Reformer Component Management

After an Overheating Incident that

Resulted in Tube Failures
A significant overheating incident occurred in CSBP’s reformer in May 2011 after 12 years of
operation. Prior to the incident, the reformer was restarted after a hot trip and the furnace was
inadvertently restarted using an incorrect procedure. Seven reformer tubes failed as a result of gross
overheating. The outlet bull tee to transfer line welds had a low to moderate level of cracking. After
the incident, conventional eddy current inspection indicated that the remaining intact reformer tubes
were badly cracked. However, subsequent metallographic assessment showed this was not the case
and LifeQuestTM reformer assessment showed all the remaining tubes could operate safely to the next
planned shutdown. The outlet bull tee to transfer line welds were successfully repaired using a
procedure that included high temperature heat treatment and temper bead welding.

D. M. Firth, Metallurgist
Quest Integrity Group New Zealand

C. W. Thomas, Metallurgist
Quest Integrity Group New Zealand

Manoj Thakur
CSBP Limited Australia

Ron Caporn
CSBP Limited Australia

hot trip and the furnace was inadvertently re-

started using an incorrect procedure. The plant
Introduction was immediately shutdown to identify the dam-
age and ensure it was fit-for-service before the

A
significant overheating incident oc-
curred in the primary reformer at plant was restarted. Reformer failures have oc-
CSBP’s ammonia production facility in curred on a number of occasions in other plants
Western Australia in May 2011 that re- when they are restarted [1], [2].
sulted in reformer tube failures. The failure oc-
curred when the plant was being restarted after a

2013 187 AMMONIA TECHNICAL MANUAL

 In late 2008 a similar overheating incident oc- The Furnace Operation
curred to the reformer at CSBP, and although no
tubes failed at the time, three were replaced The primary reformer had operated for approx-
based on inspection findings. imately 12 years at the time of the incident in
2011. The furnace has two cells (East & West)
The CSBP ammonia plant was designed by each with 36 tubes. Preheated gas/steam enters
Haldor Topsoe and built under contract as a the top of the tubes through inlet pigtails and
turn-key project. It was commissioned in 2000. syngas exits the bottom through pigtails to two
Initially operating at the design capacity of hot manifolds. These are then connected via
650 MTPD, the plant was later upgraded to bull tees to a refractory-lined transfer line. The
745 MTPD capacity in 2004. CSBP also has reformer is side-fired by four rows of eight
two nitric acid plants, one ammonium nitrate burners on either side. The flue gas exits from
plant and two sodium cyanide plants. The re- the top of the furnace and then descends through
former in the ammonia plant is shown in Figure the convection section.
1.
The following materials are used in the radiant
and convection sections:
 Catalyst tubes are Modified HP 50 (134 mm,
5.3 in OD).
 Outlet pigtails are Inco 800H.
 Outlet manifolds (4) are cast 20Cr-32Ni-
1Nb (PG 20/32Nb-16).
 Bull tees (2) are cast 20Cr-32Ni-1Nb (PG
20/32Nb-16).
 Transfer line is carbon steel refractory lined.
 1st coil mixed feed stainless steel 321 + P22
(finned).
 2nd coil process air stainless steel 321H +
P22 (finned).
 3rd coil steam superheat stainless steel 321H
+ P22 (finned).
 4th coil natural gas preheater is P11.
 5th coil HP steam is P11 (finned).
 6th coil process/combustion air is A106 GrB
(finned).
 7th coil natural gas preheater is A106 GrB
Figure 1. Primary Reformer (finned).
 8th coil BFW preheater is A106 GrB
This paper outlines the cause of the overheating, (finned).
the resulting damage, and the actions taken to
ensure the plant was fit-for-service while mini- Normal Operation of the Reformer
mizing the downtime.
The reformer normally operates at a very steady
rate with the following conditions:
 Reformer outlet temperature of 760 °C to
770 °C (1400 °F to 1418 °F).
 Outlet pressure of 3070 kPa (445 psi).

AMMONIA TECHNICAL MANUAL 188 2013

  Flue gas temperature at the first coil in the Overheating in 2008
convection section is 1023 °C (1873 °F).
 Catalyst tube metal temperature measured An overheat incident occurred in November
by pyrometer is 850 °C to 900 °C (1562 °F 2008 during the restart of the reformer after a
to 1652 °F). trip. At the time, the steam flow to the natural
gas feed to the reformer tubes (Figure 2) was in-
 Steam / Feed NG ratio is 2.2 - 2.35.
sufficient. The lack of steam flow resulted in
carbon buildup on the catalyst in the tubes, re-
These conditions are typical of many ammonia
striction of flow in the tubes and overheating of
plants where a secondary reformer is also part of
the tubes. The steam flow to the gas feed was
the process path. Given the relatively modest
zero as it was being vented, but the flow meter
operating condition, the catalyst tubes and outlet
was showing flow as it was installed before the
components are expected to have relatively long
vent valve.
lives.

Figure 2. 2008 overheat incident showing important process parameters

troduced within 10 minutes after introducing

Overheating in 2011 steam. The manual valve on the gas line was
opened without realizing that tubes had failed.
On May 23, 2011, the plant tripped as a result of
a low gas supply. Approximately 1¾ hours af- The timing of the change in feed to the reformer
ter lighting the burners, steam was introduced to tubes and the measured process gas exit temper-
the catalyst tubes. Tubes ruptured as soon as atures are shown in Figure 3. During the inci-
steam was introduced, 18 minutes after lighting dent, the following temperatures were recorded:
the burners. A fire occurred in the firebox when  Reformer outlet temperature is normally
the natural gas feed was introduced. Operating 760 °C to 770 °C (1400 °F to 1418 °F). The
procedure governs that natural gas should be in- maximum detected was 939 °C (1722 °F) at

2013 189 AMMONIA TECHNICAL MANUAL

 the bull tee and 868 °C (1594 °F) in the probably low as cold air was sucked into the
transfer line. air damper.
 Flue gas temperature (prior to first coil) is  Flue gas at the natural gas first preheat coil
normally 1023 °C (1873 °F). The maximum is normally 551 °C (1023 °F). The maxi-
was 1030 °C (1886 °F). It is believed that mum measured was 575 °C (1067 °F).
the convection section temperature was

Figure 3. 2011 overheat showing important process parameters

Apart from the catalyst tubes, the only compo- ognizing that the tubes were overheating prior to
nents that were considered to have been dam- introduction of the steam feed to the catalyst
aged due to overheating were the outlet pigtails tubes. A trip or interlock function to prevent the
and manifold. incorrect startup sequence did not exist.

Root Causes of Each Incident Mitigation in Place to Prevent

Overheating and to Even Out Tube
Overheating Incident in 2008 Temperatures
The root cause of the 2008 incident was human The following steps were taken after the 2011
error made by operations, as they did not recog- incident to prevent overheating and to even out
nize that the steam feed was still venting in by- tube temperatures:
pass and had not yet been introduced to the re-  The bottom 5 meters of the reformer tubes
former tubes before introducing the natural gas were insulated, where higher creep strain
feed to tubes. damage was measured.
Overheating Incident in 2011
 To maintain the maximum flow through the
The root cause of the 2011 incident was also tubes, the exit temperature was sacrificed.
human error by operations in not visually rec-

AMMONIA TECHNICAL MANUAL 190 2013

  Increased the process steam:gas ratio to a  The bull tees.
maximum (steam:gas is now 2.35) to reduce
the tube temperatures. As result of these risks, the components had
 Continuous monitoring of the tube skin tem- been inspected on a regular basis at up to a four
perature and turn off burners where hot year interval by:
tubes are identified.  Monitoring the creep strain in the reformer
tubes.
In addition, new interlock logic functions were  Inspection for cracking at accessible welds.
implemented including the following:
 Steam flow conditions (including minimum The following detailed inspections had not been
steam temperature, pressure and flow to the carried out:
catalyst tubes) are now permissives for the  Comprehensive measurement of the strain in
natural gas feed to re-open. the pigtails as the calculated life indicated
 A startup trip on the fuel to the reformer if that a low life fraction was used.
there is no steam feed to catalyst tubes and  Extensive dye penetrant testing of the welds
the flue gas temperature exiting the fire box on the bull tees.
is too high (in the case of this plant > 650 °C  Measurement of the movement of the outlet
or 1202 °F). headers during startup and shutdown.

After two-years in service (in 2002) and due to a

Damage Predicted and Damage catalyst change, the reformer was inspected by
Found during Normal Shutdown Quest Integrity using their proprietary LOTISTM
Inspections tube inspection method. This inspection
showed that the strain was very low and was ef-
A risk based assessment and inspection plan fectively a baseline. It also showed the highest
(RBA/RBI) of the pressure equipment in the strain was at the bottom of the tubes. Manual
ammonia plant was completed in 2004. This measurement of the circumference of the re-
plan indicated that there were no significant former tubes using a pi tape and calculation of
high risk items in the reformer or the convec- the growth strain in the tube was carried out at
tions section. The key issues identified were the bottom and 1 m from the floor of the re-
high temperature creep cracking of the follow- former in 1999, 2001, 2004, 2006, 2008, 2009
ing components: (after the first overheating incident) and 2011
 Reformer tubes, especially if the reformer (after the second overheating incident). The re-
was unevenly balanced. sults are summarized in Figure 4.
 The outlet pigtails.
 The outlet manifolds.

2013 191 AMMONIA TECHNICAL MANUAL

 Estimated Tube Growth taking acount step changes due to
overheating WEST
5.00%

4.00% WEST Minimum

Growth (%)

3.00% WEST Average

2.00% WEST Maximum

1.00%

0.00%
00

01

02

03

04

05

06

07

08

09

10

11

12

13

14

15
0

0
-2

-2

-2

-2

-2

-2

-2

-2

-2

-2

-2

-2

-2

-2

-2

-2
n

n
Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja
Date of Test

Estimated - Tube Growth taking acount step changes due to

overheating EAST
5.00%

4.00% EAST Minimum

Growth (%)

3.00% EAST Average

2.00% EAST Maximum

1.00%

0.00%
00

01

02

03

04

05

06

07

08

09

10

11

12

13

14

15
0

0
-2

-2

-2

-2

-2

-2

-2

-2

-2

-2

-2

-2

-2

-2

-2

-2
n

n
Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Ja

Date of Test

Figure 4. Strain measurement in the bottom of the reformer tubes. The data directly prior to the inci-
dents in 2008, 2011, and 2015 were estimated.

Damage Observed After the 2008 outside diameter strain of 2-2.5%. The inspec-
Incident tion also indicated 14 tubes had over 30% wall
cracking. The three tubes with the highest strain
After the overheating incident in 2008 the dam- were replaced. Following the removal of the
age found was as follows: tubes, no metallurgical assessment was carried
out to confirm the results of the inspection and
Reformer Tubes quantify the degree of creep damage.

The reformer tubes were inspected in early 2009 Bull Tee

using an external crawler to measure the tube
strain and inspect the tubes using an eddy cur- Dye penetrant testing was performed in 2009
rent technique. This showed one tube had an around the welds connecting the manifold arms
outside diameter strain >3% and four tubes with and the outlet cone to the two bull tees. This in-

AMMONIA TECHNICAL MANUAL 192 2013

 spection showed some isolated random indica-  Three failures had occurred level with the
tions of fine cracking up to 3.5 mm long adja- lower burners - 1.7 m (67 in) from the bot-
cent to the tee to outlet cone weld. Replication tom of the tube or about 1.2 m (48 in) above
showed that creep voids were present next to the the floor of the furnace.
cracking. No crack indications were present in  Four failures had occurred level with the
the bull tee to header welds. second row of burners – 3 m (10 feet) from
the bottom of the tube or about 2.5 m
After the cracking was detected, the movement (8 feet) above the floor of the furnace.
of the ends of the outlet headers and the move-  The outer surface of the reformer tubes was
ment of the bull tee were surveyed to determine more heavily oxidized in the bottom 3 m
if there was any measureable bending load ap- than the top of the tubes. When the tubes
plied to the bull tees. No difference in move- failed, the reforming catalyst within the tube
ment was detected between the bull tee and the was expelled.
ends of the outlet header.  Bowing was observed in some of the re-
former tubes, especially in the region next to
Other Items some of the failed tubes. Bowing had been
noticed on a previous shutdown but to a
The plant sustained coking damage to the cata-
lesser extent.
lyst charge. However, no other equipment or
metallurgical damage was observed in 2009.
Decoking was first tried by introducing steam
(without gas) to reduce it. It actually worked,
but loose carbon produced from this exercise
ended up increasing the pressure drop across
tubes. After few weeks of running with reduced
rates, the catalyst had to be replaced.

Damage Observed in 2011

Figure 5. Typical reformer tube failure
After the 2011 overheating incident the damage
found was as follows: The reformer tubes which had not failed were
inspected using an external crawler to measure
Reformer Tubes the tube strain and inspected using an eddy cur-
rent technique. This inspection showed 21 tubes
Seven reformer tubes were found to have failed had an OD strain >3% and over 90% of the
after the incident in 2011. The following was tubes had >30% wall cracking. 10 tubes had
observed: >45% wall cracking. The results also showed
 The failures had wide open fractures typical that the maximum strains were all located near
of overload due to short term overheating. the bottom of the tubes. The company that pro-
They ranged from 330 mm to 1200 mm (13 vided the diameter and eddy current assessment
in to 47 in) long with fracture openings of on the tubes, recommended at the time of the in-
40 mm to 350 mm (1.5 in to 13.8 in). A typ- spection, that all (100%) the tubes should be re-
ical failure is shown in Figure 5. placed as soon as possible.
 The reformer tubes were bulged at the fail-
ures. The highest OD strain in the tubes was Manual strain measurements were also repeated
200 mm (7.8 in) from a failure and was at the bottom of the reformer tubes after the in-
measured to be 3%. cident (see Figure 4). These manual measure-
 Tube failures occurred in both cells. ments correlated well with the highest meas-

2013 193 AMMONIA TECHNICAL MANUAL

 urement made by the external crawler on each overheating. Stress relief cracks (sometimes re-
tube as position of the highest strain was in the ferred to as stress relaxation cracking) in high
same locations. nickel/chromium alloys, such as 20Cr-32Ni-
1Nb, is known to occur [3]. This occurs as a re-
Bull Tee sult of localized creep in the HAZ of the weld
that is driven by residual welding stresses and
Dye penetrant testing was carried out on the bull the operating temperature. The cause is inde-
tee welds after the overheating incident. This pendent of the pressure or system stresses.
inspection showed that the extent of cracking at
the bull tee to transfer line weld had increased Other Components
from 2009. The cracking present was primarily
at the edge of the weld beads either in the weld Extensive inspection of the coils and refractory
heat affected zone (HAZ) or in the middle of the in the convection section was carried out. No
weld. The cracks ran in a range of directions. significant damage was detected.
Typical cracking is shown in Figure 6.
Examination of the outlet pigtails revealed that
they were about 1 mm (1%) larger in circumfer-
ence at the reformer end than at the manifold
end due to creep strain. It was not known as to
what proportion of this strain occurred as a re-
sult of normal service or the over-heating inci-
dents. No evidence of any creep damage or
cracking at the pigtail to manifold welds or pig-
tail to reformer tube bottom cone welds was de-
tected. The pigtails have a very flexible design
and as a result cracking at the welds was not ex-
pected.

Life Assessment and Management

of Reformer Tubes
A tube was removed from the reformer that had
been identified as having 45% wall cracking by
the eddy current inspection. The areas of high-
est wall cracking and highest strain (3.95%)
were removed from the tube. Macrographic ex-
amination of a polished ring and microscopic
examination showed that there was no evidence
Figure 6. Cracking in the bull tee welds of any cracking, but creep voiding typical for
Note: the circular punch indentations were used the level of strain measured, was present (Fig-
as reference points. ure 7).
Microstructural examination revealed that creep
voiding was present in the weld adjacent to the
cracking and that the cracks were typical of
creep voiding. It was considered that the crack-
ing probably occurred as a result of stress relief
cracking that could have been enhanced by the

AMMONIA TECHNICAL MANUAL 194 2013

 overheating by dissolving secondary carbides
and spherodising primary carbides (Figure 9).
From the results of short term heat treatment tri-
als intended to reproduce damage as observed in
the overheated tubes, it was estimated that the
tubes had experienced temperatures in excess of
1200 °C (2192 °F) in the areas of the failures.
Nevertheless, no evidence of melting was seen
indicating the tubes had not been heated to
1400 °C (2552 °F).

Figure 7. Microstructure of the worst tube

identified by eddy current testing of the tubes
showing the typical level of creep voiding

Eddy current testing was carried out in the la-

boratory using a single point probe over a range
of frequencies from 2 to 100 Hz on the follow-
ing:
 A sample taken from the top of reformer
tubes that had not seen any significant over-
heating.
 A sample from an overheated section of re- Figure 8. Typical ex-services modified HP50
former tube that had seen significant over- showing minor creep voiding
heating, but showed metallographically not
to contain cracking.

The result indicated a significant eddy current

response difference. This was probably due to
the different microstructure caused by the ef-
fects of overheating. This “conventional” eddy
current testing can readily give indications
which can be interpreted as cracking but are
due to other changes such as the oxide thick-
ness and microstructure. Techniques such as
"Cross Cancelling" eddy current testing are
probably more reliable as a backup to strain Figure 9. Typical microstructure adjacent to a
measurement (as an indicator of some form of failure in a tube
damage occurring).
The remaining life of the surviving CSBP re-
The typical microstructure of a tube removed former tubes was estimated using the Quest In-
from “normal” service after over 10 years typi- tegrity LifeQuest™ Reformer method [4]. This
cally consists of modified primary carbides, life assessment methodology has been devel-
coarsened secondary carbides, and fine creep oped specifically to process data obtained from
voids depending on the loading (Figure 8). The tube inspections based on tube diameter meas-
microstructure of the failed tubes had been mod- urement.
ified from the “normal” structure as a result of

2013 195 AMMONIA TECHNICAL MANUAL

 A number of “virtual” tubes were created, each at relatively low strain rates is an intergranular
having a different amount of strain. The levels creep process. These different damage mecha-
of strain were arranged to be close to the strain nisms are not directly additive. As the assess-
seen in the bottom of the tube where it is under- ment was made assuming that all the strain oc-
stood, the greatest measured strain was located curred by low strain rate creep, the results in
in the CSBP tubes. The past history, tube ge- Table 1 are conservative. However, it is consid-
ometry, and operating pressure applicable to the ered non-conservative to assume the additional
CSBP tubes was entered into the LifeQuest™ strain caused by the overheating incidents did
Reformer model. This model assumes that the not contribute to damage as far as the life as-
tube material ages with time and that the over- sessment was concerned.
heated material in the tubes will have a future
creep strain rate equivalent to heavily aged ma- The LifeQuest™ Reformer method does not
terial in “normal” operation. Assessment of have built-in safety factors. In accordance with
other overheated material by creep testing has the API 579 [5], the recommended retirement
shown this to be case. The results obtained are date is set at the date at 80% of the predicted
listed in Table 1. failure date (Table 1).
Maxi- Total
mum OD Rupture Predicted
80% Damage
Recommended
Creep Effec-
tive Temp
Number of
tubes above
No tubes remained in service after the incident
strain in time from Failure
2011 2000 Date
Retirement
Date
°C
(°F)
the strain
figure*
in 2011 that had strains >4%, so it was recom-
(%) (hrs)
Nov July 813
mended to limit the installation of new tubes to
1 367091 0
2041 2033 (1495) the seven failed tubes. It was later recommend-
Jan Jan 823
1.5 263575
2030 2024 (1513)
19 ed that the 14 tubes measured to have more than
2 210846
Jan March 829
20
3% OD strain in June 2011 should be replaced
2024 2019 (1524) at the planned outage for late 2011 if they are
Nov Sept 833
2.5 182646
2020 2016 (1531)
22 confidently to survive operation until June 2014.
3 159023
Feb July 831
7 A total of 20 tubes were replaced in the next
2018 2014 (1527)
Sept May 833 planned outage, during October 2011. The ad-
3.5 146824 7
2016 2013 (1531) ditional 6 tubes were replaced as a result of
March March 842
4 133631
2015 2012 (1548)
0 measurement of the tube strains and for con-
Table 1. Summary of LifeQuest™ Reformer servatism.
results
* Number of tubes in 0.5 mm strain range above Repair of Bull Tees
the strain figure (i.e., for 2% strain it is the
number of tubes with 2-2.5% strain). After the 2011 incident, it was recommended
that the bull tees be repaired. It was decided
It is believed that the two overheat incidents that these repairs should be done in the October
have contributed to the strain. The analysis of 2011 shutdown. Bull tees are known to be diffi-
tube diameter data has suggested that there was cult to repair as a result of poor weldability and
a step jump in the mean measured tube strain high temperature rapid liquation cracking in the
(Figure 4). This step change means that the HAZ [3]. As a result, it is common practice to
“normal operation creep strain” would have carry out a high temperature (1150 °C, 2103 °F)
been a little less than that used in the assess- solution treatment prior to welding. It is also
ment. It may appear inappropriate to assume known that bull tees in operation within metha-
that any strain that occurred during overheat in- nol plants with a typically higher operating tem-
cidents would not lead to damage. At high perature than ammonia plants can be successful-
strain rates under load, the damage mechanism ly repaired by orbital machining and orbital
is more like tensile deformation, leading to bulk pulsed TIG welding.
deformation of the material. Normal operation

AMMONIA TECHNICAL MANUAL 196 2013

 A repair procedure was developed to machine ing incidents. The damage was probably
out the damage orbitally and weld repair using stress relief cracking.
orbital TIG without a preheat treatment. Unfor- 6. Liquation cracking in the HAZ occurred
tunately cracking immediately occurred in the during the weld repair of the bull tees with-
HAZ of the welds, probably due to liquation out a prior heat temperature solution anneal.
cracking [3]. A high temperature solution an- Welding after a solution anneal at 1090 °C
neal was carried out with the aim of heat treat- (1994 °F) was successful.
ing at 1150 °C (2102 °F) for a short period with 7. A temper bead weld procedure was used to
a rapid fan cool. Regretfully the maximum try to minimize the risk of stress relief
achieved temperatures were 1088 °C (1990 °F) cracking of the repaired bull tees.
and 1097 °C (2006 °F) due to heating element
failures, but welding was then successfully
completed without any cracking occurring. A References
temper bead procedure was used in the capping 1. D M Firth, P Tait: Assessment of metallur-
layer to minimize the risk of stress relief crack- gical damage caused by the overheating of a
ing while the reformer is in service. catalytic reformer. New Zealand Metals and
Materials Symposium 1989.
Conclusions 2. D H Timbres and M McConnell: Primary
Reformer Failure. AICHE 2002.
1. The 2011 overheat in the CSBP occurred as 3. Materials, Fabrication, and Repair Consider-
a result of a failure of field operations to ations for Hydrogen Reformer. Furnace
visually recognize that the reformer tubes Outlet Pigtails and Manifolds. API Publica-
were overheating prior to admitting steam to tion Revised Draft 03, APRIL 2011.
the reformer feed. The need for protective 4. Charles Thomas et al: Reformer Life As-
trip function has been recognized to prevent sessment - Development of a Materials Da-
further failure. tabase. AICHE 2008.
2. Seven tubes failed as a result of overload at 5. API 579 Fitness for Service (2007).
temperatures above 1200 °C (2192 °F).
3. Conventional eddy current inspection of re-
former tubes indicated that the majority of
tubes in the reformer were badly cracked.
Subsequent laboratory examination showed
that no cracking was present and indicated
that the eddy current response was probably
due to microstructural changes that had oc-
curred due to the overheat.
4. LifeQuestTM Reformer, reformer tube life
assessment tool, indicated that a large pro-
portion of the tubes in service had long life
expectancy and that the plant could be oper-
ated immediately after the seven failed tubes
were replaced. In addition, only 14 tubes
were recommended to be replaced at the
next shutdown.
5. Cracking was detected in the bull tees that
may have been exacerbated by the overheat-

2013 197 AMMONIA TECHNICAL MANUAL

AMMONIA TECHNICAL MANUAL 198 2013