An important aspect in corrosion prediction for oil and gas wells and pipelines is to obtain a realistic estimate of the corrosion rate. Corrosion rate prediction involves developing a predictive model that utilizes commonly available operational parameters, existing lab/field data and theoretical models to obtain realistic assessments of corrosion rates. The Case-based Reasoning (CBR) model for CO2 corrosion prediction is designed to mimic the approach of experienced field corrosion personnel. The model takes knowledge of corrosion rates for existing cases and uses CBR techniques and Taylor series expansion to predict corrosion rates for new fields having somewhat similar parameters. The corrosion prediction using CBR model is developed in three phases: case retrieval, case ranking, and case revision. In case retrieval phase, the database of existing cases is queried in order to identify the group of cases with similar values of critical corrosion parameters. Those cases are ranked in the second phase, using a modified Taylor series expansion of the corrosion function around each case. The most similar case is passed to the third phase: case revision. The correction of the corrosion rate by using a mechanistic corrosion model is utilized in order to predict the corrosion rate of the problem under consideration. The (CBR) model has been implemented as a prototype and verified on a large hypothetical case database and a small field database with real data.
An important aspect in corrosion prediction for oil and gas wells and pipelines is to obtain a realistic estimate of the corrosion rate based on the existing experience. During the last few years, several corrosion rate prediction models have been developed by oil companies and research institutes and are still currently being developed. Most of the available predictive models tend to be either very conservative in their interpretation of results or focus on a narrow range of parametric effects. The models have different approaches in calculating corrosion rates focusing to different extents on various factors such as CO2 corrosion, iron carbonate film development, oil wetting, localized corrosion etc. thereby limiting the scope of the model's application in realistic assessment of corrosion rates. Most common models cannot be used in situations where H2S or organic acids dominate the corrosion process. For corrosion rate prediction there are currently two types of computer programs: programs based on the field related data and programs based on the laboratory results. Neither group of programs can predict the order of magnitude of the corrosion rate for all field conditions.1 The prediction of corrosion rate involves developing a model that utilizes commonly available operational parameters, utilizing existing lab/field data and theoretical models to obtain realistic assessments of corrosion rates while developing a computational approach that integrates both numerical (lab trends) and heuristic (field data and experience) information and knowledge about corrosion prediction.
There are several approaches utilized in developing corrosion prediction models in the literature: mechanistic, empirical, and hybrid.1 Mechanistic models provide mathematical formulation of the chemical and electrochemical phenomena of the corrosion using mass, energy and charge balances.
