Corrosion defect modelling of aged pipelines with a feed-forward multi-layer neural network for leak and burst failure estimation

Determination of the future Corrosion Defect Depth (CDD) growth of the oil and gas pipelines is vital for the management of the integrity and mitigation of failures that can affect health, safety, and the environment. To this end, this work uses the historical operating parameters for establishing the time-dependent CDD growth of corroded pipelines based on machine learning. This data-driven machine learning relies on feed-forward Subspace Clustered Neural Network (SSCN) and Particle Swarm Optimization (PSO) to estimate the CDDs of a single-SSCN by treating the first Subspace Cluster (SSC) as a regression model that comprises of the hidden and bias layers and the input variables. The multi-SSCN model is linked to the single-SSCN model through individual values decoupling, transformations and modifications of the hyperspace of the deeper layers in the SSCN model. The CDDs estimated with the SSCN models are used for a Weibull distribution dependent leak and burst failure probability estimation to compute the integrity of the pipelines at discrete sections. The results obtained demonstrate the potentials of this technique for the integrity management of corroded aged pipelines.