External corrosion in oil and gas pipelines is a very complex failure mechanism phenomena that involves the interaction of many intrinsic and extrinsic factors (soil properties wet and dry cycles coating type cathodic protection effectiveness operational temperature organic decay products etc.). These factors are either difficult to quantify or if quantified then values may have high degrees of uncertainty and a meaning that is not fully understood. A mathematical approach using Bayesian networks was developed to model the external corrosion mechanism in buried pipelines. The complexity of the corrosion mechanism was developed by identifying the complex cause-consequence relationships of multiple variables and by quantifying the degree of the variables with probability distribution functions obtained from mechanistic/physic-based models using Monte Carlo method collected data and/or subject matter expert opinion. An exploratory case study of a 42 km. buried pipeline segment is presented where the Bayesian network model was used as a supplemental methodology to the External Corrosion Direct Assessment (ECDA) method. Limited data was provided by the pipeline operator specifically on the historical survey of the cathodic protection system the repair/excavation practices external corrosion indirect assessment tools operating pressure soil composition etc. The corrosion-related probability of failure was computed by discretizing the pipeline in numerous sections running the modeling framework (the Bayesian network) with the known and unknown information/data at each location from year zero (initial operations) to year 25. Data from an in-line inspection run were used in the model. Despite the limited data and the high uncertainty of the values the Bayesian network model outcome estimated the locations with the relative highest likelihood of external corrosion failure and identified the factors that contribute to it. Results showed that locations having a higher probability of failure may be due to shielding by disbonded coating a condition that cannot be detected by ECDA. Further data collection and mitigation actions to decrease the uncertainty of the probability of failure of the results were recommended.Keywords: external corrosion uncertainty modeling Bayesian networks probabilistic modeling.External corrosion in oil and gas pipelines is a very complex failure mechanism phenomena that involves the interaction of many intrinsic and extrinsic factors (soil properties wet and dry cycles coating type cathodic protection effectiveness operational temperature organic decay products etc.). These factors are either difficult to quantify or if quantified then values may have high degrees of uncertainty and a meaning that is not fully understood. A mathematical approach using Bayesian networks was developed to model the external corrosion mechanism in buried pipelines. The complexity of the corrosion mechanism was developed by identifying the complex cause-consequence relationships of multiple variables and by quantifying the degree of the variables with probability distribution functions obtained from mechanistic/physic-based models using Monte Carlo method collected data and/or subject matter expert opinion. An exploratory case study of a 42 km. buried pipeline segment is presented where the Bayesian network model was used as a supplemental methodology to the External Corrosion Direct Assessment (ECDA) method. Limited data was provided by the pipeline operator specifically on the historical survey of the cathodic protection system the repair/excavation practices external corrosion indirect assessment tools operating pressure soil composition etc. The corrosion-related probability of failure was computed by discretizing the pipeline in numerous sections running the modeling framework (the Bayesian network) with the known and unknown information/data at each location from year zero (initial operations) to year 25. Data from an in-line inspection run were used in the model. Despite the limited data and the high uncertainty of the values the Bayesian network model outcome estimated the locations with the relative highest likelihood of external corrosion failure and identified the factors that contribute to it. Results showed that locations having a higher probability of failure may be due to shielding by disbonded coating a condition that cannot be detected by ECDA. Further data collection and mitigation actions to decrease the uncertainty of the probability of failure of the results were recommended.Keywords: external corrosion uncertainty modeling Bayesian networks probabilistic modeling.
