To design mechanical systems, functional dimensioning and tolerancing consists in allocating a target value and a tolerance to part dimensions. More precisely, tolerance analysis consists in checking that chosen tolerances allow the mechanism to be functional. In the opposite, the tolerance synthesis goal is to determine those tolerances, optimaly if possible, such as functional requirements are respected. Traditional statistical approaches are very useful in pre-design phases, but are incapable of estimating precisely non-quality risks. This PhD thesis adresses this problem from a different point of view. The objective is no longer to respect functional requirements but to guarantee the Non-Conformity Rate (NCR) of the mechanism. It is the probability that the functional requirement is not respected. Reliability methods, based on probabilistic theory, allow a precise and efficient calculation of the NCR. The main goal of this thesis is to explore tolerance analysis and synthesis domains in order to identify potential contributions of reliability methods to these issues. In this work, different reliability tools are provided enabling tolerance analysis of all kind of mechanisms with or without gaps. The probability theory allows a detailed modeling of parts dimensions. In particular, the APTA (Advanced Probability-based Tolerance Analysis of products) approach is designed to take into account random variations of some parameters such as mean shifts which influence highly the NCR. As tolerance analysis is more complex for mechanisms with gaps, a specific method is designed based on a decomposition of contact points situations and the use of a reliability system method. Several sensitivity indexes are also proposed to identify the leading dimensions on the NCR. The optimization of the production cost, under a NCR constraint, enables an optimal tolerance synthesis. The potential benefits in terms of costs exceed 50% compared to initial designs while keeping under control quality levels of products. Companies RADIALL SA and VALEO Wiping Systems, designers and manufacturers of products for the automotive and aeronautic industries, have provided case studies on which is demonstrated the relevance of the presented work. Based on this work, the company Phimeca Engineering, specialized in uncertainties, is developing and commercializing a professional computer tool.
