A large number of tools in turning-milling machines run the risk of collisions during the machining process owing to their wrong disposition, mistakes in their recognition, and lack of proper communication feedback. As the system is unable to intelligently identify the tools, it fails to avoid collisions, already from the early steps of machining. This study is aimed at developing a fast, precise, and robust automatic identification method for cutting tools in turning-milling machines. To classify the large number of types of tools in those machines, the applicability of deep convolutional neural networks is explored, employing images of the tools as data input. Subsequently, feature detection, description, and matching are performed to improve accuracy. In the second phase, on-machine dimension measurement is performed by utilizing a contact-based displacement sensor with considering the output of identification phase. The proposed approach results in high accuracy of tool identification and accurately measures the correct dimensions of the tools.
|ジャーナル||Seimitsu Kogaku Kaishi/Journal of the Japan Society for Precision Engineering|
|出版ステータス||Published - 2019 1 1|
ASJC Scopus subject areas
- Mechanical Engineering