The Study of Artificial Neural Network with Taguchi Method on Surface Characteristics of the Micro-Hole EDM for the High Tensile Steel Material

• Main Authors: Tzong-hong Wu, 吳宗鴻
• Other Authors: Pei-tsang Wu
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/50156583387560225259

碩士 === 義守大學 === 工業工程與管理學系碩士班 === 97 === The study of this research is to study the characteriss of the HY80 high tensile stell material. The HY80 high tensile steel material is employed to work piece and then precedes the machining process of micro-hole EDM. In the beginning, we selecte pulse current(Ip), pulse-on duration(Ton), pulse-off duration(Toff) and gap voltage(Vg) as the main influencing factors of micro-hole EDM. Three output factors, such as material remove rate(MRR) (mg/min), electrode remove ratio(ERR)(%), hole enlargement(HR)(mm), are selected for the machining parameters of micro-hole EDM to obtain the 72 groups of experimental values. And the relation between machining characteristics and combinations of the above machining parameters with orthogonal away L9 of Taguchi method is investigeted. The results of Taguchi Method are further confirmed by 19 groups of the single-factor experimental method to research the relation between the quality of the products and machining characteristics. Meam while, the predicting models by developing the artificial neural networks are established with the 72 groups of experimental values with training data. Then, they are tested with the 9 groups of experimental parameters from the above Taguchi Method, and proved with the 19 groups of the single-factor experimental parameters. The material remove rate(MRR), electrode remove ratio(ERR), and hole enlargement(HR) from the predictions of the artificial neural networks are obtained. We compare these outputs with the above experimental values and get the error rates. We believe the error rates from the artificial neural networks are smaller than from the Taguchi method. The rates from the artificial neural networks are clearly proved to predict the machining process of the micro-hole EDM. The results of this research can provide the consultation to the affiliated industry while machining.