1993). This capability is an advance in that the author has pre-programmed expected search requests.

A higher level of sophistication of electronic representation of technical memory would be the equivalent of an educated assistant or technician that can search the library and retrieve the pertinent information in the background without direct user involvement. Representing the knowledge that an electronic assistant contains and representing the knowledge that the user requires are two different problems, however. For the materials selection problem in structural design, the intelligent electronic assistant would have to understand, at some level of competence, the information provided by the human experts. That information would consist not only of concepts used in materials science and engineering but concepts related to the entire life-cycle of design, manufacturing, inspection, and disposal or recycling. In the ultimate case, the electronic assistant will have to know the languages of many different pertinent databases and then be capable of representing that knowledge in a consistent form. This leads to the need for the intelligent integration of information from these multiple sources.

The highest level of sophistication envisioned would provide a full-function, computer-aided electronic assistant or technician who could not only find the correct reference material but also apply the results to the query to the design problem at hand. Just as we can imagine human assistants of different levels of skill, so we can also imagine electronic assistants at different levels of utility. As mentioned earlier, full automation of databases or knowledge bases to perform the complete design task is not currently feasible. However, certain select routines or computationally intensive tasks now performed by people can be performed by electronic assistants to provide particular advice or to critique selected aspects of the design. For instance, there has been considerable research into Agent Technology, whereby a user can specify an agent to roam the Internet to obtain appropriate information (ACM, 1994). This technology requires the use of data dictionaries or mediators, however, to recognize and translate the relevant information in different databases and knowledge bases and to integrate possibly conflicting data from multiple sources. This level of capability has been shown to greatly benefit the overall performance of the design teams in several limited instances (Klahr et al., 1987; Famili et al., 1992). For example, the three areas of expertise—product design, materials selection, and manufacturing—cannot be entirely separated for highly engineered products. Materials properties depend to an extent on the processing route, and processing considerations can be influenced by design constraints. Similarly, the design must reflect the reality of available material properties, and the properties are not completely independent of the design application (e.g., high loading rates can reduce a material's fracture toughness).

Reliance on standard databases that contain physical and mechanical properties is inadequate to support materials selection processes fully. Knowledge bases are required that capture the advantages and limitations of materials, their processability, and their application histories—all of which are critical to the design process. There are problems related to the definition, development, and construction of knowledge bases, however.

While much has been written about knowledge bases, there is little agreement on the scope of what exactly constitutes one. There are wide variations in the literature describing designs and implementations of computer-aided systems.

Many companies have custom-built basic systems for their own applications. Table 4-1 summarizes representative computer-aided system application areas that relate to the materials information used in the design process in a manner consistent with the vision discussed in Chapter 3 . These applications are considered state of the art in the sense that examples can be found either in use or under development at major government and industrial sites.

Table 4-1 shows that there is a wide breadth of knowledge base applications. The list is also incomplete, since it only represents what is currently possible in the design and engineering phase of product development. If the scope were broadened to other phases of the product life-cycle, more applications could be listed that would require knowledge (e.g., diagnosis of the manufacturing process). Further work is required to determine what constitutes a knowledge base and how it differs from simple databases.

Front Matter (R1-R14)

Executive Summary (1-6)

1 Introduction (7-10)

2 Materials Selection in Structural Design (11-18)

3 Enhancing the Materials Selection Process in Design: A Vision (19-24)

4 Information Technologies Pertinent to the Materials Selection Process (25-36)

5 Conclusions and Recommendations (37-40)

References (41-44)

Appendix A: Glossary of Acronyms (45-46)

Appendix B: Case Studies Reviewed by the Committee (47-48)

Appendix C: Review of Selected Knowledge-Representation Techniques and Tools (49-56)

Appendix D: Knowledge-Based Integrated Design System (57-62)

Appendix E: An Intelligent Knowledge System for Selection of Materials for Critical Aerospace Applications (63-68)

Appendix F: Biographical Sketches of Committee Members (69-69)