Session: 03-04-02 Inline Inspection Performance I

Paper Number: 133807

133807 - Technical ILI System – Verification Versus Validation Based on Sound Engineering Practice Solely on Industry Recognized and Standardized Methodology 

Abstract: 

When it comes to qualifying technical ILI systems, the American Petroleum Institute publication API 1163 is still the gold standard, due to its quasi-legal nature in the USA (Title 49 §192.7 and §192.493).

This standard allows the operator to choose between different methods to demonstrate the performance of the ILI systems used or to be used for pipeline inspections – the verification or validation of the ILI system based on tests or simulations, and the system’s validation using historical data, i.e. verified field measurements from excavations.

Verification ensures evidence of the entire performance specification in regards to POD, POI and Sizing capabilities, while validation defends the performance specification on a sample basis respectively derives individual aspects of the as-run performance.

It is obvious that a successful verification of the technical ILI system is a prerequisite for carrying out the validation.

For ROSEN, verification can only be carried out as pull/- or pump tests on the test field under controlled conditions in order to test the essential variables in a Small-scale or full-scale test and thus prove every aspect of the performance specification in regards to POD, POI and Sizing.

As tests under controlled conditions, we define tests:

·         in pipes with controlled and known conditions, like type, roughness, material, shape or grade,

·         with controlled and specified pull/- or pump conditions, like velocity, acceleration, uniformity of the movement, and

·         Artificial or real features with specified dimension (ground truth) and shape and known ground truth uncertainty.

API 1163 provides statistical methods for the practical implementation of measurement performance qualification. However these methods relate entirely to validation of the ILI system, not verification. It therefore remains unclear whether these are suitable or which methods can be used for verification.

This paper closes the gap by presenting industry recognized and standardized, statistical methodologies and adapting them to the problem of verifying technical ILI systems in terms of their detection and sizing capabilities.

For the verification of the detection capability we rely on the Probability Of Detection (POD) concept, just like API 1163 does for the validation, and use binomial proportion confidence intervals – the inverse of binomial hypothesis tests – to determine the POD for features with a specified size and a certain confidence level. But unlike API, ROSEN uses a one-sided Clopper-Pearson interval with a lower boundary instead of a two-sided Agresti-Coull interval. This additional effort ensures to obtain statistically significant results with a guaranteed confidence level for our customers, which is not the case when using Agresti-Coull.

This method using the one-sided Clopper-Pearson interval is also known as the 29/29-method and is frequently used in the aerospace industry since decades, such as NASA and Airbus. It is named after the minimum number of detection attempts (trials) and successful detections to demonstrate a probability of 90% with a confidence level of 95%.

To quantify the sizing capability of the technical ILI system, API uses two different concepts – the Probability Of Sizing (POS) concept for a level 2 validation and Tolerance Intervals for a level 3 validation.

We recommend a different concept, called Expanded Measurement Uncertainty”, based on the GUM principles, which is the “Guide to the Expression of Uncertainty in Measurement” – the standard in science and industry.

Basically, we show how to determine reliably the sizing capability of the technical ILI system for features with the same error distribution through 30 trials – (repeated) measurements of the same or nominally identical features, taking into account other influences and uncertainty shares.

In addition, we discuss why GUM is better suited for verification purposes than POS and Tolerance Intervals.

Concluding, in this paper we present industry recognized and standardized statistical methods adapted for the Qualification of technical ILI systems for all ROSEN services, which are more reliable and statistically significant than the methods described in API 1163 for validation.

Presenting Author: Thomas Stubbe Rosen Technology and Research Center GmbH

Presenting Author Biography: N/A

Authors:

Patrick Hoening Rosen Technology and Research Center GmbH
Thomas Stubbe Rosen Technology and Research Center GmbH