In today’s manufacturing industry, new product development is one of the most critical ingredients of success for companies to survive in the fierce competition. Companies, especially their product design teams are urged to keep introducing new products in response to the frequent fluctuations in markets and technology and at the same time maintain a large product variety to cater for the diversified customer needs....[ Read more ]

In today’s manufacturing industry, new product development is one of the most critical ingredients of success for companies to survive in the fierce competition. Companies, especially their product design teams are urged to keep introducing new products in response to the frequent fluctuations in markets and technology and at the same time maintain a large product variety to cater for the diversified customer needs.

But on the other hand, the whole value chain of product development runs with continuity. In order to maintain low cost and risks in new product development, existing manufacturing capabilities should be utilized, previous experiences should be inherited and consistency in customer needs should be considered. Thus challenges are imposed to the product design team to manage continuity, responsiveness, and variety simultaneously in designing new products.

Design in an evolutionary way, which means to design new product through gradual modifications to existing designs, suggest itself as a good strategy in this context. And natural evolution as the greatest system in the universe that manages continuity, responsiveness and variety in designing lives, sets a perfect example for the implementation of evolutionary design. Many works have been done in design search and optimization using evolutionary computation techniques, which are inspired by the classical theories of biological evolution. However, there are some unresolved deficiencies in applying evolutionary computation in design problem solving, which are also observed in classical evolution theories in biology. Specifically, without a systematic way to utilize generic domain knowledge in search, most existing methods of evolutionary design focus on specific tasks of design optimization and fail to support product design evolution in a continuous way.

This research project is inspired by the recent advance in evolutionary developmental biology (evo-devo), which identifies the importance of development process in evolution and contributes to an evo-devo framework in explaining natural evolution. Using the evo-devo framework of evolution, we propose an evo-devo framework of evolutionary design with a regulation process added to the traditional frameworks. Due to the specialties of product design problems, domain knowledge is introduced in the process to guide the random evolutionary search, so that the deficiencies in traditional evolutionary design are repaired.

Corresponding theories in biology, simple evo-devo models and related methods in the design domain are reviewed. By representing domain knowledge as a causal network, we implement the regulation process in evolutionary design as an agent-based constraint satisfaction process to locally accommodate the randomly-generated design variants. Illustrative case study is given, showing how the evo-devo framework integrates a random evolutionary search with knowledge-driven design refinements. With systematic representation of domain knowledge, this approach not only better supports specific design optimization tasks, but also continuously facilitates new product design in an evolutionary way.