Failure assessment of corroded offshore pipelines using code-based approaches and a combination of numerical analysis and artificial neural network

The main purpose of the present paper is to investigate the failure pressure and the remaining life of the corroded offshore pipelines utilizing code-based methods such as ASME B31G, modified ASME B31G, DNV RP-F101 and FFS Level-1. Besides, the results of these approaches are compared with the responses of Finite Element Method (FEM) using Artificial Neural Network (ANN) model. To this purpose, the probability distribution of the corrosion defect length and depth as well as the corrosion rate values are obtained from an In-Line Inspection (ILI) survey of an offshore pipeline project in the Persian Gulf. In addition, a 3D verified numerical model of a corroded pipeline is constructed to conduct the numerical analyses. Using the numerical analysis data, a MultiLayer Perceptron (MLP) model is created to estimate the failure pressure of the pipeline with arbitrary defect dimensions. Moreover, the remaining life of the corroded pipeline is quantified in order to prevent the catastrophic consequences of the pipeline failure. The results show that the MLP model is capable of estimating the failure pressure of the corroded offshore pipeline with quite acceptable accuracy compared with the mentioned code-based methods. Furthermore, the calculated failure pressure of the corroded pipeline can be sorted in ascending order as: (1) FFS-Level 1 (2) ASME modified B31G (3 & 4) ASME B31G & DNV RP-F101 (5) FEM + MLP. In fact, the failure pressure and the remaining life calculated by DNV RP-F101 are closer to the FEM + MLP results for the defects with small and medium dimensions. Diversely, the DNV RP-F101 method underestimates the failure pressure of the corroded pipeline when the defect dimensions become larger.