Application of neural network and genetic algorithms in the machining-parameters optimization for WEDM

• Main Authors: Tsai, Tzyy-Chyi, 蔡子琦
• Other Authors: Y. S. Liao
• Format: Others
• Language: zh-TW
• Published: 1999
• Online Access: http://ndltd.ncl.edu.tw/handle/73213481525401675609

碩士 === 國立臺灣大學 === 機械工程學研究所 === 87 === In the processing of WEDM, choosing appropriate machining parameters is a vital job for obtaining expected productivity and quality. Since the machining parameters are not only of great number, but also interacted between themselves, it is obvious that only experienced and skilled machine operators are capable of handling the job. Although the manufacturers of the WEDM machine usually provide the users with a set of machining-parameter table, it doesn''''t always serve the needs of the users, because the table is based on hundreds of thousands costly experiments merely under specific conditions. To save the money of the manufacturers and the troubles of the users, this paper has proposed a effective and efficient procedure to obtain the optimal machining parameters. We first used Taguchi quality design method incorporated with analysis of variance to figure out the influence of every parameter. With the information and data collected from the experiments, we therefore established a neural network which could provide accurate estimation. Then by utilizing genetic algorithms, the optimal combination of the machining parameters was obtained. In our study, the optimization of machining parameters was conducted with different thickness of the work pieces. The machining parameters discussed were pulse-on time, pulse-off time, servo voltage, wire speed, flushing, are-on time, arc-off time, while the machining performances considered were gap width, surface roughness, table feed, sparking frequency, abnormal ratio(ratio of abnormal sparks to total sparks).