Simulation-based POD study for welded pipe inspection

Ribay, Guillemette; CEA; France

Ribay, G.; CEA, LIST; France
Chapuis, B.; CEA, LIST; France
Mesnil, O.; CEA, LIST; France
Miorelli, R.; CEA, LIST; France
Rouhan, A.; Bureau Veritas; France
Bala, R.; Bureau Veritas Exploitation; France
Pinier, L.; TechnipFMC; France
Pomie, L.; TechnipFMC; United Kingdom

ID: ECNDT-0513-2018
Download: PDF
Session: Oil & Gas 5
Room: H2
Date: 2018-06-12
Time: 13:30 - 13:50

Flowlines used in the Oil and Gas industry are composed of several pipelines welded all together. The girth weld shall withstand pressure and temperature conditions during the life cycle of the flowline and thus has to be inspected for unacceptable defects during production and before installation. One reference nondestructive testing method of such girth welds is the Zonal Discrimination Method, which uses ultrasound phased array transducers. For a given weld geometry and transducer, the performance of the inspection has to be demonstrated during qualification using statistical quantification tools such as Probability of Detection (POD) and sizing accuracy curves.

So far, the qualification is based on a substantial number of experiments on representative welded mock-ups, some with machined calibration defects, and others with realistic defects induced intentionally by artificial means. Such realistic defects are very difficult to create and have to be destructively characterized to measure their real height and location. Thus, the whole procedure is costly and the number of defects is limited.

In this communication, we present the results of a study performed in collaboration between CEA, Bureau Veritas and TechnipFMC. The aim is to use numerical simulations to support and complete experiments and to provide robust qualification procedures. State-of-the-art numerical tools based on CIVA 2017 are used to simulate phased array inspection on various defects located in the weld and to rank the criticality of the different variables of AUT set up and procedures and compute model-assisted POD curves.