Feature Selection in Milling Process Utilizing Wavelet Analysis

Prognostic and health management (PHM) is a new maintenance philosophy that deals with the diagnosis and prognosis of machine failure. PHM in rotary machines usually is done by analyzing vibrational signals, acoustic emission, temperature or oil analysis. In signal analysis, feature selection is one of the most important parts in both diagnostic and prognostic techniques. Doing so, will allow an expert or at times, even artificial intelligence of machine learning to be able to detect system failure. In this paper, wavelet analysis is first introduced to reduce the noise of spindle acoustic emission signals of a milling machine. In the next step, a variety of signal properties and features were obtained and compared with each other. Finally, features were selected that enables to detect the status of the system appropriately. At the end, discussion of the selected features with similar works done by others is carried out and suggestion of the most suitable ones was made.