Abstract:
This research studied using curvature and support vector machines (SVMs) to identify manufacturing processes of mechanical parts. Prior developments in computer vision and graphics has been able to recognize shapes of 2D and 3D models. This work contributes a technique for recognizing manufacturing processes from CAD models. Prismatic machined and cast-then-machined are two typical processes for making mechanical parts; being able to recognize them respectively assists in manufacturing cost estimation. This work integrates surface curvature estimation (from computer graphics) and machine learning to perform manufacturing recognition of artifacts. Surface curvature is used as the discriminating feature; SVMs learn the separation in between the two manufacturing processes.