Professional Documents
Culture Documents
Model For Microbiologically Influenced Corrosion PDF
Model For Microbiologically Influenced Corrosion PDF
To cite this article: Mohammed Taleb-Berrouane, Faisal Khan, Kelly Hawboldt, Richard Eckert &
Torben Lund Skovhus (2018) Model for microbiologically influenced corrosion potential assessment
for the oil and gas industry, Corrosion Engineering, Science and Technology, 53:5, 378-392, DOI:
10.1080/1478422X.2018.1483221
Model for microbiologically influenced corrosion potential assessment for the oil
and gas industry
a a a b
Mohammed Taleb-Berrouane , Faisal Khan , Kelly Hawboldt , Richard Eckert and
Torben Lund Skovhus c
a
Centre for Risk, Integrity and Safety Engineering (C-RISE), Faculty of Engineering and Applied Science, Memorial University of Newfoundland,
St. John’s, NL, Canada; bDNV GL, Dublin, OH, USA; cCentre for Applied Research and Development in Building, Energy & Environment, VIA
University College, Horsens, Denmark
List of acronyms: APB: acid producing bacteria; Aw: water activity; BN: Bayesian network; MIC:
microbiologically influenced corrosion; MMMs: molecular microbiological methods; NRB: nitrate-
reducing bacteria; OOBN: object-oriented Bayesian network; PWRI: produced water re-injection; SPs:
screening parameters; SRB: sulphate-reducing bacteria; SRPs: sulphate-reducing prokaryotes; TDSs:
total dissolved solids
CONTACT Faisal Khan fikhan@mun.ca Centre for Risk, Integrity and Safety Engineering (C-RISE), Faculty of Engineering and Applied Science, Memorial
University of Newfoundland, St. John’s, NL A1B 3X5, Canada
© 2018 Institute of Materials, Minerals and Mining Published by Taylor & Francis on behalf of the Institute
CORROSION ENGINEERING, SCIENCE AND TECHNOLOGY 379
In BN modelling, dependence is presented in two ways: MIC influencing factors and SPs
vertical dependence where the intermediate nodes depend
The decision-making process for MIC diagnosis and manage-
on the basic or the root cause nodes, and horizontal depen-
ment lacks the availability of practical tools. The proposed
dence where the basic nodes depend on each other. This
model provides 20 SPs to help the analyst/operator assess
horizontal dependence is what differentiates the BN from
the MIC potential. SPs are probabilistic metrics used to
the logic diagram methods such as fault tree and event
measure real-time conditions and trends. These metrics assist
tree, where the structure is based on basic event indepen-
the operator in identifying the weakest elements (or links)
dency. These dependences, vertical and horizontal, are all
within the system that impact MIC potential. Based on the
dictated in the form of a conditional probabilities table
SPs assessment, the mitigation strategy can target then
based on the domain expert knowledge. To consider the
those identified factors to reduce the potential for MIC.
uncertainties, the conditional probability tables are built
Measuring these variables or factors in real-time would pro-
on the concept of noisy-OR and leaky noisy-OR gates
vide an on-line systematic screening tool to support the
[32,33].
decision-making process. If monitoring of the SPs cannot
The object-oriented Bayesian network (OOBN) provides a
be performed in real-time, a regular update could be
simple graphical interface, where the complexity is hidden
defined based on the periodicity of laboratory analysis, for
within the objects. The objects are instance nodes that contain
example. Some parameters, such as metallurgical and design
sub-structures (sub-networks) formed by interconnections of
parameters, are not practically modifiable if determined as a
usual chance node (input and output nodes). The nodes are
major contributor to MIC potential. However, most of the
connected to each other, within and without the sub-struc-
SPs, such as deposition and mitigation parameters, have
tures. An instance node can be seen as a virtual node repre-
dynamic variation and relatively easy to adapt if determined
senting an instance of another network. Following standard
as active contributors.
object-oriented terminology, an object-oriented network is
MIC influencing factors are basic variables that can be
often referred to as a class. Describing a BN network in a hier-
monitored and recorded. In the proposed model, these
archical model often makes the network much less crowded,
influencing factors are presented as leaf nodes where direct
and thus provides a much better understanding of the graphi-
input is required. As can be seen from Table 2, the SPs are
cal structure. An instance node can contain another instance
the outcome of these inputs after processing. The SPs are
node inside the subnet, an object-oriented network can be
summarised in Table 3. In the model, they represent inter-
viewed as a hierarchical description (or model) of a problem
mediate nodes; however, not all the intermediate nodes are
domain.
SPs, only those that have a physical meaning are used as
SPs. At the last stage of the modelling, the output will be
The proposed model the probability of MIC occurrence (MIC potential) and the
impact assessment of the MIC SPs. All the OOBN modelling
The proposed model takes into consideration factors affecting is run using HUGIN software [34].
the potential for MIC. These factors are grouped, based on
nature and their implications with other factors resulting in
seven object-oriented sub-networks: OOBN sub-networks
. Operating parameters In the OOBN figures (Figures 2–8), the nodes with grey and
. Fluid chemistry dotted bounding are OOBN input nodes, and the nodes with
. Settlement parameters a continuous grey bounding are OOBN output nodes. The
. Material parameters OOBN input and output nodes allow the communication
. Operating history among instance nodes (OOBN sub-networks). Detailed
. Mitigation parameters structures of the sub-networks are illustrated below in
. MIC presence symptoms Figures 2–8.
The sub-networks contain MIC influencing factors and
MIC SPs connected to the MIC potential. The MIC SPs are Operating parameter sub-network
metrics used to capture the performances of different com- The proposed sub-network considers nine operating factors,
ponents of the system, from the design to the mitigation strat- including four process variables: temperature, pressure, flow
egy. The MIC presence symptoms are those factors whom and pH; and two SPs: deposition, and flowing parameters.
their concomitant presence in a specific layout can be inter- Figure 2 presents the OOBN sub-network of the operating
preted as a strong sign of MIC occurrence such as, the con- factors that influence the MIC potential and their inter-
centration of microbiological activity products, and the actions, and Table 4 summarises the variance and relevance
biofilm content and geometry. of each factor. The operating temperature has a significant
Figure 2. OOBN sub-network of the operating factors that influence the MIC potential and their interactions.
CORROSION ENGINEERING, SCIENCE AND TECHNOLOGY 383
Figure 3. OOBN sub-network of the fluid chemical factors that influence the MIC potential and their interactions.
Figure 4. OOBN sub-network of the metallurgy and the surface factors that influence the MIC potential and their interactions.
Figure 5. OOBN sub-network of the operating history factors that influence the MIC potential and their interactions.
384 M. TALEB-BERROUANE ET AL.
Figure 6. OOBN sub-network of the MIC presence symptoms and their interactions.
Figure 7. OOBN sub-network of the mitigation strategies and factors that influence the MIC potential and their interactions.
total dissolved solids and the temperature are required to number (PREN) value is given by the formula as follows:
assess the solubility of the dissolved oxygen.
PREN = %Cr + 3.3 × %Mo + 16 × %N (1)
Material parameter sub-network A general review of literature in which MIC is cited as the
The proposed material parameter sub-network considers cause of corrosion shows that as the PREN value increases,
nine factors and three SPs covering the metallurgy and design the frequency of MIC decreases [37].
aspects, where the third parameter encompasses the surface
features such as the roughness and the presence of welding Operating history sub-network
marks. The proposed operating history network considers six
Figure 4 presents the OOBN sub-network of the metal- influencing factors and one SP ‘operating history’.
lurgy and surface factors that influence the MIC potential Figure 5 presents the OOBN sub-network of the operating
and their interactions, and Table 6 summarises the variance history factors that influence the MIC potential and their
and relevance of each factor. The Pitting resistance equivalent interactions, and Table 7 summarises the variance and
Figure 8. OOBN sub-network of the settlement factors that influence the MIC potential and their interactions.
CORROSION ENGINEERING, SCIENCE AND TECHNOLOGY 385
Table 4. Leaf nodes description for the operating parameter sub-network presented in Figure 2.
Variance
Influencing Low/Medium/ Low/ Yes/
Class Subclass factors High High No Specific Relevance/impact
Operating Flowing Flow velocity [0, 1], [1, 2.5], Impacts the microbiological deposition and
parameter parameter above 2.5 m/s migration. Low velocity is the best condition for
Flow type Stagnant, the microbiological growth
Intermittent,
Continuous
Deposits Debris X Their accumulation promotes the biofilm settlement
Presence presence
Parameter Sand presence X
Deposit None, [1–3] per Counters the accumulation process of the deposits
elimination year, over 3 on the metal surface
times/year
– Operating [15, 70]°C, [71– Major role in intensifying or restraining the
temperature 120] °C, others microbiological growth based on the range
Operating [0–3], [3–103] Impacts the microbiological activity. Most
pressure Above microorganisms are killed at high pressure
103 MPa
Operating pH [0–5], [5–9.5], A pH range between 5 and 9.5 is the optimum range
[9.5–14] for the microbiological growth. However, the
impact of the operating pH is depending upon the
maturity of the biofilm consortium
Multiphase X A multiphase fluid offers weak spots for corrosion
fluid
relevance of each factor. Intermittent operations or downtime The stress, either generated by applied loads or residual stress,
are mostly due to emergency shutdown or scheduled shut- can cause a shift of the electrochemical potential by increasing
down for inspection and maintenance; both duration and fre- the internal energy level of the metal. Another mechanism
quency of the downtime are considered in the model. The that can be observed more likely on long transmission pipe-
record of the draining frequency for the last five years of oper- lines, is the generation of micro-cracks on the metal surface,
ations is also considered in this model. Some abnormal situ- or damage to the protective surface coating. The generated
ations such as the excessive internal and external stress are spots can potentially host the early microbiological deposits
also considered as factors affecting the MIC occurrence. to form the biofilm consortium.
Table 5. Leaf nodes description for fluid chemistry sub-network presented in Figure 3.
Variance
Low/Med./
Class Subclass Influencing factors High Avail/ not-avail. Specific Relevance/impact
Fluid Nutritional Carbon dioxide level X Corrosive gas. Common factor in corrosion and
Chemistry parameter presence of microbiological growth
Organic carbon Threshold: Important nutrients for microorganisms
20 mg/l
Other mineral carbon X Nutrients for microorganisms
Polysulphides level X Essential nutrient for MIC related
microorganisms
Oxygen level X Corrosive gas. If present in naturally anaerobic
environments, can promote microbiological
activity
Lactate level X Rich source of organic carbon for MIC related
microorganisms
Sulphate presence Threshold: Electron acceptor for MIC related
10 mg/l microorganisms
Redox potential Electron acceptors X Enhance activity of MIC related microorganisms
presence
Electron donors X
presence
Oxygen level X See above
Hydrogen level X Major electron donor, essential for the
electrochemical activity of the MIC related
microorganisms
Solubility of Salinity or TDS Threshold: 60 g/l Impacts the form of the microbiological growth
dissolved (type of microorganisms)
oxygen Temperature [15, 70], [71– Key factor in inhibiting or enhancing the
120], others microbiological growth and corrosion
Chemical Propor- C:N ratio Threshold: 10 Ratio key in microbiological growth
tions Water activity (Aw) [0–0.59], [0.6– A boundary for microbiological life. At low
0.89], [0.89– water activity (below 0.6) microorganisms
1] cannot survive
N-S ratio Threshold: 1 Ratio key in microbiological growth
CO2–H2S ratio (sour/ Threshold: pCO2/ Ratio is indicator for degree of souring and
sweet regime) pH2S = 20 microbiological growth
TDS: total dissolved solids.
386 M. TALEB-BERROUANE ET AL.
Table 6. Leaf nodes description for the material parameter sub-network presented in Figure 4.
Variance
Low/Medium/ Applied/
Class Subclass Influencing factors High not-applied Specific Relevance/impact
Material Surface Welding marks X Indicator of predisposition for microbiological
parameter parameter Surface waviness X attachment to metal surface
Surface roughness X
Surface coating Not existing, Protects metal surface
damaged, non-
damaged
Metallurgy Metal surface [−50, +150] mV, Plays a major role in the electrochemical activity
parameter conductivity other of the metal surface
Cathodic protection X Reduces the conductivity on the metal surface
PREN Value [0–32], [33–38], Indicator of estimate of the corrosion resistance.
higher than The PREN value is proportional to the corrosion
38 resistance of the steel
Design Bends and X Weak spots where the MIC is most likely to
Parameter inclination manifest
Discontinuities and X
dead legs
a
Pitting resistance equivalent number.
Table 7. Leaf nodes description for the operating history sub-network presented in Figure 5.
Variance
Class Subclass Influencing factors Low/Medium/High Specific Relevance/impact
Operating Downtime Downtime duration X Downtime provides suitable
history Downtime frequency X conditions for the microbiological
growth
Abnormal Internal stress X Shifts the electrochemical potential
situations External stress X by increasing the internal energy
level of the metal
Draining frequency None, [1–3] per year, over 3 Counters the accumulation process
times/year of the deposits on the metal
surface
Process equipment [0–5] years, [5–15] years, The wearing and deterioration
age Over 15 years process provides weak spots
favourable for the
microbiological growth
Table 8. Leaf nodes description for the MIC symptoms sub-network presented in Figure 6.
Variance
Low/
Med./ High/
Class Subclass Influencing factors High Low Specific Relevance/impact
MIC presence Microbiological activity Ferrous sulphide X Indicators of the activity of the MIC
symptoms products concentration related microorganisms
Thio-sulphate concentration X
Hydrogen sulphide X
concentration
Microbiological activity Concentration of planktonic X Acts as a regeneration source for the
microorganisms sessile microorganisms
Biofilm presence X Creates an environment where the
MIC process is hosted
Biofilm thickness X Indicates the stability and maturity of
the biofilm structure
SRB concentration X Play a major role as MIC-related
APB concentration X microorganisms
Other microbiological Can promote/ Have a role in either promoting or
presence inhibit MIC, other inhibiting MIC
MIC presence symptoms sub-network Figure 6 presents the OOBN sub-network of the MIC
The proposed MIC presence symptoms network considers 12 symptoms and their interactions, and Table 8 summarises
factors and 4 SPs. Microorganisms are presented in two cat- the variance and relevance of each factor. The MIC symptoms
egories. The planktonic microorganisms are floating microor- are divided into two classes. In the class of the microbiological
ganisms in the process fluid. The sessile microorganisms are activity products, it is very challenging to distinguish the ori-
the microorganisms attached to the metal surface in a biofilm gin of some products, either from the process fluid or the
structure. microbiological activity. The microbiological activity class
CORROSION ENGINEERING, SCIENCE AND TECHNOLOGY 387
Table 9. Leaf nodes description for the mitigation parameter sub-network presented in Figure 7.
Variance
Class Subclass Influencing factors Low/Med./High Specific Relevance/impact
Mitigation Reactive mitigation Pigging frequency None, [1, 6], above 6 Most common method for mechanical
parameter parameter times/year mitigation against biofilm development
Use of biocides X Chemical treatment method to prevent/mitigate
Use of molybdate X biofilm development
Proactive mitigation Cathodic protection X Reduces conductivity on the metal surface
parameter Nitrate injection X Anti-souring treatment. Enhance growth of
nitrate-reducing bacteria (NRB) to outcompete
SRB
Pigging frequency None, [1, 6], above 6 See above
times/year
Use of biocides X See above
Periodic draining None, [1–3] per year, See above
over 3 times/year
Preventive mitigation Internal Coating Damaged, non- Protects the metal surface
parameter damaged
Periodic draining None, [1–3] per year, Counters the accumulation of deposits on the
over 3 times/year metal surface
Microbiological Inspection Periodic, non- Provides a clear picture of the wall
monitoring parameter periodic characteristics, pits and biofilm presence
Use of bio-probes None, annually, over System monitoring to capture any change in the
1 time/year corrosive process and corrosion rate
Weight-loss None, annually, over
Coupons 1 time/year
Electrochemical Periodic, non-
monitoring periodic
Use of MMM Applicable, not tracks the microorganisms considered
applicable responsible for the MIC potential
Table 10. Leaf nodes description for the settlement parameter sub-network presented in Figure 8.
Variance
Low/ High/
Class Subclass Influencing factors Med./High Low Specific Relevance/impact
Settlement Water wetting Water presence X Essential and limiting element for the
parameter parameter microbiological growth
Surface X See ‘Surface Parameter sub-class’ in Table 6
predisposition
Surface coating Not applicable, damaged, Protects the metal surface
non-damaged
Anchorage ability Rust bubble X Provides surface for the microbiological
presence attachment on the metal surface
Abiotic general X
corrosion
Pit presence X
Surface X See ‘Surface Parameter sub-class’ in Table 6
predisposition
Nutritional Favorable, non-favorable Essential and limiting parameter for
parameter microbiological growth
Figure 9. The overall OOBN model for the MIC potential assessment showing the interactions among the sub-networks.
examination of the pipeline also revealed a few other pits model. For more details about this case study, the reader
that were similar to but smaller than the one that leaked. is referred to [9].
At the end of the investigation, the experts’ diagnosis con- The available field and laboratory data were input to the
cluded that MIC caused the damage. For this reason and OOBN model and the generated results are presented in
the data availability, in this case, the model validation Tables 12 and 13. The data from the case study was provided
was built based on data from this case study. Table 11 sum- in detail; however, some data necessary for the model were
marises the main field and laboratory parameters for this not specified, for example, the biofilm thickness, usage of bio-
case and the pieces of evidence used to validate the cides, etc. In those cases where the information is not
Figure 10. (a) Percentages of the abiotic SPs favourable to MIC potential (Part 1). (b) Percentages of the abiotic SPs favourable to MIC potential (Part 2).
390 M. TALEB-BERROUANE ET AL.
available, the model assumes equal probability distribution of Abiotic parameters are represented by process variables
all the node states of the missing information. For example, a and operations, design, and fluid chemistry aspects. Biotic
node with two states will have a 50% chance of being in state 1 parameters are the microbiological-related parameters. The
and a 50% chance of being in state 2. The same rule is applied mitigation inefficiency assesses all the different types of miti-
to a node with three states where the chances are eventually gation strategies along with the microbiological monitoring
divided among the states. For example, the information and the biofilm degradation parameter. It is worth noting
about the biofilm thickness is not available in this case. The that the microbiological monitoring can be performed by
model assumes a 33% chance of having a biofilm with a using different techniques; the most efficient method reported
high thickness, a 33% chance of a medium biofilm thickness, in the literature is the MMM such as the qPCR [40].
and a 33% chance of a low biofilm thickness. The equally dis- Figure 10(a,b) shows the percentages of the abiotic SPs
tributed probability is considered as uncertainty in the model. being favourable to MIC potential. From those figures, the
Thereby, the results of the case, in Table 13, are built using critical parameters can be extracted as follows:
this averaging method. To quantify the impact of these uncer-
tainties, the model calculates a lower limit, the ‘Ideal case’,
where the unavailable information is assumed at the levels 1 The water wetting parameter (99% favourable to MIC
that cause the lowest chance of MIC potential. The upper occurrence). The water wetting can be mainly improved
limit, the ‘worst case’, considers the unavailable information by applying a coating to the metal surface and reducing
is assumed at the levels that cause the highest chance of the presence of water by water purging or draining.
MIC potential. Consequently, the more information that is 2 The attachment parameter (90% favourable to MIC occur-
available for the model, the narrower the difference is rence). The microbiological attachment is mainly due to
between the upper and lower limits, which reflects the accu- the ability of the microorganisms causing MIC to settle
racy of the model. and remain attached to the metal surface. Acting to mini-
The results in Table 12 show the MIC potential in this case mise the deposition process by periodic draining and pre-
study to be 82% with the worst-case scenario to be 96%. Com- treatment along with water filtration and pigging could be
paring these results with the field data, where the MIC process appropriate strategies to lower the microbiological attach-
has been identified with certainty to be the main cause ment capability.
confirms that the 82% reflects a high likelihood of MIC Figure 11 shows the percentages of the microbiological SPs
which was confirmed as the source of failure by [9]. being favourable to the MIC occurrence. From this figure, the
critical parameters can be extracted as follows:
Sensitivity analysis of the SPs
The SPs were further analysed for their sensitivities towards 1 Sessile microbiological presence (96% favourable to MIC
MIC potential assessment. The results of their sensitivity occurrence). Targeting the biofilm structure hosting the
analysis are shown in Table 13. The SPs in each of the cat- sessile microorganisms would be the appropriate strategy
egories are further analysed in Figure 10(a,b) and Figure 11. to lower the sessile microbiological presence.
2 Microbiological activity (91% favourable to MIC occur- being 96%. This provided initial validity of the model and
rence). The microbiological activity can be reduced by tar- projects its usability in real life situations. This model will
geting the microbiological regeneration in sessile and be further tested and validated against several types of
planktonic forms. environmental archetypes such as
Figure 12 shows the percentages of the mitigation SPs . Crude systems
being efficient in attenuating the MIC. The lower in the . Produced water re-injection – systems
efficiency of the mitigation, the more critical this parameter . Sea water
becomes. The critical parameters are: . Multiphase
. Storage
1 Preventive mitigation parameter (0% efficiency). The pre- . Transmission
ventive mitigation can be improved by applying an internal
coating and performing period draining to the pipeline. The accuracy of the model is highly dependent on the
2 Biofilm degradation parameter (5% efficiency). The success reliability of data from the field and the laboratory tests.
of the biofilm degradation depends on two factors; (i) assess- Nevertheless, the model is able to adapt to missing data and
ment of the location and solidity of the biofilms, (ii) a proper also able to consider new data as evidence to update an earlier
correlation between the mechanical mitigation (pigging) and prediction. Application of the model will help promote better
the chemical mitigation (use of biocides). Thereby, the understanding and management of MIC in onshore and
improvement of the biofilm degradation parameter should offshore process operations. Based on the promising findings
be based on a proper analysis of those three factors as a sys- of this work, efforts related to the evaluation of the impact of
tematic strategy to struggle the biofilm development. the MIC on processing systems is continuing and will be pre-
sented in future papers.
Conclusions
Acknowledgements
This paper presented a new model for assessing the potential
Authors thankfully acknowledge the financial support provided by Gen-
for MIC. The model is built upon 60 influencing factors that ome Canada and supporting partners such as Suncor, Husky, Research
form 20 SPs. The synergies and dependences among the par- and Development Corporation of Newfoundland (known as Innovate
ameters are considered in modelling the MIC potential. The NL) through large-scale applied research project grant. Author Khan
model is developed in an object-oriented Bayesian framework acknowledges the support provided by Canada Research Chair (Tier I)
programme.
that is adaptive and easy to follow. The graphical illustration of
the model as interconnected instance nodes provides a clear
understanding of interactions of factors and SPs. The con- Disclosure statement
ditional dependence of parameters in a node is defined consid- No potential conflict of interest was reported by the authors.
ering the opinion of subject matter experts and past studies.
The model was tested against most cited MIC induced fail-
ure of a pipeline study available in the public domain. The Funding
model estimated MIC potential of the given case study This work was supported by Canada Research Chairs Offshore Safety
(using the available data) was 82% with the worst scenario and Risk Engineering; Genome Canada.
392 M. TALEB-BERROUANE ET AL.