Abstract
Accurate knowledge of corrosion location, severity, cause and growth rate is critical to pipeline integrity, and in line inspection (ILI) is widely regarded as the most reliable and convenient method of obtaining such knowledge. Much industry effort has therefore centered on improving the metal loss detection and sizing capabilities of ILI tools.
However, when ILI data are lacking or unattainable, operators must seek alternative ways to monitor the integrity of an asset. For managing internal pipeline corrosion, Internal Corrosion Direct Assessment (ICDA) is perhaps the best known alternative. ICDA employs the engineering analyses of corrosion and flow modelling to identify areas at high risk from internal corrosion. The highest priority areas are then excavated and directly examined in order to establish the condition of the pipeline. This combination of corrosion and flow modelling can be used to provide detailed corrosion predictions, but in the absence of ILI data, selection of excavation sites can be problematic. The inherent randomness and uncertainty in the models means that the outputs must often be overly conservative; ICDA can therefore become an extremely inefficient process unless these uncertainties are understood and effectively managed.
The shortcomings of ICDA create a need for a more reliable and accurate corrosion prediction solution which does not require a pipeline to be inspected using ILI. This paper explores the use of Bayesian networks for this purpose. Bayesian networks are graphical models capable of integrating expert knowledge and data into a single system; ‘expert knowledge’ is captured through industry standard corrosion modelling techniques, while ‘data’ is captured through historical ILIs for piggable pipelines. A trained Bayesian network can then be used to make predictions for pipelines without ILI data, based on a knowledge of their operational conditions alone.
With a case study utilizing real pipeline data, it is demonstrated that Bayesian networks can make more intelligent and less conservative predictions of internal corrosion behavior. This in turn can lead to improved pipeline integrity management decisions and more cost-efficient maintenance regimes.
