A probabilistic integrity assessment quantifies the chance of pipeline failure. Statistical distributions are used to account for uncertainties in pipeline integrity engineering models, such as the rate of degradation and resistance to failure. In a corrosion assessment, these uncertainties include those associated with feature sizing and detection as well as growth. In-line inspection (ILI) is regularly used for the detection and sizing of corrosion features, and the accuracy and reliability of tool performance is included in any assessment.

The worldwide focus on data analytics and digital transformation has led to the development of more robust and efficient software that can be used to assess large data sets, such as ILI results. A probability of failure for each individual anomaly reported can be calculated in accordance with pipeline standards, such as Canadian Standard Z662, that have recently been updated to provide guidance on probabilistic assessment. However, assumptions are often needed when selecting the distribution parameters to best represent inputs, such as material properties and corrosion feature sizing, which have historically been derived from minimum performance specifications rather than representative empirical data.

This paper describes our approach for performing probabilistic assessments for large and varied data sets. A case study is presented to illustrate a practical application for corrosion to inform performance-led integrity decision-making, such as when selecting the most appropriate inspection technology and frequency. The results are compared with probabilistic assessment results generated for pipelines with similar characteristics, selected from our database of over 10,000 pipelines, to improve the understanding of the probability of failure calculated (how does it compare with expectations) and explore where improvements in the input distribution parameters may be appropriate.