The Ammonia Process – A Challenge

for Materials, Fabrication and Design

of the Components
Even after 100 years of industrial experience, the ammonia process still represents a challenge for the
reactors, apparatus and materials involved. The “physical” service conditions (operating temperature
and pressure) comprise very high temperatures up to about 1000 °C in the secondary reformer and
pressures up to about 200 bars in the reactors and waste heat recovery system. In addition to these
physical conditions, gas atmospheres are handled which are known to cause carburization and metal
dusting, nitration and hydrogen attack. The physical conditions by themselves are known to instigate
creep and relaxation cracking. Due to the high gas velocities, erosion and fatigue may also be a
problem.

Past failures caused by some of the above mentioned damage mechanisms in the ammonia plants of
BASF and remedies which were taken, have been reported. In this paper some basic principles will
be referenced, which need to be taken into account when judging the remaining life of reformer
tubes, and some unusual failures which occurred in the high temperature converter and the boiler
feedwater preheater of the CO shift unit of one of our plants.

Dr. Jürgen Korkhaus

BASF SE

Dr. Michael Bachtler

BASF SE

thought that nitration was the reason for the brit-

tle failure of the pilot plant units after service
Introduction times of about 50 hours. Nevertheless the prob-

T
he development of the industrial ammo- lem was solved by the right conclusions drawn
nia process by Fritz Haber and Carl from metallographic sections, which indicated
Bosch 100 years ago was based on the the loss of pearlite and internal fissuring in the
solution of a materials problem caused steel’s microstructure after the short service.
by hydrogen attack on the pressure bearing
shells and tubes of the reactors and heat ex- Figure 1 demonstrates the changes within the
changers. Initially Bosch and his coworkers microstructure that occurred during the short

2013 167 AMMONIA TECHNICAL MANUAL

 term pilot plant service. The pictures are histor-
ical ones and have probably been prepared by
Carl Bosch, who was not only a chemist but also
a metallurgist by education. The conclusion
drawn by Carl Bosch was that if the material
tends to lose carbon by depletion of the pearlite,
the right material would be steel free of pearlite.
As carbon is the cheapest alloying element to
increase the strength of steel, the carbon free
material would not be able to withstand the
stresses in the high pressure process. The solu-
tion was a liner design with the so-called Bosch
holes in the pressure bearing shell, by which the
hydrogen diffusing through the soft iron liner
was released before it could accumulate and
create a high pressure at the interface. In this
way the pressure bearing reactor wall was pro-
tected from hydrogen attack. Figure 2 shows Figure 2. Design of a high pressure contact
the design of the ammonia reactor which went furnace
into service in Ludwigshafen in 1913 with a dai-
ly capacity of 30 tonnes (66,000 lbs). It is not known at what time within the project it
was realized that the hot hydrogen, and not the
nitrogen, was the critical process medium. In
any case, the safe control of the physical service
conditions of the process with temperatures be-
tween 400 and 500 °C (750 and 930 °F) and
pressures of about 200 bar (2900 psi) represent-
ed a technological quantum leap, when com-
pared to the fact that conditions in a normal
steam system at that time were 180 °C (360 °F)
Figure 1. Hydrogen attack in a catalyst tube and 10 bar (145 psi).
made of mild steel
What must be borne in mind is that Carl Bosch
and his coworkers had no access to technical
standards, codes and specifications of materials
as we have it today. Thus, it was very logical
that during the project a materials testing labora-
tory was founded at BASF. Initially the me-
chanical testing of the steels at different temper-
atures was conducted, and material
specifications for high pressure applications un-
der hydrogen attack conditions were elaborated.
These are still applied and have had a strong in-
fluence on the material development and stand-
ardization.

AMMONIA TECHNICAL MANUAL 168 2013

 The material testing laboratory was the nucleus that the service conditions remained similar to
for the materials engineering group at BASF, those in the past.
which is still closely cooperating with the am-
monia plants today. In the last years it has Nevertheless it was planned to replace the single
turned out that the primary reformer’s remaining damaged tube in April 2011 during a regular
life evaluation is a challenging task not only in shut-down. Some weeks before the shut-down,
ammonia, but also in syngas plants. Further- the tube failed (see Figure 3).
more, failure cases occurred in the high temper-
ature converter and the boiler feedwater pre-
heater of the CO-shift part of one of our
ammonia plants.

It will become evident that the ammonia process

still is a challenging one, not only for the mate-
rials, but also for the design and fabrication of
vessels.

Primary Reformers Residual Life

In one of its plants, BASF is running a reformer
furnace with 176 tubes in 4 rows so that each
row comprises 44 tubes. The furnace is top
fired and the furnace tubes consist of the cen-
tricast material G-X45NiCrNbTi 32 25 micro
alloy. They have an outer diameter of
123.2 mm (4.850 inches) and a minimum wall-
thickness of 10.8 mm (0.425 inches). The fired Figure 3. Ruptured centricast tube from the
tube length comprises about 12000 mm reformer
(≈39 feet). The design temperature for 100,000
hours of service is 970 °C (1778 °F), with a de- The inspection of the other tubes of the reformer
sign pressure of 25 bar (363 psi). revealed that a further five tubes had experi-
enced expansions of more than 2.5%. The ex-
After about 9 years of operation (71,000 hrs) the pansions were partly combined with internal in-
furnace tubes have been thoroughly inspected dications in the eddy current testing.
by eddy current and diameter measurement from
the outside. While one tube showed a damage In discussion with the plant manager and the
indication of 30% in combination with a local technical staff, it was discovered that problems
expansion of ~ 2%, which were attributed to a had occurred with the temperature distribution
possible local overheating - all the other furnace in the furnace. It was assumed that this was
tubes were free of indications. A survey of the caused by the catalyst filling. If the pressure
service conditions in the 9 years of operation drop in the individual tubes varies, the gas flow
demonstrated that the mean tube temperature will be uneven. In some of the tubes, very high
was 915 °C (1679 °F), which is considerably temperatures were measured near the bottom of
lower than the design temperature. It was con- the furnace which indicated a low gas flow, or
cluded that the tubes could be run for the next that the reaction occurred in the upper section of
two years without an operational risk, assuming the tubes.

2013 169 AMMONIA TECHNICAL MANUAL

 To counteract the flow maldistribution, the fol- mation of isolated and oriented cavities. In
lowing measures were tried: removing some of stage III, micro- and macro-cracks are formed
the catalyst from individual tubes, throttling the which lead to an accelerated damage evolution
burners, and installing insulation sleeves at the and failure of the component.
bottom part of the critical tubes. While these
measures were partly effective, they were not
able to reduce the temperature of the critically
affected tubes effectively. Hot spots were still
measured that considerably exceeded the design
temperature. Temperatures of 980 °C to
1000 °C (1796 °F to 1832 °F) were measured by
infrared camera (which can be somewhat inac-
curate, especially if there are other radiation
sources near the target as there are in case of
measurements near the bottom).

The metallurgical investigation identified that Figure 5. Creep curve with damage develop-
the failed tube ruptured due to creep damage as ment
can be recognized in Figure 4. No indication of
a fabrication defect was detected that could have In stage II the slope of the curve is more or less
contributed to the failure. constant, and here the Norton equation is some-
times applied in order to express the creep ve-
locity as a function of the applied stress.

min = K x Ơn

where n and K are temperature dependent mate-

0,2 mm
rial constants which are experimentally deter-
Figure 4. Creep fracture in the cross section mined.
Creep is a failure mode that happens at high Figure 6 demonstrates that the evolution of
temperatures above 400 °C (750 °F) with carbon creep damage in reformer tubes is very specific.
steels and 600 °C (1100 °F) with austenitic It is determined by the maximum stress/strain
steels. It is a diffusion controlled process af- resulting from the superposition of the internal
fected by the stress and strain acting on the pressure and the extreme heat flux through the
component and is consequently a steady, time- wall thickness. The damage evolution as
dependent damage process. demonstrated in Figure 6 is the result of a finite
element simulation which conforms quite well
Figure 5 shows a creep curve of a material with the practical experiences. Creep damage
where the creep strain is plotted over the expo- starts at about 1/3 of the wall thickness from the
sure time, while load and temperature are kept inner surface. At a later stage, when the void
constant. Three different damage stages can be damage has already reached the inner surface,
distinguished in the curve. In stage I, at the be- fissuring occurs in the most critically loaded ar-
ginning of the curve, reversible damage happens ea – at about 40% of the wall thickness from the
by a redistribution of cementite or dislocation inner surface. This kind of damage spreads out
reactions. In the second stage of the curve as well, until the whole cross section of the tube
(stage II) irreversible damage starts by the for-

AMMONIA TECHNICAL MANUAL 170 2013

 is affected. Before this stage is reached, the
load bearing capacity of the tube will be ex-
hausted and it will fail.

0,1 mm

Figure 8. Creep void chains and microcracks

Figure 9 demonstrates the final stage of creep

damage in a reformer tube with internal mac-
rocracks and void formation at the inner surface.
Figure 6. Creep damage evolution in a reformer
furnace tube [1]

Figures 7 and 8 show the different damage stag- 3 mm

es, oriented voids (Figure 7), void chains and
microcracks (Figure 8) as they were found in
metallographic investigations of centricast
tubes.

Figure 9. Final stage of creep damage in a tube

It is evident that this kind of failure evolution

represents a challenge when trying to determine
the state of the tubes at a point in time. In prin-
ciple, it is preferable to do an internal measure-
0,1 mm ment. This method has been promoted by
BASF in the past because the internal surface is
more defined for dimensional measurements of
the diameter compared with the external surfac-
Figure 7. Creep voids in the microstructure of a es where scaling appears. Furthermore, non-
centricast tube destructive testing from the inside of the tube
would be conducted closer to the area of the
first damage. Unfortunately, internal measure-
ments can only be performed if the catalyst is
removed, and therefore the time interval be-

2013 171 AMMONIA TECHNICAL MANUAL

 tween internal measurements easily exceeds 5 mm, the minimum sound wall thickness for a
years. This length of time is too long if the sim- design lifetime of 100,000 hr at 900 °C
ulation results shown in Figure 6 are taken into (1652 °F) was calculated to be 16.2 mm. In-
consideration. For this reason, more and more creasing the wall temperature from 900 to
the inspection of the reformer-tubes is conduct- 930 °C (1652 to 1706 °F) reduces the lifetime
ed with the external measuring tools provided from 100,000 to about 30,000 hr. Higher tem-
by the market. Most of them utilize a set of ed- peratures very easily lead to lifetimes which are
dy current probes and laser dimensional meas- much reduced.
urement devices.
In principle this curve at least qualitatively de-
What can be demonstrated by the analysis in scribes the situation of the failed tube. At first it
Figure 6 is that the time elapsed between the needed a very long time for the first creep void
first appearance of creep voids (about 1200 hr) formation due to moderate service temperatures
and the nearly complete damage (about 60,000 in relation to design. Then, the increased tem-
hr) is quite long. Prerequisite in this case is that peratures observed since 2009 promoted the ac-
the thermal and mechanical loading of the tubes celerated evolution of creep damage that ended
remains constant. In case of the failed tube in up in the tube failure in 2011.
the reformer (see Figure 4), the increased tem-
perature in the hot spots caused a dramatic in- Since that time the reformer has been retubed
crease in life-time consumption. As diffusion is and the reforming catalyst has been changed.
the fundamental process leading to creep, its Temperature measurements with different
time dependence is logarithmic, whereas the methods (infrared and thermocouples) indicate
temperature impact on the process is linear, as that there is no maldistribution of the tempera-
shown in Figure 10. As an example, the tem- ture in the furnace and that there are no hot
perature-time dependence of the creep process is spots. The maximum temperature is about
plotted here (Larson-Miller-plot) for a centricast 30 °C (86 °F) below design. The differential
tube out of the steel HK 40, material no. 1.4848 pressure between the tubes amounts to less than
(25 Cr/20Ni/0,4C–grade) at a constant pressure ± 5 % after loading the new catalyst. It is
load planned to conduct tube inspections by geomet-
rical and eddy current measurements in the fur-
nace every 2 to 3 years.

Failure of a Weld Lip Seal in a High

Temperature CO Shift Converter
Weld lip seals are applied quite often when
flange connections are used and hot gases at
high pressures are handled with high tightness
requirements. Figure 11 shows a standard de-
sign of a weld lip seal. The metallic gasket
blocks are pressed together by the flanges, and
the small sealing weld on top provides the tight-
Figure 10. Larson-Miller-plot for the centricast
ness. Most often the gasket, as shown in Figure
steel HK 40 [2]
11, is carbon steel. It has a good weldability
and can be ground open and rewelded several
According to the design equation in API RP
times before the gasket needs to be replaced.
530, for a tube with an outer diameter of 125.2

AMMONIA TECHNICAL MANUAL 172 2013

 ing during shut-down due to hydrogen embrit-
tlement and/or hydrogen-induced cracking.

Claddings are in principle quite susceptible to

this kind of damage because the solubility for
hydrogen in the austenitic and in the ferritic lat-
tice is quite different. The higher hydrogen sol-
ubility in the austenitic cladding can easily
cause an oversaturation in the adjacent ferritic
steel during cooling down. Below 150 °C
(302 °F) disbonding may occur due to the em-
brittling effect of the hydrogen on the ferritic
steel.

Figure 11. Design of a weld lip seal In case of a weld cladding, hard martensitic is-
lands appear along the fusion line in the ferritic
When hot hydrogen-containing gases at higher steel in areas which are partly molten and also
pressures play a role, steels resistant to hydro- in areas where austenitizing has occurred during
gen attack are applied such as P12 and P22 for the welding process. The high hardness is
the gaskets. The problem in this case is that caused by carbon diffusion in the direction of
these steels need to be heat treated after weld- the higher chromium alloyed austenitic steel.
ing, therefore the weld lip seal is cladded. The
blocks in direct contact with the gas atmosphere An explosion cladding does not exhibit such
consist of the P12 or P22 material. On top of hard areas and seams. The explosion causes a
this, relatively thick claddings made of austenit- very high energy release rate and acceleration of
ic steels are applied. In this way the sealing the clad steel. In the collision point of the two
weld can be made with low heat input and it can sheets, an extreme high temperature arises due
be assured that the base metal is not affected by to internal friction by which surface layers at the
the temperature of the welding process. This interface are removed so that the two materials
design also allows these sealing welds to be come into an intensive contact and are bonded
opened and rewelded several times. together. Undesired carbon diffusion leading to
hard spots and seams (such as for a weld clad-
In the specific case of the high temperature con- ding) cannot develop in an explosion cladding
verter, the inlet and outlet nozzles are connected during fabrication, because the time temperature
to the piping system by flanges with weld lip cycle even at the fusion line is not sufficient. It
seals. The gas temperature in the converter in is therefore believed that the explosion cladding
service is about 475 °C (887 °F) and the pres- is safer against disbonding.
sure is 33 bar (478 psi). The gas has a high hy-
drogen partial pressure, which made the selec- During a shut-down of the plant, the gas lip seal
tion of a P12 for the metallic gasket necessary. was opened at the gas exit pipe flange in order
On top of this, a 10 mm thick explosion clad- to conduct an internal vessel inspection. At that
ding of 321 SS (material no.1.4541) was ap- time the weld lip sealing was about 10 years in
plied. This design is shown in Figure 11. The service. After the shut-down the seal was re-
seal weld with a bevel preparation of 3 mm welded. During the leak testing, a leak was de-
depth will not affect the base material. The ex- tected not in the seal weld, but in the cladding
plosion cladding was selected in this case to interface between the two materials. The whole
provide a higher safety margin against disbond-

2013 173 AMMONIA TECHNICAL MANUAL

 weld lip seal was removed afterwards and inves- generate passivity. Hardness measurements in-
tigated in the materials engineering lab. dicate the decarburization in the P12 near the in-
terface. The average hardness in the P12 having
Figure 12 shows metallographic sections taken an ordinary ferritic-bainitic microstructure was
from the cladding. On the right side the section measured at 165 to 170 HV0,1. Values below
plane was laid perpendicular to the explosion 150 HV0,1 found near the interface are an indi-
extension direction. On the left side the section cator of decreased carbon content.
was taken parallel. The right side shows the fu-
sion line is straight, whereas on the left side the
fusion line has the typical wash-board like shape
indicating that the parameters selected for the
explosion cladding process have been in order.

0,05 mm
Figure 13. Fusion line in the explosion clad-
ding with hardness values

0,5 mm On the austenitic side of the interface, very high

hardness values exceeding 500 HV0,1 were
measured. These values are much higher than
Figure 12. Metallographic sections of an explo- the average values in the 321 SS cladding (240-
sion cladding 270 HV0,1). By the reduction of the solved ac-
tive chromium at the interface in the austenitic
Figure 13 shows the intact cladding parallel to part, it appears that the austenitic lattice is no
the explosion extension direction at a higher longer stable and a martensitic seam with chro-
magnification. The etching of the section was mium carbides has formed, made visible by the
conducted with a 3% alcoholic nitric acid solu- soft etching.
tion. It is remarkable that this solution, which
can only develop the microstructure in unal- The carbon diffusion representing the root cause
loyed or low alloyed steels, also etches the area for the metallurgical changes cannot have hap-
along the fusion line in the austenitic part. This pened during the cladding production process.
etching indicates that changes have happened It must have happened during the long term ser-
here which reduce the active chromium content vice time at high temperatures.
below the one necessary for passivation (about
11% by mass). Obviously carbon diffusion has Figure 14 demonstrates that the crack propaga-
taken place from the low alloyed P12 which re- tion follows the wash-board shape of the fusion
sembles decarburized into the austenite. Here line. The crack does not remain in one material,
chromium carbides have been formed which re- but switches from the base material to the clad-
duce the active chromium content necessary to ding side. The crack changes from the decarbu-

AMMONIA TECHNICAL MANUAL 174 2013

 rized soft base material to the extremely hard An investigation at higher magnification reveals
austenitic cladding at the interface. In a scan- that the crack is predominantly intergranular in
ning electron microscope the wash board shape the soft base metal, whereas in the hard cladding
of the crack surface is visible (see Figure 15). seam it appears to be a cleavage crack. In be-
tween some dimples are also detectable.

This fracture appearance is not typical for a hy-

drogen crack occurring at temperatures below
150 °C (302 °F). A typical hydrogen crack
would lead to a cleavage fracture appearance in
the softened P12 and to an intergranular crack-
ing in the hard martensitic seam in the cladding.
Therefore, it was concluded that hydrogen did
not play a dominant role in this case. The frac-
ture appearance indicates that predominantly
mechanical forces have caused the cracking.
These forces arise from the different thermal
0,1 mm expansion rates of the austenitic cladding and
Figure 14. Cracking along the wash-board the ferritic base material by which constant
shaped fusion line strains are induced (i.e., during any shut-down
and start-up). Due to the diffusion process, the
mechanical properties were changed over time
until the secondary stresses and strains were suf-
ficient to crack the hardened zone in the clad-
ding and the softened zone in the base metal.
Nevertheless, Figure 15 also shows some craters
that can be attributed to a synergistic action of
hydrogen.

In order to avoid similar cases, a program was

established in which all the weld lip seals with
service temperatures above 350 °C (660 °F) will
be investigated for this type of disbonding. We
assume that ultrasonic testing will be the appro-
priate non-destructive examination (NDE)-
procedure, although it may be not sensitive
enough. To ensure the capability of the ultra-
sonic testing, the first tests will be accompanied
by metallographic investigations.

Leaks in the Boiler Feedwater

0,5 mm Preheater of the Ammonia Plant IV
Figure 15. Fracture surface of the explosion in Ludwigshafen
cladding in an SEM
The boiler feedwater preheater belongs to the
CO-shift unit and is installed directly down-

2013 175 AMMONIA TECHNICAL MANUAL

 stream of the high temperature converter with wall by fillet welds. The design aims towards
the weld lip seal failure described above. The an optimization of the gas distribution on the
preheater consists of a 6500 mm (21 feet, shell side. The supports of the sleeve, which are
4 inch) long shell with an outer diameter of made of P1 grade steel, also act as guidance for
1570 mm (5 feet, 2 inch) and is in the horizontal the plug-in module.
position. On one end there is a large flange
where the U-tube bundle is inserted and fixed.
The U-tubes are made of the austenitic 321SS-
grade. The flange and the vessel are made of
high strength, low alloyed steel WB 36, material
no. 1.6368, for which there is no equivalent in
the ASME Code. The material is quite low in
the carbon content (< 0.17% by mass) and is al-
loyed with Ni (1.0-1.3% by mass), Mo (0.25-
0.45% by mass), Cu (0.5-0.8% by mass) and Nb
(<0.2% by mass). The steel is normally re-
quired to be hardened and tempered. Figure 16. Principle sketch of the feedwater
preheater
The alloying with Cu and the microalloying
with Nb result in considerable strength - yield During service the gas atmosphere on the shell
strength > 430 MPa, (62.4 kpi) at room tem- side consists of HTS effluent gas (38 mol% H2,
perature and > 343 MPa (49.8 kpi) at 400 °C 19 mol% N2, 12 mol% CO2, 2.4 mol% CO, 29
(752 °F). There is no restricted weldability due mol% H2O and traces of methane, Ar, He). The
to the low carbon content. Nevertheless the gas entering the vessel has a temperature of
steel has a quite high potential for high hardness 360 °C (680 °F) at about 29 bar (420 psi) pres-
values in the heat affected zones of welds which sure, leaves the shell at 215 °C (419 °F) and
necessitates a post weld heat treatment inde- heats the feedwater from 140 °C (284 °F) to
pendent of the wall thickness. This heat treat- 265 °C (509 °F).
ment is required to be conducted at 530 to
620 °C (986 to 1148 °F) at least 30 °C (86 °F) In 2010 a leak occurred in the shell of the vessel
below the tempering temperature. Besides the which was in service since 1980. The leak was
release of residual stresses, the main goal of this caused by a 400 mm (≈16 inch) axial crack in
heat treatment is the reduction of hardness in the the 7°o’clock-position near the transition to the
welds. bottom. The crack also covered the circumfer-
ential weld between the cylindrical part and the
Due to the excellent strength properties of the bottom.
material, the wall thickness of the shell for the
design parameters of 33 bars at 400 °C (478 psi After removal of the internals it was discovered
at 752 °F) could be restricted to 15 mm that the crack followed the heat affected zone
(0.59 inches) in the undisturbed part and to (HAZ) of the fillet weld to one of the supports
22 mm (0.87 inches) in the area of the gas inlet made of P1. Internal inspections with magnetic
and outlet where reinforcing was required. particle testing revealed that the fillet welds of
other supports were also affected. For the re-
Figure 16 shows a sketch of the vessel. In its in- pair, a window was cut out of the shell with a
terior, a plug–in module is inserted consisting of height of 200 mm (≈8 inches) and a length of
a perforated sheet (indicated red). This sheet is 600 mm (≈24 inches) and a new plate fitted to
fixed to the wall by steel supports welded to the this was welded in. The cracking along the fillet

AMMONIA TECHNICAL MANUAL 176 2013

 weld of the supports was ground out. Where too
little wall thickness of the shell remained, re-
welding was conducted followed by a local heat
treatment.

Figure 17 shows the extracted plate from the

outside (A) and inside (B). At the inside, the
support welded to the shell is visible. In the
small section in Figure 17(C) the course of the 1 mm
crack along the fillet weld is demonstrated. 0,1 mm

Figure 17. Cracking in the piece extracted from

the shell of the feedwater preheater
Figure 18. Intergranular cracking in the heat
Figure 18 shows the results of the metallograph- affected zone of the fillet weld, thick oxide lay-
ic investigation. Here the crack starting at the ers and side cracking
HAZ of the weld is shown in the section of the
plate where the crack had not penetrated the In the base material and in the transition from
shell’s wall. The crack propagates under a HAZ to base material, the crack path is predom-
slightly pointed angle along the HAZ before it inantly transgranular with some intergranular
deviates into the unaffected base metal of the parts, mainly in areas where branching occurs.
shell. Figure 18(A) gives an overview and Fig- Hardness measurements conducted in the HAZ
ure 18(B) shows that the crack path is intergran- in the crack initiation region revealed very high
ular in the HAZ with considerable oxide layers hardness values of 332 to 407 HV0,2. After the
on the crack borders. In addition, side cracking repair of the vessel, hardness measurements
along the grain boundaries is present (see red ar- were performed and the hardness measured in
rows). the rewelded fillet weldments was 200 –
220 HV0,2 in the base metal, 200 – 240 HV0,2
in the weld metal and 260 – 280 HV0,2 in the
HAZ. In the patch weld, similar hardness val-
ues were measured in the weld and base materi-
al on the inside and outside. In the HAZ, hard-
ness values of up to 300 HV0,2 were measured.
With this repair the vessel was returned to ser-
vice with the intention to replace it in 2013.

2013 177 AMMONIA TECHNICAL MANUAL

 In March 2012 a pressure test being conducted HAZ of the circumferential weld was intergran-
during a shut-down indicated a leak in the noz- ular and that the crack surface was covered with
zle tube of the vessel flange. No NDE had been quite thick oxide layers. These oxide layers
performed in this area during the failure in 2010 clearly indicate that the cracking had been pre-
because the internals had not been removed sent for some time.
here. Steel supports are not present here, there-
fore only external UT-testing had been done
with no indication of issues.

Figure 19(A) gives a front view of the flange.

The leak was in the 10 o’clock position. Crack-
ing was also found in the 2 o’clock position. In
both cases the damage was linked to internal
welds.

Figure 20. Cracking mainly along the HAZ of

the circumferential weld on the flange side lead-
Figure 19. View of the flange in feedwater pre- ing to a leakage (A), intergranular cracking and
heater (A) and cracking at the inside of the thick oxide layers at the crack borders (B, C)
10 o’clock position along the weld up to the
flange (B) Hardness measurements in the welds revealed
values of up to 415 HV0,2 in the HAZ of the
Figure 19(B) shows the internal surface after a circumferential weld where most of the cracking
magnetic particle test. Cracking appears along happened. A simulated stress relief heat treat-
the welds and has crossed the circumferential ment of the specimens of the first and second
weld to the flange on the left side of the picture, failure case resulted in a considerable reduction
as indicated by the arrows. In order not to dis- of the hardness values by more than 60 Vickers
tort the flange by grinding and rewelding, two units. It was concluded that the vessel had not
windows were cut into the shell - one in the 10 been properly heat treated after welding during
o’clock and the other in the 2 o’clock position - fabrication, especially in the case of the fillet
which could be replaced in a more controlled welds to the bars and supports made of P1. Due
manner. During the metallographic investiga- to the extreme hardness, especially in the HAZ
tion of the pieces, it was found that the cracking of the welds, intergranular cracking was initiat-
started from the inside and outside in areas of ed during shut-down. At that time condensation
welds. The penetration through the wall result- of water occurred and the CO2 acidified the
ing in the leak, however, was in the circumfer- condensate so that a more active corrosion oc-
ential weld. curred. The hydrogen produced by the corro-
sion process caused the cracking. It cannot be
Figure 20(A) demonstrates the cracking leading completely excluded that hydrogen from the
to the leak at the circumferential weld to the Schikorr reaction also played a role, however
flange (right side in the picture). Figure 20(B) the long term failure evolution made this expla-
and 20(C) make evident that the cracking in the nation less likely.

AMMONIA TECHNICAL MANUAL 178 2013

 Meanwhile a new feedwater preheater has been start-up when hydrogen is formed due to a cor-
installed which has been built out of P22 mate- rosion process.
rial (2¼ Cr, 1 Mo steel). Welding and post weld
heat treatment have been accurately monitored Finally, as demonstrated in the case of the
and documented during vessel fabrication. feedwater preheater, the importance of quality
surveillance in the vessel fabrication shop, espe-
Conclusion cially if high strength materials are applied, is
very high.
The failure cases described cover the wide range
of physical and chemical service conditions that
occur in an ammonia plant, and which need to Literature
be considered in design and fabrication. In the
high temperature range as in a reformer, creep is
1. Materials for Steam Reformer Furnaces
the dominating damage mechanism, whereas in
Seventh Status Report, Battelle Memorial
the mid temperature range, weldability, hydro-
Institute, Columbus, Ohio, October 1972
gen embrittlement, and resistance to hydrogen
attack are critical.
2. John J. Jones; Re-Tubing Reformer Furnac-
es; Technical Paper, Doncasters Paralloy,
For high temperature creep service such as fur-
Billingham, U.K., April 1992
nace tubes, there are quite sensitive NDE tools
available on the market that enables a reliable
evaluation of the remaining life. What must be
kept in mind by plant personnel is the very high
sensitivity of the creep damage process regard-
ing the temperature. A temperature maldistribu-
tion with hot spots or individual tubes being
overheated may result in service life being easi-
ly reduced by a factor of three or more. Extrap-
olations regarding the life time of the furnace
can therefore be completely wrong if tempera-
ture increases. It is necessary to carefully moni-
tor and control the temperature and temperature
distribution in the furnace on a regular basis.

The failures that have occurred in vessels in the

CO-shift conversion demonstrate that not only
should hydrogen attack be considered within
ammonia plants (as historically has been the
case) but also hydrogen embrittlement and
cracking. Secondary stresses which arise due to
different thermal expansion rates in claddings
and dissimilar welds must also be considered.
High material hardness - which can be caused
by an insufficient fabricational process or by
operational conditions - may easily result in crit-
ical cracking and leakage during shut-down or

2013 179 AMMONIA TECHNICAL MANUAL