Stochastic process corrosion growth models for pipeline reliability

[Display omitted] •Novel non-linear stochastic process corrosion growth model is proposed.•Corrosion rate modeled as random Poisson pulses.•Time to corrosion initiation and inherent time-variability properly represented.•Continuous corrosion growth histories obtained.•Model is shown to precisely fit actual corrosion data at two time points. Linear random variable corrosion models are extensively employed in reliability analysis of pipelines. However, linear models grossly neglect well-known characteristics of the corrosion process. Herein, a non-linear model is proposed, where corrosion rate is represented as a Poisson square wave process. The resulting model represents inherent time-variability of corrosion growth, produces continuous growth and leads to mean growth at less-than-one power of time. Different corrosion models are adjusted to the same set of actual corrosion data for two inspections. The proposed non-linear random process corrosion growth model leads to the best fit to the data, while better representing problem physics.