 | 2004 |
Abstract:
The configuration design stage is the first and most important step in the product development cycle. Within this step the general structural layout as well as the spatial arrangement of all components and the final product’s properties, quality and appearance are defined. This step determines the basis and also limits the potential for further development and optimization. Conventional engineering optimization methods mostly do not consider this step. Sizing, shape and topology optimization methods are applied when the overall configuration design is already available serving as a starting point. To utilize the potential of a design in an optimal way, it is preferable to expand optimization capabilities to early design stages where also the configuration of the system is under consideration. This leads to discrete, multiscale and combinatorial tasks for the optimal configuration of components, their dimensioning and also materials to be selected as well. Particular, this process include:
- topology, geometry and material of the basic geometric and structural components
- proper placement of components and equipment
- multiple criteria, constraints and design parameters giving rise to multi-disciplinary interaction
- consideration of a certain fuzziness in requirements (constraints) including qualitative statements
and (at least partly) lack of precise models The output of this process shall be a set of feasible or nearly feasible designs satisfying the basic requirements regarding geometrical, mechanical (or general physical) and experience based requirements. In this paper we present an approach for computational configuration design in the early product development phase. The geometric layout of components, material selection as well as the presence or not presence of certain structural members is taken into account. The concept is based on two level approach generating feasible geometric designs first and then performing a detailed structural optimization. It includes a evolutionary algorithm GAME capable of handling continuous and discrete design variables simultaneously as well as multiple objectives. Also response surface approximation (RSA) functionalities are integrated in the EA for improved efficiency and accuracy. For a proper and exact geometric description of the components and assembly as well as for evaluation of geometric properties a parametric CAD software is used and implemented in the configuration optimization loop. Additionally, some specific aspects regarding discrete-continuous design optimization as well as possibilities to include fuzzy and qualitative knowledge into the optimization models is discussed. Some demonstration examples for configuration optimization show the potential of our approach.