Identifying manufacturability and machining processes using deep 3D convolutional networks. (April 2021)

• Record Type: Journal Article
• Authors: Peddireddy, Dheeraj; Fu, Xingyu; Shankar, Anirudh; Wang, Haobo; Joung, Byung Gun; Aggarwal, Vaneet; Sutherland, John W.; Jun, Martin Byung-Guk
• Is Part Of: Journal of manufacturing processes. Volume 64(2021)
• Page Start: 1336
• Page End: 1348
• Publication Date: 2021-04
• Subjects: Deep learning -- CAD data -- Machining -- 3D voxel representation -- Transfer learning; Production management -- Data processing -- Periodicals; Manufacturing processes -- Periodicals; Procestechnologie; Productietechniek; Production -- Gestion -- Informatique -- Périodiques; Fabrication -- Périodiques; Manufacturing processes; Production management -- Data processing; Periodicals; 670.5
• DOI: 10.1016/j.jmapro.2021.02.034
• Languages: English
• ISSNs: 1526-6125

Abstract

Graphical abstract: Highlights: Novel machining process identification system using convolutional neural networks and transfer learning. Classifying machinability of synthetic workpieces using deep learning. Feature recognition using convolutional neural networks. Abstract: Today's manufacturing industry still requires human labor and knowledge for Machining Process Identification (MPI). Before manufacturing a part, engineers need to judge the manufacturability and identify the machining processes according to CAD model, and then the customer needs to source for qualified suppliers based on the identified machining processes. In order to realize an efficient automated industrial logistic system, developing a competent Machining Process Identification system becomes important. In this paper, a novel two-step MPI system is presented based on 3D Convolutional Neural Networks (CNN) and transfer learning. The proposed system admits triangularly tessellated surface (STL) models as inputs and outputs the manufacturability of the CAD design and machining process labels (e.g., milling, turning) as the results of classification from the neural networks. Computer-synthesized workpiece models are utilized in training the networks. In addition to the MPI system, a pre-trained framework was developed for future applications in related fields. The MPI system shows more than 98% accuracy in identifying manufacturability of a part and about 98% accuracy in identifying the manufacturing processes of synthesized workpiece models which validates the robustness of the model.