Statistical Modeling of Pitting Corrosion and Pipeline Reliability

Abstract

Pitting corrosion is treated as a time-dependent stochastic damage process characterized by an exponential or logarithmic pit growth. Data from water injection pipeline systems and from the published literature are used to simulate the sample functions of pit growth on metal surfaces. Perforation occurs when the deepest pit extends through the thickness of the pipe. Because the growth of the deepest pit is of stochastic nature, the time-to-perforation is modeled as a random variable that can be characterized by a suitable reliability model. It is further demonstrated that this time-to-first-leak is often distributed according to the Weibull extreme value reliability model.