Characterization of microcracking in a particles reinforced composite through image analysis

The damage behaviour of a particulate metal matrix composite strongly depen ds on the distribution of particles in size, number or position. A simulati on of this material in view of an utilization in a code has therefore to ve rify some of these characteristics, before and after damage. In the followi ng, images from a particulate composite are then studied in order to know t hese characteristics. The material is a Si reinforced aluminium, obtained through powder metallur gy. Images, obtained through a serial sectioning realized on a tensile spec imen after a 6 % deformation test, are analysed. Three sections are used, w ith two different magnifications x250 and x500 (fig. 1). The objective of this study is to determine an image size representative of the material for several characteristics. For this purpose an image is div ided into adjacent and not connected parts, several times each time corresp onding to a different size For each size, the mean Value and the standard d eviation of the characteristic on the partition is determined With the assu mption that the distribution of the character over the partition obeys to a normal law, it is possible to obtain at a given risk, the size of the part ition, which gives the character in a given interval around the mean value. For this purpose, a fitting of the standard deviation as a function of the size is used. In the first part, the areal density of particles is studied The sum of the three images x250 is used for this study A mean value of 0.22 is found, an d considered as the areal density for the material. A power law could fit t he standard deviation. Then, an image of sire of 0,39 mm(2) is found to be representative of the material with a 5% risk and with a possible variation less than 4% of the volumetric particles fraction. In a second pan: the crack repartition is studied on the three images x500. These images show variation of the crack grey level with the position of t he pixels on the image (fig, 4). A simple threshold is then not possible. A n algorithm is develop which : 1) finds the minimum grey level on each particle; 2) keeps as crack, all pixels whose grey level does not exceed this minimum from a chosen value s; 3) eliminates the set of pixels recognized as crack but not belonging to a crack (fig, 5 and 6). The repartition of the cracked and uncracked particles allows a definition of three classes of particles based on their size, and on their ability to crack. In the last part, the size of the image representative for the repartition into the three classes, is determined on the same base as in the first part Then, the same image of 0.39 mm(2) could be representative of the material at a 5% risk, but with a variation of three hundred of particles per mmr, for the middle particle size class, and of fifty particles for the highest particle size class, which represent the particle which are almost always c racked at 6% deformation.