Prediction of wire-EDM Process Parameters for Surface Roughness using Artificial Neural Network and Response Surface Methodology

Abstract

The prediction of machining process capabilities is important in process parameter optimization and improvement of machining performance characteristics. This paper, presents the prediction of Wire-EDM input parameters for surface roughness using Artificial Neural Network and Response Surface Methodology. Ti-6Al-4V is an alpha-beta alloy widely applied in industry due of its excellent combination of mechanical properties. However, this alloy is found to be difficult to machine by means of conventional machining processes because of its high melting temperature, high chemical reactivity, and low thermal conductivity. Nevertheless, nonconventional machining processes such Wire-EDM are able to overcome the challenge in machining Ti-6Al-4V. Response Surface Methodology (RSM) based on Central Composite design is used to evaluate and optimize the effect of pulse on time (Ton), discharge current (I) and open circuit voltage (UHP) on surface roughness (SR). Analysis of Variance revealed that open-circuit voltage is the most significant parameter affecting the obtained surface roughness followed by the discharge current. Parametric variation shows that lower surface roughness can be obtained at lower levels of UHP and I. The main contribution of this paper is the prediction of wire-EDM machining process parameters for a given surface roughness using Artificial Neural Network (ANN). The developed ANN model revealed to be 97.155% accurate with an average prediction error of 2.845%. The predictive capability of the developed ANN model is found to be satisfactory and the model can be successfully used for predicting machining process parameters for desired surface roughness in wire electrical discharge machining process.