Intelligent approach for process modelling and optimization on electrical discharge machining of polycrystalline diamond

Abstract

Polycrystalline diamond (PCD) is increasingly becomes an important material used in the industry for cutting tools of difficult-to-machine materials due to its excellent characteristics such as hardness, toughness and wear resistance. However, its applications are restricted because of the PCD material is difficult to machine. Therefore, electrical discharge machining (EDM) is an ideal method suitable for PCD materials due to its non-contact process nature. The performance of EDM, however, is significantly influenced by its process parameters and type of electrode. In this study, soft computing technique was utilized to optimize the performance of the EDM in roughing condition for eroding PCD with copper tungsten or copper nickel electrode. Central composite design with five levels of three machining parameters viz. peak current, pulse interval and pulse duration has been used to design the experimental matrix. The EDM experiment was conducted based on the design experimental matrix. Subsequently, the effectiveness of EDM on shaping PCD with copper tungsten and copper nickel was evaluated in terms of material removal rate (MRR) and electrode wear rate (EWR). It was found that copper tungsten electrode gave lower EWR, in comparison with the copper nickel electrode. The predictive model of radial basis function neural network (RBFNN) was developed to predict the MRR and EWR of the EDM process. The prominent predictive ability of RBFNN was confirmed as the prediction errors in terms of mean-squared error were found within the range of 6.47E−05 to 7.29E−06. Response surface plot was drawn to study the influences of machining parameters of EDM for shaping PCD with copper tungsten and copper nickel. Subsequently, moth search algorithm (MSA) was used to determine the optimal machining parameters, such that the MRR was maximized and EWR was minimized. Based on the obtained optimal parameters, confirmation test with the absolute error within the range of 1.41E−06 to 5.10E−05 validated the optimization capability of MSA.