Bimetallic and Alloy Welds in HP

Hydrogen and Nitrogen Service

Three failures of bimetallic welded joints in ammonia plants occurred in 1999 and in 2000 at one
inlet and in two shells of Ammonia Converter Effluent Exchangers (123Cs). Some of the
metallurgical failure investigation completed for one of the welded joints is reviewed. These failures
indicated prudent action was required in order to provide safe repairs for return to service. An
'Advisory' was issued to other operators in order to avoid failures of similar welded joints in other
plants.

Jim B. Sievert
Kellogg Brown & Root, Inc., Houston, TX 77001

Introduction ducted by Metallurgical Consultants, Inc. in Houston,

TX. What the investigation found was a small zone of

I
n June of 2000, a weldment failure caused a high hardness adjacent to the nickel butterpass in the
process side leak in an ammonia converter efflu- low-chrome ferritic shell forging on one side of the cir-
ent boiler feed water exchanger (item 123-C1) at cumference of the heat exchanger. This very small,
Palembang, Indonesia (Pusri IB). The failure will be localized hard zone rendered the area susceptible to
discussed within this article. A similar weldment failure cracking and subsequent separation due to hydrogen
had occurred earlier at the P.T. Petrokemia Gresik embrittlement and crack separation. Generally, these
Indonesia at an inlet nozzle in 1999, and a subsequent types of failures are associated with temperatures lower
similar fusion line separation was found in the Gresik than plant operating temperatures. The report further
heat exchanger shell in 2000. The Pusri failure consist- indicated that the thermal history of the forgings in the
ed of a circumferential fracture along one side of a girth exchanger apparently resulted in a different micro-
weld that joined the heavy wall (145 mm, 5.7-in. thick) structure for the cracked side vs. the uncracked side.
shell forging of the vessel to the tubesheet forging.
Verbal and digital photograph information from the
plant site indicated that the fracture occurred on the Background
shell side of the weld. This separation encompassed
one-half of the circumference of the vessel shell. The exchanger which was investigated and discussed
KBR identified sampling for metallurgical investiga- in this article for the Pusri failure was fabricated by
tion of the Pusri failure, which was subsequently con- Belleli in Italy in 1991. Note that a similar unit for P. T.

AMMONIA TECHNICAL MANUAL 77 2002

Petrokimia Gresik, had also been fabricated in Italy by
Industrie Meccanihe Bagnalo. The height of the Pusri
exchanger is approximately 7,733 mm (25.4 ft), with
an outside diameter of 1,601 mm (5.25 ft) and a shell
thickness of 145 mm (5.7 in.). The shellside inlet and
outlet nozzles have an internal diameter of approxi-
mately 408-420 mm (16-16.5 in.), and a 690-700 mm
(27-27.5 in.) outside diameter. A general cross-section
arrangement of the exchanger noting the crack location
is shown in Figure 1.

Metallurgy and Process

The shell and tubesheet both met the composition

requirements for l-l/4Cr-l/2Mo steel forgings,
(ASME SA 336 Fil). The clad composition on the
inside surface was confirmed to be a stainless steel,
AWS SFA-5.4 E 347 or SFA-5.9 ER 347. It was report-
ed that the closing weld was made with Inconel 82,
AWS SFA-5.14 ERNiCr-3, after buttering each side
with Inconel 600 followed by thermal stress relief.
The process gas inlet stream temperature of the vessel
was 458°C (856°F) and the outlet temperature was
357°C (675°F). Process design for the upstream con-
verter effluent was for about 177 bar (2,567 psi) total
pressure while the exchanger inlet operating conditions
included high hydrogen and nitrogen partial pressures Figure 1. Ammonia converter effluent/boiler
due to the hydrogen content, about 53 vol. %, and the feedwater exchanger general arrangement.
ammonia content, about 17.6 vol. %. The hydrogen
laboratory (Metallurgical Consultants, Inc.) for analy-
partial pressure was reported to be 92.66 kg/cm2 (1,318
sis. Figure 2 shows the general zone of this failure
psi). Due to the boiler feedwater supply of 266.5°C
along with the five sample locations, which had been
(511.7°F) to the channel which exits at 328°C (622°F),
provided for investigation.
the normal operating temperature near the weldment
which failed operates at approximately 427°C (800°F).
Fabrication Detail and Crack Separation
Sampling and Repairs
Area

Figure 3 shows a cross-sectional detail from the fab-

Kellogg Brown & Root, Inc. provided an equipment
rication drawings of the shell-closing seam with the
engineer followed by a welding engineer on-site in
failed weldment.
order to assist in repairs. The repairs were completed
Figure 4 provides a cross-sectional photograph from
by a contracted crew from Engineers and Fabricators
the item following section removal, rough polishing,
Company from Houston, TX. Samples were requested
and etching. The photograph shows the interior stain-
and taken from the failure by EFCO, then boxed and
less steel weld overlay (at the left), the butterpass to the
shipped to KBR in Houston. These samples were sub-
tubesheet forging (top), and the butterpass to the shell
sequently provided to an independent metallurgical

AMMONIA TECHNICAL MANUAL 78 2002

 123-C Ammonia Converter Effluent/BF W Heat Exe hanaer

Zone of assimila r weld separation from

shell plate

#1and#2

180 degrees of Sample 4

circumference was is the top and bottom
separated at bottom of weld mating zone in an
dis similar weld area 180 degrees from
the center! ine of the
major separation/crack
area

Sample 1 and 2
is the top (I ow chrome'
forging with butterpass and
inconel weld) and the
bottom half of a mating Sample 3
zone (any remnant inconel is the bottom half of
butterpass weld with the another zone in the
low chrome shellplate) in separation/crack area Sample 5- is across an
the crack area (low chrome with unfailed dissimilar weld in
stainless on the internal the s hells ide 20-inch
diameter) diameter inlet nozzle
(inconel to low chrome
with internal diameter ss
clad or weld overlay)
(about 12-inch long x 3-
inch wide x thickness)

Figure 2. General fracture zone and sample removal locations.

AMMONIA TECHNICAL MANUAL 79 2002

forging (bottom). The through-thickness (145 mm or Summary Findings and Conclusions
about 5.7-in.) separation can be seen at the bottom.
During repairs to the Pusri unit, five samples were
removed by EFCO and subsequently sent from KBR to
Fracture Face Metallurgical Consultants, Inc. (MC) for examination.
The purpose of the examination was to determine the
As shown in Figure 4, the shell forging alloy butter- cause of the failure.
pass weld had separated along the ferritic to nickel The vessel fractured along the fusion line of the nick-
alloy weld fusion zone. This is further evidenced when el-alloy weld, on only the shell side of the weld. In
the cross-section is opened and viewed over the sepa- addition to the fracture along the weld fusion line, there
ration thickness facing the previously mating fracture were intergranular cracks extending into the base metal
surfaces. Figure 5 represents a section of the circum-
ference from the 145 mm (5.7-in.) thickness at the
interface fusion line separating the ferritic shell
from the butterpass nickel weld. The top of the pho-
tograph shows the outside diameter of the exchang-
er shell, and the bottom of the photograph shows the
inside diameter of the exchanger shell at the stain-
less weld overlay thickness. The numerous weld-
ment butter passes for the shell thickness are por-
trayed in the left and right sample sections of each
face, which had previously been mated together. The SA33CF11+W.O.
fracture appeared to follow the contour of the many
individual passes of the weld in the heavy-wall
exchanger.

Similar Fracture from a Nozzle Failure

For reference only, Figures 6 and 7 show an exam-

ple of a similar-looking failure which had occurred SA33fF11+347 w,
earlier in the inlet nozzle of the same service Figure 3. Cross-section of closing seam from fabrication
exchanger at another processing unit. drawing.
The weldment, however, joined an
austenitic to a low chrome clad nozzle.
This appeared with the face of the frac-
ture in an approximately 50.8 mm (2-
in.) thick section of the 50.8 mm (20-
in.) diameter on the left and the cross
section of the nickel weldment (with
essentially no remaining ferritic avail-
able for investigation).

Figure 4. Photo at cross-section of closing seam weldment, separation at

shell forging interface.

AMMONIA TECHNICAL MANUAL 80 2002

of the shell. No cracks or fractures were observed on shows the weld profile in the same orientation as in the
the tubesheet side of the weld. design. After etching, this shows the area with a crack
Sample 2 had a crack that extended along most of along the bottom fusion zone with the shell forging.
one side of the weld, but had not completely separated; Specimens of the cracked and intact sides of the weld
no cracks were apparent along the other side. Figure 4 near the inside surface (Sample 2) were prepared for

Figure 5. Photograph mating weldment separation faces.

Figure 7. Cross section of the

weldment for a prior inlet noz-
Figure 6. Internal diameter cladding with the remainder of zle failure in another unit.
the face showing separation.

AMMONIA TECHNICAL MANUAL 81 2002

metallographic examination. Photomacrographs of the The features of consequence in Figures 8 and 9 are
two sides are shown in Figure 4 (cross-section) and distinct differences in thermal history which result in
Figure 5 (facings). At this radial location of the varying grain size for the different forgings.
exchanger, the crack extended along the fusion line and Metallographic examination of the weld near the
arrested in the stainless steel cladding. As presented inside surface showed narrow light-etching microstruc-
later within the hardness survey, smaller cracks were tures next to the fusion line on each side of the weld,
also found adjacent to the butterpass extending into the with darker-etching weld passes joining them. The dif-
low-alloy steel. There were no cracks along the fusion ferences in weld microstructure indicated that the shell
line on the weld opposite the major fracture at 180 and tubesheet had been "buttered" with an alloy, as
degrees around the girth weld or in the tubesheet side reported.
of the weldment. Examination of the inside stainless clad surface
Metallurgical acid etching of the cross-sections of the showed a dark-etching layer with cracks. As the layer
girth weld in both shellside and tubeside metallograph- was too hard for saw cutting, and it was brittle, the
ic specimens showed a light-etching zone near the layer appeared to be nitriding from service (Figures 10
fusion lines and darker-etching passes towards the cen- and 11).
terline of the weld, which indicated that the two sides
had been buttered before the closing weld was made. Hardness
Although the two sides of the weld were both report-
ed to be SA 336 Fil, there was a difference in the Vickers tests showed that the hardness of the weld
respective microstructures. In the figures to follow, and low-alloy steel next to the weld fusion line were
microstructures show the grain size on the nonfailed significantly harder on the fractured shell side than on
tubesheet side of the weld to be uniformly fine: a num- the intact tubesheet side of the weld. The differences in
ber 8 according to ASTM E 112. The microstructure of hardness between the two sides of the weld indicated
the cracked side had a mixed grain size. Some of the that at the fracture, the shell side had not been suffi-
grains were very large, 4 according to ASTM E 112, ciently stress-relieved, but that the tubesheet side had
and other grains were finer, 8 according to E 112. There been stress-relieved.
is no grain-size requirement for SA 336 Fil forgings. Similar hardness tests on specimens from both sides
Photomicrographs at 100X magnification (after of the girth weld at an uncracked location of the vessel
etching) showed a basic difference in the microstruc- opposite the failure showed a hardness range corre-
ture from the tubeside (top) and shellside (bottom) sponding to that of the tubesheet weld. The lower hard-
forgings near the weldments (Figures 8 and 9).

Figure 8. A336F11 tubesheet (intact side) Figure 9. A336F11 shellside (cracked side) near
near dissimilar weldment. dissimilar weldment.

AMMONIA TECHNICAL MANUAL 82 2002

ness on both sides of the girth weld on the opposite side Analysis by Energy-Dispersive X-Ray
of the vessel indicated satisfactory stress relief of the Spectroscopy
shell and tubesheet at this radial location.
Scans across the fusion lines of the cracked and intact
Elemental Analysis sides of the weld were performed by energy-dispersive
X-ray Spectroscopy (EDS) at the approximate locations
Elemental analyses across the fractured side and of the hardness surveys and mapped, as shown in
intact side of the weld showed generally similar distri- Figure 12. Each scan was centered on the crack or
bution of chemical composition, but there were wider fusion line, and consisted of 128 points at an approxi-
compositional gradients on the cracked side compared mate interval of 0.0007-in. (0.02 mm) between points.
with the intact side of the weld. For example, there was Scans included the base metal and weld metal on each
carbon detected in the weld near the fractured fusion side of the fusion lines. The results were plotted with
line, but virtually none in the weld next to the tubesheet points on the abscissa and the approximate weight per-
fusion line. The compositional gradients are attributed cent for each point, as determined by an iterative semi-
to greater dilution on the shell side from buttering. The quantitative standardless procedure, on the ordinate. A
chemical compositions of the weld and of the low-alloy similar scan was performed near the inside surface of
steel forgings on both sides of the weld satisfied the the weld cladding on the cracked side of the girth weld.
requirements for the reported alloys of construction. Some of the results of the EDS analyses were shown
paired, from the shell to the tubesheet. An abrupt
change in chemical composition indicates the crack or
the fusion line, as is apparent from the zero readings.
One of the plots provided the approximate percent-
age of nickel (Figure 12). As anticipated, the nickel
content of the girth weld was high and that of the low-
alloy steel on each side of the weld was low. Near the
fusion line of the cracked side of the weld, the nickel
content within the weld increased over a distance of
approximately 0.015 in. (0.38 mm) to the maximum
detected. On the intact side, the nickel content of the
weld increased to the maximum detected within 0.004
in. (0.1 mm) from the fusion line. There was also a
slightly higher average nickel content on the cracked
side than on the intact side. The difference in the rate of

Figure 10. Cracking detail.

Photomicrograph of small intergranular cracks,
indicated by dark arrows, and fracture along the Figure 11. Nitriding Layer.
weld fusion line with weld-metal remnanats, white Photomicrograph of dark nitrided layer, arrows, in the weld
arrows. overlay on the inside surface.

AMMONIA TECHNICAL MANUAL 83 2002

change of nickel content for the two sides of the weld as shown in Figure 14.
probably indicated different dilution during buttering. The manganese content of the weld was higher than
Three points of high nickel content were apparent on that of the low-alloy steels (Figure 15).
the low-alloy steel side of the crack, which probably The silicon content showed wide variations within
represented a narrow weld fragment attached to the the weld and within the low-alloy steel, but the average
shell. content appeared similar for both.
The plotted results for chromium were also provided Predictably, the iron content varied little within the
(Figure 13), with high chromium content found in the weld and within the steel, as shown in Figure 16. The
weld and lower chromium content in the low-alloy niobium content of the weld locations was found high-
steel. The chromium content on the cracked side of the er than that of the low-alloy steels because this element
weld was slightly higher than on the intact side. As was comprises part of the composition for the welding con-
the case for the nickel content, on the cracked side of sumable.
the weld, there was a gradual increase of chromium Aluminum was distributed throughout the weld and
content from the fusion line to the maximum observed. low-alloy steels at normal concentrations.
From the intact weld fusion line the chromium content The element of greatest interest in these surveys was
increased to the maximum within 0.004-in. (0.1 mm). carbon, which showed a relatively high concentration
Although there were significant local variations in on the side of the weld with higher hardnesses and the
molybdenum content in both the weld and low-alloy crack (Figure 17). Carbon was detected in the weld
steel, this element showed generally equal distribution metal up to 0.050 in. (1.25 mm) away from the edge of
when averaged for the weld metal and low-alloy steel, the crack. Carbon was detected at only one point in the

Hicftgl • CrackMl Shall W*M MfeUI - Intact rufanliMi Vftt*

TulMdwn

Figure 12. Line scans of nickel composition at 256 points, from left to right, across the shell, the
cracked shell weld fusion line, the two sides of the weld, the intract tubesheet weld fusion Une and the
tubesheet.
- O»e*ed ShW w«*d

"f TubWIWK
«Ml

^yuv*

Figure 13. Line scans of chromium composition at 256 points, from left to right, across the shell, the
cracked shell weld fusion line, the two sides of the weld, the intact tubesheet weld fusion Une and the
tubesheet.

AMMONIA TECHNICAL MANUAL 84 2002

 Oactad Shut W«M • hm«t Tttwftut W*W
WrtJ TulmtWif

WM»

Figure 14. Line scans of molybdenum composition at 256 points, from left to right, across the shell, the cracked
shell weld fusion line, the two sides of the weld, the intact tubesheet weld fusion line and the tubesheet.
Mang«*** - Cricted S**H W«lij

Figure 15. Line scans of manganese composition at 256 points, from left to right, across the shell, the
cracked shell weld fusion line, the two sides of the weld, the intact tubesheet weld fusion line and the
tubesheet.
Iron - Ciackinl Sti«n Wold Iron • Iniftet Tubesho«! WeM

\\ f
m-
30
SO
^SfojiM^V^M**»"»
SO
9 — (0 — «Î «* 0 •- O

Figure 16. Line scans of iron composition at 256 points, from left to right, across the shell, the cracked
shell weld fusion line, the two sides of the weld, the intact tubesheet weld fusion line and the tubesheet.

weld metal on the intact side. possibility of diffusion or dilution, whereas the analy-
There were differences between the semi-quantita- sis of the weld metal was performed at the middle of
tive EDS (Energy Dispersive) analyses and the quanti- the weld to determine its composition.
tative OES (Optical Emission) and AA (Auger) analy- The line scan across the weld cladding at the inside
ses of the two steels and of the weld metal in particu- surface showed a chromium-nickel stainless steel with
lar. The EDS gives semi-quantitative results only. It is significant niobium, confirming the use of a consum-
also important to remember that the line scans were able such as ER 347 for cladding. Figure 18 shows the
performed along the sides of the weld to determine the niobium line scan of the weld cladding, with the inside

AMMONIA TECHNICAL MANUAL 85 2002

 Orten - CwcMd Sta» VMä Caibon •

SMI

j
slfill i!.'. i\tw '
'2SSsS53ï,SS8s r- f t* « « n U

Figure 17. Line scans of carbon composition at 256 points, from left to right, the cracked shell
fusion line, the two sides of the weld, the intact tubesheet weld fusion line and the tubesheet.

surface. inside surface of both sides of the girth weld was nor-
High nitrogen content was detected in a surface gra- mal for this service and was not a factor in the failure.
dient of the cladding. Figure 19 shows the nitrogen dis- The inside surface of the clad metal on both sides of
tribution in the cladding. The first 19 points, starting at the girth weld showed a very hard, brittle layer with a
the inside surface (at 0.02 mm per point), show on depth of 0.007 in. (0.18 mm). Nitriding during service
average much higher nitrogen content than the remain- caused the hard, brittle layer, but the intruded layer was
der of the line scan. These 19 points correspond to the not found significant to the failure.
dark-etching layer observed during metallographic In addition to the 7 mils (0.007 in.) depth following
observation. The nitrogen line scan confirmed that a nearly nine years of service, a somewhat confirming
dark-etching layer was a nitrided layer. indicator of nitriding rate by a prior formulation can be
applied to produce the graph shown in Figure 20. The
lower curve shows the rate of depth progression for a
Nitriding relatively thick member; whereas, a higher rate is
denoted for relatively thin-walled members or where
Chemistry scans for nitrogen showed higher contents cracking has begun.
to a depth of 0.42 mm (0.017-in.). However, an actual The formula (from van Greiken) for this provides a
nitride brittle layer of only 0.18 mm (0.007-in.) was nitrogen diffusion rate of (Nt depth in \im for t years)
found. Since a cover of austenitic material at the inner for stainless steels according to process conditions of
surface of the weld repair was not included, that small temperature (T, kelvin), partial pressure of NH3 (bars),
section will be subject to surface hardening by nitriding nickel content (wt. %) and time (t, years). Log(Nt) =
due to diffusion by nitrogen. This surface hardening 7.69 - 2900/T + 0.67Log(Pnh3) -2.37Log(Ni) +
may lead to embrittlement and would occur over some C*Log(t), where C is 0.5 for thicker or 1.0 or thin (or
length of time, for which accurate estimates of rates or surface cracking) members.
depths would be unlikely. The MCI report showed the The sample results for the above suggest a slightly
depth of nitriding within the first 19 to 21 points of the greater depth (2 mils/year for 9 years or 18 mils) for
cladding (at 0.02 mm or 0.0007 in. per reading). These thick stainless steel than actually found, as stated
nitrogen percentage values are only indicative of point above. Note also that the ammonia content would have
measurements following the current service history of been about 17%. However, the bare steel exposure left
about 9 years, and do not reflect bulk nitrogen chem- during the repair is the concern.
istry. These readings begin at the internal cladding and In the case of bare steel, the formula is only valid for
extend into the ferritic, and, as noted above, do not pro- stainless, although other experience has shown bare
vide an indication of hardness and/or of cracking. steels, such as 2 l/4Cr, had a 90 mil (0.090 in.) nitride
Nitriding observed along the weld overlay on the depth after nine years exposure at 482°C (900°F) hi

AMMONIA TECHNICAL MANUAL 86 2002

 Niobium at ID

Points
Figure 18. Niobium line scan of the cladding on the cracked side of the girth weld,
with the inside surface at the left, confirming the use of E347 for cladding.

Nitrogen at ID

Figure 19. Nitrogen line scan of the cladding on the cracked side of the girth weld,
showing decreasing nitrogen gradient from the inside surface at left.

AMMONIA TECHNICAL MANUAL 87 2002

16% ammonia, while a 5 Cr Nitriding Rate of 8%Ni at 825F/1500psi (total)
received 48 mils in 11% ammonia vs NH3 Content
in a similar exposure time. Of
course, neither of these (penetra- E
«•
,10.00

tion depths) reflects crack occur- n

rence.
1.00
o
Hardness Tests ô

Vickers hardness tests were per-

formed in surveys across the
cracked and intact fusion lines,
Ii
0.10 4
1 2 4 5 6 7 B 9 10 11 12 13 14 15 16 17 18 19 20

%NH3
and across the fusion line of the •Thick •Thin or Cracking
stainless overlay on each side of
the weld, using a 5-kgf load. Figure 20. Rate of depth progression for thick- vs. thin-walled members.
The hardness test results for
Sample 2 are tabulated below and mapped in Table 1. much higher than its counterparts on and near the intact
Except as noted, the interval between test indentations fusion line. The hardness in the weld next to the
was 1 mm (0.040 in.). For ease of reference, the cracked fusion line was also somewhat higher than its
Vickers readings have been converted to Rockwell B or counterparts on the intact side of the weld. No tests
C according to ASTM E 140. were performed in the nitrided layer.
The hardness near the cracked fusion line was high, Cross-sectional views of each side of the weldment

Table 1. Hardness Test Results for Sample 2

Cracked Fusion Line Side Intact Fusion Line Side
Test HV5 Test HV5
No. Location (Rockwell) No. Location (Rockwell)

1 Shell 204 (94B) 16 Tubesheet 181 (89B)

2 Shell 201 (93B) 17 Tubesheet 179 (89)
3 Shell 202 (93B) 18 Tubesheet 196 (92B)
4 0.30 mm from Crack 373 (38C) 19 Weld Fusion Line 243 (21C)
5 0.15 mm from Crack 449 (45C) 20 Weld Fusion Line 229 (98B)
6 0. 15 mm from Crack 368 (38C) 21 Weld Fusion Line 225 (97B)
7 Weld 217 (96B) 22 Weld Fusion Line 226 (98B)
8 Weld 215 (96B) 23 Weld Fusion Line 232 (99B)
9 Weld 216 (96B) 24 Weld 191 (91B)
10 Steel 221 (97B) 25 Weld 195 (92B)
11 Steel 245 (21C) 26 Weld 200 (93B)
12 Cladding Fus. Line 247 (22C) 27 Steel 193 (92B)
13 Cladding 270 (26C) 28 Steel 193 (92B)
14 Cladding 275 (26C) 29 Cladding Fus. Line 233 (99B)
15 Cladding 261 (24C) 30 Cladding 251 (22C)
31 Cladding 246 (22C)
32 Cladding 240 (20C)

AMMONIA TECHNICAL MANUAL 88 2002

to the upper tubesheet (on the right) and the shellside Discussion
(on the left), with corresponding hardness readings
below, indicate the basic problem. The work performed by Metallurgical Consultants,
Figure 21 shows Photomacrographs (5 1/2X) for Inc., confirmed fracture and cracks along the shell side
shellside (left) and for tubeside (right) with correspon- of the girth weld in Exchanger 123-C1. The cracks gen-
ding Vickers hardness traverses below. erally followed the fusion line, and some extended in
an intergranular manner into the shell, at the hardest
locations on the shell side. The microstructure of the
Additional Hardness Testing low-alloy steel forging near the girth-weld fusion line
on the shell side showed much higher hardness than the
As a result of the finding that hardness in the frac- corresponding tubesheet side, and the bulk carbon con-
tured sample 2 varied from the forging on one side of tent of the girth-weld metal at the cracked side was
the weld to the forging on the other side, additional higher than the noncracked side. The microstructure
Vickers tests were performed across the fusion lines with high hardness was susceptible to cracking, which
near the inside surface of the nonfailed portion of the is attributed to hydrogen.
girth seam sample 4. The test interval between indenta- The characteristics of the fracture and cracks and
tions was again 1 mm. Individual test indentations were high hardness of the cracked side indicated that the fail-
also made subsequently along the fusion lines. Since ure was caused by hydrogen at low or intermediate
the shell and tubesheet sides of Sample 4 were not temperatures. It appeared that the apparent absence of
identified or apparent, we referred to the two sides as A stress-relief predisposed the weld fusion line on the
and B. The readings at all locations indicated similar shell side to failure. There was no evidence of high-
hardness, generally below 248 HV 5, which indicated temperature hydrogen attack (HTHA).
that both sides of each weld had been stress-relieved. The relationship of hydrogen cracking to the con-
See Table 2. No hardness testing of the girth weld in the struction and operating conditions of the exchanger
nozzle sample 5 was performed, because metallograph- was presented by MCI as follows: "Diffusion of atom-
ic examination of this sample had shown that it con- ic hydrogen through steel promoted by high hydrogen
sisted of austenitic weld metal joining an austenitic partial pressures may crack microstructures that are
tube. unusually hard or highly stressed at low or intermedi-
ate temperatures. In this case, a narrow zone along the

Table 2. Hardness Test Results for Sample 4

Sample 4. Side A Sample 4. Side B
Test HV5 Test HV5
No. Location (Rockwell) No. Location (Rockwell)

1 Base metal 188 (9 IB) 12 Base metal 207 (94B)

2 Base metal 186 (90B) 13 Base metal 206 (94B)
3 Base metal 187 (90B) 14 Base metal 229 (98B)
4 Fusion line 249 (22C) 15 Fusion line 252 (23C)
5 Weld 195 (92B) 16 Weld 198 (93B)
6 Weld 200 (93B) 17 Weld 195 (92B)
7 Weld 194 (92B) 18 Weld 198 (93B)
8 Fusion line 230 (98B) 19 Fusion line 226 (98B)
9 Fusion line 240 (20C) 20 Fusion line 235 (99B)
10 Fusion line 219 (96B) 21 Fusion line 222 (97B)
11 Fusion line 227 (97B) 22 Fusion line 222 (97B)

AMMONIA TECHNICAL MANUAL 89 2002

 l

Figure 21. Photomacrographs for shellside (at left) and for tubeside (at right), with corresponding Vickers
hardness traverses.

AMMONIA TECHNICAL MANUAL 90 2002

fusion line on the shell side of the exchanger weld was bon content of the low-alloy steel on each side was
sufficiently hard to be susceptible to hydrogen crack- similar.
ing. As the operating temperature of the exchanger was The more gradual change in weld-nickel content on
above the normal temperature range for hydrogen the cracked side than on the intact side also indicated
cracking, it appeared that the cracks initiated and grew more dilution, but the significance was not clear. It is
during cool-down periods when hydrogen is less our opinion that the MCI conclusion, stated as wider
mobile." Similar failures have been documented by compositional gradients on the cracked side, related to
other investigators for these conditions. carbon and not nickel. Nickel results showed reason-
Cracks occur preferentially along the weld fusion ably uniform results while that the variation was dis-
line of dissimilar welds not only because of hard tinctively different only for the carbon. The carbon on
microstructures, but also as a result of the difference in the weld side specifically shows more point readings
the solubility of hydrogen in austenitic weld metal involving carbon on the cracked side vs. the uncracked
compared to that in ferritic microstructure. Hydrogen side. This relates to carbon dilution of the weldment
solubility is many tunes higher in austenite than in the more than the results for dilution by nickel.
neighboring low-alloy steel. The resulting concentra- While there have been many published sources
tion of hydrogen along this interface between austenitic which relate the suitability in applying dissimilar weld-
and ferritic microstructures can cause cracks in suscep- ments and broadening the range in which these welds
tible microstructures upon cool-down. Generally, many should be suitable, it is also useful to acknowledge that
cycles of exposure to high hydrogen partial pressures some sources do recognize problems associated with
and cool-down are required to cause cracking of dis- these. For example, in a proposal for industry study of
similar welds to the extent that they leak. dissimilar weldments, The Materials Property Council
The differences in hardness between the shell side stated the following: "When nickel alloy weld metal is
and tubesheet side of the weld are attributed to differ- used to join a low alloy steel boiler tube to an austenitic
ences in thermal history from construction. Although stainless steel boiler tube, failure of the joint occurs by
buttering and stress-relief were reportedly specified for the formation of a planar accumulation of chromium
both sides, it appeared that the shell side of the weld carbide particles along the low alloy steel/nickel alloy
had been prepared for the final closing weld by butter- fusion line. These carbides develop slowly and subse-
ing, but had been only partially stress-relieved. The quently develop creep cavities surrounding each one.
partial stress-relief was indicated by the hardness read- These cavities grow in size until the cavities link. The
ings, generally below 248 HV 5 from the shell side research previously accomplished by MPC has shown
opposite the fracture. In contrast, the hardness and that this weld type shows little or no metallographical-
appearance of the tubesheet side indicated buttering ly detectable damage at 85% life consumed or later in
and stress-relief at both the fractured side and side component life. Note: The presence of undetectable
opposite the fracture. damage this late in the expected life of the weld results
Normally, a difference in carbon content between the hi NDE damage detection at significantly greater than
two sides of such a weld with different thermal histo- 90% life consumed. This situation results in nickel
ries would indicate carbon diffusion from the low-alloy based welds being nearly impossible to inspect and
side into the low-carbon austenitic weld metal. manage using conventional ultrasonic inspection meth-
However, in this case, the higher carbon content was ods. The development of damage, its detection with
observed on the hard, cracked side, which apparently current methods and tube failure, occurs over a short
was not stress-relieved. It is possible that there was time interval. This interval can be less than 3 years for
greater dilution by the low-alloy steel of the buttering tubes with a life expectancy of 30 years."
metal on the cracked side than on the intact side, which For this reason, it is suggested that detection of
resulted in the higher carbon content. There did not cracking may require timely, rather than delayed,
appear to be any influence of the higher carbon content repairs.
in the weld on susceptibility to cracking, since the car- The investigation, completed with metallurgical

AMMONIA TECHNICAL MANUAL 91 2002

work provided by Metallurgical Consultants, Inc., ana- ating in excess of 441°C (825 °F). These overall repairs
lyzed available samples and reached the conclusion left a small area on the inside diameter exposed to a
that there was apparently insufficient heat treatment to nitriding environment, which will harden over time. It
reduce and/or avoid small local zones of high hardness. was considered much more manageable than the poten-
Based on the information provided to us, neither KBR tial abrupt separation involving a dissimilar weldment.
nor MCI could conclude that, as one prior theory for a For the repaired exchanger, in KBR's opinion, moni-
similar separation stated, there was a "nickel migration toring for either microfissures due to decarburization
that led to formation of austenite that was subsequent- or for cracking due to high hardness and fissuring
ly strained to convert to martensite." In either event, the from nitrogen diffusion are essentially equivalent.
resultant limitation was that hard zones did develop Monitoring techniques generally involve ultrasonic
within a narrow fusion line for this case, and, in com- flaw detection and looking for the development of
bination with hydrogen, this caused cracking and sub- cracking. Since surface hardening by nitriding (or by
sequent separation. other mechanisms) does not ensure development of
Furthermore, the low chrome base material was not through thickness cracking, it is believed that the prop-
found to be damaged by "HTHA" (High Temperature er approach would be to monitor these areas. An exam-
Hydrogen Attack) as described or defined within the ple is monitoring dissimilar metal weld seams (or at
criteria of API 941 "Recommended Practice," other- repairs near prior dissimilar weldments which left a
wise known as "Nelson Curves". This form of attack small zone bare, rather than back-clad with stainless)
results in decarburization of the steel with subsequent for signs of crack initiation. If cracks are observed, they
development of fissures. The MCI investigation should be repaired. It should be noted, however, that
showed that decarburization was not detected and was some companies offer evaluations wherein cracks
not involved hi the mechanism that resulted in failure. could be evaluated using fitness for service type of
However, it is advised that using a chrome steel oper- analysis.
ating hi a hydrogen environment at a level above the H. D. Marsch did an interesting study hi 1982 regard-
current API Recommended Practice does require mon- ing nitriding for bare low alloy steels (without stainless
itoring in order to assure long-term safe operation. protection), and found severe effects in excess of about
Again, decarburization was not detected during this 410°C (770°F) in tests up to about 5.6 years. While he
failure investigation. noted disbonding of nickel weldments in these envi-
ronments, the testing also showed that nitriding rates
accelerated by a factor of 3 to 5 under higher stress sit-
Repairs uations.

The repairs undertaken on the Pusri exchanger did

not repeat the nickel butterpass weldment. The repair Preventative Actions Taken
was done by applying a low chrome weldment, which
was subsequently postweld heat-treated. Note that There have been prior studies on the suitability of
about six months following these repairs, some shallow dissimilar weldments. In particular, these have been in
(4 mm maximum depth) outside diameter cracking was the application of clad or weld overlayed piping for
again observed in the fusion zone of the shell to weld- refinery applications. One recent study noted that
ment repair. The plant performed grinding and reweld- "inspection of the most critical welds using automated
ing and re-PWHT was recommended. Due to the boil- ultrasonic methods during fabrication and shutdowns
er feedwater supply to the channel, the normal operat- are considered necessary to ensure freedom from in-
ing temperature near the weldment which failed is service cracking". This reference suggests that the
approximately 427°C (800°F). This is therefore microstructural width of the "Local Hard Zone" is a
believed lower than reports such as by Welding determinant in the suitability for similar services. This
Research Council studies regarding repair cracking in shows a narrow zone using nickel base weldments and
coarse grain heat affected zones for low chromes oper- wider zones with kon based austenitic butter welds.

AMMONIA TECHNICAL MANUAL 92 2002

However, it would be suggested that it will be difficult Pressures in Petroleum Refineries and
to control microstructure patterns within normal fabri- Petrochemical Plants, 5th ed., Supplement 1 (April,
cation techniques, especially for heavy wall thickness- 1998).
es. AWS Welding Handbook, "Materials and Applications-
While details of heat exchanger fabrication drawings Part 2," Vol. 4, AWS, pp. 353-362 (1998).
have not been historically available, Kellogg Brown & Buehler, W. M., "Examination of Failed Weld from
Root, Inc. made a decision to develop and send out an 123-C1 Ammonia Converter Effluent Steam
"Advisory" to operating plants which may have had Exchanger," Pusri-lB, Palembang, Indonesia,
similar designs so that appropriate inspections could be Metallurgical Consultants, Inc., MCI Report 20
undertaken. Although KBR received responses from a (Nov. 2000).
few operators, the specific shellside dissimilar weld- Craig, B. D., and R. B. Setterlund, "Catastrophic SSC
ment was apparently only involved in two designs pro- Failure of a Dissimilar Metal Weld in High-Pressure
vided for the two facilities in Indonesia by different Hydrogen Vessel," Corrosion/91, Cincinnati, OH
fabricators in Italy. In response to the survey, one oper- (Mar. 1991).
ator advised KBR that they have similar weldments in Grotz, B. J., and G. R. Prescott, "Disbonding between
synthesis loop services which are undergoing routine 5Cr-l/2Mo and Inconel and Nitriding of
inspection programs. They noted that they may be fol- 5Cr-l/2Mo," Ammonia Plant Safety & Related
lowing similar studies by TNO, The Netherlands, in a Facilities, Vol. 33, AIChE, New York (1993).
report entitled "Integrity of Austenitic/Ferritic dissimi- Hotman, E., and Ismudjoko, "Failure of
lar welded joints in hydrogen services" (Apr. 27,1998). Inlet Nozzle 123-C1 Ammonia Effleunt Steam
KBR is also aware of some separations involving Generator," Ammonia Plant Safety & Related
reformer (fired heater) tube tops, which also employ a Facilities, Vol. 41, AIChE, New York (2001).
dissimilar weldment transition. http://www.forengineers.org/mpc/dis_metal_welds/dis
KBR has not been advised of any further shellside _metal_welds.htm (May 2000).
dissimilar weldments or problems with shells or inlet Lang, M. C, E. Yelverton, D. Hansen, and J. Sievert,
transitions for these exchangers. Recent and current "Inspection of C-l/2Mo Components for Hydrogen
Kellogg Brown & Root, Inc. designs have not and do Damage," Ammonia Plant Safety & Related
not apply the specific heavy wall thickness closing Facilities, Vol. 32, AIChE, New York (1992).
seam detail as were applied in the exchangers investi- Lewis, K. R., R. W. J. Koers, H. C. van Wortel, and
gated. Where similar transitions cannot be avoided, A. H. M. Krom, "Assessment of Operating Limits
particular attention needs to be paid to thermal history for Critical Dissimilar Metal Welds in a
of the materials, welding parameters, and final heat Hydrocracker," NACE International Paper 01532,
treatment. CORROSION/2001.
Lochhead, J. C., and Ken Rodgers, The Practical
Welding Engineer, AWS, p. 95 (2000).
Acknowledgement Lundin, C., P. Liu, C. Y. P Qiao, G. Zhou, K. K. Khan,
and M. Prager, "An experimental Study of Causes
A substantial portion of this report, including all of and Repair of Cracking of 1 l/4Cr-l/2Mo Steel
the metallurgical laboratory work, was taken from an Equipment," WRC Bulletin, 411 (May 1996).
investigation completed for the failure and reported by Marsch, H. D., "Nitriding of Steels," Plant Operations,
Mr. Mark Buehler of Metallurgical Consultants, Inc. in 1(3), (July 1982).
Houston, TX. Penuela, L. E., J. G. Chirinos, and J. D. Dobis,
"Analysis of Dissimilar Welds Exposed to High
Literature Cited Temperature H2/H2S Conditions in a
Hydrodesulfurizing (HDS) Unit," NACE
APIRP941 Recommended Practice 941, Steels for International, Paper 383, CORROSION/1999.
Hydrogen Service at Elevated Temperatures and

AMMONIA TECHNICAL MANUAL 93 2002

Prescott, G. R., and Brian Shannon, "Process van Greiken, C. A., "Factors Controlling the Nitriding
Equipment Problems Caused by Interaction with of Stainless Steels in Ammonia Synthesis Loops,"
Hydrogen," Ammonia Plant Safety Ammonia Plant Safety & Related Facilities, Vol. 29,
& Related Facilities, Vol. 41, AIChE, New York AIChE, New York (1989).
(2001).

QUESTIONS AND ANSWERS

Pan Orphanides, Orphanco: Heat input during butter- and/or sufficiently detect susceptible microstructures in
ing shall be considered: according to the Schaeffler dia- production welds that would later be applied in envi-
gram, high dilution of base metal forms martensitic ronments that might lead to abrupt and catastrophic
structures having high hardness values and is not easy failure. It is believed that bi-metallic weldments in high
to eliminate by PWHT. Has this aspect been considered temperature/pressure hydrogen environments may
in the analysis? eventually be more susceptible to embrittlement and/or
Jim B. Sievert, Kellogg Brown & Root: Thank you for disbonding. However, there are many instances of dis-
your comment/question regarding whether we consid- similar weldments in HP/HT hydrogen environments in
ered the possibility of predicting a result of martensite refinery processes connecting piping to reactor pressure
formation in relation to the Schaeffler diagram due to vessels, which apparently have not experienced similar
high heat input welding. The short answer is "not effects. We believe, where dissimilar welds are found in
directly". However, the undertaking of EDS Chemical synthesis gas environments, they should be routinely
analysis plotting across the separation zone was specif- examined or monitored by appropriate means, such as
ically an attempt to discover this type of result as was UT. In addition, while the article specifically deals with
examining microstructures for narrow or wide hard a thick-walled dissimilar weldment for the shell of the
bands in the interface zone. The metallurgical laborato- heat exchanger, we presented information related to
ry determined no specific Schaeffler diagram correla- other similar cracking for three different forging
tion. Hard zones were determined at the failure, and sources and fabrications by two different vendors. The
nonhard zones in unfailed zones. We could not conclu- separations found only involved one-half of one zone of
sively suggest that, following nine years of service, the the nickel butterpass welds on shell-side forgings and
original welding heat-input was excessive which might on a nozzle forging. We believed it would be rather
have lead to dilution effects that resulted in martensite. coincidental to suggest that different fabricators
Nor could we also conclusively suggest that the forging achieved excessively high heat inputs in nearly identi-
was properly heat treated originally, or that the original cal locations at different fabrication shops for some-
weld heat treatment was satisfactory. Rather, it was what similar designs. Had that been the case, it would
suggested that the high hardness remained owing to have seemed likely that the nearly identical weld zones
poor prior thermal history. The point of the presentation on the tubesheet forging may also have been affected,
was to advise of a potential inherent inability to control but these were not.

AMMONIA TECHNICAL MANUAL 94 2002