J.C.Guild: Southern African Institute of Welding, Johannesburg, South Africa



Key Words: Failure Investigation, Pipeline, Molybdenum steel, Carbon Monoxide Conversion, Hydrogen Cracking, Decarburisation





The investigation into the failure of a Mo steel pipeline in carbon monoxide conversion service is described. A previous failure was attributed to creep damage, but the present failure is attributed to hydrogen damage. Both failures occurred in weld repairs. Recommendations are given to assess plant condition and prevent future failures.





A crack was discovered in 500 mm diameter schedule 60 piping at the bottom of a heat exchanger. This pipeline carries convertor feed gas from the bottom of the exchanger to the top(inlet) of two CO convertors. The plant was immediately shut dawn and repair, which involved cutting out the weld and re-welding with 1Cr Mo material, was carried out. This was the second failure to have occurred in that region since the piping was installed during a major shutdown some three years earlier. The first failure occurred about a year before the second failure. An investigation of the first failure was carried out, but only a small section of weld metal, representing the outer 25% of the weld thickness, had been made available for that examination. After investigation of the sample and scrutiny of a full thickness radiograph, showing numerous irregular cracks, it was felt that the most likely cause of the failure was creep. However, creep cracking could not be satisfactorily explained in terms of either stress or temperature, both of which appeared to be too low to constitute a real creep hazard.





The crack, shown in Figure 1, had initiated and propagated in what appeared to be a repair weld. The run of weld at that position was short and grinding had been carried out on the cap The crack ran from the centre of the weld to the edge of the cap and continued in that direction. There was no indication of abnormal deformation associated with the failure. As in the case of the first failure, a V section boat sample containing a portion of the crack was cut out of the weld for metallurgical examination. The sample thus represented about a half to two thirds of the thickness of the weld metal. The fracture is shown in Figure 2, the outer portion having an irregular columnar appearance and the inner being smooth and fine grained. The fracture surface was coated with dark gray oxide deposit.


Numerous microsections were then cut from the sample for microscopic examination. Although low power examination of these samples had shown little evidence of other defects associated with the fracture, microscopic examination of the samples in the unetched condition showed small amounts of fine intergranular cracking usually associated with the main crack. (This can be seen in Figure 3, in the etched condition.) In the etched condition much more of the fine cracking was apparent and it was noticeable that cracking had also occurred remote from the main fracture. Several samples showed evidence of decarburisation associated with some of the cracking. (See Figures 3 and 4). The decarburisation was not evident for the fun length of the main crack but was found in the bottom portion. That is, from that part of the sample nearest the inside surface. Not all of the fine intergranular cracking had obvious decarburisation associated with it.

Text Box:  
Figure 3: Fine intergranular cracking associated with main fracture surface. Note decarburisation associated with main fracture. (190X)

The sample also showed a short crack propagating from the outside surface This crack is thought to have been an original weld crack and is not considered to have been related to the failure


The weld metal was found to have contained a large amount of globular oxide inclusions most of which were relatively small in size. The microstructure in the cap of the weld was columnar in nature. Below the cap, the structure had been refined and consisted of fine grained pearlite in ferrite.


Text Box:  
Figure 4: Fine intergranular cracking remote from main fracture. Note decarburisation which has occurred. (480X)
Chemical analysis was carried out on the weld metal with the following results:


C: 0,10

Si: 0,19

Mn: 0,34

P: 0,023

S: 0,016

Cr: 0,03

Ni: 0,04

Mo: 0,53


It is believed that Bohler Fox DMo electrodes were used for welding these pipes. These are available with either a basic (Kb) or rutile (Ti) coating. The low Si content of the weld metal could indicate that a rutile coated electrode was used rather than the basic coated which would normally be recommended. However, on the basis of the analysis only, this most be considered speculation.


Hardness tests gave the following results:


Hv 30kg = 162 / 243 (14 tests)


The highest results were obtained in the cap of the weld, as might be expected. These values are considered satisfactory.





The cracking of this failure and the previous failure were similar in that both occurred in what appeared to be repair welds. The type of cracking was similar in that in both cases it was of an intergranular nature.


The intergranular nature of the cracking and the decarburisation associated with it indicate that hydrogen embrittlement has probably occurred in service. Although no visible decarburisation was observed with the first failure this does not preclude it from being due to the same type of attack. Decarburisation would not be associated with creep cracking although this would also be intergranular in nature. In hydrogen cracking molecular hydrogen dissociates on contact with the steel surface and the nascent hydrogen is then free to diffuse through the iron lattice. It can then react with carbon (or impurities) in the steel to form methane (or other gaseous products). The methane is unable to diffuse out and as it accumulates and adds to any hydrogen pressure, from the reformation of molecular hydrogen in void spaces, it creates high internal stresses that ultimately crack the metal. Intergranular fissuring with associated decarburisation is regarded as typical of hydrogen damage although it can occur without carbide reduction in alloyed steels.


The understanding of the service conditions of this pipeline at the time of the previous failure was incorrect. It was thought that the service temperature was in the range of 365 395C, but in fact the temperature is not monitored in this section of pipe. It is only measured just prior to entry into the CO convertors and cool gas can be introduced prior to the measuring point. The actual temperature of the section in which failure occurred is likely to be considerably higher. Gas coming from the CO convertors at 500C is used as a heating medium for the gas/gas exchanger and it is fed in at the bottom of that vessel. Thus convertor feed gas coming out of the bottom of the exchanger into the section of line where failure occurred is likely to be as high as 450 500C in temperature. The composition of the convertor feed gas is approximately as follows:


H2O: 64.0%

CO: 20.5%

H2: 10.4%

CO2: 4.3%

N2 + others: Balance


The pressure in the line is 50 bar (725 psi) and the hydrogen partial pressure is about 75 psi. Plotting this point on a revised (1977) Nelson curve indicates that with the low hydrogen partial pressure, the line would appear to be in the safe region. The exit pipelines from the CO convertors are also understood to be fabricated in Mo steel and these would appear to be more at risk to hydrogen damage than those where failures have already occurred. The exit lines are known to serve at 500C since the temperature is monitored and the gas composition is approximately as follows:


H2O: 50%

CO: 1.5%

H2: 26.5%

CO2: 21.5%

N2 + others: Balance


The pressure in these lines is also 50 bar and the hydrogen partial pressure is l92 psi. When plotted on the Nelson curves, these lines would appear to be in an unsafe region and must therefore be considered susceptible to failure by hydrogen damage.


The question arises as to why failures have occurred in the convertor feed gas lines but not in the convertor exit lines. The following comments can be made in this respect:


1)      The Nelson curves are determined empirically and are based on failures reported to API. They are periodically revised and the revision in 1976/77 shows that this can mean a significant change in what is considered safe or unsafe.

2)      Hydrogen damage is time related in that there is an incubation period before serious deterioration occurs. Current awareness indicates that this could be a substantial period for Mo steel at the hydrogen partial pressures under consideration.

3)      Weld metal is known to be more susceptible to failure than wrought metal. Weld metal could be expected to have more segregation and thus more chemical composition variation than wrought metal. It is also likely to contain defects which act as stress raisers and increase the susceptibility to hydrogen attack. In addition, as is shown by the chemical analysis the chromium content is very low and even the residual levels found in the pipe material would be beneficial

4)      The fact that the two failures that have occurred to date were in what appeared to be repair welds could be significant. If the repairs were not properly carried out it is feasible that hydrogen cracking (cold cracking} could have occurred in the metal shortly after welding. Radiography after welding would not necessarily detect fine cracks. Pre-existing micro cracks could be expected to aggravate the situation. However, the amount of decarburisation observed in a sample which represented at the most the outer two thirds of the weld thickness, suggests that hydrogen damage would probably have caused failure even in the absence of pre-existing defects.


During the period when this last failure was repaired several other welds were checked by ultrasonic examination for cracking and no other defective welds were found. Ultrasonic testing if properly carried out should be capable of detecting this type of cracking even on a relatively fine scale.





The failure is considered to be attributable to hydrogen damage that had occurred in service. Other welds in the same section and even the pipe material itself may also be suffering from the same type of attack.


The pipelines from the convertors to the heat exchangers are also considered to be operating in conditions which could render them liable to failure by hydrogen damage.


Since the consequences of a failure in CO conversion could be disastrous, it is necessary to establish as quickly as possible if damage has occurred in the above mentioned pipelines and how serious the damage is. During the next planned shutdown scheduled to take place in two months time, all the relevant lines and welds should be examined by non-destructive methods. Full diameter sections of pipe at least 12 inches in length, and containing a weld should be cut out for metallurgical examination. Replacement of these samples should be made with 1CrMo steel if possible. If, for any reason the plant comes off line before the scheduled shut down, these pipelines should preferably be kept at temperature. If they are allowed to cool down it would be advisable to check the welds non- destructively. The reason for this is that cracking due to hydrogen embrittlement is thought to be more likely to occur during cooling to lower temperatures because the material is less ductile and it is: also being subjected to contraction stresses.


Depending on the results of the proposed investigation it may be necessary to change these lines to CrMo steel in the future.


It would also seem advisable that process personnel, who are familiar with this section of the plant, should investigate any other areas which may be susceptible to this problem.