Heterogeneous Feature Selection with an Application in Multi-Sensor- Based Condition Monitoring of a Tool Used In Rotary Ultrasonic Machining

Abstract

Effective monitoring and diagnosis of tool wear condition plays a critical role in improving rotary ultrasonic machining quality and system reliability. In this study, force and vibration signals were first collected during rotary ultrasonic machining operation, next processed by time/frequency domain analysis or wavelet packet decomposition(WPD) for feature extraction, and then the best features are selected by a newly proposed heterogeneous feature selection method. Physical experiment data of the BK7 glass grinding process are implemented to evaluate the proposed method. The results were shown to illustrate the effectiveness of the proposed methods.