This issue of AIEDAM returns for the third time to the subject of evolutionary design. As in the previous two “mini” special issues (Volume 13, Numbers 3 and 5, 1999), this issue contains three papers focussing on the use of evolutionary algorithms to aid designers by automating part of the design process. These papers provide a showcase of the work of some of the original pioneers of this field as they demonstrate that their research is still at the forefront of developments in evolutionary design. As is fitting for this final special issue, the three papers all concentrate on what most designers recognize as the trickiest stage of the design process for computers: nonroutine, conceptual or innovative design.

The paper discusses the requirements of the engineering designer during the higher, conceptual levels of the design process in terms of search within initial predefined design spaces and subsequent exploration of transformed design spaces. Such transformation is based upon information gathered from initial search using evolutionary techniques that rapidly identify high-performance regions of complex design spaces. It is proposed that off-line processing of such information and associated design team discussion in an iterative designer/evolutionary search procedure can strengthen the knowledge base relating to the problem at hand and initiate significant change in the design environment. Research relating to appropriate evolutionary search and exploratory procedures is described and illustrated with results from simple test functions and real-world design problems. These results support discussion related to the manner in which the techniques may best support innovative and creative design activity in an interactive design team/evolutionary search environment.

The paper provides a solution to the current problem of adaptation in case-based design. The aim of this project is to develop a case-based model of design, using an evolutionary approach, for the adaptation of previously stored design solutions. It is argued that such a knowledge-lean methodology is more general in its applicability than conventional case-based design. A prototypical example in the context of 2-D spatial design of houses is used to test the efficacy and efficiency of this approach.

In designing a state space of possible designs is implied by the representation used and the computational processes that operate on that representation. GAs are a means of effectively searching that state space which is defined by the length of the genotype's bit string. Of particular interest in design computing are processes that enlarge that state space to change the set of possible designs. This paper presents one such process based on the generalization of the genetic crossover operation. A crossover operation of genetic algorithms is reinterpreted as a random sampling of interpolating phenotypes, produced by a particular case of phenotypic interpolation. Its generalization is constructed by using a more general version of interpolation and/or by adding extrapolation to interpolation. This generalized crossover has a potential to move the current population outside of the original state space. An adaptive strategy for state space enlargement, which is based on this generalization, is designed. This strategy can be used for computational support of creative designing. An example is given.

In part-selection problems, parts are selected from catalogs and connected to meet the following problem requirements: functionality, specifications, and constraints. This paper formally defines the part-selection problem, enumerates a set of design properties that are useful during a search for a design solution, and provides an algorithm for solving part-selection problems based on a novel set of operators for manipulating portions of the design space.

The application of Bayesian networks for monitoring and diagnosis of a multistage manufacturing process is described. Bayesian network “part models” were designed to represent individual parts in-process. These were combined to form a “process model,” a Bayesian network model of the entire manufacturing process. An efficient procedure is designed for managing the “process network.” Simulated data is used to test the validity of diagnosis made from this method. In addition, a critical analysis of this method is given, including computation speed concerns, accuracy of results, and ease of implementation. Finally, a discussion on future research in the area is given.

Since the operation of technical systems can be explained using physical laws, why then might we not use them explicitly in designing these systems? The characteristic initial binding variable, with which appropriate physical laws are sought, first needs to be extracted from the function of the future technical system. If there are several appropriate physical laws (i.e., operators), we evaluate them using the Analytic Hierarchy Process (AHP) method. The most suitable is then selected with regard to the chosen criteria based on design requirements. When one physical law is not sufficient for the design of a technical system, several laws are linked together using binding variables to form a conceptual chain (i.e., macro-operator). Such a chain does not only contain supporting physical laws; physical laws indirectly introduce basic models of shape, their basic topology, geometry, and basic material properties into the chain. A prototype computer-aided design system is based on the prescriptive conceptual design model presented below.