A constrained evolutionary search formalism for remote design with modules

The potential benefits of modular design over conventional design include e conomies of scale, increased feasibility of product/component change, incre ased product variety, and reduced order lead-time. The benefits of modular design have not gone unnoticed, and it is becoming evident that companies a re increasingly seeking to use this modular approach. Important research is sues in the area of design with modules center around the central theme of selecting modules to satisfy customer requirements. However, while modular product design would appear to be an attractive proposition, little work ha s been done on these research issues, particularly for the relatively commo n design environment where modules may be available from one or more geogra phically dispersed sources, and where data concerning the modules may be in a multitude of databases scattered across the globe. Under these circumsta nces, it becomes necessary: to search through the databases using a network , such as the Internet, for feasible subsets of the set of candidate module s. Such a search call be difficult due to the heavily constrained environme nt that encompasses much design activity. This paper addresses these issues by proposing a design with modules algebra and new constrained crossover a nd constrained mutation operators that aim to permit an evolutional approac h to operate effectively in such an environment, and thus to produce a popu lation of feasible designs that aim to maximize (or minimize) a given objec tive function. The implementation of this approach is described with an exa mple of personal computer design, where the designer is remote from the dat abases and communication is carried out over the Internet. Some results on the convergence of the approach are shown.