Akademik Veri Yönetim Sistemi
Sadhana - Academy Proceedings in Engineering Sciences, cilt.51, sa.2, 2026 (SCI-Expanded, Scopus)
In the present study, we investigated the prediction of depth-dependent residual stress variations during the turning of Inconel 625, a nickel-based superalloy renowned for its exceptional high-temperature strength and thermal resistance, under high-pressure jet-cooling conditions using machine learning (ML) algorithms. The experimental set-up was designed using a three-level factorial approach, varying cutting speed, feed rate, and cooling pressure, while maintaining a constant depth of cut. Turning experiments were conducted under these specified conditions. Residual stress measurements were performed using X-ray diffraction (XRD) method, with data collected at 10 incremental depths from the machined surface to 150 µm in both radial and axial directions. During these measurements, residual stress was determined incrementally by removing material layers via electropolishing. The experimental data were used to evaluate the predictive capabilities of six ML algorithms for modeling depth-dependent residual stress profiles. Each ML model employed distinct strategies to address the complex relationships between cutting parameters and residual stress distributions. Additionally, three performance metrics – mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R2) were used to evaluate the models’ accuracy and reliability. The findings revealed that the boosted tree algorithm, XGBoostR, demonstrated superior performance, achieving the lowest MAE and RMSE and thus outperforming other models in terms of predictive accuracy and reliability.