A RISK-BASED APPROACH TO MAINTENANCE PLANNING UTILIZING IN-LINE INSPECTION DATA

摘要

A common approach to the management of external corrosion in the pipeline industry is to perform an In-Line Inspection, followed by repairs of defects that fail a deterministic criterion, and then leave the line in service until a prescribed time interval has elapsed, at which point another re-inspection is performed. However, many companies have found that as a result of the uncertainty associated with MFL defect sizing and corrosion growth rates, a deterministic repair and re-inspection process may often result in unnecessary maintenance expenditures while occasionally failing to identify and address critical features. When the rare feature 'slips through' the deterministic process, companies often respond by adding conservatism to the process, leading to increased spending with little additional benefit. A better approach for evaluating corrosion defects is to view the process as an analysis of a set of stochastic variables instead of deterministic values. Through such an approach, the sensitivity of a defect's failure probability can be more effectively evaluated, facilitating a decision process that is better able to find the 'exceptions' that are not addressed by a deterministic process. This paper outlines an approach to analyzing MFL data with stochastic variables using computer simulation, along with a process for continuously improving the characterization of each variable through a feedback loop. Alternative methods to Monte Carlo, such as Importance Sampling are briefly outlined to minimize the analysis time required without sacrificing simulation accuracy. Finally, acceptance criteria are required to interpret the calculated failure probability in order to inform maintenance decision making. This is presented in a risk-based context using a previously published risk management framework. Through this process, defect repair decisions and the evaluation of the benefit of MFL re-inspection can be better optimized. Examples are drawn from actual maintenance programs to illustrate this approach.