Owing to its porous cellular structure and polymeric components, wood can be easily shaped in its transversal direction when it is heated and has sufficient water content. Thanks to this thermohydrous behaviour many manufacturing processes such as densification, moulding, shaping or bending can be considered. Thus, some mechanical properties like the hardness, the shear and tensile strength as well as the wood durability facing microorganisms are enhanced. However, the shape of this new material is not stable: it recovers its initial shape when remoistened. This phenomenon is called "shape memory " or "set-recovery" of shaped wood. Nevertheless, this problem can be solved by chemical and thermal treatments. This study focuses on thermo-hydrous treatments that allow the preservation of the ecological aspect of wood. The main objectives of this study are on one hand the better understanding of cellular and molecular mechanisms leading to decrease or to eliminate the densified wood during set-recovery thermo-hydro-mechanical (THM) post-treatments, and on the other hand the optimization of the densification and THM post-treatment processes parameters and conditions. Preliminary, to better apprehend the elimination of the set-recovery, the origin of densified wood shape memory are revised. Many studies showed that the shape memory of wood originates from the inner stresses induced by the deformation of cellulose microfibrils and other componants. This deformation generates an increase of the material elastic energy. Under moisture and heat conditions the inner stresses are relaxed by the quasi-overall recovery of microfibrils set. During the set-recovery the lignin follows the microfibrils movement due to their bonding to cellulose microfibrils by hemicelluloses. So the cellular walls come almost to its initial shape, allowing stored elastic energy to be dissipated. The hemicelluloses take a fundamental part during set-recovery as they assure the cohesion between cellulose microfibrils and lignin by physico-chemical bonds. Consequently the dissipation of elastic energy during THM post-treatments is facilitated by the rupture of these bonds causing the set-recovery of cellulose microfibrils and lignin. A model of the hemicelluloses hydrolysis reaction has been carried out. A high correlation between the polymerisation degree of the hemicelluloses and the set-recovery of densified wood has been calculated by this model. The strong hydrolytic action on the hemicelluloses has also been confirmed by our physico-chemical analysis. Thus our model indicates that lignin must be in its rubber state to be able to eliminate totally the set-recovery. At higher temperatures ( 180°C) the middle lamela becomes fragile due to the thermal degration and causes microcracks between the wood cells during the abrupt release of steam at the end of the post-treatment and during drying. These microcracks are also susceptible to be the source of the dissipation of elastic energy stored during the densification process. Finally, our results indicate that steam post-treatments are able to increase the cellulose crystallinity and the crystallites diameter. The decrease of the setrecovery can be also partially explained by these phenomena. With regard to densification and post-treatment THM conditions, the recovery tests results show an optimal temperature of densification at 110°C under saturated vapour condition. Our mechanical tests testify a significant thermal damage from 180°C. Thus the set-recovery elimination has to be done around 160°C to avoid the intercellular microcracking problems when cooling and drying the samples. The post-treatment THM can be then a promising procedure to eliminate the set recovery for the industry. Nevertheless, further studies must be carried out for better understanding the cracking problem due to the drying process for bigger samples.