submerged-arc series arc welding

* IIW Ferrite Reference Atlas
- = not determined
Blank= not identifiable
+ = as measured on a transverse section

Fig.9. Microstructural zones at the cladding interface of 309L/347 submerged-arc cladding made using low current density welding conditions and postweld heat treated for 30h at 690°C:

a) Etched electrolytically in 2% nital   

b) Electrolytically polished

In material heat treated at 690°C, microhardness in excess of 400µHV was observed in the tempered and carburised martensite. After a second stage PWHT at 600°C, the maximum microhardness of 309L/347 SAW cladding, measured in this band, fell from 412 to 376µHV. However, there was no overall correlation between the microhardness measured in the interface regions and the disbonding susceptibility.

These observations were general for the stainless steel overlays, the major variation between processes being that the fusion boundary geometry was very much more irregular with MMA welding than with the other methods. In the first-layer stainless steel claddings, the bulk weld metal microstructure was predominantly austenite and ferrite ( Table 6), with some degree of sigma formation after PWHT. The highest as-deposited ferrite content was found with the electroslag overlay, while substantial martensite was noted in the first-layer strip cladding produced with a 308L consumable. To some extent, similar comments can be made for the nickel alloy cladding. However, in this case the compositional gradient across the interface was very much steeper and the extent of martensite formation was, consequently, greatly reduced, Fig.10.

Fig.10. Typical microstructure for the Ni-Cr consumable, after postweld heat treatment, show ing very narrow band of carbide precipitation at the fusion line, electrolytically polished

Hydrogen distributions

The calculated distributions of hydrogen in standard test blocks 3 hours after reaching ambient temperature, are given in Fig.11 for charging at 150bar and 50bar with the standard test conditions. It was predicted that hydrogen would accumulate in cladding near the interface and a larger peak would be generated when a higher hydrogen pressure was used. The model predicted that a similar distribution would be formed when a reactor vessel wall was cooled to ambient temperature, Fig.12.

Fig.11. Hydrogen distributions through the thickness of the centre of an autoclave test blocks 6.7h after charging at:

a) 50bar

b) 150bar

Fig.12. Predicted hydrogen distribution through the thickness of a vessel wall 16.2h after initiation of shutdown. The calculation assumes service at 425°C and 150bar hydrogen cooling at 25°C/h to 20°C

Direct measurements of hydrogen were made near the cladding interface of an autoclave test block. The results are compared with the calculated distribution in Fig.13. It can be seen that hydrogen did build up in the cladding, near the interface, and that the measured values were higher than those calculated.

Fig.13. Hydrogen concentration close to the cladding interface:

6.7h after initiating cooling from charging at 425°C and 150bar hydrogen pressure:

a) Points plotted represent the mean concentration in slices of the indicated thickness

b) Calculated profile at the interface

Discussion

Mechanism of disbonding

From the present and other work it is clear that disbonding is a manifestation of hydrogen embrittlement. Cracking occurs at temperatures close to normal ambient, with a clear incubation period, and it did not occur in laboratory tests in which helium was substituted for hydrogen in the autoclave.

Other workers have drawn attention to the potentially embrittling effect of intergranular impurities ^[3] and carbide precipitation ^[7] on disbonding susceptibility. In the current work intergranular segregation was not investigated and no obvious effect of niobium stabilisation was seen. Since disbonding is predominantly intergranular, any mechanism causing grain boundary embrittlement will presumably have a deleterious effect on resistance to disbonding. However, metallographic observation suggests that cracking initiates in the tempered/carburised material along the fusion line and was virtually confined to the compositional/microstructural transition zone at the interface between the cladding and the parent steel. This region has a complex microstructure within which hardnesses, in excess of 400µHV, develop after normal PWHT. Such hardness levels clearly indicate the presence of martensite. Metallographic observation suggests that cracking initiates in the tempered/carburised material along the fusion line and four 'sources' formartensite formation within this transitional zone may be invoked:

1. Martensite forms on cooling from welding close to the fusion boundary where the alloy content of the weld metal is reduced by dilution with parent metal.
2. High alloy dilution regions which just transformed to martensite on cooling from welding may re-austenitise during PWHT, reverting to virgin martensite on cooling from PWHT.
3. Carbon migration during PWHT from the parent metal into the cladding may destabilise austenite by causing chromium carbide precipitation with consequent elevation of the Ms temperature, again, resulting in martensite on cooling after PWHT. ^[9]
4. A further possible mechanism for martensite formation could be hydrogen-induced transformation caused by the high hydrogen concentrations which develop at the interface during and subsequent to hydrogen charging. In this case, martensite formation would occur on cooling from hydrogen charging.

Although much of the disbonding was within identifiable martensitic regions, intergranular cracking also developed in material in which martensite was not clearly evident on etching. These regions may nevertheless be martensitic but it is also possible that the martensite is confined to the grain boundaries, having formed by mechanisms 3 and 4 above, as a result of more rapid diffusion of carbon (and hydrogen) along the grain boundaries. In this respect, it is remarked that formation of martensite at the fusion line effectively shifts the interface between ferritic and austenitic material so that, as far as hydrogen distribution is concerned, the peak concentration was predicted to occur inaustenitic material {possibly regions 3 and 4 above where cracking has been observed ^[3] }. The levels of hydrogen which are achieved in this austenitic material are higher than can be produced in homogeneous samples but even tests on material, with much lower hydrogen levels, ^[7] suggest that material of the composition of regions 3 and 4 would be susceptible to intergranular fracture.

Disbonding cracks form approximately parallel to the fusion boundary. It is difficult to identify a stress system which could act to cause the observed cracking. However, residual stresses exist in the clad materials being considered in the form of long range stresses in the plane of cladding and short range stresses at the weld bead overlap positions. Some correlation between disbonding and overlap positions was noted, but, in view of the extensive propagation in susceptible material, it can be presumed that both these sources of stress influence the incidence of disbonding.

Factors influencing disbonding

Welding process variables

The testing procedure has enabled ranking of the various processes employed in terms of disbonding susceptibility. The most obvious distinction to be made is between the high resistance of manual metal arc cladding and the relatively inferior performance of all the high deposition rate processes. However, within the latter group there is a gradation in performance between conventional submerged-arc strip cladding processes and the improved resistance of the particular variant of electroslag cladding which was tested and the high speed submerged-arc strip cladding technique. The wire consumable methods used gave results broadly comparable with those of the conventional stripsubmerged-arc tests.

Considering the strip submerged-arc process, the test results in Fig.2 show there to be some dependency of disbonding on welding conditions. However, it would not seem that the phenomenon can be obviated entirely by control of welding conditions alone. This is consistent with the concept that the 'partially mixed' or 'unmixed' zone at the fusion lines of dissimilar welds cannot be avoided, ^[10] although other workers have shown some influence of preheat ^[11] and cooling rate ^[10] on the extent of the zone.

Accepting the above model of disbonding, some sensitivity to disbonding will be inherent with weld-deposited and heat treated cladding, so that a general similarity of results from the high deposition rate processes could well beexpected. The higher disbonding resistance of MMA claddings is therefore striking. From Table 6, this cannot be directly associated with changes in deposit composition stemming, perhaps, from different bulk dilution with the MMA process, although there may well be variations in the compositional gradient across the interface between MMA welding and the other overlaying methods. However, a major difference between MMA and the other processes lies in the interface geometry. For MMA, it was very much more irregular and it is considered probable that this was the major cause of the higher disbonding resistance with this process, possibly as a result of its restriction of lateral crack extension.

Even within single process/consumable type, the results show variation, the extent being at least comparable to that associated with different high deposition rate methods. This would seem to be associated with differences in deposit composition and microstructure as considered below, indicating that choice of high deposition rate welding method is not overriding in determining disbonding sensitivity.

Consumable type

The cladding made with the Ni-Cr alloy consumable had a much lower sensitivity to disbonding than any of the stainless steel variants which were tested. This difference in behaviour has also been observed by other workers ^[3,6,12] and can be linked to a clear difference in microstructure: the steeper compositional gradient across the interface, in the case of the Ni-Cr cladding, modified the microstructure, greatly reducing the extent of martensite formation. Despite this, other workers have positively identified a thin band of martensite at similar interfaces. ^[13]

Considering the stainless steel consumables, the major distinction was between those claddings giving a high ferrite or martensite content in the bulk cladding and those in which the matrix was austenitic with circa 5-10% ferrite. In agreement with the results of other workers, ^[14-16] disbonding. resistance was increased by higher ferrite levels or the presence of 'bulk' martensite but no particular effect of niobium stabilisation was apparent. The molybdenum-containing deposit, produced with 309LMo strip, did appear to show somewhat higher sensitivity than other types, but this was also associated with a lower ferrite content.

Postweld heat treatment (PWHT)

It has been shown that disbonding can occur in both as-welded and postweld heat treated microstructures, Fig.2e. From both the present and other studies, ^[2,4,17] sensitivity to disbonding is clearly increased with extended time or more elevated temperature of postweld heat treatment and this would be expected, since martensite formation would be promoted either by increased austenitereversion or destabilisation during heat treatment.

During postweld heat treatment, the original martensite at the fusion boundary clearly becomes tempered. Equally clearly, the heat treatment cycle can induce the formation of fresh martensite and trials were carried out to indicate the extent to which secondary martensite could be tempered by a two stage heat treatment cycle. No benefit in disbonding resistance was found using a second temperature higher than the first. However, a considerable reduction indisbonding sensitivity developed when the second heat treatment was at a lower temperature, Fig.6. This benefit, which has been confirmed by other workers, ^[18] probably stems from two causes: firstly, the tempering of the martensite present close to the fusion boundary is achieved and, secondly, the low temperature minimises the formation of fresh martensite by either reversion oraustenite destabilisation.

Cladding thickness

Consideration of hydrogen diffusion through the wall of a vessel in service indicates that cladding of increased thickness will tend to reduce the peak hydrogen content developed at the interface on shutdown. The laboratory trials on small test blocks showed that increased cladding thickness reduced the extent of disbonding, Fig.4. However, the autoclave exposure was designed to saturate sample thicknesses of the range studied with hydrogen and, therefore, it is believed that the measured effect on disbonding was an artefact of autoclave testing of small test blocks, associated with the reduction in tensile residual stresses in those test blocks with thicker claddings. Despite this, in practice, increased cladding thickness may have some effect because the reduced steady state hydrogen concentration gradient ^[19] will lower the amount of hydrogen near the interface and may, therefore, reduce the peak concentration developed on shutdown.

Cooling conditions

From the present and other studies and practical experience, disbonding can be controlled by restricting the cooling rate from peak temperature. This would be anticipated on the basis of slow cool out permitting hydrogen diffusion away from the interface region. A correlation between disbonding test results and predicted hydrogen interface peak concentrations has been presented by Coudreuse. ^[20]

Implications

There is considerable disagreement in the literature concerning the roles of the variables which affect the risk of disbonding. It is likely that these stem, largely, from differences in test methods employed and in criteria for assessment of disbonding sensitivity. The present study has sought to refine testing technique and assessment to minimise such uncertainties, and two aspects are considered particularly important. The hydrogen charging time used waslong enough to achieve levels of hydrogen close to saturation in the cladding (estimated at 93%). This is important because, when shorter times are used for charging, the actual levels of hydrogen which are attained in the test blocks will be much more sensitive to any small variations in charging conditions, thereby increasing scatter in the results. Secondly, the use of a 'critical threshold pressure' to initiate disbonding is considered a more useful parameter, for either comparative or quantitative purposes, than arbitrary levels of disbonding used by other workers.

A major finding for the present study was that, with austenitic stainless steel consumables and the application of conventional postweld heat treatments, some sensitivity to disbonding in high pressure high temperature hydrogen service is inevitable. Nonetheless, a number of ways of reducing disbonding sensitivity can be identified, although it should be emphasised that all aspects relating to susceptibility should be carefully considered before a particular approach is adopted in practice.

The current results and those of other workers suggest that the following techniques increase disbonding resistance within the range of variables studied:

Ni-Cr cladding
Manual metal arc deposited cladding
Finishing temperature for PWHT below 650°C
High ferrite in the first-layer cladding
Increased cladding thickness
Martensite in the first-layer cladding
Use of vanadium-modified parent steel ^[21-24]

For existing vessels the following procedures are available for minimising susceptibility to disbonding:

Employing hydrogen release treatments before cooling
Applying an additional low temperature PWHT.

In the case of the fabrication of new vessels, all of the above techniques could be used, but for most applications it is considered that the following recommendations will be most efficacious:

Avoiding low ferrite in the first-layer deposits
Applying duplex PWHT
Use of V-modified parent steels

Summary and conclusions

Isothermal hydrogen charging of small weld clad test blocks in an autoclave followed by rapid cooling to ambient temperature has been used to assess the susceptibility of claddings, made using .various welding processes, and consumables to disbonding. The effect of postweld heat treatment has been determined and the mechanisms of disbonding have been investigated by metallographic and microanalytical techniques. The diffusion of hydrogen in test blocks and reactor vessel walls has been modelled using finite element analysis. From the results obtained in this work, the following conclusions may be drawn:

1. Disbonding is due to hydrogen embrittlement and is associated with the presence of martensite in the interface region, the martensite forming both during welding and following postweld heat treatment. Cracking occurs almost exclusively on the cladding side of the interface with only a few isolated excursions (<1 grain diameter) into the parent metal.
2. The Ni-Cr cladding had the lowest disbonding susceptibility of all the process/consumable/heat treatment combinations tested.
3. For stainless steel claddings, manual metal arc welding was the process most resistant to disbonding. As a group, high deposition rate processes showed greater sensitivity to disbonding.
4. The modified structure at the cladding interface associated with high ferrite or martensite contents in the bulk first-layer cladding conferred an increased disbonding resistance in stainless steel overlays deposited by high deposition rate processes.
5. For submerged-arc strip welding with 309L consumables, high current, high travel speed conditions produced claddings with improved resistance to disbonding compared to the slower, lower current welding conditions.
6. For stainless steel claddings produced by high deposition rate processes, welding conditions, consumable composition and resultant microstructure were more important than welding processes per se.
7. Disbonding resistance was improved by postweld heat treatments which minimised (a) time at temperature and (b) the final tempering temperature.
8. As a result of this work it can be concluded that the incidence of disbonding can be reduced, in many cases, by careful selection and control of welding process and consumable. However, experiments involving slow cooling following hydrogen charging were successful in eliminating disbonding in the test blocks employed. This is consistent with the employment of shut down schedules which enable sufficient diffusive removal of hydrogen to avoid disbonding.

Acknowledgements

The following contributed towards the work reported here - D.S. Rhead, R.W. Gant, H. McMenamin, N.J. Farrant, N.A. Kennedy, R.L. Smith, C.R. Dye, K. Butcher, D.F. Pargeter and R.H. Leggatt. The authors would also like to thank the following companies for financial support: BP, CB & I, Chevron, Framatome, JSW, Kawasaki, Kobe, NKK, Nooter, Nuovo Pignone, Shell and Sumitomo.

References

1. Ohnishi K., Chiba R., Watanabe J. Naito K. and Okada H. 'Hydrogen induced disbonding of stainless steel overlay weld'. Presented at PVRC Meeting, New York, September, 1980.
2. Sims, J. E. and Bruck, G. J. 'Factors influencing the performance of stainless steel overlays in high temperature hydrogen service.' Presented at MPC/ ASME Symposium on performance of pressure vessels with clad and overlaid stainless steel linings, Denver, June 1981.
3. Sakai, T., Asami, K., Katsamata, M., Takada, T. and Tanaka, O. 'Hydrogen induced disbonding of weld overlay in pressure vessels and its prevention', 118th Chemical Plant Welding Research Committee of Japan Welding Engineering Society, Tokyo, August 1981.
4. Anon. 'Study of disbonding in overlay weldment.' NKK Technical Research Centre Report, October, 1982.
5. Pressouyre, G. M., Chaillet, J. M. and Valette, G. 'Parameters affecting the hydrogen disbonding of austenitic stainless cladding steels'. In Proc. 'Current solutions to hydrogen problems in steels'. ASM Washington, November, 1982.
6. Nakano, M., Yasuda, K., Imanaka, T. and Nishiyama, N. 'Weld metal composition effective to prevent overlay weld cladding from hydrogen induced disbonding'. Kawasaki Steel Corporation research report, June 1983.
7. Anon. 'Hydrogen embrittlement of bond structure between stainless steel overlay weld and base metals'. Final report of Task Group III Joint Study Subcommittee on Hydrogen Embrittlement Materials.
8. ASTM G 146-01. 'Standard practice for evaluation of disbonding of bimetallic alloy/steel plate for use in high-pressure, high temperature refinery hydrogen service.' ASTM, 2001.
9. Gittos M. F. and Gooch T. G. 'The interface below stainless steel and nickel-alloy claddings.' Welding Journal 71, (12), 1992, 461s-472s.
10. Karlsson L. and Arcini H. 'Formation of unmixed zones in stainless welds.' Innovation Stainless Steel, Florence, Italy October 1993, 3.273-3.278.
11. Omar A. A. 'Effects of welding parameters on hard zone formation at dissimilar metal welds.' Welding Journal 77, (2), 1998, 86s-93s.
12. Lewis K. R., Koers R. W. J., van Wortel H. C., and Krom A. H. M. 'Assessment of operating limits for critical dissimilar welds in a hydrocracker.' Corrosion 2001, Paper 01532, NACE, 2001.
13. Olden V., Kvaale P. E., Simesen P. A., Aaldstedt S. and Solberg J. K: 'The effect of PWHT on the properties and microstructure in Inconel 625 and Inconel 725 buttered joints.' Proc. OMAE 2003, Cancun Mexico, June 2003.
14. Anon. 'On the disbonding characteristics of weld overlay by the Maglay process.' Kawasaki Steel Corporation Research Laboratories Report, 1983.
15. Tanaka O., Takeba K. and Matsashita Y. 'High speed overlay welding with strip electrodes.' Welding review, 3, (2), 1984, 58-62.
16. Ohnishi K., Fuji A., Chiba R. Adachi T., Naito K. and Okada H. 'Effects of strip overlay welding conditions on resistance to hydrogen-induced disbonding.' Quarterly Journal JWS, November 1983, 75-82.
17. Ohnishi K., Chiba R., Murakami Y., Fuji A., Nomura T., Naito K. and Okada H. 'Effect of PWHT conditions on overlay/weldmetal interface microstructure'. Translation of preprints presented at April and October meetings of JWS, Japan Steelworks Ltd., December 1982.
18. Vignes A., Palengat R. and Bocquet P. 'Disbonding mechanism and its prevention.' Interaction of steels with hydrogen in petroleum industry pressure vessel service, MPC, New York, 1993.
19. Johnson E. W. and Hudak S. J. 'Hydrogen embrittlement of austenitic stainless steel weld metal with special consideration given to the effects of sigma phase.' WRC Bulletin 240, August 1978.
20. Coudreuse L. 'Meaning of disbonding tests: results of an European joint research project.' Corrosion 2001, Paper 01533, NACE, 2001.
21. Shimomura J., Nakano Y. and Ueda S. 'Improvement in resistance to disbonding of stainless steel-overlayed 2¼Cr-1Mo steels'. ISIJ International, 4, 1991, 379-386.
22. Brouwer R. C. 'Hydrogen distribution through the wall of pressure vessels made of conventional and V-modified steels.' International Journal of Pressure Vessels and Piping, 56, 1993, 133-148.
23. Cayward M. S., Kane R. D. and Stevens C. E. 'Evaluation of hydrogen disbonding of stainless cladding for high temperature hydrogen service.' Corrosion 94, Paper 518, NACE, 1994.
24. Fusari F., Coudreuse L., Lefebre Y., Crowther M., Ødegård L., and Blaauw A. 'Properties of strip surfaced overlays on the new generation of parent metals for the petrochemical industry.' Welding in the World, 36, 1995, 173-180.