The conventional methods like linear or polynomial regression, despite their overwhelming accuracy on training data, often fail to achieve the same accuracy on independent test data. In this research, a comparative study of three different machine learning techniques (linear regression, random forest regression, and AdaBoost) is carried out to build predictive models for dry electric discharge machining process. Six different process parameters namely voltage gap, discharge current, pulse-on-time, duty factor, air inlet pressure, and spindle speed are considered to predict the material removal rate. Statistical tests on independent test data show that despite linear regression's considerable accuracy on training data, it fails to achieve the same on independent test data. Random forest regression is seen to have the best performance among the three predictive models.
TopIntroduction
Electrical Discharge Machining (EDM) is a non-traditional machining process used to process electrically conductive materials with the help of material erosion by spark bombardment on the surface. Melting and vaporization of material from the surface of part produces replicate of the tool on the part surface. A little gap is present between the tool and workpiece. The dielectric fluid is used to take away the material debris produced due to material removal from the surface and also cools the workpiece and tool. The electrolyte disposal after machining, fire hazard during handling, etc. are the major issue with the conventional EDM process. Dry EDM uses an envelope of gases like compressed air, argon, helium, nitrogen, oxygen, etc. in place of dielectric fluid, which makes it green machining process. The friction of the workpiece and tool due to dielectric, lesser heat affected zone, lesser residual stresses, any direction used for machining due to absence of dielectric tank, etc. makes dry EDM superior to conventional EDM. Dry EDM has the controlling parameters similar to conventional EDM such as applied voltage, current, pulse on and off time and frequency, the gap between tool and workpiece etc. Apart from these, types and conditions of gas used, the flow rate of gas etc. are the controlling process parameters, with typical output such as material removal rate (MRR), surface finish, overcut, etc. Apart from using gaseous medium solely in machining, combined dielectric fluid and gas can also be used in the EDM process and methodology is named as near dry-EDM process which combines advantages of the conventional and dry-EDM process (Kao, Tao, & Shih, 2007). In dry-EDM use of flat face electrode or tool leads to uncontrolled plasma extension hence the material removal rate reduces, also the material removed debris carrying becomes ineffective, which can be overcome by the use of different slotted electrodes (Puthumana & Joshi, 2011). Gholipoor et al. (Gholipoor, Baseri, & Shabgard, 2015) have compared the performance of near dry machining with the conventional and dry-EDM process in the drilling of SPK steels considering toll wear ratio, MRR and surface roughness and concluded that surface quality obtained by near-dry machining is better than other EDM processes. Also, the near-dry machining gives lower surface crack on the machined surface. Dhakar et al. (Dhakar & Dvivedi, 2016) evaluated the process parameters (i.e. applied current, gas pressure, duty factor, and lift of tool) in case of near dry machining of high-speed steels for MRR, tool wear rate and surface roughness.
Researchers are interested to study the process parameters of dry-EDM for different applications. Saha et al (Saha & Choudhury, 2009) studied the influence of the dry-EDM parameters on machining of mild steel such as gap voltage, discharge current, pulse-on time, duty factor, air pressure and spindle speed considering output response as MRR, surface roughness and tool wear rate and reported that applied current is the most significant factor followed by duty factor, air pressure and machining speed. Masanori Kunleda et al. (Kunleda, et al., 2003) investigated dry-EDM milling process parameters of steel specimen for MRR and observed that when the applied electric power goes beyond certain values the MRR increases drastically over the surface because of the chemical reaction between the oxygen gas and object material. Islam et al. (Islam, Li, & Ko, 2017) studied the process parameters of deburring of drilled holes in carbon fibre reinforced plastic (CFRP) using dry-EDM process with remarks that dry EDM with oxygen gas produces better machining than using air. Also, the dry EDM is more productive than that of conventional EDM. Jia Tao et al. (Tao, Shih, & Ni, 2008) studied experimentally the process parameters of dry and near dry-EDM mailing for maximum MRR and minimum surface finish.
Optimization of dry-EDM process parameters for best output is the prime concern of analysts. Pragadish et al. (Pragadish & Kumar, 2016) optimized the process parameters of dry-EDM such as applied current, ON time of pulse, applied voltage, gas pressure and cutting speed for maximum MRR and minimum surface finish with remarks that the applied current is the most significant process parameter.