F. Plouraboué (email@example.com) and G. Debenest (firstname.lastname@example.org) Institut de Mécanique des Fluides de Toulouse, www.imft.fr
Pore-scale modeling has considerably improved, over the last twenty years, from constant progress of 3D tomography imaging. Ranging from sub-micron (or even tens of nanometer) to several centimeter scale, tomographs have the ability to perform huge image acquisitions, (e.g. 40003 voxels) over a wide variety of rocks samples with various mineral structures.
Even if the image post-processing associated with these very large X-ray 3D scans can be technical (i.e, denoining, ring artefact corrections, thresholding), especially when designing a parallel processing workflow, there is no major challenge to obtain reliable segmentation of the solid phase.
On the contrary, the numerical estimation of the transport properties for such large images as conductances and permeability, remains a bottleneck due to the computational time.
This is why we aim at developping and characterizing an innovative method for very fast estimate of transport properties of very-large samples.
The method is based upon the image-based Effective Medium Approximation (EMA) first described in 2D in Plouraboué et al (2006), which only requires an accurate identication of saddle-points and theirs associated local diameters.
The aim of the postdoc is to efficiently implement the proposed methodand to test/validate the approach with both numerically generated samples and experimental ones provided by the experimental collaborators of the project. The steps of the workflow are the following
To validate the approach, the method have to be compared with classical numerical approaches such as direct finite-volume simulations (using OpenFOAM) and pore network modeling (using OpenPNM), which will involve interactions with the support team. The post-doc will not develop this numerical step, but will use it, as the final step of the workflow.The core of the expected outcome of this post-doc is to develop and improve the steps 2 to 6 of the above mentioned workflow. Publication in leading international journals of the domain is also expected.This position is multidisciplinary from informatics and image processing (major part) to modelling and numerical computations that could lead to publication in geo-sciences field
Segmentation, Skeletonization, Topology, recovering, Graph/Network
P. Creux (LFC/UPPA/Stanford University), P. Moonen (LFC/UPPA), P. Poncet (LMAP/UPPA)
PhD required in image processing, computer vision or applied mathematics.
Very good programmation level in C++, GPU programming knowledge would be a welcomed.
Image processing, experiences with image librairies and numerical computation.
Autonomous, communication skills.
The post-doc will be located @ IMFT, Toulouse, France
12 months contract, available from January 2017
Net salary approx. 2400 euros/month
 F. Plouraboué, F. Flukiger, M. Prat and P. Crispel (2006) Geodesic network method for flows between two rough surfaces in contact, Phyical Review E.
 L. Risser, F. Plouraboué and X. Descombes (2008) Gap filling of 3D micro-vascular networks by tensor voting, IEEE Transactions in Medical imaging
 P. Horgue, R. Guibert, H. Gross, P. Creux and G. Debenest (2016) Efficiency of a two-step upscaling method for permeability evaluation at Darcy and pore scales, Computational Geosciences.
(c) GdR 720 ISIS - CNRS - 2011-2018.