Reasoning with geometry: Predicting stress concentration factors

Fundamental to any AI-based design and analysis tool will be an understanding of geometrical form and function. This paper describes some current research into representing automotive components, and reasoning about such representations in the domain of stress concentration analysis. The process of predicting stress concentration is described and identified as a suitable subject for knowledge-based geometrical reasoning. A broadly feature-based representation is derived which attempts to model the conceptual representation maintained by an analyst. Object-oriented and rule-based methodologies are applied to this representation, and two different approaches in the content and style of the representation are highlighted. Some examples of the reasoning process applied to the representation are presented, which demonstrate the effectiveness of this technique in its ability to manipulate and simplify geometrical descriptions, based on an understanding of their significance in engineering terms.