| Failure Modes and Effects Analysis (FMEA) software has not significantly exploited artificial intelligence (AI) techniques, nor has the knowledge possessed by a team of FMEA experts been thoroughly investigated. Attempts have been made to automate the FMEA process; however, these implementations and techniques have utilized little or no AI methods. Instead, they have emphasized clerical functions, data collection, database manipulations, and automatic report generation. This dissertation investigates the knowledge organization and distribution of an expert system for failure modes and effects analysis (XFMEA). The primary contribution of this work is the specification of the organization, distribution, and nature of the requisite, problem-solving knowledge necessary to migrate computer-aided FMEA programs toward an AI approach.; The design of the XFMEA system is approached from a knowledge-use level perspective to provide a complete understanding of the problem. The blackboard paradigm is utilized to functionally decompose XFMEA into a set of knowledge sources, each containing the knowledge associated with a subfunction of the FMEA task. This research seeks to establish the theoretical background that is required for integrating and traversing diverse knowledge representation formalisms and inference procedures which organize and focus a simulation subsystem for XFMEA. Functional FMEA methodology is emphasized with the motivation to include FMEA as an integral constituent in the design process. Furthermore, the research argues that employing a blackboard framework as the computational construct for XFMEA is a key to successful automation. The research specifies an experimental XFMEA framework for relating a failure mode to an effect. Experiments are conducted that illustrate the utility of the developed theory. Although pragmatic extensions to the proposed paradigm are necessary to build a significant prototype system; the organizational model, problem-solving heuristics, and symbol-level details developed in this research provide a substrate for building knowledge-based, computer-aided FMEA systems. |
