Non-destructive examination of NPP primary circuit components and reliability of inspection (RACOON)

Tuomas Koskinen, Iikka Virkkunen, Oskari Jessen-Juhler, Jari Rinta-aho, Oskar Siljama

Tutkimustuotos: Artikkeli kirjassa/konferenssijulkaisussaChapterProfessional

Abstrakti

Non-destructive testing is vitally important in ensuring the safe and economic operation of aging nuclear power plants and other industrial systems. Reliable detection of service induced damage as early as possible allows for improved safety and effective maintenance. At the same time, finding flaws early pushes the NDT methods to the limits of detectability. Thus, effective estimation of NDE reliability is vital to successful application of NDE. While much progress has been made with NDE qualification, quantitative estimation of NDE reliability (namely, probability of detection, POD) is still not commonly used in the nuclear industry. The inspections vary, and providing sufficient empirical data for POD has remained infeasible.

In the RACOON project, inspection reliability and POD is studied with new tools
(eFlaw’s) that allow better empirical data gathering. Based on previous work, in the current project a first of a kind virtual round robin (VRR) with world-wide participation was conducted in collaboration with an international project. The target was the ultrasonic inspection of dissimilar metal welds (DMWs). The VRR demonstrated the capability and efficiency of the approach to determine NDT reliability and performance.

In addition to theoretical reliability, the consistency and repeatability of inspections are crucial to actual field reliability. In a short project ANDIE, now merged to the RACOON project, the application of machine learning to automated data analysis and defect detection was studied. Machine learning (ML) powered ultrasonic inspection has proven highly feasible approach to increase reliability, repeatability and efficiency of mechanized ultrasonic inspection. Originally short project ANDIE demonstrated the capability to use virtual flaws for training the ML models. In year 2020 the effect of different flaw types and sizes used in training the ML model was studied to have impact on detectability.
AlkuperäiskieliEnglanti
OtsikkoSAFIR2022 - The Finnish Research Programme on Nuclear Power Plant Safety 2019-2022
KustantajaVTT
Sivut300-312
Sivumäärä13
ISBN (elektroninen)978-951-38-8743-8
TilaJulkaistu - maaliskuuta 2021
OKM-julkaisutyyppiD2 Artikkeli ammatillisessa käyttöoppaassa tai oppaassa tai ammattilaistietojärjestelmässä tai käsikirjamateriaalissa

Julkaisusarja

NimiVTT Technology
Vuosikerta383
ISSN (elektroninen)2242-1211

Sormenjälki Sukella tutkimusaiheisiin 'Non-destructive examination of NPP primary circuit components and reliability of inspection (RACOON)'. Ne muodostavat yhdessä ainutlaatuisen sormenjäljen.

Siteeraa tätä