What is it about?
Oil and gas pipelines are affected by a complex set of factors, including soil properties, pipeline properties, and pipeline operations. These factors influence each other in a variety of ways. For example, pipelines are protected using different types of coatings and by the application of a negative voltage (called cathodic protection). The ability of cathodic protection is affected by different coatings and soils. Cathodic protection voltage also affects whether a coating peels off from the pipeline (called disbonding). Additionally, there is a lot of uncertainty in the data related to these factors. Not all the necessary data for decision making are collected, some are lost, and there is scatter in the data. A lot of knowledge regarding corrosion and other failure mechanisms has been acquired over the years. This paper discusses an approach to combine the uncertain data with collective knowledge of failure mechanisms to make predictions regarding pipeline failures. It uses a methodology called Bayesian network that links the factors leading to ultimate pipeline failure in a logical fashion. The Bayesian network provides a way to handle uncertain data through probabilities and links knowledge through conditional probabilities (i.e. given the probability of a causative factor, what is the probability of a consequence?). The paper provides case studies from several pipeline failure mechanisms.
Featured Image
Why is it important?
Until now, pipeline risk has been assessed in a qualitative or in a semi-quantitative manner by assigning subjective ranking or failure probabilities based on past inspection data. both these approaches have serious limitations. Subjective ranking (also called risk indexing) do not consider interactions among factors andare not transparent. Assigning probabilities based on previous inspections assume that the future is the same as the past and cannot be justified. Many major pipeline failures were not anticipated because inspection data did not indicate any problems. Neither of these methods can be used to quantitatively identify the effectiveness of corrective actions. Bayesian methods are practical, can be updated by inspection and other information, but also consider what we know about failure mechanisms. They are transparent and can be used to perform cost/benefit analyses.
Perspectives
Bayesian network models can be powerful way to use the collective knowledge of a system and also provide a framework for continually incorporating our improved knowledge. By linking sensors to Bayesian network models, we can do real time risk assessments. However, Bayesian networks are a hard sell to engineers, who are more used to thinking in terms of flow charts (i.e. a sequential approach).
narasi sridhar
DNV GL AS
Read the Original
This page is a summary of: Quantitive Assessment of Corrosion Probability—A Bayesian Network Approach, CORROSION, November 2014, NACE International,
DOI: 10.5006/1226.
You can read the full text:
Contributors
The following have contributed to this page
