Development of a Probabilistic Interaction Rule and Failure Pressure Model for Pipelines with a Colony of Corrosion Defects

Abstract

The interaction effect originating from a defect colony can pose a higher pipeline failure risk than the case when such possible interaction is not considered. Recognizing the limitations in existing models that can lead to inaccurate pipeline integrity assessment, the goal of this study is to develop a probabilistic interaction rule and predictive burst failure pressure model for pipelines with a colony of corrosion defects. First, a comprehensive database of burst failure pressure for pipes with a colony of corrosion is established that encompasses a wide range of influencing factors (e.g., spacing, defect sizes, and material properties) on the defect interaction. The performance of existing interaction rules compared in this study using probability of correct defect interaction shows an inconsistency in the existing rules and need for further improvement that considers all influencing factors (including the defect geometries and pipeline properties) holistically. In addition, the results of the performance comparison of existing burst pressure models for pipelines with a colony of corrosion defects have revealed that most existing models are conservative. The proposed interaction rule is then developed by adopting a logistic regression algorithm and holistically considering all possible influencing factors (including pipe properties and colony configurations) as independent variables. In comparison to existing interaction rules, the proposed interaction rule has the most accurate interaction predictions. Next, the failure pressure prediction model is developed by adding a correction factor to the Mixed-Type Interaction (MTI) method, which is revealed as the best method by the performance comparison of existing models conducted in this study; and the correction factor is modeled through the adoption of a multivariate linear regression, using the pipe properties and adjacent defects characteristics as the independent variables. The proposed failure prediction model is shown to be unbiased with less prediction variation compared to the best existing model. Lastly, the case study of an example pipeline with colony defects illustrates the significance of considering interaction effect on the failure prediction of the pipeline.