An Ammonia Shift Converter Failure

Brief case history of a metal failure in a high temperature shift con-

verter resulting from hydrogen cracking, and what was done to solve
it.

M. Appl, K. Feind, and W. Liebe,

BASF A.G.,
Ludwigshafen/Rhine, Germany

An extensive investigation into a failure in a high-tempera- number of smaller cracks were noticed where fittings were
ture shift converter in an ammonia plant in Rotterdam- welded to the wall in the upper part of the converter.
Pernis late in 1975 has led to several useful lessons and The most important design data and operating conditions
ultimately to process improvements for safer operation. are presented in Table 1.
The ammonia plant is a joint venture of the Dutch com- Table 1. HT Shift converter, ammonia plant at Ferais
pany, UKF BV, and the German company, BASF AG. It (design data and operating conditions)
has been producing 1,000 short ton/day of ammonia from
natural gas since 1967. Material 16 Mo 5 . . . . ASTMA 204 B
At the beginning of November, 1975, the plant was being Design pressure 34.2 atm. . . . 486 lb./sq.in.gauge
restarted after a turnaround and was being heated with Design metal temperature 482JC 900°F
process steam. It was then noticed that steam was escaping Operating pressure 29-30 atm. .. 413-427 lb./sq.in.gauge
from the insulation of the high-temperature (HT) shift Gas temperature, inlet . . . 350-385JC .. 662-725JF
converter. On removing the insulation, it was seen that the Gas temperature, outlet ..415-450"C .. 779-842JF
steam was flowing through a crack in the HT converter. Operating hours 63,000
Number of shutdowns, approx.. 20
Thereupon, the entire plant was shut down.
Magnetic and ultrasonic examinations of the converter's
interior revealed one large crack and two adjacent small Figure 2 shows the weld between the shell and the head
cracks in the wall, which has a thickness of 80 mm. They with two cracks on the inside of the. converter. Ultrasonic1
were located at the lower end of the cylindrical shell, examinations revealed several of these cracks in the weld
perpendicular to the weld between the shell and the head. metal. It was therefore an obvious assumption that the large
The position and dimensions of the cracks are shown in crack also originated from an initially small crack within the
Figure 1. The large crack was 400 mm. long, with a weld.
maximum gap of 0.1 mm. All the cracks run from the inside A cross-section at right angles to the weld can be seen in
to the outside. The small cracks are 50 mm. long and have Figure 3. The weld was automatically welded from the out-
not yet penetrated to the outer wall. None of the cracks side and built up to about 60% of the wall thickness (sub-
starts from a welded part on the inner wall. In addition, a merged arc process, Ulme 3 rod, Arcosite Mo powder).

Figure 1. Cracks in the HT shift converter. Figure 2. Cracks in the weld.

40
After the root had been machined, the inside of the con-
verter was welded manually with a Philips coated kV2
electrode. The weld was stress relieved at 620 - 630°C, not
in a furnace but by local induction heating. According to the
manufacturer's report it must be assumed that the local
stress relief operation was carried out correctly.
Hardness was measured on four lines drawn over the
cross-section of the weld. Fairly high.values, i.e., 240 - 270
Hv., were found for the weld metal, as opposed to 170 Hv
for the parent metal. The figures for the strength of the weld
metal were also close to the upper limit.
Several microscopic studies of the manual weld in the
vicinity of the cracks revealed accumulations of pores and
minute gas channels in the weld metal of the root passes.
Figure 4 shows a horizontal cross-section through the weld.
The pores and a crack issuing from one of them can be
clearly recognized. A microsection perpendicular to the
surface of another crack also revealed pores in the welding
material, as seen in Fiture 5. A crack running parallel to the
edge of the fracture commences at a large pore. In Figure 6,
the details are reproduced of a crack in the weld metal. It
can be seen from the polished sections that the cracks are

Automatic Weld Manual Weld

Figure 5. Porous weld.

intercrystalline and that their edges are lined with oxides

(magnetite). All the cracks in the weld that connect with the
inner wall of the converter display this layer of magnetite,
whereas the other fine cracks do not.
Origin of the cracks in the weld
The accumulation of pores in the manual weld indicates
that the electrodes were moist. In such cases, hydrogen is
formed during welding and diffuses into the welding
material. The result is "hydrogen cracking." The initial
cracks in the weld were probably formed immediately or
within a few hours after welding.
60% 40% » » According to the Materials Testing Department of the
—«
Dutch State Mines (DSM), crack formation was favored by
Figure 3. Cross section of the weld. high stresses that were induced during welding in the rigid
construction. It is likely that all the stresses were not
eliminated in the thick-walled converter by the local stress
relieving adopted. The surface of a fracture that was
induced in a compact tension specimen during the DSM
fracture mechanics investigation shows that there were
residual stresses in the weld. Further growth of the cracks
could have been caused by strain-aging at the extremities of
the cracks, low-cycle fatigue (starting up and shutting
down), or local hydrogen embrittlement at the extremities.
The BASF Materials Testing Department considers that
the cracks could have arisen in the following manner. If
atomic hydrogen that has been dissolved during welding
diffuses and moves to the pores and to other inclusions,
Figure 4. Horizontal section through the weid. molecular hydrogen is formed. Since this hydrogen is no

41
Figure 6. Details of a crack. Figure 7. Surfaces of the big crack.
longer capable of diffusing, very high pressures are built up of the plate metal for the shell and head. The properties of
in the small cavities. This results in intercrystalline cracks the plate metal and the weld metal were determined by
in the material of the weld. various means (chemical analysis, metallography, and sev-
It is also possible that the fine cracks in the vicinity of the eral mechanical tests including fracture mechanics). Per-
pore accumulations did not occur until after a period of fectly satisfactory values were obtained for the plate metal
operation; for instance, during the first shutdown. In this by chemical analysis, the tensile test, and the DVM
case, the formation of molecular hydrogen at flaws, to- notched-bar impact test, DIN 50115 (20°C). However, the
gether with embrittlement caused by the diffusing hydro- Charpy-V-impact test at various temperatures revealed that
gen, would be responsible for the cracks. It is possible that the transition temperature (abrupt drop in impact strength) is
hydrogen from the process gas diffused into the material about +40°C. This is too high for this grade of plate metal
after the converter had gone on stream. In the course of and indicates that the quality is not up to standard.
further shutdowns, a few of the small primary cracks would
then be gradually enlarged by the mechanism described Changes In structure of plate metal
until they reached the inner surface of the converter. The structure of the plate metal in the original state and
The two fracture surfaces of the large crack in the after normalizing in the laboratory (930°C/'/2 hr. cooling in
converter are reproduced in Figure 7. Two zones can be furnace) is shown in Figure 8. The original material has a
clearly distinguished: a) a grey section that lies within the Widmannstaetten structure and consists of ferrite and bain-
weld metal and on the inside of the converter: and b) a ite. After normalizing in the laboratory, the structure
reddish brown and black section that is very probably the consists of a linear arrangement of ferrite and perlite such as
extension of the crack. can be expected from a material of this nature that has been
The shape and size of the grey section are roughly similar subjected to optimum heat treatment. The hardness of the
to those of the other cracks in the weld. With the aid of plate metal decreased from 180 Hv. before normalizing to
X-ray diffraction pictures, it was determined that the very 125 Hv. after normalizing. The conclusion is that the
adherent and thick grey layer of oxide consisted mainly of material had not received optimum heat treatment.
magnetite. The thickness of this layer proves that the grey The fracture mechanics work carried out in the DSM
surface is the initial crack that was in contact with the laboratory revealed the following:
process gas for a long time and that it is much older than the 1. The initial flaws in the weld are large enough to allow
extension of the crack. cracks to be initiated from the weld.
A factor that affects propagation of cracks is the quality 2. The plate metal is of inferior quality, with the result

42
 converter, because it was not at all certain if it could then be
reused, what the costs of repair would be, and how long the
repair would last.
Inasmuch as the old HT shift converter was no longer
available and since the delivery time for a new one is six to
eight months, the problem was how the ammonia plant
could be put quickly on stream again at as high a production
rate as possible without the HT shift converter. The flow-
sheet of the Pernis ammonia plant with the CO conversion is
shown in Figure 9.
Fortunately, the two LT shift reactors and the HT shift
converter have the same design metal temperature, i.e.,
—
original ümmmHB normalize
^^^^^^^^^^^^^^^^^^^^^^^^^^^ufrjjgjjgg^^BjgaEB,
Secondary Reformer
Figure 8. Structure of the plate material.
that instable brittle cracks could be propagated.
In the opinion of the DSM Materials Testing Department,
the only reasons that the brittle crack did not finally proceed
any further were that a low figure had been taken for the
design stress in the material, and that the crack probably
occurred during a startup when full operating pressure had
not yet been reached.
The BASF Materials Testing Department is of the opin-
ion that oxidation, i.e., the formation of magnetite, played a
part in the crack propagation. The process steam penetrating
Methojiator
into the crack caused magnetite to be formed at the edges of
the crack and atomic hydrogen to be liberated. Figure 9. CO shift conversion ammonia plant at
Formation of magnetite within the crack is associated Pernis.
with a considerable increase in volume. Thus a wedging 482°C. Therefore, it is possible to use one or both of the two
effect takes place, forcing the edges of the crack further LT shift reactors instead of the HT shift converter.
apart in the material that has been embrittled by hydrogen, Five means were discussed for the rapid resumption of
e.g., during shutting down. The crack is thus enlarged in production without the old HT shift converter.
stages. After it has attained a critical length, the final phase
sets in, consisting of a spontaneous extension of the crack. /. Without HT shift conversion. The old HT shift con-
verter should be dismantled and replaced by a length of
The magnetite embedded in the crack exerted an explo- pipe. Only the LT shift conversion should remain in opera
sive force probably while the shell was cooling during tion.
shutting down. The effect was reinforced by the hydrogen The advantage would be that due to the comparatively
embrittlement of the material. This also explains the re- simple assembly work, production could be resumed quite
stricted length of the crack; because if the length of a crack
quickly.
were to become critical while the full internal pressure is
being exerted on the vessel, the crack would not come to a Disadvantages would be as follows. The CO content at
halt. the LT shift converter outlet would be comparatively high
Neither on the inner wall of the converter nor on the sides (0.8 - 1.2%). Production would be only about 90% of the
of the crack were there any signs of decarburization wuch as design capacity. At the outlet of the LT shift conversion,
that which occurs together with methane formation when a high temperatures would be reached, viz., 285ÛC, as a
metal is attacked at high temperatures by hydrogen under result of which the activity of the catalyst would be im-
pressure according to the Nelson curves. paired more rapidly than usual. If this technique were
adopted, the bypass for the waste heat boiler E 103 would
Possible provisions to stay onstream have to be opened, and the amount of water for quenching
at the outlet of the LT shift conversion would have to be
After the crack in the HT shift converter had been de- increased.
tected, and all decisions had to be made immediately and all
the necessary measures had to be taken to ensure rapid re- 2. Without LT shift conversion. The two reactors in the
sumption of production. It was decided to reject the catalyst, LT shift conversion would be filled with HT shift catalyst
which had been in use for two years. Careful oxidation of and would be made to act as the HT shift conversion by
this catalyst would have taken at least four days, incurred installing new pipes.
additional costs, and would not have yielded a very active The advantage — Experience on this had already been
catalyst. It was also decided not to repair the old HT shift obtained, because the plant had been run once before with

43
the HT shift conversion only when the catalyst was being longer than six months, the total amount of ammonia pro-
changed in the LT shift converter. duced would be more than in the first case.
Disadvantages — The CO content at the methanator inlet In view of the long delivery time for the new HT shift
would be very high (1.8 - 2.4%). The temperature rise in converter, the fifth solution offered the most advantages. It
the methanator would amount to about 170°C, and the pro- was decided to use the small LT shift reactor as a temporary
duction would be only about 75% of the design capacity. HT shift converter. Therefore, the upper LT shift vessel was
The assembly work is more difficult and would require removed and installed in place of the defective HT shift
more time than in the first case. reactor.
3. Filling the LT shift converters with HT shift catalyst
and replacing the HT shift converter by a pipe. The assem- Operation with a small HT shift converter
bly for this would be the same as in the first case and could
thus be carried out simply and rapidly. However, this par- A period of 31 days was required for dismantling the old
HT shift converter and connecting the small reactor. After
ticular case is technically infeasible, because the pipes up-
stream and downstream of the LT shift conversion are the HT and LT converters had been filled with fresh
catalyst, the plant was taken on stream again.
unsuitable for the high temperatures and the heat balance
would no longer be valid. With a high steam/carbon ratio and a temperature of
4. A total a/48 cu.m. of HT shift catalyst plus 17.5 375°C at the inlet of the HT shift conversion, the CO content
at the converter outlet was 3 - 3.5%. Owing to the small
cu.m. of LT shift catalyst. The advantage would be that the
volume of catalyst, quite a high temperature was maintained
CO content at the outlet of the HT shift conversion would be
at the inlet of the HT shift conversion, in order to achieve a
practically normal.
high rate of reaction. As a result of the debottlenecking
Disadvantages — The CO content at the outlet of the LT
effected during the turnaround, the production was 114-
shift conversion would be comparatively high, i.e., about
119% of the design capacity.
0.6%. Since the LT shift catalyst is gradually poisoned
during operation, the life that could be expected from such a A new HT shift converter made of 13 Cr Mo 44 (ASTM
small volume would be quite short. A 387-65 B) with a correspondingly thinner wall was
5. A total of 17.5 cu.m. of HT shift catalyst plus 48 ordered. By virtue of the thinner wal it is now cheaper to
cu.m. of LT shift catalyst. Advantages would be production make a converter from the high-grade material 13 Cr Mo 44
at more than 100% of the design capacity could be than from 16 Mo 5. The new HT shift converter was to be
achieved, the CO content at the outlet of the LT conversion installed in the autumn of 1976, and 19 days will be
would be low, and the LT shift catalyst could be expected to required for this purpose. Thus the crack in the old HT shift
have an adequately long life. converter resulted in a total loss of production of 50 days.
Disadvantage — Assembly would be more difficult and
would require seven days more than in the first case. Acknowledgment
However, this seven days' loss of production would be The authors wish to thank the Materials Testing Depart-
more than compensated by the higher production capacity. ments of DSM and BASF and the colleagues at UKF in
Thus, since the delivery time for a new HT shift converter is Pernis and Ijmuiden for their assistance. #

LIEBE, W. D, K. PPL, M.

DISCUSSION
D.U. TRUAX, Standard Oil Co. of California: Your can answer your question. I don't know if he is informed
Charpy impact test showed a transition temperature about this.
of 40°C. I recognize that the poor microstructure of VAN GRIEKEN: We discussed temper and brittleness
carbon half-moly plate contributed to that. Did any of but we didn't go further on in that study.
your tests look at the possibility of temper and brittle-
ness for showing a higher transition temperature? Q. Your report suggests that hydrogen was in the
LIEBE: The tests were made by the Materials Testing vicinity of the weld. I'm surprised that hydrogen would
Department of DSM (Dutch State Mines) and perhaps stay around at that temperature.
Mr. van Grieken of UKF Ijmuiden, The Netherlands, LIEBE: A chemical analysis was made in Ijmuiden and

44
hydrogen was found in the plate material and in the To my astonishment after we took a sample out of
weld metal. the middle of the thick wall of the vessel, doing machin-
VAN GRIEKEN: Welding with damp electrodes gives ing at low temperature, we still found several milliliter
hydrogen, which can give you a crack that can initiate hydrogen per 100 g.
further cracking. Further on you have the hydrogen This blocked in hydrogen could impair toughness
from the process gas that can diffuse into the metal. and contribute to the stress pattern already existing
This hydrogen only gives decrease of toughness by bad local stress relieve conditions and results in
below say 200°C. low critical crack dimension.

45