Advanced Reliability Modeling And Analysis Of Pipeline Networks Considering Degradation Behavior

Abstract

The pipelines are critical infrastructure in the network oil and gas distribution system, which are designed to transport massive amounts of oil and gases from the source to the enduser. Pipeline corrosion is a major issue facing pipeline operators because most of the pipelines are made of copper, steel, or cast iron which are subject to corrosion. It grows over time and threatens the safety and reliability of pipelines. Therefore, an accurate prediction of corrosion growth is crucial to prevent leakage or rupture in pipelines that might lead to severe consequences such as human and environmental disasters and economic losses. Due to the highly stochastic and uncertainty associated with the corrosion growth nature, intensive research efforts propose predictive models to capture corrosion behavior in the pipeline by estimating corrosion depth over time, which can be used to assess the reliability and remaining useful life (RUL) of the corroded pipe. Therefore, the purpose of this dissertation contributes to the field of corrosion prediction by developing and evaluating novel models for predicting corrosion growth in oil and gas pipelines. First, a time-dependent corrosion rate model called multiple in-line inspections (M-ILIs) based-linear corrosion rate model was developed to predict the corrosion growth process in pipelines, particularly when dealing with limited corrosion data. Second, the isotonic regression (IR) model was proposed to predict corrosion depth in oil refinery piping. Third, the multi-sensor corrosion growth model with latent variables was proposed by combining the agglomerative hierarchical clustering (AHC) algorithm and the vector autoregression (VAR) model. Fourth, the AHC algorithm and multivariate gaussian process regression (MGPR) model with latent variables was developed for predicting multivariate corrosion depth sensor data. Monte Carlo simulation (MCS) and limit state function (LSF) are also incorporated with the corrosion growth models to estimate the reliability of pipe in oil refinery. Finally, this dissertation contributes to the field of pipeline corrosion prediction by offering accurate, stable, and reliable models for corrosion growth prediction in oil refinery piping systems. These models demonstrated promising corrosion growth prediction results, making them useful for effective maintenance planning and corrosion management in oil refineries.