Image analysis of corrosion pit damage


Journal Article


Journaux, S., Guillaumin, C., Gouton, P., Paindavoine, M., Thauvin, G.




Image analysis of corrosion pit damage

Journal / Livre / Conférence

Revue de Métallurgie


Corrosion contributes frequently to the damage of metallic components in the low-pressure part of steam turbines working with wet steam. The most commonly observed mechanism of damage is initiated by the forming of pits, their growth and then the formation of cracks. The development and the propagation of the crack under stress corrosion or fatigue corrosion provoke the rupture of the metallic part. To act on the significant parameters of the damage, it is necessary to know the importance of the different phases, their contribution in the global mechanism and also the conditions of passage between phases of pitting, growth and cracking. This quantification involves the observation of many damaged specimens in aqueous solution, with growing lengths. T he image analysis is a relevant tool to characterize qualitatively and quantitatively the pitting and the first stages of the damage mechanism during time. The present work focuses on the development of a method of image analysis following the pit growth in the course of time on rough machined specimens after different times of immersion in chemically corrosion matter. This article is organized as follows : In the first pan: we describe methods of the measures of kinetic growth of corrosion pits by image analysis, the phenomenon of corrosion, as well as of experimental device of our study In the second part we expose the developed technical framework about technical constraints, linked to acquisition and measure. In the third part, we present the totality of the technique of image analysis. This phase contains two stages where mathematical morphology algorithms are described In the last part, we expose some experimental results. The steady kinetic of pit growth rests on the realization of cartography of the useful surface of the specimen. We have developed a method of image analysis to follow the growth of pits in the course of time on rough machined specimens, after different duration of immersion evolving naturally at free potential in a medium of test. This quantification has required the development of a mechanical device to hold the specimen in position, an original methodology for its locating between each period of immersion, a device of lighting in low-angled and incident. The measurement software has two phases that comprise a double automatic scanning of the surface. One undertakes an optimized scanning of the totality of the surface to G = 100 (2,04 mu m/pixel), which represents approximately 300 fields of observation (314, 16 mm(2)). The choice of this magnification rests on the analysis duration, which is approximately 1 h 20 min, and also on a good digital definition of objects. One measures then the position of the center of gravity the width and the length of pits called potential. Once the 300 observations finished, one changes the magnification so that G = 200 (1.04 mu m/pixel). One undertakes a second analysis from the position of the potential pit. The chain of image analysis far each scanning comprises a phase of pre-treatment, numerical processing threshold. binary processing and measure. The software of image analysis is essentially composed of algorithms with a mathematical morphology basis. The presented method allows the quantification of pits from their germination (diameter greater than or equal to 10 mu m). Numerical processing is undertaken by a directional sequential alternated morphological filter to grayscale image, according to the orientation of machined strias. This filter is built by sequential application of directional closings and openings. Directions are defined by the orientation linear structuring elements, of growing size and angle 0 degrees, 27 degrees, 45 degrees 63 degrees, 90 degrees, 117 degrees, 133 degrees, 153 degrees. To undertake the cartography of the pitted surface and to establish the steady kinetic of the growth of pits, we have implemented a method of shape recognition. The measured pits are then described try a discriminating curve and classified by neural networks. This will have allow us to access information on the geometry and on the nature of objects to be measured in order to classify them by types of population. The totality of measured information will allow the realization of the cartography of a specimen surface to a given immersion time. The kinetic of growth of a pit will be deducted by superposition of information of each cartography to a time.







‹ Retour à la page précédente