TY - JOUR
T1 - Multi-objective optimization framework for five-pass wire-drawing process
AU - Wang, Jen Hung
AU - Lo, Yu Lung
AU - Wang, Hung Yu
AU - Tran, Hong Chuong
N1 - Funding Information:
This study received financial support by the Ministry of Science and Technology of Taiwan under Grant No. MOST 107-2218-E-006-051. The study was also supported in part by the Ministry of Education, Taiwan, through funding provided to the Intelligent Manufacturing Research Center (iMRC) at National Cheng Kung University (NCKU).
Publisher Copyright:
© 2020, Springer-Verlag London Ltd., part of Springer Nature.
PY - 2020/4/1
Y1 - 2020/4/1
N2 - A systematic method is proposed for optimizing the die geometry and processing conditions of a multi-pass wire-drawing process. In the proposed approach, FEM simulations based on the robust Taguchi design method are first performed to determine the drawing force, maximum surface axial stress, and maximum die stress for given values of the reduction angle, bearing length, and back tension properties of the die. The Taguchi analysis results are then used to train a neural network (NN) to predict the wire-drawing outcome for any given values of the input parameters. Finally, a genetic algorithm (GA) based on the NN is employed to determine the optimal settings of the input parameters. The validity of the GA optimization results is confirmed by means of FEM simulations. A simple method is proposed for visualizing the optimization results in 3-D space. The feasibility of the proposed framework is demonstrated by means of a practical case study involving a five-pass wire-drawing process for AISI 1022 low carbon steel.
AB - A systematic method is proposed for optimizing the die geometry and processing conditions of a multi-pass wire-drawing process. In the proposed approach, FEM simulations based on the robust Taguchi design method are first performed to determine the drawing force, maximum surface axial stress, and maximum die stress for given values of the reduction angle, bearing length, and back tension properties of the die. The Taguchi analysis results are then used to train a neural network (NN) to predict the wire-drawing outcome for any given values of the input parameters. Finally, a genetic algorithm (GA) based on the NN is employed to determine the optimal settings of the input parameters. The validity of the GA optimization results is confirmed by means of FEM simulations. A simple method is proposed for visualizing the optimization results in 3-D space. The feasibility of the proposed framework is demonstrated by means of a practical case study involving a five-pass wire-drawing process for AISI 1022 low carbon steel.
UR - http://www.scopus.com/inward/record.url?scp=85084133127&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85084133127&partnerID=8YFLogxK
U2 - 10.1007/s00170-020-05226-9
DO - 10.1007/s00170-020-05226-9
M3 - Article
AN - SCOPUS:85084133127
SN - 0268-3768
VL - 107
SP - 3049
EP - 3063
JO - International Journal of Advanced Manufacturing Technology
JF - International Journal of Advanced Manufacturing Technology
IS - 7-8
ER -