Empirical Mode Decomposition/Hilbert-Huang Transform Analysis and Machine Tool Prognostics
Gary G. Leisk
Economic competition in the manufacturing industry is generating demand for more advanced diagnostic and prognostic capabilities in machine tools. While machine tool builders have offered systems for both machine and process monitoring for years, there is a clear need for sensor-based diagnostic and prognostic capabilities. The aim of this research is to explore the role that Empirical Mode Decomposition/Hilbert-Huang Transform (EMD/HHT) analysis could play in such systems. EMD/HHT, developed by Dr. Norden Huang of NASA/Goddard, is a two-stage analysis method for analyzing time-series signals, with a conspicuous capability for analyzing nonlinear, non-stationary signals. The first stage consists of a decomposition of time-series signals into intrinsic oscillatory modes, known as intrinsic mode functions, through a sifting process. The second stage involves the application of the Hilbert transform to the intrinsic mode functions and the construction of a time-frequency-energy distribution of the signal. The resulting Hilbert spectrum shows instantaneous frequencies as a function of time, giving sharp identifications of embedded structures. In this research, EMD/HHT is used to analyze acoustic emission signals generated by an operating computer numerical control milling machine.
February 15, 2001
Annual NIST/Sigma Xi Postdoctoral Poster Presentation