Machine Learning approach to corrosion assessment in subsea pipelines

Integrity of pipelines transporting hydrocarbons over long distances is a growing and challenging problem for Oil&Gas companies, since the age of plants and components is worldwide increasing. Internal corrosion is one of the most dangerous damage mechanisms active in pipelines. Since it is due to interaction of different mechanisms, a large degree of uncertainty is associated with the attempt of quantifying a prediction for the future evolution of damage. Existing models rarely reproduce field data. Given high nonlinearity of the corrosion process, a Machine Learning approach has been investigated, focusing on Artificial Neural Networks (ANN). In particular, an ensemble of ANNs is generated. This strategy strongly improves the results obtained not only by deterministic models, usually considered in literature, but also by single ANN models. Given the high uncertainty inherent to real internal corrosion problem, results from Machine Learning approach are promising.