Vers la maîtrise de la qualité des assemblages de précision

In several industrial case studies, we put in evidence that microsystem manufacturing costs skyrocket because of low yields. Sometimes product functionalities are badly defined ; sometimes metrology systems, in spite of their high cost, happen not to measure the desired functions of the assembly ; at other times process off-centerings (mean distance to target) measured on batches required systematic adjustment, or else targets are not in the middle of the tolerances range. Resulting products present lack of profitability. Often, quality engineers first reaction is to fight process variability. We have shown that this is not the right policy in precision engineering. The first step is to verify that measurements are reliable, then insure the relevance of the functional requirements and after that, control the process off-centerings. It is then time to search ways to reduce variability. We provide engineers with a set of questions to analyse problems, so that product cost and profitability are considered from the beginning. For the low yields seen in precision assembly, we formulate how to calculate the ratio of conforming assemblies whatever part probability density functions might be. Low yields are due to three main sources : functional interval, off-centering and variability. Measurement variability is a fourth one, not negligible, often included in, and mixed up with process variability. A first cure for low-yields is illustrated by an example where measurement variability and tight tolerances are overturned by functional tolerancing and measurement. In precision engineering, where tolerance intervals are specified in the range of some microns, off-centering effects greatly influence assembly yields. Inertial tolerancing allows designers to assure that assemblies are produced at expected cost. Inertia is a joint specification including both process variability and off-centering. If high costs are due to process variabilities, selective assembly (sorting and mating parts) may be a cost-effective option. Moreover, it affords to reach very small variabilities, unreachable with existing manufacturing machines, or to avoid heavy investment in new equipment. Few make the most of this method, which is more practiced than confessed. Our state of the art, the most important to our knowledge, has pointed out the lack of cost perspective on this subject, on which some work is presented. Through an in-depth study of submillimetric press-fit, different techniques to reduce variability are illustrated. It all starts with a clear definition of press-fit functional requirements, before trying to optimize its design. Parameters of the Lamé-Clapeyron model are identified to be the relevant ones to design press-fits. Among them, interference is the most sensitive and critical one in press-fit, because submillimetric holes are hard to produce precisely. Knowing the functional requirements and the main sources of variability affords the optimization of assembly ro