The reliability of Non-destructive testing (NDT) is an on-going challenge. The consequences of failed inspections can be dire, and thus the requirements for NDT reliability are very high. The work is technically demanding and requires skilled use of the available equipment and keen judgement to properly discern flaw signals from noise. Somewhat paradoxically, the work is also very tedious and repetitive. Most of the inspected targets do not contain any flaws but the inspectors need to be constantly alert for the possibility. The recent studies in human factors have brought advances in (among other things) improved readability of inspection procedures and procedures of reviews and redundant inspections widely used in order to improve overall inspection reliability. In present paper, the effect of feedback and variation on inspector performance is studied. To test this, a small empirical study was completed. An online tool was created with simplified UT set-up: B-scan image of the data and software gain control and tools to indicate cracks with point and click. The system generates random flawed data images on the fly. The user than analyses the images indicates found flaws by clicking them. After 150 images have been analysed (many of them without flaws), the system uses the provided hits and misses to compute a POD curve and confidence bounds using standard (ASTM E2862) techniques. Additional "learning" version of the tool was created. In this "learning mode", after user requests next image, the system shows results of the current images (i.e. hits, misses and false calls in the current image). This set-up thus provides the inspector with direct feedback of his success and should thus better facilitate learning this particular inspection task. The tool was presented to small group of inspector in level-III inspector training and results gathered from trainees both before and after training (and with and without feedback). The results from this small group of inspectors indicated, that direct feedback on achieved reliability can quickly improve POD values. However, the study group was small and thus the results need further investigation and confirmation.
[1]
S. R. Doctor,et al.
Summary of the PISC round robin results on wrought and cast austenitic steel weldments, part II: wrought-to-cast capability study
,
1996
.
[2]
Ference Marton,et al.
Necessary Conditions of Learning
,
2014
.
[3]
Marija Bertovic,et al.
Human factors in non-destructive testing (NDT): risks and challenges of mechanised NDT
,
2015
.
[4]
Daniel Kanzler,et al.
Paradigm Shift in the Holistic Evaluation of the Reliability of NDE Systems
,
2013
.
[5]
Seldis Thomas,et al.
European Methodology for Qualification of Non-destrutive Testing (Third Issue)
,
2007
.
[6]
Steven R. Doctor,et al.
Evaluation of Manual Ultrasonic Examinations Applied to Detect Flaws in Primary System Dissimilar Metal Welds at North Anna Power Station
,
2012
.
[7]
J Swanson,et al.
Business Dynamics—Systems Thinking and Modeling for a Complex World
,
2002,
J. Oper. Res. Soc..
[8]
C. D. Cowfer.
Basis/background for ASME code section XI proposed Appendix VIII : ultrasonic examination performance demonstration
,
1991
.
[9]
Tuomas Koskinen,et al.
Producing a POD curve with emulated signal response data
,
2018
.