Bayesian Modeling of External Corrosion in Underground Pipelines Based on the Integration of Markov Chain Monte Carlo Techniques and Clustered Inspection Data

In this study, a model is developed to assess external corrosion in buried pipelines based on the unification of Bayesian inferential structure derived from Markov chain Monte Carlo techniques using clustered inspection data. This proposed stochastic model combines clustering algorithms that can ascertain the similarity of corrosion defects and Monte Carlo simulation that can give an accurate probability density function estimation of the corrosion rate. The metal loss rate is chosen as the indicator of corrosion damage propagation, obeying a generalized extreme value (GEV) distribution. Bayesian theory was employed to update the probability distribution of metal loss rate as well as the GEV parameters in order to account for the model uncertainty. The proposed model was validated with direct and indirect inspection data extracted from a 110-km buried pipeline