Knowledge based engineering analysis

This dissertation is an investigation of the application of knowledge-based programming to engineering analysis. The goal of the work is to increase the availability of specialized analytic knowledge to mechanical designers. Conventional computer-aids are procedural, which means that the knowledge they contain cannot be directly inspected. We may conclude that procedural programs contain knowledge only by observing their actions. Since the knowledge is less visible in procedural analysis tools, expert users, who have confidence in the tool, are required. Designers typically cannot afford to become experts in using a particular analysis tool because they often have a very broad range of activities. In contrast to procedural knowledge, knowledge encoded in knowledge-based programs is declarative. Knowledge built on declarative representations is readable by designers as well as by computers, making it easier to understand and trust computed results. In order to investigate the applicability of the knowledge-based approach to engineering analysis, a declarative program for rigid-body mechanics was built. It was learned that metalevel control is an important part of such a program, and several ways of applying it are discussed. Incremental model building is another way in which the feasibility of the knowledge-based approach is improved. It was also discovered that knowledge-based analysis tools can be used to model devices from several viewpoints using contradictory physical theories. Several ways of resolving the contradictions are covered. From this work, we conclude that engineering analysis tools built on a knowledge-based foundation are feasible, and that among many other advantages, this approach can be used to model devices from different viewpoints.