A reliability-based limit states design (LSD) method for assessment of corrosion and crack features has been developed for onshore transmission pipelines as part of a joint industry project. The rule-based LSD approach is a simplified form of the reliability-based approach that reduces the latter to a set of deterministic checks. The LSD corrosion assessment method and a comparison of its performance against one operator’s reliability approach were published in previous IPC papers [1,3]. This paper compares the LSD corrosion and crack assessment methods to another simplified reliability-based approach, namely the Pipeline Integrity Reliability Analysis (PIRA) Level I, that was published in a previous IPC paper . The PIRA model is staged into three levels, where Level I analysis is a simplified form of the reliability-based approach where the probability of failures of predefined features’ sizes are precalculated using conservative assumptions for a fast turnaround screening analysis of the entire pipeline system in order to identify areas requiring more in-depth full probabilistic Level II or III analyses. This paper describes the application of both LSD and Level I methods on two low vapour pressure (LVP) liquid pipelines and provides a comparison of the results.
The comparative analysis was based on the number of features not meeting reliability targets in both methods and, thus, are eligible for repair according to each method. Out of the two pipelines considered, one has corrosion and the other has cracks as the dominant threat. The results show that there are noticeable differences between the outputs of the two methods. The root cause of these differences was investigated by conducting a sensitivity analysis on the input parameters, including: reliability target, wall thickness distribution, feature depth and length distributions, pipeline sectioning procedure and the usage of model error (which is used in the LSD method but not in the PIRA Level I method). For cracks, differences between the Modified Ln-Secant model used in the LSD method and the CorLAS model used in PIRA Level I were also considered.
It was observed that the discrepancies in the required repairs resulting from the two methods can be mostly attributed to the feature depth distribution, wall thickness distribution and the reliability target. For cracks, the burst pressure model selection also had a significant impact on the results.