Data-driven model to predict burst pressure in the presence of interacting corrosion pits

Abstract

This paper presents a data-driven approach to predict the pipelines’ corrosion-induced Burst failure. In this approach, different aspects of pit growth progression and spatial distribution of pits are simulated. The proposed approach takes advantage of population characteristics to model these aspects of the degradation paths for each pipe section down to the size of single joints. The insights obtained from simulations are used to project the degradation of each pipe section. Understanding corrosion behavior and field data are used to model the corrosion-related parameters such as corrosion pit dimensions, probability and time of initiation, and location. The failure is modeled using the probabilistic simulation considering degradation rate, interactions among pits, and material properties as stochastic variables. The proposed approach and included models are tested using multiple real-life inline inspection datasets. Validation of predicted properties shows prediction errors ranging from 3%–10% depending on the three remaining strength calculation approaches. This work aimed to serve as an important tool for risk-based maintenance prioritization, inspection interval assessment, and the fitness of service assessment of pipelines.

Keywords

• Pitting corrosion
• Failure analysis
• Extreme value analysis
• Monte Carlo simulation
• Inline inspection
• Pipeline failure analysis