Off-centering robust calibration of Eddy Current Array Probe used in Power Plants Steam generators

Speaker:
bausson, sebastien; EDF CEIDRE; France

Authors:
Bausson, S.; EDF CEIDRE; France
Mistral, Q.; EDF CEIDRE; France
Le-Moing, Y.; WORTEST; France
Zorni, C.; EDF CEIDRE; France

ID: ECNDT-0370-2018
Download: PDF
Session: Eddy Current-Techniques 3
Room: H2
Date: 2018-06-14
Time: 15:20 - 15:40

The Eddy Current examination of Nuclear Plants steam generator tubes is a very demanding process. Challenges include: high number of tubes (5000 per steam generator), small flaws and complex signal analysis, data to be reviewed quickly with precision and accuracy… The tubes are made of Inconel alloy 600TT, with outside diameter 19.05mm (3/4’’) and width 1.09mm: the flaws looked upon can be cracking, thinning, corrosion buildup… To keep up a good tradeoff between accuracy and speed inspection, EDF started using an inside tube array probe which consists of 32 pancake coils (16 on the circumferential and 2 rows of coils in the axial direction). The outside diameter of the probe is 15.5mm (0.61”).
In EDF standard, each steam generator tube is inspected in line with a calibration tube containing artificial flaws allowing for standardization of probe signals and checking of its functionality. The array probe being an emitter to receiver type of probe, it is sensitive to off-centering, specially with probe ageing. A typical solution to limit off-centering effect on the calibration phase is to use two different calibration flaws: 30% outside groove for amplitude setup and dimensional variation for phase setup.
In this paper, we present signal processing algorithms improving calibration with a probe affected by off-centering effects. We start by exploring the cyclic noise related to off-centering on a flawless part of the calibration tube to derive a robust estimator of the average electrical mean. Then, we compare two adaptive filters allowing simultaneous amplitude and phase calibration while minimizing the cyclic noise contribution. Both filters are based on assumptions on the noise phenomenon: the first one assume it to be sinusoidal while the second takes a broader physical model. Results of filtering are evaluated on the probe signal over other calibration tube flaws.