Corroded pipeline assessment using neural networks, the Finite Element Method and discrete wavelet transforms

An essential task in the oil and gas industry is establishing an efficient way to assess corroded pipeline integrity. The literature shows that integrity analysis with Finite Elements simulations is the most effective. However, when faced with solving practical problems, the inconvenience of the high computational cost arises. This work aims to develop an efficient system to accurately predict the burst pressure of corroded pipelines with complex corrosion profiles through hybrid models combining multiresolution analysis, numerical simulations, and metamodels. The corroded region will be captured from ultrasonic inspections. Subsequently, the representation of corroded zones is parameterized with a discrete wavelet transform to reduce the amount of data representing the defect. The metamodel is built by training a neural network with the coefficients obtained from the wavelet transform and the pipeline material properties. The training data for the neural network are the failure pressures computed with non-linear finite element analysis of three-dimensional synthetic models with similar statistics to real corrosion profiles. The results obtained with the neural networks are accurate for all the test cases presented in this work.