Abstract
The requirement to validate the performance of in-line inspection (ILI) is outlined in the API 1163 standard. Appendix E of that standard is primarily focused on using statistical hypothesis tests on field validation measurements.
An initial focus on hypothesis testing masks the more important pipeline integrity question at hand: Based upon the field validation measurements obtained, what is a defensible estimate of the uncertainty in this set of ILI measurements? The general theory of statistical tolerance intervals provides a framework for developing these uncertainty estimates and for assigning confidence based upon the number of validations measurements obtained.
This paper provides a brief introduction to statistical tolerance intervals and provides some guidance on their application to ILI performance assessment. The basic concepts of interval length, proportion within the interval, and confidence are discussed and related to terminology used in API 1163. Methods which make no assumptions about the underlying distribution are presented as well as methods based upon normal distribution assumptions. Two-sided intervals (“The ILI error is ±10% … ”) as well as one-sided lower bounds (“The ILI undercall is no larger than 15% … ”) are considered.
