Hit/Miss POD With Model Assisted and Emulated Flaws

Research output: Chapter in Book/Report/Conference proceedingConference contributionScientific

Researchers

Research units

  • VTT Technical Research Centre of Finland

Abstract

The determination of the reliability of an inspection is of high significance. In particular, it is important to determine what is the largest crack, which could conceivably be missed during the in-service inspection. This information is utilized in order to choose the most effective method for different situations. Probability of detection (POD) curves are used to quantify the inspection effectiveness. However, major obstruction for POD curves is the requirement for a lot of data points in order to give reliable estimates of the lower limit performance, rendering reliable POD curves highly expensive to produce.
In this study, POD curve is estimated using only few thermal fatigue cracks, which is insufficient to produce a POD curve alone. In the present study the idea is to emulate the amplitude response from the measured crack in a way that represents an amplitude response from a certain crack size and in addition CIVA simulation is used to produce amplitude response data from similar simulated cracks. Both amplitude responses are then in turn converted to a B-scan image for inspectors to evaluate whether there is a crack or not. Then a POD curve is generated from the achieved hit/miss data. The idea is to decrease the amount of needed real flaws and also to determine if an inspector can tell the difference between real, emulated and a simulated flaw.

Details

Original languageEnglish
Title of host publication12th European Conference on Non-Destructive Testing (ECNDT 2018)
Publication statusPublished - 1 Aug 2018
MoE publication typeB3 Non-refereed article in conference proceedings
EventEuropean Conference on Non-Destructive Testing - Gothenburg, Sweden
Duration: 11 Jun 201815 Jun 2018
Conference number: 12

Publication series

NameThe e-Journal of Nondestructive Testing
PublisherNDT.net
Number8
Volume23
ISSN (Electronic)1435-4934

Conference

ConferenceEuropean Conference on Non-Destructive Testing
Abbreviated titleECNDT 2018
CountrySweden
CityGothenburg
Period11/06/201815/06/2018

Download statistics

No data available

ID: 27183633