Ultrasonic machining of hard-to-machine materials involves material removal through the fracture of the workpiece surface induced by vibrating abrasive particles. This process not only governs material removal but also directly affects surface roughness, which is critical for precision manufacturing. In this study, the ultrasonic machining of Inconel 718, a high-performance superalloy, is investigated with a focus on predicting surface roughness using acoustic emission (AE) signals. An AE sensor integrated into the machining setup captures real-time signals, which are analyzed in both the time and time–frequency domains to extract informative features such as RMS and kurtosis. These features, combined with process parameters, serve as inputs to a support vector regression (SVR) model for surface roughness prediction. The SVR model that incorporates signal features achieves a prediction accuracy of 99.72%, outperforming the model developed without signal features, which achieves an accuracy of 96.54%. The study demonstrates that incorporating AE signal features into predictive modeling provides a highly accurate and reliable method for real-time surface roughness monitoring, offering strong potential for intelligent process control in ultrasonic machining.
Mirad et al. (Wed,) studied this question.