Analysis of ILI vendor performance on Enable Midstream's pipeline system

摘要

With all the vendors in a common sample, there is no bias to over or undercalling anomalies. This indicates that there is no expectation that our ILI program is biased to one direction or the other as a whole. However, at the more granular vendor and diameter specific level, differences are evident. This could be because of the division of the samples into subgroups will increase the variance due to the reduced effective sample size. Most of the vendors' distribution of error is roughly centered around zero but some of them showed a bias to over or undercalling anomalies. Vendor C has a disproportionate number of overcalls and Vendors E and F undercalls. While overcalls are conservative, they can trigger unnecessary excavations; diverting resources from other efforts. Vendor A's error distribution is roughly centered around zero but there are significant number of negative outliers that are evident in Figure 3. Reviewing the percentage of calls that are within-specification, the data exhibits a noticeably less proportion of in-specification than would be predicted based on advertised tool tolerances. The average within tolerance performance of all vendors combined is about 69% of the time vs. the 80% vendor specification.