Pipeline integrity assessment using mass inspection data

腐蚀对几乎所有管道来说都是一种时间依赖性危害。由于腐蚀引起的管道失效会对社会、经济和环境造成严重后果。管道的在线检测（ILI）用于确定管道中腐蚀缺陷的尺寸，随后对测量的缺陷进行极限状态评估，以识别安全关键缺陷。ILI通常会产生大量数据，特别是对于遭受高密度内部腐蚀的管道。基于风险和可靠性的管道完整性管理在过去的几十年中获得了越来越多的相关性。它依赖于概率模型来推断实际的腐蚀增长的ILI结果，是受检查的不确定性。当前的概率腐蚀增长模型要么只关注可用检查数据的子集，要么没有被设计成有效地处理大量检查数据。因此，管道完整性评估和决策可能会由于所需的数据截断而产生偏差，从而导致次优结论。本研究计划的目标是开发基于大量ILI数据的腐蚀能源管道完整性评估的有效概率模型。提出了三个研究任务，将导致两个腐蚀生长模型，一个不匹配的缺陷和匹配的缺陷。 在第一个任务中，使用基于种群的方法开发了不匹配缺陷的概率腐蚀生长模型。将来自ILI的所有测量缺陷尺寸视为一个总体，通过调整尺寸、可检测性和错误判定不确定性来确定每个ILI的实际缺陷尺寸总体。然后推断当前和未来的缺陷尺寸，用于整个管道段的完整性评估。 第二个任务的重点是发展一个概率匹配算法的腐蚀缺陷，从大量的ILI数据。由于测量误差、高密度腐蚀和新缺陷的生长，来自两个或更多ILI的报告缺陷位置很少完全匹配。现有的方法，如迭代最近点法是从计算机视觉领域。它们针对增加的错误调用数量以及先前检查中的几个缺陷在一个缺陷中一起增长的可能性进行调整。一个概率模型进行调查，以描述匹配不确定性的可能的缺陷组合的排名。 在第三个任务中，开发了匹配缺陷的分级腐蚀生长模型。除了现有的能力，尺寸误差调整和随机过程来模拟实际的缺陷增长，分层模型包括开发的匹配不确定性。将其推广到分析轴向、周向和径向缺陷的生长，并研究了未知随机变量的有效估计方法。

Corrosion is a time-dependent hazard for virtually all pipelines. Pipeline failures due to corrosion can lead to severe consequences for society, the economy and environment. In-line inspections (ILIs) of pipelines are used to size corrosion defects in pipelines and subsequent limit state assessment on the measured defects is performed to identify safety-critical defects. ILIs often lead to mass data, particularly for pipelines that are subject to high density internal corrosion. Risk- and reliability-based pipeline integrity management has gained increasing relevance over the last decades. It relies on probabilistic models to infer the actual corrosion growth from the ILI results that are subject to inspection uncertainties. Current probabilistic corrosion growth models either focus only on a subset of the available inspection data or are not designed to efficiently process mass inspection data. As a consequence, pipeline integrity assessment and decision making can become biased due to required data truncation leading to sub-optimal conclusions. The objective of this research program is to develop efficient probabilistic models for the integrity assessment of corroded energy pipelines based on mass ILI data. Three research tasks are proposed that will result in two corrosion growth models, one for unmatched defects and one for matched defects. A probabilistic corrosion growth model for unmatched defects is developed in the first task using a population-based approach. Considering all measured defect sizes from an ILI as one population, the population of actual defect sizes is determined per ILI by making adjustments for sizing, detectability and false call uncertainties. The current and future defect sizes are then inferred for the integrity assessment of entire pipeline segments. The second task focuses on the development of a probabilistic matching algorithms for corrosion defects from mass ILI data. Reported defect locations from two or more ILIs rarely match exactly due to measurement errors, high-density corrosion and growth of new defects. Existing methods such as iterated-closest-point method are adopted from the field of computer visions. They are adjusted for the increased number of false calls and the possibility that several defects from previous inspections grew together in one defect. A probabilistic model is investigated to rank the possible defect combinations for describing the matching uncertainties. A hierarchical corrosion growth model for matched defects is developed in the third task. In addition to the existing capabilities of sizing error adjustment and a stochastic process to model the actual defect growth, the hierarchical model includes the developed matching uncertainties. It is extended to analyze the defect growth in axial, circumferential, and radial direction and an efficient approach to estimate the unknown random variables is investigated.