A failure prediction model for corrosion in gas transmission pipelines

Abstract

Transmission pipelines comprise a major part of a gas network, conveying natural gas within jurisdictions, and across international boundaries. In the United States, more than 10,000 failure incidents have been reported in gas transmission pipelines in a 20-year period from 1996 to 2016 leading to a cumulative property damage of more than $748 million. Among different failure sources, corrosion is ranked as the most frequent one, corresponding to approximately a quarter of total failures. Though in-line inspection is counted as the most frequently applied corrosion monitoring technique for oil and gas pipelines, it imposes considerable costs due to the necessity of implementing frequent inspections using smart devices. For this reason, several failure prediction models have been developed to estimate the corrosion failure. However, the majorities of these prediction models rely solely on experimental tests or limited historical records which undermine the extent of their applicability and ignore pipeline environmental and geographical circumstances. The objective of this research is to develop failure prediction models for external corrosion in underground gas transmission pipelines by considering both conventional and environmental/geographical variables. For this objective, multiple regression analysis was performed on the accessible historical data reported for gas transmission pipelines. Two main climate regions of Great Plains and South East in the US were selected, and their corresponding failure prediction models were developed. Such development was based on a step by step procedure analyzing different scenarios. Considering diagnostic measures, null hypothesis and residual analysis, scenario 3 was selected as satisfactory. The validation tests of the developed models present a root mean square error (RMSE) of 0.04 and 0.07 and R-Sq of 0.93 and 0.75, respectively. The results of this research can be applied in maintenance planning of gas transmission pipeline to estimate the critical time in which a pipeline may encounter external corrosion failure, and to accordingly schedule the maintenance activities.

Asset management analytics for urban water mains: a literature review

“…There is a trend of increasing use of machine learning algorithms to predict the failure rate. Zakikhani et al (2021) provided a review of failure prediction models, including machine learning models for oil and gas pipelines. Malek Mohammadi et al ( 2021) used K-Nearest Neighbor (KNN) to predict the condition of sewer pipes.…”

Recent Advances in Nondestructive Method and Assessment of Corrosion Undercoating in Carbon–Steel Pipelines

“…Based on the corrosion and failure data collected from the experiments, statistical models have been developed to assess and predict the future state of the pipeline structure. Several statistical models exist in the literature, such as linear regression, nonlinear regression, Bayesian inference, and Markov chain models for carbon-steel pipeline corrosion prediction [127][128][129][130][131][132]. Expert opinions have been used to calibrate the distribution of the related parameters.…”

Naïve Bayes Classifier for Predicting the Novel Coronavirus

“…x Naïve Bayes seems a simple classification algorithm but in case of predictive modeling [7], it performs an impressive role. It is covered under the family of probabilistic classifiers which is entirely based on Bayer's theorem.…”