Material Removal and Dimensional Accuracy in Electrical Discharge Turning of AISI 4140 Steel Under Lateral Flushing and Full Immersion


Palabıyık G. C., Özerkan H. B., Subaşı M., Mavi A., Ayhan E., Aydoğdu A. B.

Machines, cilt.14, sa.11, ss.1-22, 2026 (SCI-Expanded, Scopus)

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 14 Sayı: 11
  • Basım Tarihi: 2026
  • Doi Numarası: 10.3390/machines14111360
  • Dergi Adı: Machines
  • Derginin Tarandığı İndeksler: Engineering Source (EBSCO), Scopus, Technology Collection (ProQuest), Science Citation Index Expanded (SCI-EXPANDED), Directory of Open Access Journals
  • Sayfa Sayıları: ss.1-22
  • Gazi Üniversitesi Adresli: Evet

Özet

Electrical discharge machining (EDM) offers certain advantages as a thermal process but also has drawbacks. A hybrid system called electrical discharge turning (EDT), which combines traditional turning with EDM, has been developed to overcome these. In this study, experiments were conducted using a copper electrode and cylindrical AISI 4140 tempered steel with a diameter of 12 mm. The experimental design was planned as a full factorial analysis, and the effects of parameters such as dielectric medium, machining depth, discharge current, and workpiece rotational speed on surface roughness (Ra), material removal rate (MRR), tool wear rate (TWR) and dimensional accuracy were investigated. Pulse-on time, pulse-off time and lateral flushing pressure were kept constant. In the experiments, an increase in discharge current increased surface roughness, MRR, and TWR, while an increase in workpiece rotational speed showed similar effects. The effect of discharge current on performance outputs was found to be more pronounced compared to rotational speed. The dielectric medium had no statistically significant main effect on any response. Because the achieved machining depths fell 26–83% short of the intended values, first- and second-order regression models and an XGBoost model based only on controllable process parameters were developed to estimate the depth to be set on the machine. In leave-one-out cross-validation, the second-order regression model (R2 = 0.85, RMSE = 0.56 mm) and the XGBoost model (nested; R2 = 0.82, RMSE = 0.61 mm) clearly outperformed the first-order linear model (R2 = 0.67, RMSE = 0.83 mm); these models require verification with independent experiments before practical use.