Take the following text excerpted from a paper on the optimization of a hydraulic cylinder. While it might be relatively straightforward to extract the shape features from this drawing, we might have more difficulty relating the design objectives to the geometry without having read the design text. The point is, to fully comprehend a design, we need both the design text (memos, specifications, analyses) and the design geometry. This gives rise to the following question: would it be possible to learn the engineering function of a design based upon the design text just as we can learn manufacturing knowledge from the design geometry? The answer to this question is the thrust of my research project on the discover of terminological patterns in design text to act as a basis for building a meaningful model of the design.
Transition: Before I jump into a description of the learning strategy, it would be worthwhile to explore how the technology of learning design could be put to use in an industrial setting of engineers doing design.
| BEST Lab | Department of Mechanical Engineering | University of California at Berkeley