The work presented in this thesis deals with the robust design of products. Particularly, it focuses on the design process preliminary phase where design models and specifications are already defined. A robust optimization approach is proposed. It aims to: reduce the scattering of the objective function included in the product specifications when the uncertainties reach the design parameters, maintain good performance of the product and ensure the constraints feasibility. We propose a new “robust” product specification that changes the objective function and the constraints of the initial specification in order to integrate the concept of robustness previously defined. The second contribution is an analysis of methods found in literature to propagate uncertainties across design models. Design parameter variations are modeled by probability distributions. The theoretical analysis of these methods is completed by numerous tests to investigate the accuracy of the results and to select the method used thereafter. The robust product optimization approach proposed in this work is finally implemented and tested on two case studies. It incorporates the propagation of uncertainties within the optimization loop to automate the search of a robust optimal solution for the design product.
