Abstract :
[en] The objective of this work is to study the deformation of polypropylene foam during a dynamic crash loading. The first difficulty consisted in conceiving an experimental setup that would allow to visualise intermediate steps in the deformation of the foam, which requires a non-destructive imaging technique. Fast external surface imaging is not sufficient for an accurate study of the deformation, therefore our attention focused on X-ray tomography. Because crash loading time is much smaller than tomogaphic acquisition time (a few milliseconds vs. almost an hour), several interrupted crashes are applied (a dynamic loading with a constraint on the strain), in between which a micro-tomogram is acquired. If we assume that the foam behaviour is not modified by the interruptedness of the dynamic loading (as in quasi-static loading), then we obtain a series of tomograms showing the evolution of the foam sample during dynamic compression. The second difficulty, that is presented here, is to use this information to quantify the foam deformation (and subsequently use this experimental data for predictive modelling) at the mesoscopic scale, i.e. that of the beads. Polypropylene foam is a multi-cellular material, each bead (around 2mm in diameter) is composed of micrometric cells, making the separation of the beads a difficult image analysis problem, as compared to separation of foam bubbles in a metallic foam, for instance. The solution we propose extracts a representative volume inside each bead, in order to calculate, at each stage of the compression, values such as grain density. For this purpose, a first processing, consisting of a sequence of simple signal processing and discrete morphological operators, determines approximate bead centres. With a simple nearest neighbours approach, the position for a given bead is identified for each stage of the experiment. From each centre a deformable surface algorithm is applied: a spherical mesh is placed inside the bead and expands until reaching the bead walls. This allows us to visualise the deformation and measure the average density of each bead during the compression. Unfortunately, due to the difficulty in precisely identifying bead walls (even manually), complete bead volumes and wall thickness is not yet determined: the deformable surfaces contain only the majority of the inside of the beads, and there still remains significant intersticial volume which has no physical significance (it represents a volume in which the bead wall is located, volume that is a function of the image processing). Future work to obtain more accurate data would focus on methods such as skeletonisation of the intersticial volume.
