Pre-design of products.

Summary Our work concerns the design of products in the preliminary phases, or pre-design of products, found in the literature under the expression of conceptual design, at the crossroads between the fields of Artificial Intelligence applied to design, integrated and simultaneous design, design theories and industrial methodologies such as Value Analysis. In a first step, we proposed enriched models of specifications. These are functional models of objectives and satisfactions (or preferences) related to a product, its associated services and its design-industrialization project, trying to take into account the following aspects: uncertainty, context, multi-actors and evolution. In a second step, we worked on the way to translate these product objectives into solution concepts (or architectures) in the most automated way possible: by synthesis methods, or by semi-automatic methods which, far from taking the place of the designers, allow them on the contrary to have generic design tools giving them the means to devote themselves to creativity tasks. In particular, we have contributed to the development of a method for concept synthesis in spatial allocation (constraint programming technique) and methods for concept synthesis of planar mechanisms (neural network, multi-agent, case-based design techniques). In a third step, when concepts have been proposed, their performance must be evaluated, the difficulty residing on the one hand in the coarse definition of a concept and thus in their ability to represent entire classes of solution products, and on the other hand in the absence of detailed design. Reasoning from incompletely defined models is an important part of our work. Methods from qualitative physics have been developed, and performance modeling has been approached in a probabilistic and fuzzy way as well as in an approximated way by metamodeling techniques. Finally, we have addressed the measurement of the adequacy of the performances of a concept with the specifications, for a generic product, in order to develop functionalities of performances, costs and risks monitoring in the pre-design phase, as well as functionalities of help to the re-design and optimization of the concepts (methods based on qualitative physics or on a Monte-Carlo type simulation).