Finite Element Based Modeling of Surface Roughness in Micro Electro-Discharge Machining Process

1. Introduction

The demand of miniaturized products has been increased due to their ability to provide more functions in a limited space. The miniaturized products require the production of components with features in the range of 1 to 999 µm (Masuzawa, 2000). Although these components could be manufactured by various methods, their shaping through removal of excess material is termed as micromachining. Micro Electro-Discharge Machining (micro-EDM) is widely accepted unconventional micromachining process having ability to machine electrically conductive materials. This process is also referred as electro-discharge micro machining (EDMM) or µ-EDM. It is used for the production of various components such as fuel injection nozzles, micro-mechatronic actuator parts, electronic and optical devices, micro tools etc. (Liu et al., 2010). The micro-EDM process is an adaptation of macro or conventional Electro-Discharge Machining (EDM) for micromachining (Rajurkar et al., 2006). The performance of a component micromachined by micro-EDM is affected by the quality of the surface produced during the process. Surface roughness is an important quality characteristic of a micromachined component. The availability of a model for the prediction of surface roughness in micro-EDM process helps not only in the improvement of surface characteristics but also in the reduction of machining cost.

Many efforts have been made to model surface roughness due to EDM process using empirical and analytical methods. Jeswani (1978) presented an empirical model of EDM to estimate the surface roughness based on experimental results for tungsten carbide, high speed tool steel, high-carbon steel and mild steel workpieces. The effect of discharge energy and material properties of electrodes on surface roughness was also studied. Mamalis et al. (1987) presented an empirical model of EDM for the surface roughness using low-carbon steel and alloyed steel workpieces. They investigated the effect of pulse current, pulse duration and pulse energy on surface roughness. Tsai & Wang (2001a) presented a semi-empirical model of EDM for the surface roughness using peak current, polarity, pulse duration and material properties. The average prediction errors were found to be less than 12 percent. Tsai & Wang (2001b) subsequently developed neural-network models for the prediction of surface roughness. The experimental data based on Design of Experiment (DOE) was used for the training purpose. An average prediction error was found to be 8.58 percent. Petropoulos et al. (2004) presented single and multiple regression models of EDM for surface roughness using Ck60 steel plates. Keskin et al. (2006) did the experimental investigations on the surface roughness of steel workpieces for EDM process. An equation for surface roughness was obtained by multiple regression. Bhattacharyya et al. (2007) presented a model of EDM for surface roughness based on response surface methodology. The influence of peak current and pulse-on time on surface roughness was studied for die steel using this model. Pradhan et al. (2009) proposed neural network models of EDM based on experimental data for the prediction of surface roughness. It was found that back-propagation network provided better performance compared with the Radial basis function network. Salonitis et al. (2009) proposed an analytical model of EDM for the determination of surface roughness and material removal rate and they found that predictions were in good agreement with experimental results.