Recently, many works have been published on the use of CNC as polymer reinforcement for nanocomposite applications [1,2] but these methods of elaboration of CNC composites at a laboratory scale are often based on solvent dispersion. In order to extrapolate to an industrial scale, extrusion processes are required. Nevertheless, two points have to be considered to produce thermoplastic composite by melt-processing: - a low thermal stability of CNC (below 200°C) - a strong tendency of CNC to agglomerate when dried Direct extrusion of CNC leads to a very poor dispersion whatever extrusion processing conditions used. A two step methods have been selected [3]: a first dispersion of CNC in the water-soluble polymer in order to obtain a composite PEO/CNC, then this concentrate product is dispersed by melt-blending in polymers matrix of LDPE [4]. We have chosen this method to elaborate our composites and we have tried to improve the dispersion of CNC at a pre-industrial scale by an optimization of material parameters that can be tune to enhance dispersion of CNC Rheological measurements, optical and MEB observations have been done on these ternary blends (PEO, CNC, and LDPE) to select best conditions leading to well dispersed CNC composites: • For a 5w% of CNC in the blend, what is the optimal CNC/PEO ratio in order to have a good re-dispersion in a LDPE matrix ? • We continue by playing with the molecular weight of the PEO used and we have seen an important effect of this parameter on the CNC dispersion • Finally, we have made a rheological percolation curve by varying the CNC content in the optimized blend. The rheological percolation threshold found on this optimized blends is in agreement with a theoretical value derived from L/D value which is below 10 in our case. Other final physical properties of the optimized blends have been characterized to select possible industrial applications. 1. Aspler J et al. Biopolym. Nanocomposites. John Wiley, 461–508 (2013) 2. Mariano M et al . J Polym Sci Part B 52:791–806 (2014) 3. Ben Azouz K et al .Macro Lett 1:236–240 (2012) 4. Pereda M et al., Appl Mater Interfaces 6:9365–9375 (2014)