MODELING OF SURFACE ROUGHNESS OF ELECTRO-DISCHARGE MACHINED COMPONENTS

Abstract

Titanium alloys and Superalloys are the important material for space applications because of their compatibility with liquid hydrogen, liquid oxygen and high specific strength. Machining of these materials is more difficult because the elements (Ni, Co, Cr, Ti) that give these materials their high strength and corrosion resistance also give trouble while machining. Typically the higher the nickel content, the more difficult it is to machine because the nickel or cobalt content or the titanium constituent imparts the material tendency to work harden. Additionally because of superalloys characteristic of maintaining their hardness and strength at high temperatures, they do not break during the cutting process. All these encourage the need to adopt some unconventional machining methods to machine these materials (Inconel and Titanium alloy). EDM has been proposed as an effective solution to this problem because machining by this process is depends on thermal properties of the material (like thermal conductivity) which is more in case of Inconel as compared to stainless steel. In the present work, modeling of process parameter for predicting surface finish and material removal rate of Inconel-718 and Titanium alloy (Ti-6A1-4V) is carried out. Detailed experimental investigation has been carried out to study the effect of four process parameter i.e. pulse-on time, pulse-off time, peak current and flushing pressure on material removal rate (MRR), average surface roughness and maximum surface roughness have been studied. Pilot experiments were conducted using fractional factorial design (34-1) to know the range of the process parameters for main experiment. Main experiments were carried out using response surface methodology (RSM) approach to formulate the regression model. Response surfaces are plotted for different responses. Interaction effects between process parameters have been studied. Validation of regression models also have been made by conducting confirmation/verification experiments. It has been observed that surface roughness and MRR both are increasing with increase in pulse-on time, peak current. To achieve good surface finish pulse-on time and peak current value should be low. MRR and surface roughness both are increasing with rm pulse-off time upto 37.5 ps afterwards it shows little decrease in case of Ti-6A1-4V, whereas for Inconel-718 it shows same increasing trend. MRR and surface roughness shows increasing trend with flushing pressure upto 30 kg/cm2 beyond that both are decreasing. Combined effect of pulse-on time and peak current shows net increase in MRR and surface roughness for both the material. No significant interaction effect between process parameters has been observed for both the material. Artificial Neural Network (ANN) modeling has been done to model the surface roughness (Ra and Rt) and MRR. Multilayer Feed Forward (MLFF) network with Back Propagation (BP) learning technique is used to approximate the non-linear functional relationship between input and output process parameters. Training of ANN has been done using the results of the main experiments, while the data of four confirmation experiments have been for simulation or testing of the ANN. Simulated results by ANN show good agreement with experimental and regression model results. Using these models, pre-adjustments can be made to get the final product of the desired quality at minimum cost.