Development of Multi-Objective Prediction Model for Wire Electrical Discharge Machining of Inconel 718 Nickel-Based Superalloy

• Published in: Key engineering materials 2018-11, Vol.789, p.37-44
• Main Authors: Yang, Ching Been, Lin, Cang Ge, Zhan, Jia Lin, Chiang, Hsiu Lu
• Format: Article
• Language: English
• Subjects: Artificial neural networks; Electric wire; Machining; Mathematical models; Multiple objective analysis; Neural networks; Nickel base alloys; Parameters; Superalloys; Taguchi methods; Training
• ISSN: 1013-9826; 1662-9795; 1662-9795
• DOI: 10.4028/www.scientific.net/KEM.789.37

Owing to the fact that conventional Taguchi methods cannot predict the experimentresults of non-level values, this study developed a staged Taguchi neural network prediction modelthat combines the merits of experimental data from the Taguchi method and the learningcapabilities of artificial neural networks. We first used the optimal parameter combinations derivedfrom the L9 orthogonal array experiment data and grey relational analysis as training examples forthe Stage-1 network to construct a preliminary network. Next, we used the crucial factors in theoptimal parameter combinations derived from the grey relational analysis as additional trainingexamples for the Stage-2 network. The results of the staged Taguchi neural network predictionmodel indicate that the prediction performance of the preliminary network in Stage 1 was poor dueto an insufficient number of training examples, while the Stage-2 network produced excellentprediction results.