Estimation of corrosion damages by Bayesian stochastic models

Abstract

This work outlines a probabilistic model for predicting temporal evolution of the internal corrosion damage depth in active sites of pipelines. The evolution of damage is specified by a stochastic process with stationary independent increments and is characterised by the state of damage at a given time instant from the propagation function that describes damage velocity. The propagation function is related with a Bayesian model that can incorporate all the available information such as inspections and velocities obtained from coupons or laboratory measurements. The quality of inspection is expressed in terms of detection capacity of the inspection tool and the errors in measurements. The model robustness lies in the fact that corrosion damage is specified in terms of a physical measure and is usable under practical conditions. Corrosion damage velocities are based on measurements, which implicitly consider multiple factors such as aggressive chemical environment, microstructure, operating conditions intervening in the corrosion process, as well as their variability. The application of the model is exemplified using inspection reports of two industrially operated pipeline systems that transport oil products.