A neural network/expert system approach for design improvement of productsmanufactured by EDM

The stochastic nature of the electro discharge machining (EDM) process does not allow for a precise prediction of its effect on the machined features. However, there is a direct interrelation between feature design and the pr ocess results. The objective of this work: is to suggest a neural network b ased system to facilitate and optimize the design process of products to be machined by EDM. A comprehensive analysis by a neural network and expert s ystem is presented. Aspects of features coding and relations with the proce ss parameters are discussed. Experimental results confirm design improvemen ts and a practical system is described.