A N O F F I C I A L

P U B L I C A T I O N
 ?he NACE Ints

CAROLINE PIPELINE FAILURE:

FINDINGS ON CORROSION MECIiANISMS IN
WET SOUR GAS SYSTEMS CONTAINING SIGNIF'ICANT CO,

Nguyen N. Bich
Shell Canada Limited
P.O.Box 2506
Calgary, Alberta, T2P 3S6, Canada

Kevin Goerz
Shell Canada Limited
Caroline Complex, Bag 500
Caroline, Alberta, TOM OMO, Canada

ABSTRACT
A leak occurred in the 5-32?,flowline between Junction 1 and the South Compressor Station in the
Canada Limited Caroline Complex gathering system near Sundre, Alberta, due to internal wet &S
(hydrogen sulphide) and CO, (carbon dioxide) pitting corrosion at the bottom of the pipeline. The
significant contributing factors to the extremely high pitting corrosion rate, approximately 30 mm/y
(1200 mpy), are considered to be unexpectedly high amounts of chloride ions in the produced well
fluids, settling of the produced water under low flow conditions, high condensatdwater ratio,
inadequate inhibition and pigging, and insufficient monitoring programs. Corrosion mechanisms in
sour gas gathering systems with significant CO, concentration were reviewed. Preliminary fmdings
pointed to CO,partial pressure, not KS,as a main corrosion rate determining factor. The steps taken to
prevent similar future incidents were also reviewed.

Keywords: carbon dioxide, corrosive gas, sour gas wells, inhibition, gas transmission

INTRODUCTION

On January 6, 1994, the Caroline 5-32s flowline leaked between the Junction 1 facilities and the
South Compressor Station at the Red Deer River crossing near Sundre, Alberta. Subsequent
investigations have determined that the leak was caused by internal wet YS and CO, pitting corrosion at
the bottom of the p i p h e . The failure was not expected as the line had been in service for only six
monrhs. The unusually high corrosion rate, later determined to be approximately 30 m d y (1200 ~ P Y ) ,
was an unpleasant exception to ow normal experience where pitting corrosion in wet sour gas gathering
systems is usually less than 4 m d y (120 mpy).

A systematic program to look into the causes of corrosion in Caroline 5-32s was initiated. The
scope of the program also includes other sour gas fields as there is a need to determine whether the
corrosion mechanisms operative in Caroline 5-323 can also prevail elsewhere.

This paper describes the fmdings on the causes of failure at 5-32s. It also presents some
preliminary results from the early work on corrosion mechanisms in wet sour gas gathering systems
with significant CO, concentration.

BACKGROUND

Description of the Caroline Complex Gathering System

The Shell Canada Limited (SCL) Caroline Complex gathering system consists of f&en (15)
producing wells split into three areas. Wells in each of those areas produce into the compressor station
for that area, where the sour gas and hydrocarbon liquids are separated and transported in separate
pipelines to the SCL Caroline Gas Plant for further processing.

The 5-32s flowline runs from the 5-32s well to the South Compressor Station (SCS). A second
flowline runs along the same right of way from the 6-20 well to the SCS. Approximately two thirds of
the way along the right of way, both flowlines run through the Junction 1 facilities where production
from a third well, 8-IS, can flowdown either of the lines fiom Junction 1 to the SCS.

The 5-323 flowline is a 219.1 mm OD X 8.2 mm WT CSA 2245.1, Gr. 359, Category If sour
service material. The approximate length from Junction 1 to SCS is 3.3 km. The 8-1s flowline is a
114.3 mrn OD X 6.0 mm WT CSA 2245.1, Gr. 359, CategoryfI sour service material. The
approximate length from the 8-1s well to Junction 1 is 1.0 km. The chemical composition of the steel
pipe is given in Table 1.

A field map showing the locations of the wells and pipelines, and the location of the leak is
provided in Figure 1.

Operating History

The initial internal corrosion control program for the multiphase sour gas gathering system in
Caroline included:

Batch treatments of conosion inhibitor of all lines prior to s t . of production.

Injection of a water soluble corrosion inhibitor at the wellheads into the produced fluids to protect
surface facilities and pipelines during production flow.
* Maintenance procedures for running cleaning pigs to remove any liquid and solid holdup which
might occur in low spots in the flowlines, and.
Maintenance procedures for conducting batch treatments of corrosion inhibitor of a l l lines to
supplement the continuous corrosion inhibition program.
 The 5-32s flowline was first put in service in July, 1993 with production from the 5-32s well.
First production from the 8-1s well into the 5-32s flowline was in October, 1993. From October, 1993
until the date of the leak, January 6,1994, only production from the 8-1s well flowed through the 5-32s
flowline. The effluents from the 8-1s well were never combined with those from the 5-32s well.

From March 1993 until November 1993, produced fluids from the 8-1s well were intermittently
flowed through the 6-20 flowline. At no time was production from the 8-1s well transported through
I
the two flowlines simultaneously. When production from the 8-1s well flowed down the 6-20 flowline,
it was always combined with production from the 6-20 well.

DESCRIPTION OF THE FALURE

The following is a summary of the physical findings reported during the investigation conducted
after the incident, with reference to the period when the 8-1s well production was being transported
down the 5-32s flowline:

All pits were found in the bottom of each line (between the 5 o'clock and 7 o'clock positions). In
the 5-32 flowline, pits approximately 1.2 cm in diameter were found 20-25 centimeters apart. In
between these single pits, there was very little corrosion. The pit which resulted in the leak was
approximately 1.7 cm in diameter. Figure 2 shows the corrosion pit which resulted in the leak.
High levels of chloride ion, approximately 33,000 parts per million, were found in water samples
taken from the 8-15 well. Early records showed the chloride ion levels to be in the few hundred
parts per million range.
Elemental sulphur was found in samples taken from the bottom of the 8-1s pipeline pits near the
corrosion.
Greigite (which is a corrosion product of steel formed when oxygen or elemental sulphur is
present) was found in samples taken from the 5-323 flowline. No greigite was found in the 8-1s
flowline.
Higher than normal levels of solids (mostly iron sulphides and sulphur) were found during the
5-32s and 8-1s instrumented pipeline corrosion inspection tool (IPCTT) logging runs after the
leak.
Testing indicated there may have been occasions when there was a lower critical continuous
inhibitor concentration level than the increased corrosive environment in the 5-32s flowline
required.
The 5-32s flowline was PCIT inspected. The portion of the line from Junction 1 to the South
Compressor Station was the portion with the through-wall corrosion pit. Logging of that portion
of the line found 31 locations with greater than 50% wall loss, 37 sites with w@l loss between
25% and 50%, and 21 sites with wall loss less than 25%. No indications were found in the portion
of the line upstream of Junction 1.
The PCIT results on the flowline from the 8-15 well to Junction 1 confumed intermittent
corrosion was present throughout the flowline, with 2 areas ranging from 20% to 40% wall loss.
The well bore tubulars of the 8-1s well were logged and found to have 0% to 20% wall loss in
portions of the string.
These more corrosive conditions were unique to the 8-1s and 5-32s flowlines. AU. other operating
sour service pipelines in the Caroline Complex, including the 6-20 flowline which as previously
indicated had carried productions fluids from both 8-1s and 6-20 wells, were inspected and found
free eom corrosion.
 CAUSES OF THE FAILURE

Overview

Investigation by Shell Canada has indicated that the leak was the result of internal wet &S and
CO, pitting corrosion at the bottom of the pipeline. Cut-outs from the leak segment and additional
segments verified the mechanism to be internal wet KS and CO, pitting corrosion in the bottom of the
pipe (5 o'clock to 7 o'clock position). Findings from the investigation indicate that the contributing
factors were:

The flow regime in the line segment allowed water to remain stagnant in low lying locations.
The high chloride ion content in the produced formation water, in combination with produced
solids settling on the bottom of the pipeline in low lying locations, weakened or chemically
damaged the protective iron sulphide scales and corrosion inhibitor f h .
The continuous corrosion inhibitor, selected and tested for medium chloride ion content and lower
condensate/water conditions, was not able to repair the degraded protective scales and frlm in the
high chloride ion content water, high condensate/water ratio and presence of solids.
When localized breaks in the protective iron sulphide scale and inhibitor film occurred, conosion
pits fonned due to wet &S and CO, pitting conosion.
Pigging and batch inhibition procedures, which were insufficient to deal witfi'the more corrosive
conditions involved, were unable to prevent the pits from initiating and growing.

Details of the contributing factors follow:

Well Composition Fluid

After an initial period of production from the 8-1s well during which only fresh water, or water of
condensation, was produced consistent with the performance of the other wells in the Caroline
Complex, a change in the produced well fluids composition occurred whereby formation water
containing high amounts of chloride ions (salt) was produced with the gas and condensate hydrocarbons.
This difference in chloride ion content, as well as the presence of solids including sulphur compounds
and iron sulphides, was verified by a one-week duration well production test completed on the 8-18 well
in October, 1994 as part of the investigation of the incident.

This represents a set of conditions very different from those considered in the initial corrosion
control program design for the Caroline gathering system, and for the conditions found in the other
wells in Caroline produced up to the time of leak, and since then. The conditions, used in initially
selecting the corrosion inhibitors and designing the corrosion control program for Caroline are more
representative of the conditions seen to date with the other wells in the field. A comparison of the
significant components of the well fluid compositions for both the 5-328, typical of other wells in the
Caroline Complex, and the 8-15 well highlights the differences (refer to Tables 2-3).

The information used in the initial corrosion control program design was based on the limited well
test data available at the time, which indicated that lower chloride ion levels (less than 20,000ppm) and
no significant solids production would be representative of the early producing life for the Caroline
field.
 The early breakthrough of the high chloride ion formation water in the 8-1s well can be explained
by the way the well is completed and the proximity of the producing gas zone to a non-movable
formation water aquifer. The well is completed through both the gas zone and the formation water
zone, and the distance separating the producing interval in the gas zone from the water zone is
approximately 5 m. This proximity increases the risk of formation water being produced with the gas
and condensate. All other wells in the Caroiine Complex have been reviewed for similar conditions,
and three (3) other wells have been identified as having a similar potential for early breakthrough of
formation water. The monitoring program in place in Caroline addresses these wells. No indications of
changes to the well fluid compositions have been observed in any of these wells.

Water Settling

A review of the production history through the 5-32s flowline has confirmed that the potential for
water holdup was high during periods when production from the 8-1s well only was being flowed
through the line from Junction 1 to South Compressor Station (SCS). In low-lying areas the flow was
not sufficient to consistently maintain the superficial gas velocity above the 3 mlsec minimum target
velocity specified in the initial corrosion control program. This allowed the more corrosive formation
water to remain stagnant long enough for pitting corrosion to initiate.

Another factor somewhat unique to the Caroline Complex which contributed to water settling, is
the high condensate to water ratio (greater than 50 to 1) of the well fluids in Caroline. The light
hydrocarbon liquids have a tendency to separate from the produced water, particularly in stagnant areas,
such that a thin layer of water can form under the condensate adjacent to the pipe surface. Without the
aid of pigging, this can make it more difficult to move the water along in the pipeline.

Protective Barrier Breakdown

In wet sour gas systems, when the steel pipehe material is exposed to the wet %S gas some
corrosion occurs. This forms corrosion products which provide a stable, protective iron sulphide scale.
This protective iron sulphide scale helps to reduce further corrosion. Corrosion inhibitors further
enhance the protectiveness by providing a protective inhibitor film. If these protective films and scales
are weakened locally, these weak spots can become sites for pitting. In the case of the 5-32s flowline,
the following factors are believed to have caused the protective scale and frlm breakdown:

: Chloride ions contribute to pitting through initiation of localized corrosion by

breaking down the protective iron sulphide conosion scales and by preventing formation or
re-formation of the protective corrosion inhibitor film necessary to prevent further pit growth.

Chloride ion concentration levels as high as 100,000 parts per million from the 8-1s well fluids
were fust estimated during the investigation immediately after the incident, and were verified when
additional testing of the 8-1s well was undertaken in October, 1994. Levels as high as 80,000 parts per
million were found in this one-week duration well production test. A review of the well history records
indicates that an increase in chloride ion concentration was noted from water analysis samples taken the
previous October, where levels of 33,000 parts per million were reported. Prior to that date, only fresh
water, or water of condensation, containing less than 1,000 parts per million chlorides, was being
produced as part of the well fluids. After that date, formation water, more commonly referred to as
salty or high salinity water because of the high chloride ion concentration levels, also started to be
produced with the gas and condensate. By comparison, analysis for chloride ions from the 5-32s well
and other wells in the Caroline Complex have shown levels to be in the 2,000 parts per million or less
range, indicating that only relatively fresh water, or water of condensation, is being produced.

&lij&: Solids, consisting mostly of iron sulphides and sulphur, were reported during pigging runs
after the leak occurred, and were confhmed during the 8-1s well production test conducted in October,
1994.

The role of solids in contributing to the pitting mechanism includes: adsorption of the inhibitor
(thus lowering the effectiveness of the inhibitor), settling on the bottom of the pipe, and initiation of
breakdown of the protective corrosion inhibitor ftlm.
..
of In-: Analysis of the continuous inhibitor dosage levels done as part of the
investigation has indicated that on occasions, levels in the 5-328 flowline may have been below the
levels required for the more corrosive conditions of the 8-1s well fluids. Lab testing has confirmed that
at dosage levels below the critical level, pitting corrosion could occur. It is believed that the under
dosage condition resulted from the change in water production to the more corrosive, high chloride ion
formation water, in combination with adsorption of the inhibitor by the solids being generated from the
8-1s well. Other factors which may be involved, but are thought to only play a minor role, involve
some reaction of the inhibitor with components produced with the liquid hydrocarbons, and some
partitioning of the inhibitor in the liquid hydrocarbons. The operating practice at the Caroline Complex
was to target for dosage levels substantially in excess of the recommended dosage level. Testing and
checks done on the other wells and flowlines in the Caroline gathering system have confirmed that
dosage levels are above the recommended levels.

-priate Inhibito~:The initial selection of the continuous corrosion inhibitor was based upon
extensive laboratory tests. The selected inhibitor was judged to be the best candidate under conditions
considered to represent the most extreme case for the Caroline well fluid compositions. Subsequent
review has identified the following differences in those conditions versus conditions encountered in the
leak in the 5-32s flowline:

@ CNoride ion levels from 8-1s were significantly higher than the initial test conditions
(80,000 parts per million actual versus 23,300 parts per million used in initial tests).
The condensate to water ratio used in the initial tests was set at 2 to 1. This ratio was considered
to be conservative relative to actual conditions (greater than 50 to 1) as it was believed that higher
ratios afforded more inherent corrosion protection. It is now believed that ratios in the range of
greater than 50-100 to 1 can actually decrease inhibitor efficiency by interfering with the transfer
of inhibitor to the water layer next to the steel surface.

REPAIR AND RECLAMATION STATUS

The current status of the repair and reclamation activities associated with the incident is
summarized below:
g The segment of the 5-32s flowline from Junction 1 to SCS was isolated and removed from
service. A small portion of the pipe where the leak occurred was removed for use in the
investigation, and a patch welded in place such that the pipeline segment could be left in a
suspended state with a low pressure inert gas purge.
0 Modifications were completed at the Junction 1 facilities to isolate the 5-32s flowline from
Junction 1 to SCS and provide the capabilities for production from 6-20, 5-323 and 8-1s wells to
be flowed down the 6-20 flowline only, from Junction 1 to SCS. IPCIT logging runs on the 6-20
flowline have not shown any evidence of pitting or other corrosion.
0 Since the incident, the 8-1s well has only been produced for one-week, in October, 1994,
conducted as part of the incident investigation. At this time there are no immediate p l m for
further production of the 8- 1S well.
All cleanup, environmental assessment and rechrnation activities related to the incident have been
completed.

PREVENTATIVE MEASURES

The following preventative measures have been implemented to the corrosion control program for
the Caroline sour gas gathering system:

Well design conditions for each Caroline well have been reviewed to identify those wells where
the potential for production of high chloride ion formation water conditions similar to those
experienced in 8-1s well might be anticipated. An ongoing review process is now included as part
of the well and pipeline monitoring programs.
Changes have been incorporated in the well fluid testing, analysis and reporting requirements to
identify conditions which would indicate a change to more corrosive conditions (i.e., increasing
production of high chloride ion formation water). Key target conditions are monitored, and
deviations from the specified targets are flagged and corrective action taken.
The corrosion inhibitors used at Caroline have been re-evaluated and changes to the inhibitors and
appLication practices have been implemented. Testing has taken into consideration the higher
chloride ion content, high condensate to water ratio, and presence of solids found in the 8-1s
analysis, and will be ongoing as changes to the produced fluids are identitied through the welt
fluid testing and corrosion monitoring programs.
Revisions have been made to the operating practices to address the potential for water settling.
Changes include revised criteria for minimum superficial gas velocities taking into account line
diameter, improved flow modeling to better predict flow regimes and ~ l ~ c a t i to o nthe pigging
program requirements. Reporting systems have been revised to require deviations from the
specified operating targets to be flagged and corrective action taken.
fncreased use of onstream corrosion monitoring devices, including h e a r polarization resistance
(LPR), electrical resistance (ER) type corrosion probes and corrosion coupons, to monitor for
changes in the corrosivity of the produced fluids. This will help provide more awareness of
potential changes to the corrosivity of the produced fluids and the ability to implement appropriate
comctive measures. This will be used in combination with the planned instrumented pipeline
corrosion inspection tool (IPCIT) logging programs to verify the effectiveness of the proactive
monitoring and mitigation programs outlined above.
Updating of the corrosion controi and operating procedure manuals to clarify requirements, roles
and responsibilities. The manuals and operating procedures will continue to be updated as more
operating experience and data is collected on the performance of the field.
 CORROSION MECHANISMS REVIEW PROGRAM

Background

The high corrosion rate experienced in 5-328line, estimated to be over 30 d y (1200 mpy)
prompted an internal review and assessment of other SCL sour gas pipelines. Will other lines be
subjected to the same high corrosion rate? Is Caroline 5-32s a unique case? Why do seemingly similar
systems have different corrosion rates? What are the factors that control corrosion in sour gas systems?
What are the corrosion control measures if the potential corrosion rate is 30 mmly? A research program
was initiated to provide answers to all these burning questions.

Role of CO,in Uninhibited Wet Sour Gas Conosion

Tests in the laboratory have provided us with an unusual role that CO, plays. Table 4 shows that
in the lab tests, average corrosion rate (ACR) for 1-week autoclave tests is approximately 0.5 to
3 mm/y. Longer-term corrosion rate could be lower than the 1-week test rate. Since most if not all
corrosion failures in sour gas gathering systems have been pitting, accurate prediction of average
corrosion rate might not be of significant value. The short laboratory time exposure rarely results in
severe pitting rates often observed in the field. From our experience, the worst case pitting corrosion
rate could be predicted from the laboratory data using the maximum corrosion rate (MCR), as
determined from electrochemical measurements. MCR occurred within the fust few hours of testing
when the iron sulfide scales have not been fully protective. For example, as illustrated in Figure 3,
MCR was approximately 5.5 m d y while ACR was 0.6 mdy. The rationale was that, if, for whatever
reasons, the iron sulfide scales fail to become protective, corrosion in the worst case scenarios would
proceed at the maximum rate, MCR.

Table 4 indicates that maximum corrosion rate seems to be proportional to the CO,partial pressure
but not the H$ nor the total acid gas pressures. Furthermore, MCR can be estimated to be
approximately 0.5 times the deWaard CO,corrosion rate'.

K.van(jelder2has formulated the following rule of thumb to estimate pitting corrosion in wet sour
gas pipelines containing significant CO, (3-5 bars pCO,) and in the absence of elemental sulfur: the
worst case localized corrosion rate in sour gas gathering systems is approximately one half of the
deWaard CO, corrosion rate. In most SCL gas fields, the CO, partial pressure is approximately 2-3
bars, giving a maximum corrosion rate of 4 m d y . This rule of thumb has worked well in a majority of
situations. Caroline 5-32 is a most notable exception to this rule.

Other Contributing Factors

The Caroline 5-32s flowline corrosion rate (30mdy) indicates that corrosion rate could be much
higher than the 4 m d y commonly observed if there are other contributing factors. One objective of the
research program is to quantify the contribution to corrosion rates by the factors, most of which have
been identified in the Caroline investigation:
 Fluid holdup potential which can be expressed by gas and liquid velocities, topography,
condensate-to-water ratio (CWR), etc.
The chloride contents.
The presence of sulfur.
Other solids.
Inhibition deficiency.
Others.

Based on initial results, a pitting corrosion model has been formulated:

Pitting Initiation Tendency = f (number of contributing factors)

Pitting Corrosion Rate = f (PCO,, T) x f (velocities, C1-, S, solids, etc.) x f (inhibition deficiency).

Table 5 shows some preliminary results based on known corrosion failures. Quantification of the
contributing factors is in progress.

CONCLUSIONS

An investigation of the causes of the Caroline 5-32s flowline leak has identified the contributing
factors of the extremely high pitting corrosion rate: high chloride concentration in the produced fluids,
settling of the produced water under low flow conditions, high condensate-to-water ratio, inadequate
inhibition and pigging, and insufficient monitoring programs. Study of corrosion mechanisms in sour
gas systems containing signscant CO,has revealed the significant role of CO,, not Y S , in determining
pitting corrosion rate. Other contributing factors as identified in the Caroline failures are being
quantitatively investigated.

REFERENCE

1. C. deWaard and U. Lotz, "Prediction of GO, Corrosion of Carbon Steel", CORROSIONl93,

New Orleans, paper 69.
2. K. vanGelder, Internal Shell Canada Limited communications.
 Table 1: Steel Pipe ChemrecalComposition

Table 2: Gas composition for 5-32 S and &IS wells

Table 3: Water Composition for 5-3 and 8-1sWells

Table 4: Correlation aboratory Corrosion ates and Acid Gas Pressures

Test Measured Estimated deWaard Ratio Ratio Ratio Ratio

Duration Average Maximum '93 CO, MCRl MCRl MCW MCW
hours Corrosion Corrosion Corrosion PCO, pH$ PT CQ, CR
Rate ACR Rate MCR Rate
mm/y mm/y

(7) (8)

.75 7
Yes I Yes I No 1 7.5 1 5.2
 1 ALBERTA

Figure 1 : Caroline gathering system field map

 Figure 2: Inside view of the pipe leak.

Calculated Stern-Geary constant, B = 16 rnV

V

Average polarization resistance, Rp = 303 Ohm.crn2

1 b~ Average corrosion rate = 0.6 mmly

Time, hours
Figure 3: A typical plot of corrosion rate versus
time for wet sour gas laboratory testing
(65'C, 3.5 bars CO,, 12.2 bm YS).