Optimization with multiple objectives in multiple passes of wire drawing process

Abstract

Wire drawing is a plastic deformation process to gradually reduce the diameter of the wire rod During the process a wire is drawn to pass through either a die (single-pass process) or a series of dies (multi-pass process) Reduction ratio distribution is a crucial part of design in multi-pass process Geometric designs of dies also affect the quality of product and the cost of energy Additionally the application of back tension can substantially reduce the maximum pressure of dies and incidentally extend service life of dies In this research a five passes steel (AISI 1022 Alloy) drawing process from 5 5mm to 2 75mm in diameter is designed and optimized The drawing velocities in every pass were deliberately controlled in Finite Element Method Taguchi Orthogonal arrays (L_16 (4^3)) was adopted to build up the parameter combination in different FEM cases Artificial Neural Networks (ANNs) were used to train a general model after simulations were all finish in the specific pass Then Genetic Algorithm (GA) was employed to find the best combination of parameters in the specific pass based on the model generated by Artificial Neural Networks (ANNs) There were three optimization targets in this study— 1 The drawing force relative to the cost of energy 2 The maximum axial stress on wire surface to maintain decent quality of product 3 The maximum stress on dies which influences the wear of dies For instance it is a multiple objective task In general the concept of Meta Model was taken into practice since the FEM was substituted for the real-world experiments and then the ANN model was trained to find a general solution of FEM The process of optimization must start from the first to the fifth pass in sequence due to the nonuniform deformation The parameters of previous passes need to be fixed in order to create the similar nonuniform deformation and residual stress distribution In case the parameters of previous passes alter different results appear At last a mechanism was invented to operate the optimization results

Date of Award	2019
Original language	English
Supervisor	Yu-Lung Lo (Supervisor)