All the industry from the Polymers Area, change their sourcing strategy due to 2 constraints : fossils resources availability which are the basis of the « regular » Polymers and the incentives in the reduction of the energy consumption. This double motivation drive the car makers R&D Departments and their suppliers to evaluate and use natural fibers reinforcement as an answer to this double objective : their length is by itself a powerful mechanical reinforcement media which also leads to a very interesting density decrease. Also their natural sourcing decrease the fossil resources needs. A great number of R&D collaboratives actions made this challenge happen focusing on the application, the fiber itself as raw material, its preparation, on the characteristics of the fibers based composite for such or such application even on the supply chain, but without accounting on the decohesion phenomena. This is a key step in the compounding stage. Mastering decohesion is important to keep fibers length which is their main attribute as mechanical reinforcement additives. The DEFIBREX project focus on this topic with an ambitious and generic methodology, centered of the natural fibers decohesion phenomena analysis under mechanical constraints. This analysis must lead to a decohesion behavioral model which should be capable to describe a wide types of fibers and mechanical constraints ranges. This model will be validated by experiments run on a variety of fibers.( sourcing, physico-chemical treatments...) with the ambition of proposing the widest global classification of the various used fibers in the industry. To reach these ambitious objectives, the DEFIBREX partnership associate complementary scientific skills (sourcing, pre-treatment with FRD and INRA, process and modelisation with CEMEF and µTomography morphologic analysis (I2M) , with industrial partners (K-TRON, SCC, CLEXTRAL et FAURECIA,) whose contribution will be to help reaching the project objectives and answering questions such as : how to feed a twin-screw extruder without degrading fibers length ? How to set a twin-screw extruder in such a way to maximize mixing efficiency and without degrading fibers length ? What is the functionality increase that can be expected on such injected pieces, and what is the dimensional saving that we can account ? How to deliver and disseminate the behavioral model to industrials users ? To answer those questions, the DEFIBREX project is self organised in 5 work packages : Fibers will be sourced and analyzed on the morphological, mechanical and bio-chemical point of view in WP1. WP2 is dedicated to phenomenological study of the compounding process on a lab scale level with a wide range of fibers types. The compound characterization process by these experiments will be used as input for WP3 whose objective will be to set the behavioral model and its declination as a numerical model in the Twin-Screw extruder simulation program LUDOVIC (as a media for dissemination). In parallel, a fibers classification will be be proposed to establish and sort their polymers reinforcement capability, as well as the compound processability evaluation. The WP 4 is focusing on the industrial scale : fibers feeding, productions of composites samples, compound characterization in order to wider the application range of the behavioral model. Also, the WP4 will produce compound to inject real automobile parts with the objective to evaluate the performance increase of such new parts. This will be achieved in the WP5. Classically, the WP6 will be the project coordination work package. Finally, DEFIBREX is expected to contribute to the knowledge and understanding of the natural fibers decohesion phenomenum and delivering to the manufacturing industry a key technological breaktrough.