Biorefinery of sunflower whole plant can be conducted with water using a nine modules Clextral Evolum HT 53 twin-screw extruder (TSE). Aqueous extraction of oil is an environmentally cleaner alternative technology to solvent extraction. TSE carries out three unit operations: conditioning and grinding, liquid/solid (L/S) extraction and L/S separation. The compressing action by the reverse screws (CF2C) is essential for L/S separation. Positioned in module 9, CF2C screws push part of the mixture upstream against the general movement in TSE, and this counter pressure ensures the L/S separation efficiency above the metal filter, located in eighth position. Oil is extracted in the form of two emulsions, stabilized by phospholipids and proteins, and usable as co-emulsifiers in cosmetic industry. An aqueous extract containing water-soluble components from whole plant is also generated; it could be recycled. As a mixture of fibers and proteins, the cake can be moulded by thermo-pressing into boards, usable in the furniture and building industries. In this study, fractionation was conducted from next inlet flow rates: 54 kg/h solid and 183 kg/h water (3.4 L/S ratio). The screw speed varied from 249 to 124 rpm, corresponding to a filling coefficient (ratio of the solid inlet flow rate to the screw speed) from 217 to 436 g/h rpm. The filling coefficient directly affects the L/S separation efficiency. The latter can be estimated from next experimental data: the outlet flow rates of both cake and filtrate, the cake moisture content, the residual contents of oil and water-soluble components in the cake, and the extraction yields in dry matter, lipids and water-soluble components. For low filling coefficients (i.e. high screw speed), the L/S mixture compression in CF2C screws is insufficient, not allowing a satisfactory L/S separation. Conversely, for high filling coefficients (i.e. low screw speed), solid particles accumulate more upstream from the pressing zone, obstructing part of the filtering screens and thus reducing the filtration surface. A less efficient L/S separation is then observed. From the experimental data evolution, optimal screw speed was estimated at 182 rpm using a second order polynomial regression, corresponding to a filling coefficient of 297 g/h rpm. Extraction yields in dry matter, lipids and water-soluble components were 22%, 49% and 40%, respectively. Such filling would lead to a specific mechanical energy of 103 W/h kg whole plant processed.