Though lots of numerical methods have been proposed in the literature to optimize mechanical structures at the final stage of the design process, few designers use these tools since the first stage. However, a minor modification at the first step can bring significant improvements to the global performances of the structure. To address the pre-design optimization problem, evolutionary algorithms (EAs) were selected, because of their ability to explore widely the design space. Nonetheless, they do not perform very well in the presence of constraints. Furthermore, in many industrial applications, multiple objectives are pursued together. Therefore, a new approach called PAMUC (Preferences Applied to MUltiobjectivity and Constraints) was proposed. First the user has to assign weights to the objectives. Then, an additional objective function is built by linearly aggregating the normalized constraints. Finally, a multicriteria decision aid method, PROMETHEE II, is used in order to rank the individuals of the population. PAMUC was validated on standard multiobjective test cases, as well as on the parametrical optimization of poppet valves from the VINCI engine, designed by Techspace Aero for launcher Ariane 5. The second step of the thesis consisted in incorporating an inference engine within the optimization scheme in order to take expert rules into account. First, information about dimensioning and design has to be collected among experts in a specific domain. Then, each potential design generated by the EA is tested and repaired (with a given probability) according to the expert rules. This approach seems very efficient to reduce the size of the search space and guide the EA towards the global feasible optimum.