SPE 159484

Maximising the Life of Corroding Tubing by Combining Accurate Multi-

Finger Caliper Data with Corrosion Modelling
M A Billingham, Intetech Consultancy Ltd; B King, Tullow Oil Ltd; A Murray, READ Cased Hole Ltd

Copyright 2012, Society of Petroleum Engineers

This paper was prepared for presentation at the Abu Dhabi International Petroleum Exhibition & Conference held in Abu Dhabi, UAE, 11–14 November 2012.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
The purpose of this paper is to discuss the relationship and value brought by empirical field data measurement, extensive
interpretation and analysis of that measured field data and powerful predictive modelling with the input of multiple
production parameters, for the prediction of rate of corrosion in downhole completions.

This paper involves studies conducted upon Tullow Oil’s Bangora Field in Block 9, Bangladesh. Compositional analysis of
the gas produced from discovery well Bangora 1 identified CO2 as being present at a concentration of around 0.6 mol%.
Initial corrosion modelling highlighted that carbon steel completions were not suitable for long term production in this
environment, however carbon steel completions were installed for logistical reasons and a corrosion monitoring programme
utilising multi-finger caliper surveys was put in place. These surveys identified extensive corrosion of production tubing
within a short period of production.

Accurately calibrated multi-finger sensor measurement was performed and repeated four times within a 2 year timeframe by
Read Well Services. Intetech Ltd was engaged to carry out detailed corrosion modelling and prediction with input parameters
including fluid production rates, well head and bottom hole pressures and temperatures and produced fluids composition.
This modelling and direct measurement of the corrosion rates showed a good level of agreement and have given confidence
to the life of the completion in relation to the rate of corrosion before expected failure.

The technical benefit of this work has been that it has reduced the risk that a tubing leak arises, which would require the well
to be shut-in resulting in significant deffered production and associated loss of revenue while waiting mobilization of
workover equipment. Cost savings have arisen by being able to safely defer the time to workover of the wells in a location
where workover equipment is not readily available and to reduce the frequency of caliper surveys. Tubing replacement is
planned at as late a date as possible consistent with the modelling predictions, thus maximising the useful life of the string.
2 SPE 159484

Introduction

The Bangora Field is located in Block 9, Bangladesh, approximately 100km East of Dhaka, Figure 1. The field was
discovered in 2004 with the drilling of Bangora 1. Production testing of Bangora 1 showed substantial gas flows from three
separate zones . Follow-up wells Bangora 2, 3 & 5 also encountered prolific reservoir sections.

Figure 1: Bangora Field location

Corrosion had not been previously identified as being a problem within Bangladesh and any gas discovery was expected to
be ‘sweet’. Compositional analysis of the gas produced from Bangora 1 identified CO2 as being present at a concentration of
about 0.6 mol% . The potential corrosion issues and the limitations of using carbon steel completion materials were identified
in the well design process. However, due to the lead times associated with sourcing CRA tubing and completion equipment,
as well as various commercial and regulatory issues, using alternative completion materials was not practical in the time
available. A decision was taken to install carbon steel completions, which were considered a temporary solution to allow
initial well testing and early start of production from selected zones. Bangora 1 was completed as a cemented monobore,
Bangora 2, 3 and 5 were completed conventionally, all using 12.6 ppf N80 carbon steel production tubing, although Bangora
3 was later recompleted with 13Cr tubing following a completion failure unrelated to CO2 corrosion, Figure 2.
SPE 159484 3

Bangora Well Completion Schematics (not to scale)

Key to Symbols: TRSCSSV SSD Nipple Slickplug Perfs

Bangora-1 Bangora-2 Bangora-5

Depth
15m

20" 20" 20"

200m

13 3/8" 13 3/8" 13 3/8"

750m

4 1/2"

4 1/2" 4 1/2"

9 5/8" 9 5/8"

2500m
A Sand gas gas gas

B Sand gas gas gas

C Sand gas gas gas

4 1/2" 7" 7"

3000m
Upper D Sand gas gas gas

3050m Mid-D
Lower D Sand gas gas gas

3150m
E Sand gas gas gas

Figure 2: Well Completion Schematics, Bangora 1, 2 and 5.

4 S
SPE 159484

A corrosion mo onitoring progrram was identiffied as being necessary

n from the outset andd four caliper suurvey campaiggns were
ccarried out from
m November 2006 to Decemb ber 2008. Undeertaking these surveys was chhallenging withh significant asssociated
mmobilization co ble local oilfield services and the remotenesss of the locatioon. Surveys weere very
osts due to the lack of availab
ssuccessful in sh
howing how th he condition of the tubing wass developing. H However, it is ddifficult to deriive tubing life
m inspection data when ongo
ppredictions from oing changes in n operating connditions are afffecting the corrrosion behaviouur. Tullow
OOil therefore deecided to modeel the corrosion
n behaviour of the wells in orrder to predict future corrosioon rates and tubbing life.
TThe aims were to determine iff it was possiblle to extend thee interval betw
ween caliper surrveys, and to pplan for the eveentual
wwork-over of thhe wells.

IIntetech performmed the first modelling

m exerccise in 2009, ussing productionn data up to thee end of June 22009 and data ffrom the
ffour caliper surrvey campaignss. A second mo odelling exerciise was performmed in Januaryy 2011, and a thhird modellingg exercise
inn June 2011 affter a caliper su
urvey campaignn in May 2011.

CCaliper Surve eys

RRead Well Serv vices performeed the inspectioon surveys usin ng 40-arm multti-finger calipeer (MFC) tools. Additionally,, an
eelectro-magnetiic thickness too ol (ETT) was run
r in the Octo ober 2007 camppaign. The caliiper tools surveeyed the wholee of the
pproduction tubiing strings, how wever only cerrtain parts of th
he tubing stringgs could be loggged with the E
ETT due to the impact of
ngs on the ETT measurement.. Based on the combination oof the MFC andd ETT data in tthe intervals where both
thhe casing strin
toools recorded valid
v data it waas determined that
t the OD off the tubing rem mained close too the nominal vvalue (4.5") andd there
wwas no evidencce of any thinniing from the ODO side. As corrrosion damagee is not anticipated from the aannulus side unnder
nnormal circumsstances, tubing g penetration vaalues have been n calculated froom the MFC ddata based on aan unchanged oouter
ddiameter.

IIn each instance, MFC data was

w processed tot remove indiv vidual arm drifft and to matchh the depths of reference com
mpletion
ittems with the well
w completio on diagrams. Thhis allows timee-lapse comparrison between tthe successive surveys.

CCorrosion Morphology
TThe anticipatedd morphology of o CO2 corrosio on in the Banggora conditionss is isolated pittting [Kermani 1997; Sydbergger 1996].
TThe Bangora wells
w are produccing relatively low quantities of water and pparts of the tubbing may not be wetted, or onnly
inntermittently so,
s depending on o the temperatture and pressu ure profile up tthe wells. The ccaliper surveyss confirmed the
eexpectation thaat pitting is the mode of attackk. The measureed mean ID didd not change siignificantly bettween successiive
ssurveys, indicatting that the majority
m of the tubing
t surface was
w undamageed, Figure 3.

F
Figure 3: Examp
ple showing dev
velopment of tubing mean ID data (Bangora 1))

E
Even in the lateest survey (201
11), the majoritty of the MFC measurementss were at or neaar the original ID. However, the
maximum meassured ID increaased with time, Figure 4.
m
S
SPE 159484 5

F
Figure 4: Examp
ple time-lapse caliper data show
wing developme
ent of tubing m aximum ID (Ban
ngora 1)

DDetailed examiination of the survey data ind dicated isolated

d pitting–type ffeatures. By thee 2011 surveyss, there was a cclear
tendency to loww-side pitting over
o the more highly
h deviated
d parts of the tuubing. Howeveer, even withinn the low-side ttrack the
ddepth of penetrration is very variable with th
he deeper pittin
ng features beinng short and isoolated. As an eexample, the m
maximum
ddepth feature fo
or Bangora-1 ata 2011 is show wn, Figure 5.

F
Figure 5: Extrac
ct from 2011 cam mpaign MFC datta showing the maximum deptth feature in Ban
ngora 1. Plot in
ndicates minima
al change of
ID
D over the majo
ority of the tubin
ng surface.

LLocalised pittin
ng does not hav
ve a significantt impact on preessure or load-bbearing capaciity of the tubing, because of tthe support
oof surrounding,, uncorroded, material.
m This can
c be shown by b calculation, for example uusing formulae such as those iin DNV
RRecommended Practice RP F101 [ DNV 20 004 ] . Consisteent with this annalysis, field exxperience is thhat failures from
m CO2
ccorrosion in downhole tubingg typically exhiibit as leakage into the annuluus while the tubbing remains m mechanically inntact,
6 SPE 159484

rather than as collapse or bursting due to general thinning. It is expected that the Bangora production tubing can tolerate local
pitting corrosion to virtually the full wall thickness before failure and that failure would first occur by leakage to the annulus,
rather than by burst or collapse.

Therefore, for the purpose of estimating tubing life time, the minimum allowable wall thickness as per API 5CT was used,
namely 6.02 mm for these production strings [API 2004 ] .

Corrosion Modelling
Initial Corrosion Modelling

Corrosion modelling was performed using Intetech’s software model, the Electronic Corrosion Engineer (ECE ®). This
software enables estimation of carbon steel corrosion rates in CO2- and H2S- dominated systems and additionally evaluation
of various CRAs in different fluid characteristics. The corrosion model within ECE is based upon the well-known de Waard –
Lotz corrosion equations, which have been used in the oil industry for many years. The implementation within ECE has been
further developed by Smith and de Waard and calibrated with field experience; a full description has been published [Smith
2005].

ECE version 4 was used in the 2009 study. In 2011, both ECE version 4 and the newly developed ECE version 5 were used.
For the Bangora conditions, there is no fundamental change in the model between versions 4 and 5 and no major differences
in the results obtained. However, there are changes in the implementation of the model in the software package which mean
that ECE version 5 is more convenient to use. One of these relates to modelling of water condensation or evaporation up the
tubing: ECE 5 allows the user to input the liquid-phase water rate at the well-head (which is what the field Daily Production
Reports record) and automatically back-calculates the water rate at all points down the tubing. This is useful in Bangora
conditions where water rates are very low and the parts of the well may be water dry under particular operating regimes.

The quality of input data is of critical importance in corrosion modelling, and poor quality or missing data is often one of the
main barriers to achieving useful and reliable results.

The Daily Production Reports provided complete data for production rates and for well head flowing pressures. These are the
most critical inputs to the corrosion model.

Bottom hole temperatures and pressures and well-head flowing temperatures were estimated from original well test data for
each well and the daily production data. Variations in these derived inputs affect the estimated corrosion profile down the
well to some extent (for example, the depth at which corrosion rate is greatest and the estimated corrosion rates at bottom
hole or at the surface), but had little impact on the value of the maximum corrosion rate which is the key output for
estimating the remaining tubing life.

Table 1 summarises the source of the principal inputs used in modelling.

TABLE 1: SOURCES OF INPUT DATA FOR MODELLING

Parameter Source
Tubing dimensions Well completion diagrams, caliper surveys
Gas, water & condensate rates Actual, from Daily Production Reports
CO2 content Actual, 0.59% (from well testing)
Tubing deviations Actual, from deviation survey reports
WHFP Actual, from Daily Production Reports
BHFP
BHFT Estimate, based on actual values in Well Test data and daily production reports
WHFT

No liquids analysis data was available. Water was assumed to be entirely condensed water with no salts or bicarbonate
content; this is reasonable for gas wells with very low water rates.

Corrosion modeling can be carried out on an automated, on-line basis, for example by software such as Intetech Well
Integrity Toolkit (iWIT) which can feed on-line production data into the corrosion model and thereby calculate daily
corrosion rates and wall thickness losses. However, for the small number of wells in the present study, a semi-manual
approach was more efficient. The production histories were split into several time periods with similar conditions and
corrosion rates calculated for each time period. These were combined to give total wall thickness loss values for comparison
with the caliper logging data.
S
SPE 159484 7

F
Factors affectiing the Corrossion Behaviou
ur

T he CO2 corrosion rate in thesee gas wells aree the temperatuure gradient; the partial pressuure of CO2
The key factorss influencing th
y of the fluids. These factors are incoorporaated in the ECE
aand the velocity E corrosion moodel.

TThe temperaturre gradient is tyypically from 215°F

2 at bottom
m-hole to 100-150 °F at the ssurface. Other ffactors being eequal,
ccorrosion rate typically
t increaases with increasing temperatture from ambiient, but towardds the bottom oof the wells thee
temperatures arre high enough h for protectivee iron carbonatee scale to be foormed, so corroosion rates are expected to reduce.

TThe partial presssure of CO2 iss lower at the to

op of the well (due to the low
wer total pressuure at wellheadd compared witth bottom-
hhole), which tennds to reduce the
t corrosion raate towards thee surface.

O
On the other haand, gas and liq quid velocities increase towarrds the wellheaad as the gas prressure drops, and this factorr acts to
inncrease the corrrosion rate. Th he overall prod mmscfd) is a criitical factor in ddetermining coorrosion rates bbecause of
duction rate (m
itts relation to flluid velocities in the tubing.

TThe ECE modeel predicts that corrosion ratess are quite senssitive to gas prroduction rates in these particcular conditions. Caliper
ddata from well B2 provided an n illustration of
o this, Figure 6: no significan
ant wall loss waas measured beetween surveyss 2 and 3, a
pperiod when gaas rates were lo
ow, at around 4-10
4 Mmscf/d; while higher w wall losses werre measured beetween surveyss 3 and 4
wwhen gas rates were typicallyy 20 to 25 mmsscfd. Also, corrrosion is expeccted to be minim mal for periodss when the wellls are not
pproducing.

F
Figure 6: Minim
mal wall penetra ation between surveys 2 and
d 3, a period w
with low produc
ction rates; gre
eater penetratio
on between
s
surveys 3 and 4, with higher production rates.

M
Model Calibra
ation

TThe caliper maxximum penetraation values weere used for an

nalysis: the caliiper survey meean penetrationn values were inn nearly all
ccases very close to zero.

T w thickness and % penetraation calculatedd on the basis oof nominal

The caliper reports are expresssed in terms of the nominal wall
IID. It is not kno
own what the actual
a baseline tubing dimenssions were, andd it is quite norrmal for the tubbing wall thickkness as-
8 SPE 159484

supplied to be less than the nominal on average. The first logging runs in each well were some time after the start of
production, and the data show quite large % changes from the nominal dimensions. However, this is a combination of the
initial tubing dimensions deviating from the nominal (a baseline offset) and actual metal loss. True penetration figures were
derived between repeat sets of caliper data. For periods before the first survey and after the last survey, the wall penetration
was estimated by corrosion modelling only.

First-pass corrosion rate estimates were larger than the figures derived from 2006-2008 caliper data by factors of around 2-3
times. The general trends of the estimates with depth were similar to caliper data, except that the caliper data indicated that
corrosion was minimal towards the surface, while the model estimated significant corrosion rates near the surface.

The estimated corrosion rates are quite sensitive to the liquid velocity, and hence to the quantities of liquids present and to the
liquid hold-up. The corrosion model was tuned by slightly increasing the liquid hold-up values to give lower corrosion rates
more in line with the caliper surveys. Calculations were made with several hold-up values and compared with the caliper
data. Deliberately, the model estimates were kept no lower than the mid-range caliper estimates so as to remain conservative.
After calibration, the model estimates were within -0 + 20% of the measured calliper data for the three wells; see Table 2
below.

In early 2010 the Daily Production Report data all showed a step-change in the liquid condensate rate. This was an artefact
caused by changes in operation and measurements at the Bangora gas processing plant (when a hydrocarbon dew-point
controller was installed) rather than a sudden change in the make-up of the well fluids. For the purpose of corrosion
modelling, the gas:condensate ratio from earlier periods was applied to data after this event. This ensured that the calculations
are on the same basis as the 2009 calibration and modelling. This does illustrate an important benefit of having some actual
inspection data and being able to perform a calibration, in that the callibration can account for uncertainties in the input data
for modelling and other unknown or unmeasurable factors.

TABLE 2 : COMPARISON OF MAXIMUM PENETRATION FROM TUNED ECE MODEL AND FROM CALIPER DATA
Caliper data, maximum
Well Period ECE-4 (tuned) estimate
penetration in period
Survey 1 to Survey 4
B1 1.65 mm 1.9 mm
7/11/06 to 8/12/08
Survey 1 to Survey 4
B2 0.69 mm 0.83 mm
26/11/06 to 13/12/08
Survey 1 to Survey 3
B5 1.13 mm 1.16 mm
9/10/07 to 10/12/08

Calculations were made using the tuned model settings for the periods before the first caliper survey and since the last
survey, and forwards to predict tubing lifetime.

Details of Corrosion Modeling

The following Figures 7 and 8 show typical corrosion profiles generated by the ECE model for two different operating
conditions. The peak corrosion rate is usually predicted to be at an intermediate depth in these wells, but the position can be
higher or lower with changes in operating regime. This is a consequence of the interactions between the factors influencing
corroiosn mentioned above, and how these change with depth down the well.
SPE 159484 9

Figure 7 : Corrosion rate profile estimated by ECE for conditions typical of Bangora B2, circa 17 mmscfd gas rate

Figure 7: Corrosion rate profile estimated by ECE for conditions typical of Bangora B5, circa 40 mmscfd gas rate

One possible reason for a sharp drop off in corrosion rates towards the bottom of the well is that very little or no condensed
water is present in some conditions of temperature and pressure. The ECE software calculates the amounts of condensed
water present with depth, making it easy to see if this is an important factor, Figure 9.
10 SPE 159484

Figure 9 : Water rates as a function of depth calculated for the conditions corresponding to Fig. 8 above.

Lifetime Prediction

The maximum wall thickness loss was calculated with a combination of caliper measurements and corrosion modelling data
 For the period before the first caliper survey, the penetration rate is modelled using ECE and based on the actual
production daily report data
 Between the first and last caliper survey, the actual penetration from caliper surveys is used
 From the last caliper survey up to the date of analysis, the penetration rate is modelled with ECE and based on the
actual production daily report data
 For the future period, the penetration rate is modelled with ECE based on the typical operating conditions at the time
of analysis

Corrosion rate estimations are based on the empirical corrosion model in ECE, and because of the calibration process, on the
caliper data. Allowance has to be made for margin of error and uncertainties in both caliper data and modelling. The typical
standard deviation for the corrosion predictions in the range of application of the model is about 25%. For a practical, yet
conservative estimate of remaining life-time, the estimated corrosion rate + two standard deviations has been used in
calculation of remaining life, i.e. a corrosion rate 50% greater than the headline calculated rate. Assuming a normal
distribution, then the actual life time should be greater than this prediction with about 97% probability.

The corrosion model estimates made in January 2011, 24 months after the previous caliper surveys, were in very close
agreement with the actual caliper data measured afterwards in May 2011, Figure 10. This has given good confidence in the
corrosion estiamates and the life-time predictions.
SPE 159484 11

Maximum penetration
4
Modelled to Jan
3
(mm)
2011
2
Actual to May
1
2011 (Caliper
0 Survey)
B1 B2 B5
Bangora Well

Figure 10 : Comparison of model predictions (made in Jan 2011) and actual caliper data (measured in May 2011)

The following table summarises the forward predictions made in the 2009 study, the predictions in Jan 2011 (before May
2011 caliper surveys) and finally the latest predictions using the May 2011 caliper data.

TABLE 3: SUCCESSIVE TUBING LIFETIME ESTIMATES

(97% CONFIDENCE LIMITS FOR ECE MODELING)
Well Basic De Waard ECE 4 and ECE 5 Models
model
Date of 2009 At April 2009 At Jan 2011 with updated At Jun 2011 with May 2011
estimate production data caliper data
Bangora 1 Feb-2010 Aug-2011 Apr-2012 Sep-2013
Bangora 2 NA Nov-2014 July-2016 Dec-2017
Bangora 5 Feb-2011 May-2011 Oct-2011 Jan-2013

Using the May 2011 caliper data allowed greater precision in the predictions compared with the Jan 2011 estimates.
Although the the mid-range estimates (50% confidence) did not change that much, the confidence limits are tighter and the
97% confidence estimated are extended. The reason for this difference is that actual wall loss values from caliper surveys are
now used for the period Dec 2009 – May 2011, rather than modelled corrosion rates. Hence, the +50% margin on corrosion
rates is no longer applied to this period.

The estimated life for B2 is considerably longer than for B1 and B5 due to the lower production rates (and hence lower
corrosion rates) in B2 compared with B1 and B5 wells.

Ongoing operating conditions

Reservoir pressures are expected to decline slowly, which will tend to reduce the CO2 partial pressure and corrosivity
slightly. However, it is anticipated that gas production rates will be maintained so far as possible, so fluid velocities could
rise slightly. Corrosion modelling allows the impact of alternative future production scenarios to be be investigated.
12 SPE 159484

Summary and Conclusions

The combination of caliper survey data and corrosion modeling has allowed much more accurate assessment and prediction
of tubing life than using either single technique individually.

 Inspection data confirmed the morphology of corrosion attack and potential mode of failure
 Calibration of modelling against actual inspection data can overcome shortcomings or uncertainties in the input data
and allow more precise use of corrosion modelling

Following this approach, the corrosion modelling predictions and actual corrosion losses turned out to be in very close
agreement for the three Bangora wells completed with carbon steel tubing.

An important technical benefit of this work is that it has improved the knowledge of the tubing condition and therefore
reduced the risk that a tubing leak arises due to corrosion, causing possible casing damage.

The first corrosion modelling exercise (in 2009) provided sufficient confidence to defer further caliper surveys, which had
been performed approximately every six months up until that point, for about 30 months. This alone was estimated to save
approximately $2 million.

Cost savings have also arisen by being able to safely defer the time to workover of the wells in a location where workover
equipment is not readily available. Tubing replacement can be planned at as late a date as possible consistent with the
modeling predictions, thus maximising the useful life of the string.
 Based on the initial 2009 exercise, it was possible to plan for eventual work-over and completion replacement with
13Cr materials in 2011.
 The second corrosion modelling exercise, and the third corrosion modelling exercise combined with the 2011 caliper
survey campaign, allowed the scheduling of the work-overs to be pushed back further into 2013.

References

API Specification 5CT/ISO 11960:2004, Specification for Casing and Tubing, Eighth Edition 2004, Washington, DC: API.

DNV Recommended Practice RP F101, Corroded Pipelines, October 2004, Høvik Norway: DNV

Kermani M. B., Smith L. M. eds. 1997, “Corrosion Control in Oil and Gas Production”, EFC publication 23, London:
Institute of Materials

Sydberger T et al 1996, “Modes of CO2 Corrosion Damage: Consequences for Corrosion Control Strategies”; Corrosion
1996, paper 28: NACE International

Smith, L.M. and de Waard, K. 2005, “The Electronic Corrosion Engineer – Corrosion Prediction and Materials Selection for
Oil and Gas Producing Environments”, Corrosion 2005, paper 05468: NACE International