With the increased use of automated sensor arrays, continuous online monitoring, and more-accessible data interfaces, lack of data is becoming a problem of the past. However, a new challenge is emerging: how to effectively utilize the large quantities of data, especially when complex systems or processes are involved. Internal corrosion in a natural gas storage pool is one such case, with internal corrosion susceptibility depending on interactions among a variety of fluid composition, flow, and material parameters. Bayesian network (BN) modeling was explored as a solution to this challenge due to its ability to analyze complex cause-effect relationships in large data sets while considering the variability and uncertainty in the data.
The BN modeling was employed within the context of a wet gas internal corrosion direct assessment (WG-ICDA) indirect inspection step, and served as the primary tool for assessing a natural gas storage system. A BN was developed to characterize internal corrosion in the piping system, incorporating aspects such as seasonal changes in operation, multiphase flow modeling, general and localized corrosion, and mitigation measures. Current and historical input data were gathered from sources including pipeline geographic information system (GIS) data, direct and indirect corrosion monitoring results, historical flow testing, and automated sensor arrays. Corrosion "defects" were grown over the life of each pipeline as probability distributions, retaining data variability. Several locations were selected for ICDA detailed examination based on the BN modeling results, effectively narrowing the storage field piping system to a few representative sites.
Despite several implementation challenges, BN modeling appears to be a promising approach for corrosion assessment in complex oil and gas pipeline systems due to its ability to combine a lifetime's worth of data to produce specific and justifiable results.
Key words: Internal Corrosion, WG-ICDA, Bayesian Networks, Corrosion Modeling, Natural Gas Storage Systems, Big Data