Optimisation simultanee des caracteristiques de la qualite d'un produit en environnement distribue

In an effort to respond to competition from the emergent economies, many companies have been engaged in processes of productivity improvement. Despite this, it is increasingly difficult to compete at the manufacturing level, since labour costs in these countries are very low. As an innovative strategic response, companies are concentrating increasingly on their product development process with a view to rendering this faster, attempting to utilize technologies that will require several years before being copied. Not necessarily possessing the core competencies required or the specialized design tools to accomplish this, businesses have been forced to outsource this design work to a number of different outside actors, creating a need for collaboration among these partners.;The goal of the work presented in this document is to develop an interactive algorithm amenable to use in a distributed environment and capable of executing multi-objective optimization. The algorithm must be interactive so that a number of partners may collaborate on one project, it must be able to take into account several objectives with the goal of optimizing a number of quality features of the product, and it must function in a distributed environment such that the partners are able to work simultaneously. To accomplish this, an interactive algorithm based on goal programming was twinned with a technique called Linear Physical Programming which permits work groups to input their preferences during the optimization process. The algorithm created was designated the IPOMP algorithm (algorithme Interactif pour résoudre des Problèmes à Objectifs Multiples basé sur les Préférences des groupes de travail). The algorithm is simple to use and generates Pareto-optimal solutions. However, optimization remains under the control of one single person (centralized environment). The algorithm was tested on two problems: a mathematical problem and a real design problem. The results obtained show that the algorithm functions well, and that only a small number of iterations are necessary to obtain Pareto-optimal solutions. The algorithm is useful because it avoids discussion between work groups to get a consensus.;In a second case, the IPOMP algorithm was amalgamated into a distributed optimization structure. This structure is advantageous since it permits groups to work simultaneously and in a physically decentralized fashion. The algorithm thus created is called IPOMP-ED, where ED stands for distributed environment, and permits resolution of multi-objective problems in a distributed environment. The solutions obtained are identical to the solutions found in a centralized environment. When differences are observed, the reasons that might be given to explain these differences are associated with the behaviour of the Sequential Quadratic Programming optimization algorithm of the software program DONLP2.;Finally, different methods of robustness analysis and sensitivity analysis were demystified so as to be able to define a process for choosing a robust solution. The methodology proposed permits identification of variables influencing the variation of objective functions so as to be able to take actions to reduce this. The results obtained show that, in certain cases, the deterministic solutions found are already robust, while in other cases robust solutions are obtained. The process suggested is simple but it permits study of the variability of objective functions and improvement of product quality from the design phase.;The research performed in this thesis revolved around the challenges to which companies must rise in regard to product design, that is to say: (1) developing new products in collaboration with other work groups (within and external to the company); (2) producing a product which responds to the needs and expectations of the client; (3) reducing product development time through simultaneous development and, (4) reducing the costs of development of the product.