One of the leading threats to the integrity of oil and gas transmission pipeline systems is metal-loss corrosion. This threat is commonly managed by evaluating measurements obtained with in-line inspection tools, which locate and size individual metal-loss defects in order to plan maintenance and repair activities. Both deterministic and probabilistic methods are used in the pipeline industry to evaluate the severity of these defects. Probabilistic evaluations typically utilize structural reliability, which is an approach to designing and assessing structures that focuses on the calculation and prediction of the probability that a structure may fail. In the structural reliability approach, the probability of failure is obtained from a multidimensional integral. The solution to this integral is typically estimated numerically using Direct Monte Carlo (DMC) simulation as DMC is relatively simple and robust. The downside is that DMC requires a significant amount of computational effort to estimate small probabilities.
The objective of this paper is to explore the use of a more efficient approach, called Subset Simulation (SS), to estimate the probability of burst failure for a pipeline metal-loss defect. We present comparisons between the probability of failure estimates generated for a sample defect by Direct Monte Carlo simulation and Subset Simulation for differing numbers of simulations. These cases illustrate the decreased computational effort required by Subset Simulation to produce stable probability of failure estimates, particularly for small probabilities. For defects with a burst probability in the range of 10−4 to 10−7, SS is shown to reduce the computational effort (time or cost) by 10 to 1,000 times. By significantly reducing the computational effort required to obtain stable estimates of small failure probabilities, this methodology reduces one of the major barriers to the use of reliability methods for system-wide pipeline reliability assessment.