Non-Destructive Testing '92

C. Hallai, P. Kulcsar

• Publisher: Elsevier Science
• Year: 2012
• Print ISBN: 9780444897916
• eBook ISBN: 9780444599780
• Topic: Technology & Engineering
• Subtopic: Industrial Health & Safety
• Index: Technology & Engineering

Approx.710 pages

ELETRIC TECHNIQUES

Processing of cartographies from steam generator tubes using eddy current testing with an absolute coil

P. Attaoui^a, B. Benoist^a, R. Besnard^a, T. Sollieir^a, P. Gaillard^a and R. Lengellé^a, ^aCEA CE Saclay DTA/DTM/STA/LCME 91191 GIF/YVETTE CEDEX - FRANCE; ^bUTC URA 817 BP 649 60206 COMPIEGNE - FRANCE

ABSTRACT

This paper deals with the processing of electromagnetic cartographies from steam generator tubes testing. These images are disturbed by background noise due to probe lift-off changes and by the rolling transition zone signal. Procedures which allow us to obtain a “flat” cartography will be presented in part one. Then, using mathematical morphology tools on the cartography, we will expose the first results dealing with sizing and orientation of the defects.

1 INTRODUCTION

As nuclear plants age, the inspection of steam generator tubes poses new problems. New defects [1] appear in the area of the tube sheet, the support plate and at the U-bend. The testing of the tube is undertaken by using an absolute coil which improves the detection and the signal resolution. The probe has a helicoidal movement to scan the inside of the tube. The data are represented for each of the signal complex coordinate as a cartography: this allows us to have a developped view of the tube [2].

These images are most often disturbed by background noise due to probe lift-off changes when there is a tube ovalization, a probe offset or a pilgrim pass process deformation. Dimensionnal variations, like rolling transition zones, also generate signals which disturb the defect analysis. This study implies one first phase of processing which consists of improving the signal to noise ratio of the image, by subtracting the low frequency components characteristic of background noise. (A particular procedure also allows us to get rid of the rolling transition signal). In the first part, we present this preprocessing and in a second part we approach the first results of the defect analysis (mathematical morphology-wise).

2 CARTOGRAPHY DECOMPOSITION

The decomposition principle is as follows: if an image is represented by a vector V and if the basis vectors v_i is chosen, one can always obtain a decomposition of V into a weighted sum of each v_i vector. One only has to compute the inner product between V and each v_i to know its contribution. If one reasons upon a signal, one only has to choose an appropriate base functions for the decomposition. We have taken the sinusoidal functions (1, sin(x), cos(x), sin(2x), …) from which we can construct a base of bidimensionnal functions defined by all the possible products of unidimensionnal functions (for each of the longidudinal and circumferential direction).

In order to decompose the image, the window around the defect has to be removed so that the domain is no longer continous. A new corresponding base has to be constructed with functions which are no longer orthogonal but still normalized. The rebuilding of the noisy surface is made from functions which are orthogonal but no more normalized.

We can describe it as follows:

So we get the several bases of bidimensionnal functions:

The base B1 so that	b1i, j (x, y) = fi(x) · fj(y)
The base B2 so that	b2i, j (x, y) = gi(x) gj(y)
The base B3 so that	b3i, j (x,y) = hi(x).hj(y)

Let S be the initial surface to be decomposed.

1^st step: computation of the coefficients of projection of the surface (minus the defect window) on the basic surfaces:

2^nd step: choice of the maximum coefficient

3^rd step: subtraction of the corresponding component from the entire surface.

The residue Res(x,y) becomes the new image S(x,y). The computation of the new coefficients allows us to substract the next component. This procedure is applied as many times as necessary to eliminate the background noise. This decomposition on a non-orthogonal base implies that the same component can be subtract several times (but non in succession).

NB: Using sinusoidal functions implies a periodicity which exists along y-axis but which is imposed along the x-axis (longitudinal axis). To avoid this, a rough flattering is applied so that the first and the last column of the image become equivalent. As for the rolling transition signals, they are for the most part eliminated by computing the mean profite along the generatrix.

Results: the figure 1 presents the data acquisition and an example of flattering.

Figure 1a-c la:data acquisition, lb:original cartography, lc:"flat" cartography resulting of the processing

3 FIRST DEFECT ANALYSIS: A MORPHOLOGICAL APPROACH

After this processing which improves the signal to noise ratio, a first analysis can begin. Defects are characterized by their orientation, their dimension, their depth.

Using mathematical morphology tools on the cartographies will give us information about the first two parameters: orientation and dimension. The depth information cannot be estimated from the presented electromagnetic surfaces. We need to come back to the complex plane and to the study of the Lissajous curves obtained from rows or columns extracted from the cartographies (X and Y line).

The morphological processing lies in applying a skeletonization algorithm to the image [3]. In the first instance, we have studied some simple defects whose characteristics ar...

Table of contents

• Cover image
• Title page
• Table of Contents
• Copyright
• PREFACE
• FOREWORD
• 13th World Conference on Non-Destructive Testing 13th WCNDT
• NDT RELIABILITY - A WAY TO GO
• ACOUSTICS
• ACOUSTIC EMISSION
• ALL
• ELETRIC TECHNIQUES
• HOLOGRAPHY
• LEAK TESTING
• MAGNETIC
• OTHER
• PENETRANT
• RADIATION TECHNIQUES
• RADIOGRAPHY
• AUTHOR INDEX