You are on page 1of 5

PIPING/RELIABILITY

A
failure was detected during the shutdown of an HF
alkylation unit, when a leak was discovered after chemi-
cal cleaning of the pipe. The perforation occurred at a
weld joining a 90 elbow with a straight pipe section. The failed
section was removed for replacement. Internal visual inspection
revealed preferential corrosion at the weld.
Steel samples were taken from several places including the welds
and the base metal on both sides of the weld and sent for chemical
analyses. The objective was to determine the level of residual ele-
ments, in particular Cu, Ni and Cr, that are said to reduce the HF
acid corrosion resistance of carbon steel. The weld that was cor-
roded through thickness exhibited a high content of these residual
elements. The failure was, therefore, attributed to this fact.
FAILED PIPE
The failure was detected on the recycle isobutane piping of the
depropanizer section. The affected section was a pipe circuit 12 in.
(30.5 cm) in diameter and running from the depropanizer tower
to the depropanizer feed recycle exchanger. The process stream
contains propane, iso-butane, butane and traces of hydrofluoric
acid and water. Fig. 1 shows a partial isometric drawing of the
affected pipe circuit.
The leak was discovered during chemical cleaning to remove
iron fluoride scale inside the pipe. During the last recorded
replacement of the pipe circuit, pipe spools were prefabricated in
the refinerys repair shop and installed in place by field manual
welding. The perforation occurred at a weld that appeared to
have been performed in the shop. Further inspection detected
additional welds corroded selectively, including field welds. This
showed that the affected welds were not related to a particular
welder, welding machine, electrodes or welding conditions.
Three pipe samples containing a weld in the middle were exam-
ined, including the failed one with a perforation at the weld (Fig. 1).
They were 12 in. (30.5 cm) in diameter, with a nominal wall thick-
ness of 0.375 in. (9.53 mm). The system operates at a pressure of
290 psig (2 MPa) and a temperature of 99C (210F). The sections
of pipe had been in service for at least 16 years. The material was
carbon steel piping ordered to specification ASTM A-106 Gr. B.
RESULTS
Visual inspection. Fig. 2 shows the three pipe samples sub-
mitted for this investigation. They were labeled A, B and T.
The pipe sample containing the failed weld was A. Evidence of
metal loss was also observed at the weld in sample B. The third
sample, referred to as T, did not show selective corrosion at the
weld, but the base metal on one side was significantly thinner on
the other side of the weld. The perforation in A is shown in Fig.
3. The weld was preferentially corroded, while the base metal on
both sides was not as severely corroded.
According to Fig. 1, pipe section samples B and T shared one
90 elbow. Ultrasonic thickness readings were taken from all
samples; the results are shown in Table 1. The corrosion rate was
calculated for each sample, on each side of the weld based on the
reported time in service (16 yr). The estimated corrosion rate was
between 4 and 6 mpy (0.10 0.15 mm/yr).
Chemical analysis. Samples removed from each pipe sample
were collected for chemical analysis. The analyses were performed
on the weld metal and on each base metal on both sides of the
weld. Table 2 summarizes results of the chemical analysis on these
samples. Carbon content for the welds is fairly low. It is 0.14%,
0.07% and 0.07% for the welds in samples A, B and T.
However, carbon content of the weld in sample A is high
considering the usual for mild steel electrodes. Carbon content
of wires and electrodes for steel welding is usually between 0.05%
Preferential corrosion of welds
in HF service
A high content of residual elements was the suspected cause
T. MUNSTERMAN and A. MAYORGA, Capstone Engineering Services, Inc., Houston, Texas

Isometric drawing showing affected pipe section. FIG. 1


September 2004 issue, pgs 113119
Used with permission.
www.HydrocarbonProcessing.com
Reprinted from:
HYDROCARBON PROCESSING SEPTEMBER 2004
PIPING/RELIABILITY
and 0.12%. Also, the chromium content of the weld in sample
A is high, considering that filler metals for carbon steel welding
do not contain this element. Assuming a 25% dilution, the filler
metal for sample A may have contained approximately 0.19%
carbon and 0.15% chromium. The manganese content for the
welds is higher than in the base metalabout 0.99 1.34% as
compared with 0.59 0.73% for the base metal.
Table 3 summarizes the amount of residual elements taken as
the sum of Cu, Ni and Cr in weight percentage for each analysis.
The amount of residual element was 0.51% for the weld that was
selectively corroded in sample A. For the base metals, the residual
element content was 0.03% and 0.11%. The amount of residual
elements was 0.17% for the weld in sample B that also suffered
selective corrosion. In this case, the major contribution came from
Cu content, about 0.13%. In the case of the weld in sample A, the
Cu content was also the highest, 0.31%, but Ni and Cr content
was 0.09% and 0.11%, respectively. In the weld in sample B, Ni
and Cr content was 0.02% for each element.
The carbon equivalent was calculated using the following
formula:
C C
Mn Cr Mo V Ni Cu
eq
= + +
+ +
+
+
6 5 15
Table 3 also summarizes the carbon equivalent for
each steel sample. They are all below 0.40%, except for
the weld in sample A that was 0.41%. Some refiners
specify a carbon equivalent below 0.40% to achieve
low hardness, good weldability and low susceptibility
to hydrogen stress cracking.
Metallographic examination. Cross-sections
through the weld of the three pipe samples were
removed and prepared for metallographic examina-
tion. Fig. 4 shows a cross-section through the weld
of sample A. Corrosion occurred preferentially at the
weld metal, leaving the adjacent base metal almost
intact. Fig. 5 is a cross-section through the weld of
sample B. It also showed selective corrosion at the
weld but the corrosion seems to have stopped when
reaching this weld bead.
Selective corrosion at the sample A weld left a con-
cave surface at the remaining bead, as shown in Fig.
4. The fact that the remaining weld in sample B has a
The weld was preferentially corroded while the base metal
on both sides was not as severely corroded.
FIG. 3
A was the pipe sample containing the failed weld. FIG. 2
Corrosion occurred preferentially in the weld metal,
leaving adjacent base metal almost intact.
FIG. 4
TABLE 1. Estimated corrosion rate for base metal
Base metal Remaining wall thickness Estimated corrosion rate
A1 7.9 mm (0.311 in.) 0.10 mm/y (4 mpy)
A2 7.0 mm (0.275 in.) 0.15 mm/y (6 mpy)
B1 6.93 mm (0.273 in.) 0.15 mm/y (6 mpy)
B2 7.5 mm (0.294 in.) 0.13 mm/y (5 mpy)
T1 7.06 mm (0.278 in.) 0.15 mm/y (6 mpy)
T2 7.8 mm (0.308 in.) 0.10 mm/y (4 mpy)
TABLE 2. Chemical composition of the metal samples. Locations
1 and 2 are base metal, location 3 is the weld
Sample Location C Cr Ni Mn Mo Si Cu P S
A A1 0.22 0.03 0.03 0.69 0.003 0.19 0.05 0.006 0.019
A2 0.24 0.01 0.01 0.60 0.18 0.01 0.009 0.022
A3 (Weld) 0.14 0.11 0.09 1.34 0.010 0.65 0.31 0.011 0.012
B B1 0.23 0.02 0.02 0.72 0.003 0.24 0.04 0.006 0.023
B2 0.24 0.01 0.01 0.66 0.17 0.01 0.015 0.019
B3 (Weld) 0.07 0.02 0.02 1.03 0.23 0.13 0.027 0.017
T T1 0.20 0.02 0.02 0.73 0.003 0.24 0.04 0.007 0.025
T2 0.24 0.01 0.01 0.59 0.18 0.01 0.009 0.021
T3 (Weld) 0.07 0.02 0.02 0.99 0.46 0.02 0.009 0.018
TABLE 3. Amount of residual elements and carbon equivalent.
Locations 1 and 2 are base metal; location 3 is the weld
A1 A2 Aweld B1 B2 Bweld T1 T2 Tweld
Cu + Ni + Cr 0.11 0.03 0.51 0.08 0.03 0.17 0.08 0.03 0.06
C
eq
0.35 0.34 0.41 0.36 0.35 0.26 0.33 0.34 0.24
HYDROCARBON PROCESSING SEPTEMBER 2004
PIPING/RELIABILITY
convex shape (Fig. 5) suggests a corrosion process that consumed
all the weld metal up to that point and then ceased. This is prob-
ably why the exposed weld bead maintains it shape.
Fig. 6 shows the cross-section through weld sample T. A sub-
surface weld defect was revealed in this metallographic section.
The weld metal itself looks unaffected by corrosion. However,
there is a noticeable thickness difference between the two base
metals at the location examined. The base metal on side 1 on the
left side of Fig. 6 is thinner than the base metal on side 2 on the
right side of Fig. 6.
The remaining wall thickness measured for each base metal
at the metallographic location is reported in Figs. 4, 5 and 6 for
samples A, B and T, respectively.
The microstructures observed in sample A are shown in Fig. 7
for both base metal sides. The microstructure is typical of carbon
steels that have been finished at very high temperatures and have
not been normalized. In sample A, the base metal microstructure
is coarser in one side than in the other. In both cases, the ferrite
is present as grain boundary allotriomorphs and Widmansttten
plates. In between these ferrite forms is fine pearlite, which is
the dark etching constituent. Fig. 8 shows the rather coarse steel
microstructure found in side 1 of sample T, the 90 elbow. The
Widmansttten plates are more obvious here.
Hardness testing. Fig. 9 summarizes the Vickers microhard-
ness profiles obtained from the three samples studied. The profiles
were taken across the weld, from one base metal to the other. The
heat affected zone (HAZ) is usually the harder because it tends
to cool fast during welding. This was the case for samples B and
T, where the HAZ at both sides of the weld appeared harder than
the base metal on both sides, and the weld metal itself. For sample
A, the remaining weld metal appeared to be harder than the base
metal at any location and than the other two welds examined.
DISCUSSION
The HF alkylation process licensors for refining have their own
material specification and training manuals describing the process
and many other aspects deemed necessary to get acquainted with
the unit. When defining the different likely corrosion mecha-
nisms in carbon steels in HF environments, one licensor manual
1

stated that weld decay was reported on carbon steel weld metal
exposed to both anhydrous and aqueous HF environments. The
attack was described to be either more severe on the weld than on
the base metal, or simply the corrosion rate was very high at the
weld but negligible at the adjacent base metal.
Minutes from a meeting of the Corrosion Refining Informa-
tion Exchange (currently referred to as Technology Exchange
Group 205X, Refining Industry Information Exchange, and
available from NACE International as REFIN.COR 6.0) reported
a case in 1988 that referred to a short carbon steel pipe section
which had experienced excessive uniform metal loss to less than
1.3 mm (0.05 in.). The failure was attributed to a higher than
normal residual alloy content, although the sum of Cu, Ni and
Cr was only 0.16%.
In the minutes from the NACE 1989 meeting, a question was
asked if others had seen accelerated corrosion at carbon steel pip-
ing welds in HF service. A reply indicated that preferential weld
corrosion was not really common in HF service and that when it
occurred, it was due to abnormal weld chemistry. In the minutes
of a 1995 meeting, a presentation was made of the results of an
inspection program on welds in an HF alkylation unit. The phe-
nomenon was referred to as a possible galvanic effect of coupling
high residual element carbon steels to low residual element carbon
steel. This was to account for a case where the carbon steel that
corroded the most had low residual element content.
Hashim and Valerioti
2
reported in 1993 results indicating the
effect of residual Cu, Ni and Cr on the corrosion resistance of
carbon steel in hydrofluoric acid alkylation service. They recom-
mended a maximum of 0.2% for the sum of these three residual
elements.
Another HF alkylation unit reported that a reducer experi-
enced a leak. In this case, the whole reducer corroded much faster
than adjacent carbon steel; it was found to also contain high resid-
ual Cu, Ni and Cr. The case, which occurred in a depropanizer
feed line, was published in 1997.
3
The inspection following the
failure was thorough and selective weld corrosion was found.
Leomar Peuela and Jose Chirinos
4
also reported this case and
stated that there were two failures. One occurred in one flange (the
case previously reported) and the other in a weld in the depropanizer
charge carbon steel pipe containing propane, isobutane and anhy-
drous hydrofluoric acid at 88C (190F) and 2.2 MPa (321 psig).
Examination showed severe uniform corrosion at the inside surface
of the flange while the adjacent elbow showed minimal corrosion loss.
The other failure showed preferential attack in the weld without any
significant corrosion loss in the pipe components. A total of 265 loca-
tions were inspected ultrasonically, but only three showed significant
metal loss. A total of 109 locations were inspected by radiography,
Sample B showed selective corosion at the weld, but
the corrosion seems to have stopped when reaching
this weld bead.
FIG. 5
A subsurface weld defect was revealed in this
metallographic section.
FIG. 6
HYDROCARBON PROCESSING SEPTEMBER 2004
PIPING/RELIABILITY
and 10 were found with significant metal loss. They used an optical
emission analyzer for in situ chemical analysis with the following
results: 32 locations were found with residual element content lower
than 0.2%, 100 locations with residual element content between
0.21% and 0.40%, and 24 locations with residual element content
higher than 0.41%. These 24 locations with residual element content
higher than 0.41% were inspected again with ultrasonics but no
significant metal loss was found in any of them.
The information reported during these NACE International
Information Exchange meetings is contradictory in that residual
Cu, Ni and Cr content did not always provide an explanation to all
cases. As a result, a joint industry project known as Materials Speci-
fication for HF Alkylation Plant was initiated and the results were
recently published.
5
A beneficial carbon content effect was reported,
particularly in carbon steel with very low residual element content.
A minimum carbon content of about 0.18% was recommended,
in which case the specification should be changed to 0.15% for the
sum of Cu and Ni only, excluding Cr. For weld metals, typically
having a low carbon content, a maximum of 0.15% for the three
residual elements, instead of 0.2%, was recommended.
The weld metal of sample A had a residual element content
of 0.51%, significantly above the acceptable limit of 0.15%, and
this could account for the selective corrosion. The remaining weld
metal in sample B exhibited residual element content of 0.17%.
Also, the carbon content in any of them meets the minimum sug-
gested (0.18%). Some evidence indicated that the root passes of
this samples weld metal were totally corroded. This suggests that
the chemical composition of these root passes may have greatly
differed from the remaining weld that did not corrode.
During NACE Corrosion 2003, the NACE TEG X119, Mate-
rials for Receiving, Handling and Storing Hydrofluoric Acid and
Anhydrous Hydrogen Fluoride, reported a failure occurred in an
anhydrous hydrogen fluoride (AHF) liquid and vapor phase com-
position of 9 wt% and 18 wt%, respectively, operating at 86C.
The feed line was constructed from low temperature alloy steel.
The failure point was in a weld seam. In the system, corrosion rate
of fittings and weld metal was approximately 20 mpy and 34 mpy,
respectively. Elemental composition of the fittings and weld metal
contained over 0.2 weight percent trace alloying elements. It was
0.4% for fittings and 3.79% for the welds.
No work has been published on the galvanic corrosion hypoth-
esis stating that, when carbon steel with low residual elements is
coupled to steels with high residual elements, one may tend to
corrode preferentially. Others have made reference to the possible
effect of these residual elements in disrupting the iron fluoride
protective layer, but, again, no one has yet published any research
work on this subject.
The weld of the perforated pipe also exhibited high hardness val-
ues, with a maximum of 231 HV. This could indicate that the weld
in sample A was not post-weld heat treated (PWHT), but, in this
refinery, PWHT is mandatory for all piping placed in HF service.
Some refineries still specify that weld hardness should be limited to
200 HB maximum (210 HV) for HF alkylation service and state
that PWHT is required if the weld hardness surpasses this limit.
Since hardness in the HAZ can be higher, it is customary to
specify 248 HV for the HAZ, which is equivalent to the most
commonly known limit of 22 HRC used to prevent hydrogen
Base metal microstructure for both sides (a) and (b) of
sample A (100x).
FIG. 7
Base metal microstructure for side 1 in sample T (100x). FIG. 8






Microhardness profile across the weld in the three pipe
samples.
FIG. 9
HYDROCARBON PROCESSING SEPTEMBER 2004
PIPING/RELIABILITY
stress cracking. No one has suggested that hardness has to do with
selective weld corrosion, but, in this case, it may indicate whether
the weld was properly PWHT. The carbon equivalent was, in
general, very low. Therefore, hardness level in the HAZ in all three
cases seems high. It is possible that these welds were not properly
PWHTin particular, sample A, which had the hardest weld
among the others examined. If so, the effect of residual stresses
may still account for this selective weld corrosion.
The joint industry project, Materials Specification for HF
Alkylation Plant, stated that PWHT did not provide any signifi-
cant benefits in relation to corrosion behavior in the HF environ-
ments studied.
The pipe sections have been in service for at least 16 years.
Corrosion rate at the weld metal that experienced a leak was
then approximately 0.58 mm/y (23 mpy). This compares with
an estimated 0.10 0.15 mm/y (4 6 mpy) found for the base
metal. Among the three cases studied, the failed weld had not
only the highest amount of residual elements but also the highest
carbon equivalent (0.41%), the highest carbon content (0.14%),
the highest Mn content (1.34%), and the highest Si content
(0.65%).
The selective weld corrosion in samples A and B was attributed
to the high content of residual elements Cu, Ni and Cr.
The residual element hypothesis is not yet conclusive and
definite. The truth is that the phenomenon is not yet totally
understood. HP
LITERATURE CITED
1
UOP Process Technology Training Manual, HF Alkylation, March 1990, UOP
Inc., Illinois, 1990.
2
Hashim, H. H. and W. L. Valerioti, Effect of Residual Copper, Nickel and
Chromium on the Corrosion Resistance of Carbon Steel in Hydrofluoric
Acid Alkylation Service, Paper no. 623, CORROSION 93, Houston, Texas,
NACE International, 1993.
3
Chirinos, G., S. Turgoose and R. C. Newman, Effects of Residual Elements
on the Corrosion Resistance of Steels in HF, Paper no. 513, CORROSION
97, Houston, Texas, NACE International, 1997.
4
Peuela, L. and Jose Chirinos, Carbon Steel Flanges and Weld Evaluation
on HF Alkylation Unit, 2nd Inspection and Corrosion Workshop, Valencia,
Venezuela, Dec. 3 5, 1997, PDVSA-Intevep, Los Teques, Venezuela, 1997.
5
Gysbers, A., et al., Specification for Carbon Steel Materials for Hydrofluoric
Acid Alkylation Units, Paper no. 03651, CORROSION 2003, Houston,
Texas, NACE International, 2003.
Anelsy G. Mayorga has been with Capstone Engineering
Services, Inc., since March 2001. She has 16 years of experience
in inspection, engineering consulting and metallurgical analysis in
the oil processing industry. Currently, Ms. Mayorga is involved in
metallurgical and process consulting and failure analysis. She is a materials engineer
with a master of science degree from Ohio State University in welding engineering.
Tim Munsterman has over 20 years of experience in metal-
lurgical consulting to the process industries including refining,
petrochemical, chemical and pulp and paper. He has expertise in
equipment evaluation to determine the causes of problems and
develop practical solutions. Mr. Munsterman is involved in corrosion consulting
to identify corrosion mechanisms and rates of attack and provides expert witness
services in connection with failure analysis litigation.
Article copyright 2004 by Gulf Publishing Company. All rights reserved. Printed in U.S.A.

You might also like