Hydrogen Cracking - Its Causes, Costs and Future Occurrence
P H M Hart
Weld Metal Hydrogen Cracking in Pipeline Girth Welds, Proc. 1st International Conference, Wollongong, Australia, 1-2 March 1999. Published by Welding Technology Institute of Australia (WTIA), Silverwater, NSW, Australia, 1999
The paper looks at the problem of hydrogen cracking both in general fabrication and with particular reference to pipeline welding. General consideration of its cost together with a few specific examples are given. Some of the reasons for major occurrences of the problem are considered. It is noted that while there has been a marked trend for the incidence of the problem to decrease over the last two decades this can be attributed to developments in steel production and reductions in hydrogen levels of welding consumables, together with better knowledge, which is in general more widely spread. However, another trend, now well established, for the location of the problem to tend to move from the HAZ to the weld metal is also noted as well as the recognition that it is potentially more complicated to predict conditions to avoid weld metal cracking than it has been for the HAZ.
Pipe welding, hydrogen, cracking, cost, multipass, high strength.
P H M Hart, Head of Materials Department, TWI, Granta Park, Great Abington, Cambridge, United Kingdom.
Hydrogen cracking in arc welding may be nearly as old a problem as the history of arc welding itself but it is not recorded as such until much later circa 1938  . This might seem surprising given that one of the earliest references  to hydrogen embrittlement in steels appeared over sixty years earlier and the earliest covered electrodes were anything but low hydrogen. Since its recognition as a problem, particularly during and immediately after World War II, there has been extensive study of it throughout the industrialised World. The main reason for the effort that has been put into the subject arises from the costs that originate from it, both in terms of fabrication, repairs, late delivery and consequential loss of plant (and human life) from failure primarily due to hydrogen cracking initiating fatigue or brittle failures.
Hydrogen cracking during fabrication of steel structures still occurs today although generally on a much reduced scale. Many, perhaps most, of the continuing occurrences arise not because of a lack of basic data but because of a lack of application of already existing knowledge. Some arise because adequate recognition of what constitutes a change from previous satisfactory experience is not made, in effect not understanding what constitutes an essential variable and for that variable what is a significant change. Some occurrences arise because there is an inadequate understanding and a classic example of that in the past was the influence of sulphur content [3,4] on the problem of HAZ hydrogen cracking in the late 60's.
In 1973 it was estimated  that in Britain alone annual costs amounting to £260 million were borne by industry as a result of manufacturing problems directly attributable to welding. At least £40 million of this total arose from the need to repair hydrogen induced cracks at welds. Service failure due to fatigue and brittle fracture of welded components cost the industry £140 million annually and many of these originated from hydrogen cracks at welds.
Girth welds in linepipe are no exception to the above in respect of their share of difficulties and occurrence of the problem. Indeed the most common girth welding technique, stove pipe welding, requires high hydrogen process, manual welding with cellulosic electrodes, to be successful.
The paper will look at the history of the problem in general but in the pipeline industry in particular, considering examples of its cost to the industry, current levels of understanding and future trends and needs.
2. Hydrogen cracking in welds
2.1 General causes
The primary factors involved are fundamentally those of hydrogen embrittlement in steel in general. They have been covered in detail elsewhere by, for example, Coe et al 
and Graville 
. Simply put, four basic factors need to be simultaneously satisfied: the presence of a sufficient amount of diffusible hydrogen, a susceptible microstructure, a tensile stress and a temperature near to normal ambient. Since the microstructural susceptibility is a function of temperature these factors can be viewed as interacting at room temperature as in Fig.1
. For hydrogen assisted cracking to occur all three must overlap and this, together with the influence of temperature, shows that the problem can be avoided by attention to any one or all of the potential factors.
Fig.1. Factors involved in hydrogen cracking
2.2 Girth welding
There are at least three factors which, with respect to influencing the risk of hydrogen cracking, can be considered specific to girth welds in pipe. First of all, stovepipe welding is carried out in its most traditional form with the high hydrogen cellulosic electrodes capable of depositing weld metal containing 60-80m/l H 2 /100g of deposited metal. Secondly, at least for the root pass, the heat input is characteristically low, primarily because of the high welding speed demanded by the technique and production. The heat input may be typically 0.4-0.8kJ/mm and this will give fast cooling rates and a tendency to harden both heat affected zone and the weld metal. Thirdly, and this is really particular to field welding of pipe, the partially completed weld (may be with only the root pass completed) can be strained during the lowering off process. All of these factors will contribute to increasing the risk of hydrogen cracking.
3. Some examples of major problems from fabrication hydrogen cracking
3.1 General fabrication
In general fabrication, possibly the most well known occurrence was that which contributed to the failure of the Kings Bridge in the early 1960s. The initiating defect for the brittle fracture that occurred were identified as HAZ hydrogen cracks at the toe of transverse welds. The causes were almost certainly multi-factor and included out of specification steel, i.e. a higher carbon equivalent than was expected, inadequate preheat and control of consumable condition, inadequate inspection and a brittle steel. Repair costs were at least A$0.5 million (in 1964) which could be considered something of the order of A$20 million today.
In the winter of 1981, a refinery vessel (a thermal cracker bubble tower) failed by brittle fracture during a revalidation hydrotest. The weld had low initiation toughness at the hydrotest temperature. This, combined with the locally high hoop stress and an unusually high residual stress (for a nominally stress relived vessel) was the cause of fracture initiation.
The initiating defect, which was clearly visible on the fracture surface, was a fabrication-induced transverse weld metal hydrogen crack. The replacement bottom head assembly cost £1.25 million and installation costs were about £0.75 million. However, the cost of restoration was relatively insignificant compared to the loss in production, estimated at £25 million  .
In 1982 catastrophic failure occurred during normal operation of a 7m long by 1m diameter pressure vessel (catchpot) in an ammonia plant. The failure consisted of fragmentation of the vessel into over 350 pieces, weighing from less than 1kg to over 2 tonnes. Some fragments were found more than 1km from the site. Naturally, there was considerable subsidiary damage to other components of the plant. A formidable piece of detective work was needed to piece the vessel together, with the aim of tracing the crack path and finding the primary fracture origin. In the end, this was achieved, with only 10% of the vessel wall unidentified. Initiation was traced to the toe of a fillet weld attaching a gusset plate to the shell, the gusset being one of a series supporting an internal ring.
A careful examination revealed that initiation had occurred from thumbnail-shaped cracks about 4mm deep. These had the characteristics of heat affected zone (HAZ) hydrogen cracks, and this was consistent with the HAZ hardness (353Hv2.5) and the high carbon equivalent (0.56).
Fortunately, the failure took place in the early hours of the morning when the only person on site was a night watchman, who was unharmed. However, the failure is estimated to have cost about $50 million  .
The author is aware of many other major occurrences of general fabrication hydrogen cracking that have had very substantial cost impact for those involved because of the delays these have caused during construction, changes in construction method and occasional 'dispute' costs. Sadly, from the point of view of providing hard data on costs, documented cases (and therefore lessons to be learnt) etc. these rarely get into a suitable published and referenced form - many for obvious reasons.
3.2 Pipeline girth welding
Possibly one of the most serious and costly pipeline failures, at least in terms of loss of human life, was a pipeline failure in the Middle East in 1970. The failure was in some above ground pipework which was X60 24-42" diameter mostly of 0.5" wall thickness. The pipework, which was quite extensive with valves, scrubbers, headers etc., was nearly complete and in a fenced compound. Because of its geographical location pressure testing was (unusually) carried out with gas and not water and had recently been carried out on one section so that the pipework was currently gas filled and pressurised at the time when some work was being carried out on one part of it. Suffice to say a brittle fracture occurred at a girth weld, the gas was ignited and a holocaust occurred with 17 men being burnt to death inside the compound. The failure investigation determined the initiating cause of the brittle fracture was an HAZ hydrogen cracking in the root of a girth weld from an internal back weld repair made with a cellulosic electrode. Two of the characteristics of stove pipe welding (i.e. vertical down welding with cellulosic electrodes) that of the high hydrogen and low heat input, contributed to the crack formation. However, a third characteristic, that of the rapidly applied hot pass was absence and hence the hard HAZ cracked. Although the repaired girth weld was radiographed the angle of the fusion boundary of the back weld to the pipe wall was too acute for the crack to be picked up.
The costs of this failure were not published but must have been very considerable since the pipework was located within 500m of a gas treatment plant which itself was partially damaged in the explosion. All of the above ground pipework was destroyed and the project delayed for many months.
During the construction of the initial gas feeder mains in the UK in the late '60s, following the discovery of gas in the North Sea, there were many delays due to hydrogen cracking. Much of the pipe was X60 36" diameter and often of 5/8" wall thickness. At the time this pipe would certainly have been viewed as 'big inch' and of relatively heavy wall thickness and hence quite a challenging step forward. Also at that time not all pipe of this strength level was produced by controlled rolled material and some of it was from normalised plate material with its inherently higher carbon equivalent capable of producing HAZ hardnesses up to at least 450HV in the as-deposited root pass HAZ. The combination of this heavier wall, larger diameter X60 pipe, and some with high carbon equivalent levels, was certainly relatively new to the industry and this 'new experience' contributed to increasing the risk of cracking. Certainly during this time, the importance of the cooling time between making the root pass and its tempering by the hot pass, and how this is influenced by factors such as preheating practice, wall thickness, pipe diameter and number of root welders became increasingly evident. The occurrence of cracking during these constructions certainly spurred consideration of the specification of weldable linepipe material and led to the development of the British Gas full scale pipe weldability test which incorporated important aspects of a field welding procedure such as lowering off.
Following this period of difficulties in the late '60s and very early '70s and the experience gained, the incidence of major problems appeared to decrease, and this was certainly helped by two developments. The first was the growing use of controlled rolling to produce X60, X65 and then X70 steels with reduced carbon and carbon equivalent, thus reducing the risk of producing hardened heat affected zones and hence of cracking. The second was the growing use of non-stove pipe techniques, principally the use of gas shielded metal arc welding processes with their inherently much lower hydrogen levels. Certainly these contributed to the minimal hydrogen cracking problems in major projects such as the 48" Alaskan pipeline in the early 1970s.
However, it's almost as if this success, and the reductions in linepipe carbon equivalent, through controlled rolling development lead the industry to drop its guard against the hydrogen cracking problem and forget the inherent dangers of the high hydrogen level of cellulosic electrodes, since the problem re-occurred, on a major scale, during the construction of the Moomba to Sydney pipeline in 1974. Extensive hydrogen cracking occurred in the relatively thin walled, 8.5mm, 34" X65 line. According to the recollection of those involved at the time restitution costs, excluding consequential costs, were A$15 million which is perhaps equivalent today to the sum of A$100 million. Not unusually the pipe supplied was dual sourced, the Australian supply being in general of higher carbon equivalent and it was also REM treated. Although the instance of cracking, which as well as conventional root HAZ (plus weld metal) ( Fig.2) was often initiated at root intrusions and often located in the weld metal ( Fig.3), was greater in the Australian pipe it was ultimately generally agreed that this was not due to the REM treatment but may have been due to a combination of it's higher carbon equivalent and poorer fit up. Molybdenum segregation in the root runs may also have been a contributing factor.
Fig.2. Typical example of root HAZ hydrogen crack extending into the weld metal
Fig.3. Weld metal cracking from a root intrusion initiating feature
The forms of girth weld hydrogen cracking that have been discussed up to now are the conventional longitudinal (with respect to the girth weld) in either the HAZ or weld metal, or quite commonly a combination of both. However, hydrogen cracking in fabrication welds can take other forms and in particular in the weld metal it can be of a transverse orientation. Moreover in C:Mn or C:Mn lightly alloyed deposits its common form is angled at approximately 45° to the weld surface and buried. It is often referred to as chevron cracking owing to its appearance on a longitudinal section through the weld centreline ( Fig.4). In general fabrication experience this form of cracking has been found to be more likely in thicker sections than are currently used on most land pipelines. However, perhaps the first significant occurrence of this form of cracking in linepipe girth welds was in 1982 when a section of the Moomba to Sydney pipeline was replaced by heavier wall (17.5mm) as a mitigating measure against stress corrosion cracking. This occurrence of the problem almost certainly was primarily due to the significant rise in wall thickness above that for which satisfactory experience had previously been obtained by the industry. It would be interesting to establish the details as to why this problem was not discovered during weld procedure qualification testing.
Fig.4. Longitudinal weld metal section showing 45° transverse or 'chevron' cracking
The UK has also had experience of thicker wall pipelines and girth weld metal cracking but two additional factors may have contributed to an occurrence of chevron cracking in the early '90s. These were the use of a compound bevel (the wall thickness was 19 and 25mm) and the use of more alloyed consumables, selected to meet the more stringent Charpy requirements for the heavier wall thickness. One consequence of the compound bevel would be to tend to increase the layer thickness and decrease the interpass time for each layer, compared to the shallower layers in a wider vee preparation. The expected effect of this would be to decrease the hydrogen diffusional loss from each pass/layer thereby increasing the hydrogen remaining in the weld following the final cool out to ambient temperature.
4. Prediction of safe welding procedures
The earliest methods of deriving safe welding procedures simply used the previous satisfactory experience, and in pipeline welding a crucial part of this experience was the rapid execution of the hot pass after completion of the root pass, although how much initially this was driven by production reasons rather than detailed understanding of its effect on avoidance of hydrogen cracking is very much a matter of conjecture. In general fabrication, avoidance of HAZ cracking largely relies on the use of an approach that produces a HAZ microstructure that is adequately soft (critical hardness method) so that cracking does not occur for the given level of hydrogen and strain applying, and this approach was the basis for the first nomogram prediction method 
. To adopt this approach for girth stove pipe welds is less easy because of the very low thermal input of the root pass. Moreover the rapidity of the stovepipe process allows a modification to this in using the quick application of the softening hot pass to become an integral part of the welding technique for avoiding root pass hydrogen cracking, but to the best of the author's knowledge this aspect has not been incorporated into a published prediction scheme.
The focus of the hydrogen cracking problem in pipe welding has principally in the past been on the HAZ, and this has been a driver for the reduction in the carbon and carbon equivalent levels that have taken place over the past three decades in linepipe steels. Pipeline welding is not alone in seeing a major reduction in HAZ hydrogen cracking primarily due to reduction in carbon and carbon equivalent levels since this has also been observed in general structural fabrication. Both areas of fabrication are experiencing the change in emphasis from HAZ to weld metal cracking as the greater concern. There are probably three main reasons for this, the reductions in carbon equivalent levels in steels, an increasing trend for the use of high strength steels with more alloyed weld metals and an increasing trend for thicker sections to be welded. Some data now exist in relation to weld metal hydrogen cracking in general fabrication and some of this has shown the situation is more complex than that of the heat affected zone. First of all, at least for C:Mn type weld metals, in contrast to the HAZ situation, increasing heat input is not necessarily beneficial  ( Fig.5). Secondly, whilst most occurrences of HAZ cracking occur in 'as-formed' microstructures this is not by any means always the case for weld metals since cracking is often seen in re-heated regions. Thirdly, for some consumable types, including cellulosic electrode deposits, there is considerable change in the mobility of the hydrogen. Much of that which is room temperature diffusible in an as-deposited bead becomes non-room-temperature diffusible when the bead is re-heated [10-12] . Thus in these deposits diffusible hydrogen is converted to 'residual' hydrogen with important consequences for trying to characterise the hydrogen level of a multipass weld.
Fig.5. The influence of heat input on preheat to prevent weld metal hydrogen cracking in multi-pass C:Mn submerged arc weld deposits
5. What lessons should be learnt
Looking back over nearly thirty years many lessons of course have already been learnt but two aspects in particular stand out when consideration is given as to how, in the future, to avoid similar occurrences in the future to those described earlier. The first relates to the benefits of a well designed and well proven procedure qualification test. In this respect it is clear that to be successful this must incorporate all the factors that can influence the problem, i.e. it must be as realistic as possible. A very important part of this is establishing what are the essential variables, and then the ranges of these for which the procedure remains valid must also be rigorously established. General standards e.g. ASME IX, EN 288, ISO 9956 for procedure qualification are inadequate for the field welding of pipelines because they do not recognise all the relevant factors for this application. This is why there are many national, and now being developed international (e.g. ISO/DIS 13847) standards, specifically reflecting the needs of this application.
The second aspect is adequate and appropriate inspection. Obviously, to be effective, this requires consideration of the possible orientation of cracks and the use of appropriate inspection techniques Thus as already mentioned the internal backweld HAZ crack was an unfavourable orientation for detection by normal radiographic techniques. Most chevron cracking is fine and in addition because of its 45° angle orientation is also rarely detected by radiography. It is interesting to consider why the occurrences of chevron cracking referred to above were not picked up at procedure qualification. Was it because of the use of radiography only or were they not produced, due to some inadequacy of procedure qualification testing? Alternatively was there some aspect of welding procedure that was different in the field compared to that during shop qualification testing i.e. had all the essential variables not been identified and their ranges adequately set?
The final aspect of inspection, which is not often considered in relation to pipeline welding is the fact that hydrogen cracking can be delayed and therefore inspection should not be carried out until an appropriate delay time has occurred after welding has been completed. This aspect may require much greater consideration when (if) we see pipelines in materials of >X100 strength level in the future.
6. What of the future
There is no doubt that overall the incidence of hydrogen cracking during fabrication is decreasing. Three main factors at least, are contributing to this. Steel development is constantly leading to more weldable materials. Hydrogen levels of consumables are decreasing (especially if we consider for pipe welding the low hydrogen vertical down electrode as well) and thirdly the knowledge to avoid the problem is improving and, generally, (but not always) this is being more widely disseminated. Offsetting these trends are at least two others i.e. those of increasing steel strength level and section thicknesses both of which will tend to increase the risk of cracking. The former is of increasing concern in respect of avoiding weld metal hydrogen cracking because, as yet, no means of applying the equivalent of 'TMCP' techniques to weld metals, avoiding the otherwise inevitable increase in alloying, have been devised. However, this does not mean that a trend to consider high strength grades for pipelines, for example, should be dampened by fears of hydrogen cracking. Certainly any such fears should be borne in mind when establishing welding procedures and in procedure qualification and testing but the basis of success, or failure, really does begin with detailed welding procedure design and specification, essential variable determination and acceptable range determination. This must of course be coupled with appropriate inspection techniques and adequately high levels of quality control and assurance for a successful outcome.
7. Summary and conclusions
Hydrogen cracking during fabrication of steel structures, including pipelines, has caused considerable costs to industry in the past including occasionally loss of life. The instances of this form of cracking are now much reduced compared to 30 or more years ago largely due to better knowledge of the problem, quantitative understanding of factors involved, development of more weldable steels and development in the welding consumable industry leading to lower hydrogen levels. The pipeline industry has generally learnt to live with the inherently high level of hydrogen in its traditional stovepipe welding technique which, while diminishing in extent, is probably still the most widely used process.
This relatively satisfactory position of a decreasing occurrence of fabrication hydrogen cracking is potentially under threat because of trends leading to a greater potential risk of weld metal rather than conventional HAZ cracking arising primarily from the trend to use higher strength steels and pipelines and the absence of a predictive scheme for the avoidance of weld metal hydrogen cracking at least as comprehensive as current schemes for the HAZ. However, this need not be a barrier to the satisfactory use of higher strength steels, but avoidance of hydrogen cracking in these materials does require a recognition of its possibility and the need for carefully designed and well proven weld procedure qualification testing. While the 'siting shots' for such procedure qualification testing and development would economically benefit from schemes for predicting weld metal hydrogen cracking that are at least as good as current HAZ methods, the presence of such a scheme is not considered essential for the derivation of safe procedures for welding high strength steels and pipelines.
The author wishes to acknowledge helpful discussions with several colleagues including Leigh Fletcher, in the preparation of this paper.
|Swinden T and Reeve L 1938.
|Trans Inst Weld 8 7-24.
|Johnson W H 1875.
|Proc Royal Soc 23 169-179.
|Smith N and Bagnal B I 1968.
|Brit Weld J 15 (2) 63-9.
|Hewitt J and Murray J D 1968.
|Brit Weld J 15 (4) 151-8.
|Coe F R 1973.
|Welding steels without hydrogen cracking, Publ TWI.
|Graville B A 1975.
|Cold cracking control, Publ Dominion Bridge.
|Harrison J D, Garwood S G and Dawes M G 1990.
|Weld Metal Fab 58 (2) 144-147.
|Bailey N 1970.
|Metal Const 2 (10) 442-446.
|Pargeter R J 1992.
|Effects of arc energy, plate thickness and preheat on C-Mn steel weld metal hydrogen cracking, TWI Members Report 661/1992.
|Coe F R and Chano Z 1975.
|Hydrogen distribution and removal from single bead weld during cooling. Weld Res Int 5 (1) 33-90.
|Ødegärd O and Evans G M 1971.
|Apparent diffusivity of hydrogen in multipass deposits. Metal Const 3 (2) 47-49.
|Hart P H M and Evans G M 1997.
|Hydrogen content of single and multipass steel welds. Weld J 76 (2) 74s-80s.