Wendy Dannels reminded me about a 3D search engine developed by 3-D Seek. Search Engine Journal reports that the system now handles six thousand 3D objects, like hinges, bolts, conveyor belts, motors, and so on.
The idea is that you doodle an image of the part you're looking for, and the search engine tries to make a match. You can try it out at 3d-seek.com.
It is an admirable attempt. But I think it may suffer from the same problems as speech input and handwriting input -- it's hard for computers to understand vague and inconsistent human notions (such as grammar and sketching) and may prove to be less efficient than using the keyboard to find a part.
OTOH, I wouldn't mind such a service. For example, I see a European road sign that makes no sense to me. I should be able to photograph it or sketch it, and submit it to a search engine, which tells me it meaning. But unlikely. Even Google's Images database works by examining text surrounding the picture, not the picture itself.