Abstract
Inspection data obtained from in-line inspection can be used to assess present integrity as well as to predict future integrity of corroding pipeline by using a combination of probabilistic analyses and simulation process. However, numerical errors arise when all the inspection data including non-extreme and extreme measured defects were used. In this study, a combination of an extreme value distribution and peaks-over-threshold method was adopted to analyse the effect of extreme data upon pipeline reliability. The paper focuses on the elimination of the low-risk data in the analysis in order for extreme values to be significantly emphasized. The study found that the high threshold value would lead to the high failure probability. The optimum threshold value is constrained by the number of remaining data. The selection of threshold and extreme distribution value must be cautiously carried out to minimise the possibilities of over predict the pipeline's time to failure.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
