Net Shape

Manufacturing

Laboratory

NSMLAB

KSME 2022Preform design method for uniform strain distribution based on convolution neural network


Author
Seungro Lee, Kyungmin Kim and Naksoo Kim
Title
Preform design method for uniform strain distribution based on convolution neural network
Conference
KSME 2022
Month/Year
November / 2022
Location
Gwanju, South Korea
Abstract
Preform design is important for full filling of the die and uniform deformation in forging processes. In this study, a preform design methodology for uniform strain distribution is introduced. The proposed method is based on a convolution neural network (CNN) algorithm. By convolution operation with weight arrays, the model extracts geometrical features of the forging product, forging simulation result, and strain distribution, and connects those features with the corresponding preform. For the different forging products, the saved weight arrays can extract the characteristic features, so that proper preform shape can be easily acquired without any iterations. The proposed design method is utilized for the H-shaped forgings and validated through numerical experiments for rail wheel forging and disk forging. The proposed method has shown to be reliable in preform design procedures, granting uniform deformation with full filling in forging processes.
Status of Conference
Published

Department of Mechanical Engineering, Sogang University, Seoul, Korea.
Tel:+82-070-7439-8635     Fax:+82-2-712-0799

Copyright 2015 NSMLAB. All right reserved.