Akademik Veri Yönetim Sistemi
İmalat Teknolojileri ve Uygulamaları, cilt.6, sa.3, ss.296-307, 2025 (TRDizin)
This study aims to optimize power consumption observed while milling Inconel 718 superalloy—well known for its poormachinability—and to develop machine learning-based prediction models. Experiments were carried out on a TaksanTMC 500 V CNC milling machining center at three cutting speeds (40, 60, and 90 m/min) under four distinct cuttingconditions: dry, Minimum Quantity Lubrication (MQL), cryogenic, and cryogenic+MQL. Energy consumption wasmonitored in real-time using a KAEL Multiser signal analyzer and the collected data were analyzed through ANOVAand regression approaches. The ANOVA results revealed that cutting speed is the most significant factor influencingenergy demand (p<0.001), whereas cooling/lubrication strategies exhibited no statistically significant effect. To addressclass imbalance the dataset was augmented via a SMOTE-based method and ensemble and regression-based ML models(Random Forest, Gradient Boosting, Linear Regression) were trained for power prediction. The findings indicated thatthe Gradient Boosting algorithm consistently achieved superior accuracy across all cutting environments withperformance levels reaching R²≈0.97 and RMSE≈7 W. Results indicate that combining experimental data withcomputational methods is effective for decreasing energy consumption in machining and advancing sustainableproduction goals. The proposed methodology contributes to enhancing both efficiency and environmental sustainabilityin the industrial processing of Inconel 718.