School of Chemical Engineering CHEMENG 4054 Research Project A New Risk Assessment for Microbiologically Influenced Corrosion of Metals Connor Skoss Principal supervisor: Kenneth Davey Co-supervisor: Samuel D. Collins School of Chemical Engineering, University of Adelaide, Adelaide, SA 5005 Abstract Microbiologically Influenced corrosion (MIC) is an electrochemically driven form of corrosion that is initiated by the presence of microbial activity, which can lead to failure of metals in chemical engineering processes. Existing risk assessment methods have several disadvantages in being highly dependent on specific microorganism-metal systems, and do not account for fluctuations in bacterial behaviour. A simplified MIC model was developed, and is used for a new risk assessment of MIC using a probabilistic based risk framework that can quantify random fluctuations as a distribution. This assessment was conducted for a carbon-steel pipeline at standard operating conditions. The findings of this new risk assessment framework were compared with traditional methods and found to offer significant improvements by determining the frequency that MIC would occur over a particular operating period. These findings have application to a wide range of industries involved in metal selection and fluid flow through process equipment. Keywords: carbon-steel pipe corrosion; Single Value Assessment (SVA); Fr13 risk modelling; Microbiologically Influenced Corrosion (MIC)
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
School of Chemical Engineering
CHEMENG 4054 Research Project
A New Risk Assessment for Microbiologically Influenced Corrosion of
Metals
Connor Skoss
Principal supervisor: Kenneth Davey
Co-supervisor: Samuel D. Collins
School of Chemical Engineering, University of Adelaide, Adelaide, SA 5005
Abstract
Microbiologically Influenced corrosion (MIC) is an electrochemically driven form of
corrosion that is initiated by the presence of microbial activity, which can lead to failure of
metals in chemical engineering processes. Existing risk assessment methods have several
disadvantages in being highly dependent on specific microorganism-metal systems, and do
not account for fluctuations in bacterial behaviour. A simplified MIC model was developed,
and is used for a new risk assessment of MIC using a probabilistic based risk framework that
can quantify random fluctuations as a distribution. This assessment was conducted for a
carbon-steel pipeline at standard operating conditions. The findings of this new risk
assessment framework were compared with traditional methods and found to offer significant
improvements by determining the frequency that MIC would occur over a particular operating
period. These findings have application to a wide range of industries involved in metal
selection and fluid flow through process equipment.
Keywords: carbon-steel pipe corrosion; Single Value Assessment (SVA); Fr13 risk
The 0.45 mm yr-1 corrosion output of the SVA in Table 1 is comparable to that of the value of
0.504 mm yr-1 determined by Collins et al. (2016). Similarly, it is supported by the findings of
Maxwell & Campbell (2006) that determined it is realistic that MIC has been found to induce
up to 10 mm yr-1 of corrosion to steel pipes. This validates the unit operation model as an
appropriate corrosion model when using deterministic risk assessments such as the SVA. The
Fr 13 simulations were found to be stable, where 10 000 r-MC samples were found to be
sufficient in simulating all possible practical combinations of scenarios that could occur with
MIC. If each of the simulation scenarios were thought of as one day, then Fig. 1 shows 2500
failures, where some magnitude of corrosion due to MIC occurred, this would be considered
unacceptable given (2500/10 000)*365.25 ≈ 91 failures per year or approximately one failure
every four days with a tolerance of 25 %. However, as Fig. 1 is a probabilistic distribution of
all possible outcomes, there is no reason the failure events will be equally spaced in time.
Regardless, this insight of predicting the theoretical number of occurrences of corrosion
occurring over a given time period is not available from traditional methods, such as the SVA.
Running the same simulation at lower and higher tolerances gave insight into how effectively
doubling the tolerance from 25 % to 50 % found failures reduce by approximately two-thirds
to 31 per tolerance year. Conversely, not having any tolerance level at all resulted in only a
50 % failure rate where it was anticipated to be higher. This is suggestive that in its current
mathematical expression, tolerance does not completely effectively model design safety.
Page 18 of 39
Similarly, as currently expressed within the p definition, it may be unrealistic in real operating
conditions to run a higher tolerance to achieve a more acceptable failure rate due to
limitations in implementing tighter design safety control on certain equipment.
The SVA results of the second tier studies supported literature (Maxwell and Campbell, 2006;
Javaherdashti et al., 2001) that corrosion inducing bacteria can be inactivated in basic pH
conditions, whilst thriving and therefore enhancing corrosion in more acidic pH
environments. However, the Fr 13 findings highlighted how traditional deterministic methods
such as SVA are not able to provide a complete risk analysis for processes prone to natural
fluctuations. The chlorine gas simulation at pH 7.5 showed corrosion became almost non-
existent at 0.002 mm yr-1 in the SVA, whereas the Fr 13 results found that failures still occur
48.5 % of the time. Therefore, chlorine is not as effective at fully deactivating bacteria, as the
SVA would suggest. These findings from the chlorine gas simulation are also supported from
the simulation of high acidic conditions of pH 3 where the SVA produced a very high
corrosion value of 18.79 mm yr-1, whilst the Fr 13 analysis determined a lower than
anticipated failure rate of 35 %, inferring that the bacteria cause corrosion less frequently, but
more severely. Moreover, modifying the tolerance for these studies saw only very minor
changes in failure rates. This is potentially representative of how the process should not be
run at these pH ranges long term, as no degree of design safety is fully effective. The findings
of both second tier studies are highly suggestive that the abiotic corrosion model is unable to
fully quantify fluctuations in bacterial behavior at the two extreme pH condition ranges.
Page 19 of 39
Conclusions The following conclusions were made:
1. Microbiologically influenced corrosion (MIC) of a carbon steel pipeline has been
shown to be responsive to a quantitative probabilistic risk assessment
2. The application of the probabilistic Fr 13 framework offers significant improvements
over existing corrosion risk modelling methods for conducting a comprehensive risk
assessment of the MIC of metals, and is applicable to corrosion prone environments in
Australia, such as the natural gas pipelines in Bass Strait
3. Based on a simplified abiotic corrosion model, the Fr 13 analysis found for standard
operating conditions of a pipeline, an unacceptably high frequency of corrosion
occurring, and equivalent to 91 days for each year of operation. This insight is not
available from traditional methods, such as SVA
4. The development and inclusion of the solver function to determine Ecorr improves
upon the accuracy of the original work of Collins et al. (2016) by providing a
justifiable mathematical method of calculating Ecorr at different pH and temperature
ranges
5. An abiotic corrosion model has significant limitations in quantifying fluctuations in
bacterial behaviour when using a probabilistic-based risk assessment
6. Incorporating biotic and material specific corrosion parameters into the unit-operation
model would enhance the applicability of this corrosion risk model to assess MIC for a
wider range of materials and corrosion-inducing bacterial species. This will require
conducting laboratory work that replicates that of Smith et al. (2011), but uses
additional metal samples and water solutions that replicate the environment in which
these metals are used.
Acknowledgements
I would like to thank Dr Davey, Dr Lavigne and Mr Collins, all of whom provided valuable
guidance and support throughout the project.
Page 20 of 39
Nomenclature
Numbers in parentheses after the description refer to the equation in which the symbol is first used or defined
CR Corrosion rate, mm yr-1 [12, 13]
Cb,H+ Concentration of species in bulk electrolyte = 10–pH x1000 mol m-3 [8]
Cs,H+ Concentration of species at steel surface = 10-6 mol m-3 [8]
DH+ Diffusion coefficient = 9.47 x 10-9 m2 s-1 [8]
E Potential, V [6]
Ecorr (V vs. SCE)
Free corrosion potential V [6]
Erev (V vs. SCE)
Reversible potential for species, V [7]
E°H+ (V vs. SCE)
Standard (equilibrium) potential = -0.241 V [7]
F Faraday constant = 96,485 C mol-1 [4]
ΔHH+ Enthalpy of activation = 30,000 J mol-1 for proton reduction [5]
J0,H+ Exchange current density, A m-2 [4]
jref0,H+ Reference exchange current density = 5 x 10-2 A m-2 [5]
MFe Molecular weight = 55.85 g mol-1 [12]
nH+ Number of electrons transferred in the process [4]
p Corrosion rate risk factor, dimensionless [16]
R Universal gas constant = 8.314 J mol-1 K-1 [4]
%tolerance Practical tolerance over design corrosion rate CR, % [16]
T Temperature of electrolyte, K [4]
TR Reference temperature = 293.15 K [5]
Greek Symbols
αa,H+ Anodic transfer symmetry function = 0.6 dimensionless [4]
αc,H+ Cathodic transfer symmetry function = (1 - αa,H+) = 0.4 dimensionless [4]
Page 21 of 39
δN,H+ Nernst diffusion layer thickness = 1.67 x 10-5 m [8]
ηH+ (V vs. SCE)
Overpotential, V [7]
ρFe Density of iron = 7,850 kg m-3 [12]
Subscripts
a Anodic symmetry function
c Cathodic symmetry function
T Total system parameter
References Abdul-Halim, N., Davey, K.R. (2015), A Friday 13th risk assessment of ultraviolet irradiation for potable water in turbulent flow, Food Control, 50, 770-777.
Chandrakash, S., Davey, K.R., O’Neill, B.K. (2015), An Fr 13 risk analysis of failure in a global food process – Illustration with milk processing, Asia-Pacific Journal of Chemical Engineering, 10, 526-541.
Collins, C., Davey, K.R., O’Neill, B.K., James T.G. Chu (2016), A new quantitative risk assessment of Microbiologically Influenced Corrosion (MIC) of carbon steel pipes used in chemical engineering, Chemeca 2016, 3386601
Davey, K.R. (2011), Introduction to fundamentals and benefits of Friday 13th risk modelling technology for food manufacturers, Food Australia, 63, 192-197.
Davey, K.R. (2015), A novel Friday 13th risk assessment of fuel-to-steam efficiency of a coal-fired boiler, Chemical Engineering Science, 127, 133-142.
Davey, K.R., Cerf, O. (2003), Risk modelling - An explanation of Friday13th syndrome (failure) in well-operated continuous sterilisation plant, in: Proceedings of the 31st Australasian Chemical Engineering Conference (Product and Processes for the 21st Century), Adelaide, Sept. 28 – Oct. 1, 2003.
Davey, K.R., Chandrakash, S., O'Neill, B.K. (2013), A new risk analysis of Clean-In-Place milk processing, Food Control, 29, 248-253.
Davey, K.R., Chandrakash, S., O'Neill, B.K. (2015), A Friday 13th failure assessment of clean-in-place removal of whey protein deposits from metal surfaces with auto-set cleaning times, Chemical Engineering Science, 126, 106-115.
Davey, K.R., Lavigne, O., Shah, P. (2016), Establishing an atlas of risk of pitting of metals at sea – demonstrated for stainless steel AISI 316L in the Bass Strait, Chemical Engineering Science, 140, 71-75.
Page 22 of 39
De Waard, C., Lotz, U., Miliams, D.E. (1991), Predictive model for CO2 corrosion engineering in wet natural gas pipelines, Corrosion, 47, 976–985
Gu, T., Zhao, K., Nesic, S. (2009), A New Mechanistic Model for MIC Based on a Biocatalytic Cathodic Sulfate Reduction Theory, in CORROSION 2009, Atlanta, March 22 - 26, 2009
Iverson, W.P., 1966, “Direct evidence for the cathodic depolarization theory of bacterial corrosion”, Science 151, 986-988. ISSN 00368075; 10959203
Javaherdashti, R., Raman-Singh, R.K., 2001, “Microbiologically Influenced Corrosion of Stainless Steels in Marine Environments: A Materials Engineering Approach. In: Proc. Engineering Materials”, Melbourne, Australia, Sept. 23-26.
Maxwell, S., Campbell, S. (2006), Monitoring the Mitigation of MIC Risk in Pipelines, in CORROSION 2006, San Diego, March 12 - 16, 2006.
Pots, B.F.M., John, R.C., Rippon, I.J., Thomas, M.J.J.S., Kapusta, S.D., Girgis, M.M., Whitman, T. (2002), Improvements on DeWaard-Milliams corrosion prediction and applications to corrosion management, in CORROSION 2002, Denver, April 7 – 11, 2002.
Roberge P.R., 2000, Handbook of Corrosion Engineering. McGraw-Hill, New York, pp. 35-54, 187-220, 335-336, 1047-1059. ISBN 0070765162
Von Wolzogen Kuhr C. A. H., Van Der Flught L S, 1934, “Graphitization of Cast Iron as an Electro biochemical Process in Anaerobic Soils,” Central Laboratory T. N. O., Holland
Vose, D., 2008, “Risk Analysis-A Quantitative Guide”, 2nd ed. John Wiley and Sons, Chichester, UK, pp. 64, 59, 34, 41-43, 19, 201-205 ff, 379 ff. ISBN: 9780470512845
Vose, D.J., 1998, “The application of quantitative risk assessment to microbial food safety”, J. Food Protect. 61 (5), 640-648. ISSN: 0362028X
Wanklyn, J.N., Spruit, C.J.P (1951), Influence of Sulphate-reducing Bacteria in the Corrosion Potential of Iron, Nature 169, 928-929
Weiss R., 1973, “Survival of Bacteria at Low pH and High Temperature” Indiana University, USA. Accessed 25/04/16 <http://mobile.tube.aslo.net/lo/toc/vol_18/issue_6/0877.pdf>
Xu J, Sun C, Yan M & Wang F, 2012, “Effects of Sulfate Reducing Bacteria on Corrosion of Carbon Steel Q235 in Soil-Extract Solution”, in International Journal of Electrochemical Science, Vol 7 (2012), pp. 11281 – 11296
Page 23 of 39
APPENDIX A
In Smith et al. (2011), a rotating disk electrode (RDE) was used as the working electrode
(WE) with the tip containing a sample of pipeline steel. A counter electrode (CE), made from
platinised titanium, was used as the electron sink/source, and a saturated calomel electrode
(SCE) reference electrode (RE) used for potential measurement. Synthetic produced water
was used to simulate MIC bacterial activity. This contained sulphate, chloride and hydrogen
sulphide.
It is assumed that the electrons formed by the oxidation of iron in the steel (Eq. [1]) are
consumed by the reduction of protons (Eq. [2]):
𝐹𝑒 → 𝐹𝑒!! + 2𝑒! [1]
𝐻! + 𝑒! → !!𝐻! [2]
The transfer of charge (electrons) occurs only at the steel surface between the steel and the
water (electrolyte) because of the nature of electron transport. The overall corrosion reaction
is:
𝐹𝑒 + 2𝐻! → 𝐹𝑒!! + 𝐻! [3]
The charge transfer at the steel surface can be described by the Butler-Volmer equation (Gu,
2009) for current density due to the oxidation of iron (anodic process)
𝑗!",!! = 𝑗!,!! 𝑒𝑥𝑝!!,!! ∙!!! ∙!
!∙!∙ 𝜂!! − 𝑒𝑥𝑝
!!!,!! ∙!!! ∙!
!∙!∙ 𝜂!! [4]
Page 24 of 39
The exchange current density is given by:
𝑗!,!! =!!,!!
!!∙
!!,!!
!!,!!
!!! [5]
The overpotential is given by:
𝜂!! = 𝐸!"## − 𝐸!"#,!! [6]
The reversible potential is calculated using the Nernst equation (Roberge, 2000):
𝐸!"#,!! = 𝐸°!! +!.!∙!∙!!!! ∙!
∙ log 𝑐!,!! [7]
The mass transfer at the steel surface is described by a flux balance arising from the
dependency on diffusion and charge consumption (Smith et al., 2011):
!!",!!
!!! ∙!= −𝐷!!
!!,!!!!!,!!
!!,!! [8]
This can be rearranged to isolate the current density due to mass transfer to give:
𝑗!",!! = 𝑛!! ∙ 𝐹 ∙ −𝐷!!!!,!!!!!,!!
!!,!! [9]
Because the number of electrons produced in the oxidation of iron is balanced by the
reduction of protons (shown by Eq. [3]), the total current in the system i.e. the sum of the
current due to oxidation and reduction, must be zero, namely:
Page 25 of 39
𝑗! = 𝑗!" + 𝑗!",!! + 𝑗!",!! = 𝑗!" + 𝑗!! = 0 [10]
Eq. [10] can be rearranged to give:
𝑗!" = −𝑗!! [11]
The current density due to iron oxidation can be converted to corrosion rate using molecular
weight and density (Gu, 2009) to yield:
𝐶𝑅 = !!"!!∙!!"
∙ 𝑗!" [12]
Which can be simplified to:
𝐶𝑅 = 1.155𝑗!" [13]
Eqs. [1] through to [13] define the simplified unit-operations model for synthetic MIC
corrosion rate of a steel pipeline in water.
Fr 13 Risk Simulation
In contrast to the traditional single input of the SVA, to mimic naturally occurring
fluctuations in the system the probabilistic Fr 13-risk simulation considers the input
parameters as a distribution of values, together with the probability of that value actually
physically occurring.
This means that the output will be a distribution of probabilities of particular outcomes
(Davey, 2015; Davey et al., 2015; Abdul-Halim and Davey, 2015). Because all practically
possible inputs are simulated, the output will include unwanted outcomes i.e. ‘failed’
operations in which MIC occurs. A fundamental requirement of Fr 13 risk modelling is a
practical and unambiguous definition of risk and failure (Abdul-Halim, 2016; Davey, 2011).
An off-specification of an acceptable corrosion rate can be conveniently used as a corrosion
risk factor p such that:
Page 26 of 39
𝑃 = 𝐶𝑅′− 𝐶𝑅 [14]
Where CR’ is an instantaneous rate of corrosion (or mathematically more strictly, the CR
obtained using Fr 13 simulation). Eq. [14] is convenient because for all values p > 0 the
corrosion is greater than acceptable. However, a mathematically more convenient form of the
corrosion risk factor (Abdul-Halim & Davey, 2015; Davey, 2015; Davey et al., 2015) is
𝑝 = 100 !"!
!"− 1 [15]
Eq. [15] is computationally more convenient because it is dimensionless and because
corrosion rates greater than acceptable (i.e. failures) can be readily identified for all values
p>0.
Generally however, a design specification includes some measure of tolerance – a measure of
design safety. The corrosion risk factor of Eq. [15] can then be written as
𝑝 = 100 !"!
!"− 1 −%𝑡𝑜𝑙𝑒𝑟𝑎𝑛𝑐𝑒 [16]
Eqn.’s [1] through to [16] describe the unit operation corrosion model and that of the risk
factor definition utilised in the Fr 13 framework.
APPENDIX B
This appendix will address the construction and use of the solver function that is used to
recalculate Ecorr (free corrosion potential) when one or both of the input variables for
temperature and pH are changed.
A key limitation of the original work of Collins et al. (2016) was that Ecorr was incorporated
into the corrosion model as a constant value of -0.616 as seen in the below Table B1 in Row
5. This introduces a degree of error by treating Ecorr as a constant, where in reality Ecorr is a
multivariable function of the inputted temperature and pressure. Resultantly all calculated
values that use the fixed Ecorr value would also introduce errors that affect the accuracy and
validity of the overall corrosion value output.
Page 27 of 39
Table B1: Unit Operation Model Constants (Collins et al., 2016)
Row Parameter SVA Fr 13 simulation
4 Constants
5 Ecorr (V vs. SCE) -‐0.616 -‐0.616 Constant
6 αa,H+ (dimensionless) 0.6 0.6 Constant
7 αc, H+ (dimensionless) 0.4 0.4 Constant
8 nH+ (dimensionless) 1 1 Constant
9 jref0,H+ (A m-‐2) 0.05 0.05 Constant
10 ΔHH+ (J mol-‐1) 30000 30000 Constant
11 TR (K) 293.15 293.15 Constant
12 E°H+ (V vs SCE) -‐0.241 -‐0.241 Constant
13 Cs,H+ (mol m-‐3) 0.000001 0.000001 Constant
14 DH+ (m2 s-‐1) 9.47E-‐09 9.47E-‐09 Constant
15 δN,H+ (m) 0.0000167 0.0000167 Constant
16 F (C mol-‐1) 96485 96485 Constant
17 R (J mol-‐1 K-‐1) 8.314 8.314 Constant
18 tolerance (%) -‐ 50
It was determined the most efficient method to address the limitation of using a constant value
for Ecorr was to create a solver loop based around the below Eqn. [10] for the overall charge
transfer where Ecorr is defined by when 𝑗!" + 𝑗!! are in equilibrium which yields the free
corrosion potential. Figs. B1 and B2 depict the solver parameters as inputted in the solver
function, and the cells that return the value once the loop as converged as close as possible to
zero. These parameters were created to both define Ecorr so a value is converged upon quickly
and be able to be responsive to changes in the input variables of pH (i. e.C!,!! is converted
from the input pH) and temperature.
𝑗! = 𝑗!" + 𝑗!",!! + 𝑗!",!! = 𝑗!" + 𝑗!! = 0 [10]
Page 28 of 39
Fig.1 B1 displays the five solver parameters:
1. The first parameter is 𝐶!,!" ≥ 𝐶!,!". This ensures that if corrosion is occurring, then
the metal hasn’t completely corroded, in which case concentration of iron in the bulk
solution would be likely be greater then that of the surface of the metal
2. The second parameter, 𝐶!,!" ≥ 0.001 is described by Javaherdashti et al. (2001) as the
lowest recordable surface concentration of iron
3. The third parameter, 𝐶!,!".≤ 1e-05 represents the maximum concentration of iron in
the bulk solution
4. The fourth and fifth parameters both describe the domain of values that Ecorr may take
based on the findings of Simth et al. (2011) that found the range of values to be
-0.4 ≤ Ecorr ≤ -1.
Fig. B1: Solver Parameters
Page 29 of 39
Fig. B2: Solver Loop Input Cells for equating jT = 0
Instructions for Use:
Note: Based on 2014 Microsoft Excel on Apple running OS X Yosemite. Key different details
relevant for Windows users will be highlighted when necessary
1. Changing input parameters:
SVA – Adjust one or both inputs of temperature and pH
Fr 13 – Adjust one or both inputs of temperature and pressure by changing the
RiskNormal and RiskTruncate function such that: RiskNormal (mean, stdev,
RiskTruncate (minimum, maximum)).
Figure B3: Input parameters of temperature and pressure
2. Opening Solver Function:
SVA – Solver is a default add-in on Apple running Microsoft Excel. For Windows it
will need to be opened via selectable Add-Ins. By clicking ‘Solve’ the loop will
automatically begin running and attempt to find converge to the specified constraints.
Note, if the copy of Microsoft Excel is requiring an update, then running the loop will
likely shut the program down
à Tools à Solver à ‘Open Solver’ à Click ‘Solve’
Page 30 of 39
Fr 13 – Not necessary to open and run solver when undertaking Fr 13 when changing
the input parameters. It will run a separate loop automatically for every iteration
(where the user specifies the number of iteration) once the simulation initiates.
When using Windows however the following steps must be taken to ensure that the
solver is not running a single loop and using that value for all simulated iterations.
à Simulation Settings à Macros à Select ‘Excel Tool’ instead of ‘VBA Macros’
à Click ‘Run Macros’
Figure B4: Screenshot of Guide to Opening Solver for SVA
Figure B5: Screenshot of Guide to Running Solver for SVA
Page 31 of 39
APPENDIX C
Table C1 – SVA and Fr 13 Results (T = 293.15 K, pH =5.15, Tolerance = 0 %)
Row Parameter SVA Fr 13 Simulation 1 Input 2 T (K) 293.15 293.15