Optimized Sampling Frequencies for Weld Reliability Assessments of Long Pipelines

Abstract

This work responds to a need to obtain statistically adequate quantities of input data for probabilistic assessments of girth welded pipelines. It is desirable to limit the data collection efforts to that warranted by overall reliability and confidence requirements. This is motivated by the challenge of limited and/or high-cost access to in-service pipelines for property measurement and flaw inspection. The goal of this work is to develop a methodology for specifying the minimum, cost-optimized sampling frequency for each of the model input parameters in a multiple girth weld fitness-for-service (FFS) assessment. The sampling frequency specifies the number of samples for each of the model variables to minimize the given cost function. The methodology is based on a user-specified confidence level for the computed reliability of a given pipeline segment (or segments), as well as the estimated per-sample cost of obtaining data for each parameter. The methodology is implemented as an add-on module in the GirthRel computer program. By expressing FFS in the language of risk-based inspection and maintenance, tools such as the one developed here will be invaluable to the development of a pipeline risk management system.