This grant provides funding for research that will result in techniques for automatically computing the purposes of the geometric features on the parts of a mechanical device. These purposes serve as a form of design documentation. This work will employ a combination of techniques from computer graphics, artificial intelligence, and mechanical engineering. The first step in the process is identifying the features on a part. This will be accomplished by creating new model simplification algorithms. These algorithms will remove the small geometric details (protrusions, cavities, etc.) from a part to reveal the simple underlying geometry. Subtracting this simplified geometry from the original geometry will reveal the set of features. To explain the purpose of a particular feature, a simulation of the device with the feature removed will be compared to a simulation of the original device. The differences in the simulations indicate which of the device's many behaviors are produced by that feature. Causal reasoning techniques will then be used to construct an explanation for how the feature causes its behaviors. These explanations, which constitute one of the purposes of the feature, will be automatically translated into natural language (English) and will then be incorporated into the device's documentation. If successful, the results of this research will lead to advanced design documentation tools which will increase the scope and completeness of design documentation while at the same time reducing the cost of producing documentation. The particular kind of rationales computed by this approach (the purposes of features) will facilitate a variety of common engineering tasks including coordinating collaborative design and enabling product redesign without creating unintended side-effects, thereby reducing design cost. Furthermore, if successful, this research will produce fundamental advances in model simplification and causal reasoning.
