Predicting the failure of plain dent pipeline subjected to cyclic loads using an artificial neural network

Pipelines are widely used to transport fluids over long distances. When a pipeline is subjected to variable internal pressure and external disturbances such as subsidence, structural damage caused by the presence of defects can intensify. Since 2010, the use of artificial intelligence has become a crucial means of solving complex problems, even for interpretation and decision-making using machine learning tools. The main aim of this paper is to use an artificial neural network to predict the failure of a cyclically loaded pipeline with a dent defect. A feedforward neural network has been created to predict the state of a dented pipeline using data from several experiments based on geometric and material properties and cyclic fluid pressure. It enables rapid detection of faults such as plain indentations for maintenance operatives, speeding up intervention time and reducing material and environmental damage. Two evaluation criteria were taken into account for the different models in order to select the optimum architecture in terms of accuracy.