International audience ; Thermostructural composites are increasingly used in many industrial applications that require materials with excellent thermo-mechanical properties at high temperature. These materials usually consist in the combination of a matrix embedding a reinforcement made of cylindrical fibers. The quality of the composites depends partly on the matrix infiltration step, in our case performed by Chemical Vapor Infiltration (CVI) process, because it is decisive on the residual porosity and the matrix quality. Experimental determination of the optimal processing parameters is expensive and time-consuming, so it seems interesting to model this process. The morphological evolution of the preform has an impact on the infiltration through various phenomena (gas transport, chemical reactions). To understand the influence of the pore shapes on infiltration quality, an approach has been developed, which is based on 3D images acquired by synchrotron X-ray microtomography. Image processing allows separating the raw fibers from the matrix and the remaining voids; then, morphological parameters (pore and fiber diameters, matrix thickness, etc .) become accessible. A computer code based on a Monte Carlo/Random Walks algorithm with surface discretization by Simplified Marching Cubes makes it possible to model the infiltration. The comparison of the segmented and the simulated matrix blocks gives information on the infiltrability of the preform. Then, it becomes possible to relate infiltrability to physicochemical parameters and to the fiber arrangement, and also to predict optimal infiltration conditions of a given new texture.
|