8th North America Conference on Industrial Engineering and Operations Management

Turret Punch Time Estimate for Sheet Metal Parts using Machine Learning Approach

Yearn-tzuo Hwang, Ting-Yi Zhang & Hung-Chun Hsu
Publisher: IEOM Society International
0 Paper Citations
1 Views
1 Downloads
Track: Manufacturing
Abstract

Turret punch machines have been used to produce parts in sheet metal shops since 1950s. It is common to place multiple parts of different sizes and designs from several customers on one sheet blank to manufacture. However, this practice makes it difficult to calculate the fabrication time (and therefore subsequent manufacturing cost) of each part because turret punch machines only display the total machine run time per entire sheet blank, and yet the punching process is based not only on the parts themselves but also on the punching tools used and punching routes undertaken. This paper proposes a Machine Learning approach to solve this longstanding punching time distribution problem by using the punching tools’ hitting counts, the moving distances of sheet blanks and the number of tool changes during production time. 425 sheet blanks of parts with various designs and sizes are processed, and machine run times for each sheet blank are recorded from a turret punch simulation program. Five common types of turret punching operations (“hit”), such as nibbling, single hit, line shearing and etc., along with transverse and longitudinal moving distances of sheet blanks and tool changes are considered in this research. Every part on the sheet blanks is broken down into its component geometric entities (e.g. line, arc, hole, etc.). Geometric entities are then grouped by the kind of punching hit and/or assigned punching tool itself. Three supervised Machine Learning models, namely Linear Regression, Ridge Regression, and Lasso Regression are used and compared with 10-fold cross validation process each consisting of 383 (90%) sheet blanks for training and the remaining (10%) for validation. The preliminary results indicate that the proposed Machine Learning approach is viable and consistent to determine the fabrication time for every part on sheet blanks manufactured by turret punch machines. It can be integrated with real-time data collection of the counts of punching hits, parts’ geometry, sheet blank’s motion and machine run time on any turret punch machines to determine the fabrication time and subsequent production manufacturing cost of each part in the Industry 4.0 era.

Published in: 8th North America Conference on Industrial Engineering and Operations Management , Houston, United States of America

Publisher: IEOM Society International
Date of Conference: June 13-15, 2023

ISBN: 979-8-3507-0546-1
ISSN/E-ISSN: 2169-8767