Corrosion Prediction in Industrial Systems
Quality Moment
SEEMS LEGIT
Corrosion prediction – – – – – – – – – – –
Why predict corrosion? The corrosion challenge Corrosion basics Types of prediction and their usage History and families of CO2 corrosion models The deWaard equations Other common methods and software Strengths and weaknesses Pointers for success Corrosion modelling through the design process Conclusions
The Purpose of Corrosion Prediction
Why predict corrosion? • It’s dangerous. – 200 fatalities in US from pipeline corrosion failures 1989-98 (US GAO 2000)
Gas pipeline corrosion failure in San Bruno, CA 2010. Four fatalities.
Why predict corrosion? • It’s expensive: – – – – –
Australia: A$13bn p.a. (CSIRO, 2009) USA: US$279bn p.a. (US FHA, 2001) Developed world: 3-5% of GDP Developing world: 10-20% of GDP 66% of pipeline failures (AEUB, 1998)
(most of it preventable or manageable)
Why predict corrosion? • Corrosion prediction determines the design – – – – –
Design life Materials selection / corrosion allowances Corrosion monitoring and inspection Chemical treatments, coatings, linings Repair and replacement
• Determines the ultimate risk
The Corrosion Challenge
Photo: Döbert
Photo: Döbert
• One material • One design environment • Many corrosion rates Photo: Döbert
The corrosion challenge • Corrosion is a “chaos mathematics” problem. – Very small changes in conditions can lead to very large changes in outcome – Well below the detail level of design data
• More akin to predicting the weather than conventional engineering analysis
The corrosion challenge • Even under laboratory controlled conditions, corrosion rates show huge variability • Out in the real world… – Production streams vary • Process conditions and compositions are uncertain – Particularly in raw material production such as oil and gas, mining » Reservoir predictions, changes over time
– even in many manufacturing processes » e.g. side reactions, variable feedstock, process upsets
• Contaminants
– Production data is uncertain • We don’t operate to design conditions
– External conditions are uncertain • Weather, seasonal variation, coatings…
The corrosion challenge • Our aim: – To predict a reasonable conservative case – To show sensitivity to variation in conditions – To create a start point for corrosion design
• Not: – To predict corrosion rates to 3 decimal places. • this is not a precision subject • Be wary of anyone who claims otherwise.
Corrosion Basics
Photo: Axolotl Pty.
Corrosion basics • Corrosion occurs when metal is dissolved in an aqueous solution. • Slight differences in the crystal structure of metal form +ve and –ve sites. – Metal dissolves at the -ve site (anode) to release electrons – Water is split at the +ve site (cathode) to absorb electrons – The specific reactions vary according to the corrosion mechanism
M(s) -> Mn+(aq) + ne-
H2O + ½O2
-ve Anode
M
2OH+ve Cathode
ne-
Corrosion basics • Simple CO2 Corrosion H2 (gas)
CO2(g)
Gas
CO2(g) + H2O HCO3-(aq)+ H+(aq)
Fe Fe2+(aq) + 2e-
2Fe2+(aq) + HCO3- FeCO3+ H+
Water 2H+ + 2e- H2 (gas)
Fe
2eSteel
A network of chemical reactions • Chemical Reactions – – – –
Fe Fe2+(aq) + 2e CO2(g) + H2O HCO3-+ H+ 2H+ + 2e- H2 Fe2+(aq) + HCO3- FeCO3+ H+ • This is the simplest version. – Several different Iron Oxide / Iron Hydroxide / Iron Carbonate forms – Dozens of forms of Iron Sulphide – Other reactions between iron and contaminants
A network of physical mechanisms • Diffusion Processes – CO2 from gas to liquid – HCO3 from liquid to steel through the corrosion / debris layer – H2 from liquid to gas
• Physical conditions – Flow and turbulence – Heat transfer and temperature
• Contaminants – H2S, Organic Acids – A myriad of other competing reactions
All naturally occurring processes • Not engineered reactions – Little or no control
• Corrosion Engineering is Backwards Chemical Engineering – Chemical Engineering • Trying to maximise a chosen reaction and minimise side reactions
– Corrosion Engineering • Trying to predict or control a naturally occurring phenomena • Fighting the 3rd law of thermodynamics
• Uncertainty is a fact of life. It is inherent in the corrosion process.
Predictive Approaches
Three main types of predictive approaches • Corrosion databases – lists of materials, conditions, corrosion rates
• Laboratory tests – Recreate the environment, insert a sample, measure the result
• Theoretical or empirical models – Calculated predictions based upon process conditions
Corrosion databases • Pure, historical empirical data – Material + conditions → corrosion rate – May be laboratory or inspection / experience based
• Used for… – Materials that are not carbon steel – Environments that are not: • Oil, gas • Water (seawater, boilers, produced water)
– Unusual corrosives: • Other than CO2, H2S, oxygen, seawater
• Basis of corrosion engineering for many industries – Mining and mining processing – Petrochemical and Pharmaceutical
Corrosion data sources – Corrosion societies • National Association of Corrosion Engineers (NACE Corrosion Data Survey) • European Federation of Corrosion (EFC) • Australasian Corrosion Association (ACA)
– Academic papers and research – Industry bodies – Design Standards • DNV, ASME, API, Standards Australia, NORSOK
– Corporate research units and experience • What have other plants done / experienced?
– Government research units • CSIRO, US Navy, NASA, UK HSE, US DOT
– Vendors and trade associations • Steel mills, ASSDA, NiDi (Caution! Vested interests!)
Corrosion databases • Advantages – Simple – Cheap and fast – Cover materials and conditions which do not have analytical models available
Corrosion databases • Disadvantages – Low resolution • Broad classes of material and conditions – Usually not quite the conditions you want
• No knowledge of actual test / operating conditions – Flowrates, Contaminants, Operating methods – Little traceability
– No idea of the accuracy / repeatability / sensitivity – Use with caution (last resort)
Corrosion databases • Pointers for success – Find out as much as you can about the test conditions • Read the paper, talk to the researchers • Examine track record: do they cover your industry or process?
– Look for independent validation or a second source of data • Watch for cross quoting – Identical results => comes from the same experiment
– Take only data close to your example • “Worse” conditions may not be worse. – More / higher is not always more corrosive – Consult the theory and chemistry
– Use a large safety factor – Build your own • Inspection histories, own database of material performance • Share with peers / industry societies
Laboratory Studies
Photo: Asia Scientific Apparatus
Laboratory studies • Recreate the process conditions in the laboratory – Measure the resultant corrosion
• Options – In house laboratories • Large operators only
– Private / state run laboratories • CSIRO, IFE (Norway), TWI (UK)
– Universities • Curtin, Ohio, Tulsa, Imperial College London, Heriot Watt Edinburgh
– Chemical vendors • Caution! Vested interests!
Laboratory studies • Three types of lab study – Beaker (atmospheric batch test) • Simple, quick’n’dirty screening • Limited accuracy
– Autoclave (pressurised batch test) • Control the atmosphere and conditions • No flow – Can use rotation or jet impingement to simulate flow
• Good accuracy if used properly – Appropriate corrosion types and conditions
– Flow loop (pressurised, circulating test) • By far the most representative • Also the most time consuming and expensive
Typical Flow Loop
Image: Institutt For Energiteknikk (IFE) Norway
Laboratory studies • Advantages – Can “precisely” replicate your exact process conditions – Can control and tailor the test programme
Laboratory studies • Disadvantages – – – – –
Expensive Slow (months per test) Small data set (unless in a JIP) Still cannot cover every variable Vendor labs can be biased
• Best used for verification and validation of modelling / database predictions
Laboratory studies • Pointers for success – Choose a lab with industry ties or experience in your industry – Use a progressive programme • • • • • •
Start with beaker tests for screening Autoclave to determine final candidates Flow loop if required for final validation Do as many sensitivity variations as you can Run the tests for as long as possible Use real process fluids where possible
– Use a private or in-house lab where possible • Discuss tests and process requirements with the lab operator
– Buy in to JIPs early to share costs and data • Speculative membership – testing takes a long time • Too late to start when you already have a problem or project
Corrosion Models
Photo: Bill Edwards
Corrosion models • Main varieties – CO2 corrosion (oil and gas) • With amendments for H2S, chlorides, organic acids
– – – –
Seawater (marine, power generation, oil and gas) Steam / boiler water (power generation) Dissolved oxygen (power generation, petrochemical, oil and gas) Other, specialised specific correlations
The deWaard family tree of CO2 Corrosion Models
Shell HYDROCOR
BP CASSANDRA
de Waard , Lotz, Milliams 1991
deWaard, Lotz, Dugstad 1995
Total / Elf CORPLUS
Electronic Corrosion Engineer (ECE)
Many others
de Waard and Milliams 1975
Nyborg Olsen and Halvorsen
IFE Norway
University of Tulsa SPPS CO2
NORSOK M506
Gusta et.al.
Nyborg et. al..
University of Ohio MULTICORP
FE KSE Nyborg and Dugstad IFE Vapour Corrosion Model
Gusta et.al. University of Ohio TOPCORP
deWaard Milliams 1975 •
“Carbonic Acid Corrosion of Steel”, Corrosion Vol. 31, No. 5, 1975 (NACE) – First usable general corrosion model for CO2 in oil and gas systems. – Modelled the effects of… • • • •
Partial pressure of CO2 pH Temperature Surface condition
deWaard Milliams Nomograph
deWaard Milliams 1975 • Simple, usable – Based on real design parameters – Generally adequately conservative for most oil and gas liquid systems
• Limitations and weaknesses – Sour systems, organic acids – Vapour / condensing phase (“TOL”) systems
• Many tweaks and “fudge factors” published to extend the usage
deWaard, Lotz, Milliams 1991 •
“Predictive Model for CO2 Corrosion Engineering in Wet Natural Gas Pipelines”, Paper 577, Corrosion 1991, National Association of Corrosion Engineers – Probably the most widely used and practical model available. – Improvements to 1975 model: • • • • • •
Non ideal gas behaviour Corrosion product correction Measured pH Correction Condensation rate / saturation correction Effect of Oil, Glycol Effect of corrosion inhibitors
deWaard, Lotz, Milliams 1991 • •
The deWaard Lotz, Milliams Equation Vcorr = Base corrosion rate (mm/yr) – –
Correction factors (0 – 1) Fscale = Corrosion product correction
•
FpH
•
Fcond –
•
1710 0.68 log( f CO2 ) T
T = Temperature (K) fCO2 = Fugacity of CO2 (bara)
• •
– –
log(Vcor ) 5.8
= pH correction pHsat = Saturation pH pHact = Actual pH
= Condensation correction (TOLC)
log( Fscale )
2400 0.6 log( f CO2 ) 6.7 T
log( FpH ) 0.32( pH sat pH act ) pH sat 3.71 0.00417(T 273) 0.5 log( f CO2 ) C Fcond 0.25
C = Condensation Rate (g/m2s)
Fgly 1.6(log(W ) 2) Fgly –
= Glycol correction W = Water content of Glycol Mixture (weight%)
deWaard, Lotz, Milliams 1991 • Strengths – Usable • • • •
Easy to understand Based on real empirical data Uses real design parameters Easy to put into a spreadsheet
– Generally conservative in sweet, liquid systems
deWaard, Lotz, Milliams 1991 • Weaknesses – Very poor at vapour phase corrosion • Too simplistic, linear
– Does not cover important corrosion modifiers • H2S, Organic Acids, Flow
– Not always conservative • Powerful but blunt tool • Often abused for purposes beyond intent
deWaard, Lotz, Milliams 1991 • Usage – Liquid phase only – Sweet (no H2S, low organic acids) only – “Quick and dirty” estimate • Order of magnitude results
– Validate or extrapolate other methods • E.g. lab results
• Use with caution – Beware of multi-digit accuracy predictions – Be cautious, use uncertainty factors – Validate by laboratory tests / similar operations
deWaard, Lotz, Dugstad 1995 •
Influence of Liquid Flow Velocity on CO2 Corrosion: a semi empirical model”, Paper 128, Corrosion 1995, National Association of Corrosion Engineers
• Evolution of the 1991model to include the effect of liquid flow • Based upon extensive empirical data combined with theoretical mechanisms
deWaard, Lotz, Dugstad 1995 • More accurate and realistic than 1991 model – Inclusion of flow and transport of corrosives – Usually conservative
• Somewhat less usable than 1991 – Flow parameters more complex than those often readily available to designers – Usually require a simulation or flow assurance study
• Basis of many commercial corrosion prediction packages – – – –
Hydrocor Electronic Corrosion Engineer BP Cassandra (discontinued) OLGA Corrosion Module (CORR1 Liquid Model) • and many more
deWaard, Lotz, Dugstad 1995 • Weaknesses – Still unable to model corrosion in the vapour phase with any accuracy • Simplistic multiplier
– No account for H2S, organic acids – Limitations of applicability not clearly defined • Often abused beyond the scope
– Limitations of accuracy not discussed • Order of magnitude estimate only • Common to see reports quoting thousandths of mm per year
– Seductively simple • Easy for inexperienced people to use • Needs an experienced eye to interpret the results into a reliable design
NORSOK M506 • Developed by Statoil, IFE and Norsk Hydro JIP • Adopted by Norwegian Standards (NORSOK) • Similar scope to deWaard 1995 – Basic liquid phase CO2 corrosion model • Includes flow effects • Based on more readily available parameters
• Simple software freely available from NORSOK
NORSOK M506 • Strengths – Usable • Based upon empirical data – reasonably accurate / conservative
• Uses real design parameters
– Software freely available – Contains clear limitations on method applicability • Temperature, pressure, CO2 maxima
NORSOK M506 • Weaknesses – Same as deWaard 1995 • No H2S or organic acids • Not applicable to vapour phase
– Usage • • • •
Liquid phase, sweet (no H2S) Mild conditions (no high temperature or pressure) Order of magnitude estimates Verification / sense check
Hydrocor • Proprietary software from Shell Global Solutions – Until recently, only available within Shell associated operators • Or used by SGS on contract
• Advanced development of deWaard family of models
Hydrocor • Strengths – Comprehensive pipeline model • Liquid and vapour phase – Internal heat transfer model and for vapour phase
• Accounts for H2S and organic acids – Simplified models
• Comprehensive range of outputs • Automates production profiles and sensitivity • New versions very usable
Hydrocor • Weaknesses – Proprietary • Expensive, limited availability outside Shell
– Simplified flow regime model • Often better to turn off
– “Black box” – method detail not available – Questionable reliability in vapour phase • Several examples of non-conservative design causing failure
Multicorp • Mechanistic liquid phase model developed by University of Ohio JIP – Only available to members – Anyone can join • For a price
Multicorp • Strengths – Comprehensive mechanistic model • Attempts to model most of the processes in the corrosion reaction – Physical and chemical
– Incorporates H2S, organic acids, flow effects – JIP membership allows influence of test regime / model basis – Open method – no “black box”
Multicorp • Weaknesses – Requires JIP membership (medium expensive) • Can be accessed via consultants like WGIM
– Academic model – less practical / usable than Hydrocor etc. • Less usable software (large spreadsheet)
– Liquid only • Topcorp available for vapour
– Work in progress
Topcorp • Mechanistic vapour phase model developed by University of Ohio JIP – Requires separate JIP membership to use software • Open to all – for a price
– Can commission studies from “Praecipium” • University of Ohio consulting arm
Topcorp • Strengths – Probably the most sophisticated and comprehensive vapour phase corrosion model available at this time – Open availability and method
Topcorp • Weaknesses – – – –
Expensive Not very user friendly Work in progress Does not have a long track record • Method weaknesses not known
Practical Corrosion Prediction
Photo: Axolotl Pty.
Practical corrosion prediction in oil and gas projects • Technique and resolution of the predicitons develops as a design matures – Concept Select – FEED – Detailed Design
Concept Select • Initial Screening – Inlet and outlet conditions only – Liquid only (simple vapour estimate) – Simplified parameters / estimates • Detail not available yet
– Feasibility of carbon steel • Requirement for inhibitor / corrosion resistant materials
• Methods – Simple deWaard 1991, 1995 or NORSOK spreadsheet – Corrosion Database – Feasibility laboratory tests (beaker)
Concept Select
FEED • First detailed model – Liquid and vapour modelling • Cooling condensation vapour modelling only
– Initial Laboratory studies (autoclave) • Detailed feasibility check – Identify further studies (e.g. corrosion inhibitor testing) – Justify more detailed tests
– Longitudinal temperature profile (e.g. pipeline) – Single production condition or simplified production profile • Simple sensitivity assessments
– Material selection • Carbon steel corrosion allowance • Identify areas requiring corrosion resistant alloys – Identify candidate corrosion resistant alloys
FEED – Single OLGA or Hydrocor profile
Detailed Design • Full detail corrosion model – Mature flow assurance simulation • Multiple temperature profiles (changes along the process) • Detailed production profile (process changes over time)
– Full detail liquid and vapour modelling • • • •
Cooling condensation corrosion Hungry water Mixing condensation Detailed sensitivity assessments
– Aggregate through life corrosion model • 3-400 individual models
• Detailed, extensive laboratory testing – Flow loop to validate corrosion modelling – Selection of corrosion inhibitor or chemical treatment
Detailed Design – Multiple OLGA or Hydrocor Profiles
Prediction progression as the design matures • • • •
Greater detail More confidence Less uncertainty Less conservatism
The art of Corrosion Engineering
Photo: Colin Winterbottom “Elegant Corrosion”
The art of corrosion engineering
• You will have noticed… – Fundamentals of corrosion processes are unstable • Small changes in conditions can mean big changes in outcome
– Data we use is inaccurate – Models all have accuracy and applicability caveats
The art of corrosion engineering deWaard / Hydrocor Conservative
Non-Conservative
Nesic and Vrhovac, “A neural network model for CO2 Corrosion of Carbon Steel, Journal of Corrosion Science and Engineering, Vol 1. Paper 6, March 1999
IFE / NORSOK Conservative
Non-Conservative
The art of corrosion engineering
• Experience in the vagaries of corrosion behaviour is ESSENTIAL to interpret the results of a corrosion prediction into a safe and reliable design. – The model result is only the beginning
• The most important document in your materials and corrosion design is the CV of your corrosion engineer.
• The only true wisdom is knowing that you know nothing - Socrates
Thank you for your attention.
Photo: Colin Winterbottom “Elegant Corrosion”