The genetic algorithm (GA), an optimization technique based on the theory of natural selection, is applied to structural topology design problems. After reviewing the GA and previous research in structural topology optimization, we describe a binary material/void design representation that is encoded in GA chromosome data structures. This representation is intended to approximate a material continuum as opposed to discrete truss structures. Four examples, showing the broad utility of the approach and representation, are then presented. A ®fth example suggests an alternate representation that allows continuously-variable material density. Concluding discussion suggests recommended uses of the technique and describes ongoing and possible future work. Ó 2000 Elsevier Science S.A. All rights reserved.

Successful manufacturing system designs must be capable of satisfying the strategic objectives of a company. There exist numerous tools to design manufacturing systems. Most frameworks, however, do not separate objectives from means. As a result, it is difficult to understand the interactions among different design objectives and solutions and to communicate these interactions. The research described in this paper develops an approach to help manufacturing system designers: (1) clearly separate objectives from the means of achievement, (2) relate low-level activities and decisions to high-level goals and requirements, (3) understand the interrelationships among the different elements of a system design, and (4) effectively communicate this information across a manufacturing organization. This research does so by describing a manufacturing system design decomposition (MSDD). The MSDD enables a firm to simultaneously achieve cost, quality, delivery responsiveness to the customer and flexibility objectives. The application section illustrates how the MSDD can be applied in conjunction with existing procedural manufacturing engineering.

Extending previous efforts, this article describes how a speciating genetic algorithm is used to distribute subsets of the evolving population of solutions over the design space. This distribution of solutions is analogous to different species exploiting different niches in an ecosystem. In addition to reviewing genetic algorithms with an emphasis on techniques to cause such niche exploitation, we describe how we use statistical cluster analysis techniques to quantify the extent to which a population is speciated and how this measure can be used to probabilistically encourage mating of reasonably similar designs (i.e., intraspecies mating). Results demonstrate the creation of different good designs of characteristically different topology and shape.

No abstract