Inferring Hardness From High-Speed Video of the Machining Process

This paper presents the results of a study to assess the feasibility of inferring workpiece material hardness from high-speed video data of chip formation obtained during a turning operation. The motivation for assessing hardness in situ comes from the fabrication of shaped charges, where spatial variation in hardness is known to affect the performance of the shaped charge. While other in-process data could be used for this purpose, video data are analyzed here because of the stand-off, non-contact advantages afforded. This is especially relevant for highly qualified machining processes for small-lot, high value parts where any interference with the process (e.g., introduction of cables near the machine tool) is undesirable. A multistep image processing procedure is presented which is used to extract several features from the video data. These features are then used to develop a classifier which can be used to predict work-piece hardness. Multiple classifier designs (Knn and Ratchet) are considered.