Failure Assessment of Pipelines Due to Microbiologically Influenced Corrosion

Microbiologically influenced corrosion (MIC) is a difficult degradation mechanism to diagnose in pipeline systems due to the complex interaction between biotic (i.e., microbial) and abiotic (e.g., fluid chemistry, pipe/vessel metallurgy/corrosion, and operating conditions) factors. This complexity often makes it difficult to accurately assess pipeline failures due to MIC. However, even with available data, failure investigators often face a number of challenges in diagnosing MIC such as how to properly integrate the available datasets, questions regarding data accuracy (e.g., confidence in the sampling and/or analysis method used) and lack of available information from operators (e.g., missing data). As a result, practical MIC failure assessments are most often performed by experts or specialists with significant knowledge and working experience in this topic. Based on these issues, the objectives of this thesis are three-fold: 1) to quantify the actual prevalence of MIC related pipeline failures in Alberta’s oil and gas sector, 2) to perform a gap analysis of failure investigation methods used to assess these pipeline failures, and 3) to develop a novel expert system based on machine learning to assist both experts and non-experts in assessing potential MIC related pipeline failures. The first part of this study highlights a review and analysis of MIC related pipeline incidents in the province of Alberta, Canada over a three-year period (2017-2019). This review was used to quantify the occurrence of MIC failures relative to other corrosion mechanisms, and to conduct a gap analysis of MIC failure investigation techniques being used relative to the current state of the art. Over this three-year period, MIC was found to be responsible for 13.6% and 4.8% of all pipeline leak incidents due to internal and external corrosion, respectively (either as the main failure mechanism or as a contributing factor). Most of these failures were seen to occur in small diameter upstream pipelines (with less than or equal to 220.3 mm outside diameter) carrying mainly multiphase fluids (oil-water emulsions) or produced water. In terms of the failure investigation methods currently being used, it was noted that there was some inconsistency among reports and a number of important gaps were identified. Various assessments lacked microbiological test data, in particular, tests which specifically identify microbial functional groups or speciation, which is critical to confirm observed corrosion mechanisms. Furthermore, a number of these assessments identified MIC primarily on the basis of corrosion morphology, which has been shown to be an incorrect assumption and approach without additional evidence. Details related to sampling methods were also lacking in these assessments, which created some uncertainty as to the quality of data obtained. Overall, most assessments did a reasonable job in characterizing and including chemical (solids, fluids, and corrosion products), metallurgical/ corrosion, and operating data. However, the integration of these various layers of evidence (i.e., connecting corrosion to microbiological activity, and eliminating possible abiotic corrosion mechanisms) was missing in many reports. The second part of this study highlights the modeling of an expert system for the classification of internal microbiologically influenced corrosion (MIC) failures related to pipelines in the upstream oil and gas industry. The model is based on machine learning (artificial neural network) and involves the participation of 15 MIC subject matter experts (SMEs). Each expert evaluated a number of model case studies representative of both MIC and non-MIC related upstream pipeline failures. The model accounts for variations in microbiological testing methods, microbiological sample types, degradation morphology, among others, and also incorporates cases with select missing datasets which is commonly found in actual failure assessments. The output classifications comprised elements of both potential for MIC and confidence in the data available. The results were contrasted for 5- and 3-output classification models (5OC and 3OC, respectively). The 5OC model had an overall accuracy of 62.0% while the simpler 3OC model had a better accuracy of 74.8%. This modelling exercise has demonstrated that knowledge from subject matter experts can be captured in a reasonably effective model to screen for possible MIC failures. It is hoped that this study contributes to a better understanding of the prevalence of MIC in the oil and gas sector, and highlights the key areas necessary to improve the diagnosis of MIC failures in the future.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.