Paper No.
03623
2003
CORROSION
CO2 CORROSION PREDICTION BY USE OF THE NORSOK M-506 MODEL – GUIDELINES AND LIMITATIONS Stein Olsen Statoil Research Centre 7005 Trondheim – Norway
ABSTRACT Prediction of CO2 corrosion is required for material selection and determination of the required corrosion allowance. Several models giving different results for the same input parameters are available and can be used. No international guidelines or standards exist for corrosion prediction. The NORSOK standard M-506 is used by the Norwegian operators and authorities for prediction of CO2 corrosion. This paper describes limitations and guidance for use of the model in different conditions and for different systems. All predictions are associated with very large uncertainties and the trend in the oil industry is a shift in focus from development of prediction models to more focus on effective use of inhibitors. Keywords: corrosion modeling, carbon dioxide, carbon steel INTRODUCTION CO2 corrosion of carbon steel (C-steel) materials is one of the main corrosion mechanisms encountered in the oil and gas industry. In most oil and gas production fields CO2 is present in quantities up to several percent in the gas phase. CO2 may be the only corrosive specie present in the environment (sweet service) or a combination of CO2 and H2S may be present in the well stream. H2S over a certain level will represent a risk for sulphide stress cracking (SSC) of materials and the environment is usually defined as “sour”. In addition to representing a risk for SSC, H2S will also influence the corrosion rate. In this context, the system is often categorised as CO2 or H2S dominated dependent on the ratio between CO2 and H2S in the gas phase. CO2 corrosion represents a potential problem from the production tubing until the gas is removed from the hydrocarbon (HC) phase. This paper mainly focuses on CO2 corrosion from the wellhead to the down stream process unit, i.e. flowlines, process piping and pipelines.
Copyright 2003 by NACE International. Requests for permission to publish this manuscript in any form, in part or in whole must be in writing to NACE International, Publications Division, 1440 South Creek Drive, Houston, Texas 77084-4906. The material presented and the views expressed in this paper are solely those of the author(s) and not necessarily endorsed by the Association. Printed in U.S.A.
1 Publication Right Government work published by NACE International with permission of the author(s). Requests for permission to publish this manuscript in any form, in part or in whole must be made in writing to NACE International, Publications Division, 1440 South Creek Drive, Houston, Texas 77084-4906. The material
Prediction of corrosivity and quantification of potential corrosion rates is required to be able to make a cost effective material selection. Particularly for long pipelines, the cost of material represents a significant part of the total capital cost and a correct material selection is thus important to optimise the life cycle cost for a development. There is a variety of prediction models for CO2 corrosion available. Some models are commercial products and some are described in the open literature. Nyborg1 published a comparison of different models and also tried to correlate predictions to field data. There is a large scatter in predicted corrosion rates from different models. This does not automatically imply that some models generally are better than others, but may reflect that some models are suited for specific conditions and should thus be limited for these conditions. Some models require input of many parameters, other can work with less detailed input. Models are usually developed for a defined domain, and their use outside this domain will reduce the accuracy of the prediction. Correlation of models to field data has been very difficult because accurate uninhibited corrosion rates data from field applications with all the necessary input data are very difficult to obtain. CO2 corrosion mechanisms are extremely complicated due to the effect of protective layers on the surface. Bare metal CO2 corrosion is fairly well understood, but still after many decades of detailed studies, the mechanisms related to formation of protective layers are not predictable with sufficient accuracy. The differences between various prediction models are in fact mainly due to this effect. The decision to include the protective effect of carbonate layers in a prediction model is partly technical and partly “strategic” to reflect a degree of conservatism in the prediction. Models giving different results and lack of universally accepted guidelines on CO2 prediction complicate the process of material selection. For large companies doing all the important material selection in house with their own models, this may not be a big problem. The trend in all industries, however, is to use consultants and engineering companies for material selection. Without any clear guidelines, the outcome of a prediction is thus dependent on which model is used. From an industry standpoint, it is preferable to have guidelines and standards for CO2 prediction. The argument that often is used when it is proposed to develop universal guidelines is that this is too complicated and too many commercial interests are involved. This is of course true, on the other hand, standards and guidelines have been developed in other areas where the starting point probably was as complex as it is for CO2 prediction. USE OF MODEL PREDICTION Flowlines and pipelines for new developments is the most important area for CO2 corrosion prediction. The capital cost for such lines usually represents a significant part of the total capital cost, and the decision to select a C-steel or a corrosion resistant alloy (CRA) as pipeline material is thus of great importance. The C-steel option is normally a combination of C-steel and inhibition. Such material selection is usually performed at a very early stage of a development, and the outcome of the prediction can influence the whole concept. An alternative solution to a very expensive CRA pipeline may be that some up stream processing is undertaken to reduce the corrosivity to be able to select a less expensive material. CO2 corrosion prediction is also used as input in detailed engineering for material selection in process piping in gas treatment and oil stabilisation processes. In such cases other aspects than CO2 corrosivity will normally be considered, i.e. safety and the need for regular inspection. Corrosion
2
inhibition may also be more complicated in process piping due to flow effects in complex geometries and possible stagnant conditions in dead legs. Thus, many companies prefer to use CRAs in process piping, unless the corrosivity is very low or the expected design life is very short. Predicted CO2 corrosion rates are used as input to calculate the expected total accumulated metal loss over the lifetime for the system. The results from such calculations are used in an early stage of a development to decide if C-steel or a CRA shall be used. During detailed engineering, the results are used to calculate the required corrosion allowance (CA). For a long large diameter pipeline, even a 1 mm additional CA represents a large cost. There are basically two concepts that are used for calculating the accumulated corrosion in an inhibited system; the inhibitor efficiency, and the inhibitor availability concept. Both use the uninhibited predicted CO2 corrosion rate as input. • The inhibitor efficiency is defined as the reduction of the un-inhibited predicted CO2 corrosion rate in %. This approach reflects the theory of surface coverage. Typical values used for the efficiency are 85 – 95%. The accumulated corrosion is calculated based on the inhibited corrosion rate over the whole lifetime. • The availability concept is based on the assumption that the inhibitor is not present in the system at the required concentration over the whole lifetime. The availability is the period in percent of the total lifetime that the inhibitor is present and working. The inhibited corrosion rate is defined based on field experience to be in the order of 0.1-0.3 mm/year. The total accumulated metal loss is thus given from the sum of the inhibited corrosion rate for the period the inhibitor is present and the predicted uninhibited corrosion rate for the period the inhibitor not is present. The trend in the oil and gas industry is now to use the availability concept, which intuitively better represents the real life situation. It can also be used to define requirements for the inhibitor programme, as a high corrosivity will require a high availability, which has to be reflected in how inhibition is monitored during operation. It has been a shift in industry focus from prediction to operation. Prediction is still very important in order to make the right decisions during engineering. However, CO2 corrosion prediction is only a part of a process to achieve a successful operation. After many years of intensive research, prediction of CO2 corrosion is still very uncertain. A successful operation will mainly depend on a good inhibitor programme, and this has led to more emphasis on the inhibitor selection and deployment and management issues. Prior to start of operation, inhibitors are selected, and the injection system and the monitoring equipment has to be installed. A key issue to achieve successful operation is then the management system that should consist of the following elements: • Policies / strategies • Organisational aspects • Delegated responsibilities • Procedures and standards • Reporting routes • Feedback of information • Audits
3
THE NORSOK STANDARDS The NORSOK standards2 are developed in Norway by Statoil, Norsk Hydro and Saga Petroleum (now Norsk Hydro). The NORSOK standards consist of standards for various technology areas, among them materials technology. There are mainly two materials standards that are relevant for CO2 prediction and material selection. The NORSOK M-5063 is the CO2 prediction model available as an Excel spread sheet. The NORSOK M-0014 is the main material selection document that describes how predicted corrosion rates shall be used for material selection. Also the NORSOK standard M-0025 contains parts that give the justification for the procedures selected in the other standards. The NORSOK standards are now adopted by the Norwegian oil companies and the Norwegian Petroleum Directorate (NPD) as a standard for CO2 prediction for material selection and calculation of the CA. MODEL DEVELOPMENT A detailed description of the NORSOK M-506 model is given by Halvorsen and Søntvedt6. The model is a semi-empirical model only based on laboratory data from experiments undertaken at the Institute for Energy Technology in Norway. A total of 2400 data points form the basis for the model. All experiments are performed in flow loops at various temperatures, pH values, CO2 pressures and temperatures. Neither hydrocarbon phase nor free gas phase has been present in the experiments. The selection of data was performed based on the following criteria: • Only data from one type of steel (St52 type C-steel) • Specimens with general corrosion only • Testing with low iron content • Tests without Calcium or Acetic acid • No pre-corrosion Different types of steels have been included in the tests and differences in corrosion rates have been observed for various steels. The decision to select the St 52 type steel is based on the fact that the majority of data are obtained for this material and that the St 52 type steel generally gives the highest corrosion rates. This implies that the predicted corrosion rates are conservative with respect to most steel types. The chemical composition of the St 52 steel is given in Table 1. During corrosion testing of C-steels in CO2 environment, the morphology of the corrosion attacks can vary from uniform corrosion to localized corrosion in the form of Mesa-type attacks. Mesa attacks are characterized as defined areas with apparently deep uniform corrosion. Very seldom more pitting-like attacks have been observed in the loop testing. Typically, local corrosion rates can be significantly higher than uniform corrosion rates. It is not unusual to observe Mesa attacks 5-10 times deeper than uniform attacks at the same test conditions. As only uniform corrosion rates are used as basis for the NORSOK model, it is obvious that the prediction can underestimate the corrosion rate dramatically if Mesa corrosion occurs. The mechanisms for Mesa corrosion are not fully understood, but the flow rate is clearly an important parameter. This means that accurate prediction of Mesa corrosion in fact is not possible and that all prediction models can underestimate the corrosion rate if Mesa corrosion occurs. The complexity of CO2 corrosion is particularly linked to the protectiveness of carbonate layers formed on the surface. The general trend is that the carbonate layers are more protective at
4
elevated temperatures and at high pH values. The protectiveness is, however, not only linked to the saturation level of iron in the bulk phase, but there is a strong influence of the kinetics of the formation of layers. Apparently two identical experiments can give very different results, which may be attributed to minor differences in surface preparation and how the experiments are started. A stop of circulation in the flow loop will lead to a local increase in the iron content and sometimes this can lead to the formation of more protective layers. Specimens with an oxide on the surface will normally corrode with a lower rate than polished specimens. This illustrates the complexity of the mechanisms and thus the accuracy that in fact can be obtained for a prediction model. The data used in the NORSOK model are all from experiments with low content of iron ions, no stop in circulation and only with polished specimens, which will lead to conservative results. By analysing actual corrosion attacks from in-service pipelines and piping, the morphology of the attacks seems to be more diversified than what can be observed from single phase laboratory experiments. Mesa corrosion is not very often reported, and local pitting attacks are more often seen. This may be due to different surface conditions as pipes usually have a mill scale, and the wetting conditions will be very different due to many stops and that only small amounts of water is present in the start of production. In addition local deep grooves can be observed in connection with bends and Tees where local flow effects can be important. Often corrosion attacks are observed in connection with welds. None of these parameters, which can be decisive for the final corrosion rate, can be used in a prediction in an early phase of a project as this is not known. Most experimental work is performed for temperatures above 200C as low temperature testing requires cooling. Most models are thus basically based on data for higher temperatures. For long subsea pipelines, a major part can in fact be operated at low temperatures. Extrapolation of high temperature data down to low temperatures (e.g. 40C) can lead to large errors in predicted corrosion rates. Experimental work showed corrosion rates noticeably lower at 50C than predicted corrosion rates based on extrapolation from data at higher temperatures. Due to this, the present NORSOK model is limited to prediction above 200C. There are, however, some data from low temperature experiments. A project is ongoing to generate more data and implement these into the next revision of NORSOK M-506. Flow has a large influence on CO2 corrosion, both with respect to mass transfer and fluid mechanical wear of protective layers. In principle, mass transfer effects should only be important for film-free conditions. With a corrosion layer, diffusion of reactant and corrosion products will occur in the pores with more or less stagnant conditions and will thus only be controlled by the thickness of the layer and not by the bulk flow conditions. Fluid mechanical effects of corrosion layers will influence the build up of protective layers and may partly or fully remove such layers. A mechanistic modelling of the flow effect requires detailed description of the corrosion film in terms of porosity, thickness and mechanical strength. As this cannot be known, particularly not for new pipelines under planning, it is a fact that flow effects on CO2 corrosion rates cannot be accurately predicted. The approach used in the development of the NORSOK model has been to select the shear stress as the flow parameter. The modelling of the flow effect in the NORSOK model is a pure empirical correlation from the laboratory data. Shear stress is well defined in a single liquid system and can be calculated by use of a flow model for the water phase in a multi phase system. The flow effect observed in the single-phase laboratory experiments can thus directly transferred to multi phase flow systems by use of the shear stress. However, based on the descriptions above, shear stress cannot represent all the effects of flow on the corrosion rate, which will limit the use of the model.
5
Water wetting is in many systems the most important parameter that determines the corrosion rate in a real system. For many oil production systems, even at a water cut in the order of 20-30%, no corrosion can be observed for flowing conditions. This is due to the fact that the oil represents the continuous phase and no water wetting of the steel surface takes place. There are, however, reports of water wetting in oil pipelines at very low water cuts (e.g. 1-2 %). Lighter condensates do not entrain water and wetting can be expected even at very low water cuts. Important parameters that influence the water wetting are the oil emulsion properties, flow velocity, temperature etc., but no accurate mechanistic model exists to determine if water wetting occurs. As many of the parameters that determine the water wetting are not known in a pre-qualification phase, an accurate prediction taking the water wetting into account is regarded as impossible. In the NORSOK model, no effect of possible lack of water wetting is taken into account, and 100% water wetting is regarded as the base case. In NORSOK M-001, it is recommended to use the inhibitor efficiency concept for determination of the inhibited corrosion rate and to calculate the required CA. The trend in the industry is to use the availability concept, which may better reflect the mechanisms and that puts more emphasis on the operation in terms of requirements to ensure continuous injection. In reality, using 90% availability or 90% efficiency will basically give the same result in terms of accumulated metal loss over the lifetime. For future revisions of the NORSOK, there will be a discussion about what inhibitor concept to use. NORSOK recommends a corrosion efficiency up to 90% based on the predicted corrosion rate for pipelines, but this shall be documented. In practice, this is easily obtained in laboratory testing with a good inhibitor. For process piping, a maximum efficiency of 75% is recommended. LIMITATIONS FOR USE OF THE NORSOK MODEL The NORSOK model is based on experimental data with no H2S present in the system. There is limited data from the literature on the interaction between CO2 and H2S, and quantified prediction of corrosion rates under such conditions cannot be performed. Iron sulphides are significantly less soluble than iron carbonates. Even for trace amounts of H2S, the corrosion film will consist of a combination of carbonates and sulphides. There are strong indications that trace amounts of H2S in fact have a positive effect on the corrosion rate as the sulphides improve the protectiveness of the film on the surface. The risk for pitting corrosion, however, increases with increasing H2S content and at a certain level of H2S, the local pitting rate can exceed the predicted rate only based on the CO2 content in the system. In NORSOK this has been implemented by defining that the application of the model is only valid for conditions when the ratio of partial pressure of CO2 and H2S is larger than 20, and only if the partial pressure of H2S is less than 0.5 bars. Organic acids will influence the corrosion rate and high corrosion rates have been observed due to high content of organic acids, even for very low CO2 pressures. The mechanisms related to acetic acids have been studied in several projects and the major findings are that the potential corrosivity is increased by the increased content of un-dissociated acetic acid and that the acetic acid will reduce the tendency of protective film formation. Based on the current knowledge, it is stated that the NORSOK model can under-predict the corrosion rate if the total content of organic acids exceeds 100 ppm and the partial pressure of CO2 is less than 0.5 bars.
6
Increasing the bulk pH by use of sodium hydroxide or amines is one method for corrosion control in a pipeline (pH stabilization). This is normally used in combination with glycol. The NORSOK model shall not be used to predict corrosion rates as no experimental data for such conditions are used for the model. This method represents conditions with strong buffering and high super-saturation of iron and will normally lead to very low corrosion rates. In some cases a moderate pH adjustment is undertaken to reduce the corrosivity and make the conditions more favourable for effective inhibition. Use of the NORSOK model for moderate pH adjustment is acceptable as such conditions represent a pH that can be encountered in formation waters with high content of bicarbonate. Top-of-line corrosion can occur in hot pipelines due to condensation of water. This type of corrosion is normally more severe in conditions with high content of organic acids. Top-of–line corrosion cannot be predicted with the NORSOK model. A reduction factor for glycol to be applied on the predicted corrosion rate is proposed by DeWaard7. This factor is adopted in the NORSOK model. Since reduction in corrosion rate due to film forming inhibitors is also applied, the combined effect of the glycol and the inhibitor is defined to be a maximum of 90%, i.e. the effects shall not be added. This is based on laboratory experiments that show that these effects are not fully independent. However, a limitation of 90% is regarded as conservative. The model is based on experimental data for temperatures above 200C. As a consequence of this, all predicted corrosion rates for temperatures below 200C are set equal to the predicted corrosion rate at 200C. All experiments for lower temperatures show, however, a steady decrease of the corrosion rate with reduced temperature. More experimental data is currently being generated and will be implemented in the next revision of the model. Flow effects are modelled in terms of shear stress, which is directly transferred from single flow to multi phase flow. The correlations are purely empirical and the model shall thus not be used to determine any critical flow velocities. The NORSOK model is a single point model and the conditions for each point must be manually entered into the program. For conditions with condensing water, the pH will vary along the pipeline due to increased bicarbonate from the corrosion process. The development of pH must be evaluated separately to define the conditions for various locations in a pipeline. GUIDELINES FOR USE OF THE NORSOK MODEL Process systems Unprocessed well stream. This is defined as piping and components between the wellhead and the inlet separator. The NORSOK model should not be used for choke valves or the spool down stream the choke where very erosive conditions can be expected. Generally, the shear stresses are higher in piping systems than in longer pipelines. In addition these systems contain bends, Tees, reducers, manifolds etc., which will be exposed to particularly high shear stresses. This may influence the effect of the corrosion inhibitor in addition to potential erosion corrosion. For many fields in operation it is often a request to define the maximum operation velocity to increase production. The NORSOK model shall not be used to predict erosion corrosion or the maximum 7
allowed flow rate as the risk for Mesa corrosion or loss of inhibitor efficiency is not included in the model. For low and moderate flow rates the model can be used to evaluate the potential corrosivity in the system. NORSOK has defined a maximum reduction effect of inhibitors in such systems to 75% due to many failures in field operations. Oil stabilisation process. This is defined as separators and the piping transporting the oil phase in between the separators. The NORSOK model can be used to evaluate the corrosivity in such systems. Care should be taken when predicting corrosivity in separators due to possible stagnant conditions and sand deposits that can lead to galvanic effects and reduced effect of the inhibitor. The total pressure and thus the partial pressure of CO2 are gradually reduced in such process. It is assumed that equilibrium conditions are not achieved immediately down stream of control valves for pressure reduction. NORSOK has thus defined that the predicted corrosion rate should be increased by 25% to compensate for non equilibrium conditions, which is an arbitrary chosen value not supported by any data. Produced water systems. The pressure is normally low in most parts of a produced water system. The CO2 partial pressure is thus low and the predicted corrosion rate will be low. The general field experience is that corrosion rates in such systems are much higher that predicted. This may be due to low efficiency from the corrosion inhibitor, but conditions in such systems are often not well controlled as drain systems may be dumped into the produced water system. The high corrosion rates may thus be attributed to ingress of oxygen and potential microbial corrosion. The general recommendation is thus not to use the NORSOK model for prediction of corrosion rates in produced water systems. Gas treatment systems. This will include piping, coolers and scrubbers transporting gas from the separators. Depending on the separator design and throughput, some water droplets may be entrained in the gas. The gas is normally saturated with water except down stream of compressors where the gas is heated. Some condensation will take place in the piping, but due to very limited cooling, no significant corrosion will normally occur. After cooling, larger amounts of water will condense and potential corrosion can occur. Due to rapid cooling, this water will contain only small amounts of iron and the pH value will mainly be determined by the CO2 pressure. The general recommendation is not to use the NORSOK model for gas treatment system, except for the coolers, scrubbers and piping in between where free water can be expected. The option in the model for under-saturated conditions at low pH, defined as the “pure” pH, must be used to evaluate the corrosivity and no effect of the inhibitor shall be used unless vapour phase inhibitors are used. The effect of such inhibitors shall be documented. Pipelines Full well stream gas and condensate pipelines. In a gas and condensate pipeline, the water phase can vary from only condensed water to large amounts of formation water. In the latter case, the corrosivity evaluation will be the same as for pipelines from oil production, where formation water normally will be produced. For hot pipelines with significant condensation of water, top-of-line corrosion can represent a problem. This type of corrosion cannot be predicted by use of the NORSOK model. In a pipeline with only condensed water, there are basically two conditions that should be considered, the inlet conditions with potentially low pH water present, and conditions along the pipeline with iron saturated water. Upstream a pipeline, some water will condense in the tubing, in process piping and in small gathering lines that may be made of stainless materials. At the 8
inlet of the pipeline significant amounts of under-saturated low pH water, also called “hungry water”, will be present. NORSOK recommends that the corrosion rate shall be predicted by use of the “pure” pH given in the program. When the water phase is saturated with iron ions, which is regarded as the normal situation, the pH value given for the iron-saturated conditions shall be used. Full well stream oil production pipelines. Formation water is normally the major water phase and CO2 corrosion prediction can be undertaken by use of the NORSOK model using the partial pressure of CO2, the total pressure and the formation water content as input parameters. NORSOK is not recommending any use of reduction due to lack of water wetting. Stabilised and partly stabilised oil and condensate pipelines. The NORSOK model can be used to evaluate the corrosivity in liquid transport pipelines. The partial pressure of CO2 is defined by the conditions in the last separator stage in the process up stream the pipeline. To compensate for non-equilibrium conditions in the separator, an evaluation is needed to determine if the partial pressure should be increased somewhat. As total pressure in the prediction model, the total pressure in the same separator shall be used, even if the pressure has been increased from pumping of the liquid. The temperature to be used shall be the actual temperature in the pipeline. NORSOK is recommending no reduction in the predicted corrosion rate due to lack of water wetting. Even for low water content oil transport lines, emulsion breakers are often used in the up stream process and free water is often found in such pipelines. Many companies recommend use of corrosion inhibitors in oil transport pipelines, even for low predicted corrosion rates due to several reported failures. Dry gas pipelines. Gas for export is normally dried in a glycol contactor to achieve a water dew point lower than the operating temperature in the pipeline. NORSOK is defining that the water dew point temperature shall be more than 10 degrees below the operating temperature in order to regard the conditions as non-corrosive. Some glycol will be introduced into the export pipeline, mainly as glycol vapour that condenses in the pipeline. This glycol will extract some water from the gas and a liquid phase containing glycol with 5-8% water is normally found in dry gas export lines. This liquid will be transported as a very thin liquid film in the bottom of the pipeline. The velocity of the liquid film is very low and it may take years before this is detected at the outlet of the pipeline. Such liquid film will normally not represent any problem with respect to corrosion and corrosion prediction is not regarded as necessary. No corrosion allowance is usually used to compensate for these conditions. In case of upset conditions with introduction of significant amounts of water into the pipeline, the NORSOK model can be used to estimate potential corrosion rates, but this will require an estimate of the amount of water and a flow calculation to determine the transport of the water through the pipeline. UNCERTAINTIES The total uncertainty involved in prediction of CO2 corrosion is a result of combined uncertainties for all factors involved. In addition to the inherent uncertainty of the prediction model itself, there are large uncertainties linked to input parameters, effect of water wetting, how the system will be operated, the actual efficiency of the corrosion inhibitor and the availability of the inhibitor. The parameters involved in an evaluation of the total uncertainty can be illustrated as shown in Figure 1. As described in previous paragraphs, the prediction of CO2 corrosion has a significant inherent uncertainty. This uncertainty can in principle be treated by statistical methods based on a 9
deterministic model. This requires some form of probability distribution for the different parameters involved. There are no data that can be used to establish such distributions for parameters like the effect of pH, partial pressure of CO2, temperature, flow etc., which means that these must be made based on assumptions. As the model normally is based on some conservative assumptions, there is already some conservatism built into a model like the NORSOK model. As a consequence of all this, a standard statistical analysis will not give a realistic probability distribution for a prediction model. For most prediction models input parameters like temperature, pressure, water composition, and flow rates are required. When performing prediction for new pipelines, such parameters are normally defined in a design basis made for the project. Parameters like flow, temperature and pressure in a pipeline are normally given from a fluid flow simulation, which is based on assumed production rates and pipe diameter in addition to the fluid properties. Such data can be changed and the actual condition when production starts can vary significantly from what originally was estimated. In addition condition will vary over time. Water analysis is normally based on a few samples and can be very uncertain. Generally, it can thus be expected that input parameters are very uncertain and will contribute significantly to the total uncertainty in a corrosion prediction. For pipelines transporting limited amounts of water, water wetting is probably the most important parameter that determines the actual total accumulated corrosion in a pipeline for the whole production period. Water wetting cannot be accurately predicted, but there is still possible to identify conditions for which wetting is unlikely, i.e. oil pipelines at high velocities and low water cut. As most models, including the NORSOK model, do not consider lack of water wetting, the prediction can be very conservative for some cases. The corrosion rate for inhibited pipelines is mainly determined by the efficiency and availability of the inhibitor. When a corrosion efficiency factor is applied to a predicted corrosion rate, a fixed value of typical 90% is applied. This is normally done independently of the temperature and flow conditions. When the availability is applied in a prediction, values for the availability and the inhibited corrosion rate are defined. All this illustrates the uncertainty involved in prediction of accumulated corrosion in an inhibited system. When evaluating the total uncertainty in prediction of corrosion rates, all factors mentioned above have to be considered. To spend a lot of effort in establishing the uncertainty for one of the factors has no practical effect unless this factor is the main factor dominating the total uncertainty. The consequence of this is that a very large uncertainty always will be involved in prediction of CO2 corrosion. This has led to a shift in focus by the oil industry from development of prediction models to more focus on effective use of inhibitors. A successful operation will mainly depend on a good inhibitor programme, and this has led to more emphasis on the inhibitor selection and deployment and management issues. SUMMARY AND CONCLUSIONS There is no universally accepted standard for prediction of CO2 corrosion. As there are several models available in the market giving different results for the same input parameters, the decision regarding material selection and determination of the corrosion allowance will depend on the model that is used.
10
The NORSOK prediction model is one model available on the Internet that is used by the Norwegian oil companies. This is a simple model based on laboratory data. The main limitations for this model are the following: • The model cannot be used for ratio between CO2 and H2S partial pressure less than 20 and when the H2S partial pressure is above 0.5 bars. • The model cannot be used for total organic acid content above 100 ppm when the partial pressure of CO2 is less than 0.5 bars. • The model cannot be used for prediction of corrosion rates for pH-stabilised systems. • The model cannot be used for prediction of top-of-line corrosion. • The model is a point model and change in pH along a pipeline due to corrosion has to be determined separately. Guidance for the use of the model is given for various systems and pipelines. A very high uncertainty will always be linked to prediction of CO2 corrosion rates, and consequently the oil companies have shifted focus from development of prediction models to more focus on effective use of inhibitors, which in fact is the most important factor to achieve successful operation with no failures or leaks. REFERENCES 1.
R. Nyborg, “Overview of CO2 Corrosion Models for Wells and Pipelines”, paper no 233, CORROSION/02, (Houston, TX; NACE International, 2002)
2.
NORSOK Material Standards, http://www.nts.no/norsok/
3.
NORSOK Standard M-506, ”CO2 Corrosion Rate Calculation Model”, http://www.nts.no/norsok/m/m50601/m50601.htm
4.
NORSOK Standard M-001, “Materials Selection”, http://www.nts.no/norsok/m/M-001r2.pdf
5.
NORSOK Standard M-002, “Materials Requirement Justification”, private communication.
6.
A. M. K. Halvorsen and T. Søntvedt, ”CO2 Corrosion Model for Carbon Steel Including a Wall Shear Stress model for Multiphase Flow and Limits for Production Rate to Avoid Mesa Attack”, paper no 42, CORROSION/99, (Houston, TX; NACE International, 1999)
7.
C. DeWaard, U. Lotz, D.E. Milliams, “Predictive Model for CO2 Corrosion Engineering in Wet Natural Gas Pipelines”, paper no 577, CORROSION/91, (Houston, TX; NACE International, 1991)
11
TABLE 1 CHEMICAL COMPOSITION AND MICROSTRUCTURE OF St 52 CARBON STEEL Materia l St52 batch 1 St 52 batch 2
C
Si
Mn
S
P
Cr
Ni
Cu
Microstructure
0.18
0.34
1.50
0.017
0.023
0.08
0.03
0.010
0.15
0.18
1.57
0.011
0.014
0.03
0.04
0.015
Ferriticpearlitic Ferriticpearlitic
Input parameters: Temperature Pressure
Prediction model: Protective layers pH effect Flow effect
Water wetting
Inhibition: Efficiency Availability
FIGURE 1- List of parameters involved in a total assessment of uncertainties associated with CO2 corrosion prediction.
12
