Title: Using neural connection and management of the die design database

Authors: Jui-Chang Lin, Kingsun Lee

Addresses: Department of Mechanical Design Engineering, National Formosa University, 632 Yunlin, Taiwan. ' Department of Mechanical Engineering, ChienKuo Technology University, 500 Changhua, Taiwan

Abstract: This research uses neural networks to manage the primitive technological database and link to analytic data of mould. The relative theory was established by means of linking network; it can be extended to the case of design by the network, and to reduce a lot of design time. This research used design theory to experiment and to set up 27 sets of 3D-CUP contours of database of the drawing die, its design parameter includes punch/die dimension, punch/die radius on different material thickness. The 27 sets of die drafts were drawn by the CAD software, and then to simulate actual design situation with database of design parameter by the FEM software. The relationship between database and network can be established through the training for linking the network. The database after linking can predict directly the processing situation in any design parameter without the route of CAD and FEM. The error of this research is about 6%, it can be less if it collects more design information.

Keywords: deep drawing; 3D square cup; neural networks; simulated annealing; die design database; CAD; simulation; design parameters; FEM; finite element method.

DOI: 10.1504/IJMTM.2007.011857

International Journal of Manufacturing Technology and Management, 2007 Vol.10 No.2/3, pp.323 - 341

Published online: 28 Dec 2006 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article