Application of multiresolution analysis and deep learning to obtain failure pressure of corroded pipelines

Abstract

The assessment of corroded pipelines is considered a very important task in the oil and gas industry. The present work aims to develop an efficient system to accurately predict the burst pressure of corroded pipelines with complex corrosion profiles through hybrid models using multiresolution analysis, numerical analysis, and metamodels. The corrosion profile is obtained from ultrasonic inspections and the data is provided as a river bottom profile. The real corrosion shapes are parametrized considering a discrete wavelet transform to reduce the amount of data that describes the defect. The coefficients obtained from the wavelet transform are used as inputs to feed a deep neural network system for quickly and accurately predict the burst pipeline pressure. Eight different steel materials are considered in the NN build process. Synthetic models that have similar statistics to real corrosion profiles are created and submitted to non-linear FEM analysis, for the different materials. The failure pressures obtained from the synthetic defects are used to train a neural network to predict the burst pressure of the pipelines. The results obtained with the deep neural networks are very accurate for all cases presented in this work.