1.0 INTRODUCTION
Mounting costs and increased competition within the manufacturing industries demand that products complete the cycle from design to production in less time and with higher quality than previously attainable. Rapid product realization is a goal that can be achieved through an understanding of stages and cycles of engineering design, an examination of paradigms within which designers effectively work, and the development of an effective environments and tools with which to work.
The traditional physical prototyping, testing, and redesign of products is slow and costly, and unacceptable in light of the computational software and hardware resources currently available. Even before physical prototypes are created, Computer Aided Engineering (CAE) permits substantial product analysis to occur within the design cycle, thereby reducing overall cycle time and costs. With computer based tools, designers may actually increase the number of design iterations, just as accountants are more likely to increase their "what-if" scenarios with electronic spreadsheets than if they had to do manual calculations. However, with proper tools and models, the total cost of exploration within the virtual environment of the computer is substantially reduced while the likelihood of a better solution being reached is increased.
Good engineering design practice has always been more than the drafting of a part which meets functional requirements. Designers always relied on the experience and proactive insights into the artifact's creation to create producible designs. But there are limits to the breadth of a single designer's experience. Now there are aspects of the total product life cycle, including marketing, manufacturing, and transportation, as well as environmentally-based disposal and recycling issues that can -- and often must -- be considered. Information technology, of which computers play a key role, allows to be present in the design phase an abundance of information that even the most ambitious of design teams cannot master.
The provisions of information technology can potentially overwhelm the designer or design team. For example, safety-based, functional strength requirements of a part must be considered along with materials selection for recycling, cost-reduction dictates from finance, and weight limits from transportation. But this same technology can also provide a medium to effectively coordinate and assist this comprehensive, concurrent engineering. Toward these ends, research commenced in exploring the requirements of an environment which can at best automate engineering design and at least intelligently assist engineers in concurrent design.
Design advisors were originally conceived as computer utilities that brought downstream concerns to the design stage for feedback and active guidance during the design process. Design advisors play a role as intelligent assistants, providing abstractions of downstream concerns which make the knowledge more readily assimilated into the design stage. They can act as proxies for experts, and may ideally be continually maintained and updated by experts in a given domain.
Not limited to downstream considerations, advisors may incorporate upstream concerns as well. With design intent and other design synthesis concepts available for the analysis process, results of advisors at some stage in the design-to-manufacturing process can be routed both back to design and ahead to manufacturing.
With a freer flow of information, it is reasonable to replace the analogy of streams, an artifact of the traditional manufacturing flow, with a bi-directional piping of knowledge throughout a design process. However, to break away from the concepts entrenched in traditional design-to-manufacturing conduits, an advisor is simply defined here as an "intelligent" agent that uses abstracted knowledge from one domain to guide activity in the same domain or another domain.
Design advisors are computer agents that play an increasingly important, and anthropomorphic, role. Agents can at minimum be considered modular components of a system while ideally they act as independent entities. Agents can be used for information preparation, reducing the search space and manifesting the data in a more tractable manner to the user or other agents of the system. These agents may be of a numerical-analytical nature, knowledge based, or of any other model. They may be initiated by the user or autonomously invoke themselves, and they may be used to automate repetitive and mundane tasks.