A new TFA-BD method for processing non-stationary mechanical vibration signals

This paper begins with reviewing past research, leading to proposing a new TFA-BD method, where TFA is time-frequency analysis and BD is bind de-convolution. Subsection 2.2 of the full paper explains mathematically the two signal processing steps of TFA-BD method, including two theorems in Step Two. Subsection 3.1 gives the results of simulation experiments, whose performance index is usually given by eq.(25) of the full paper; the performance index results of our new TFA-BD method (Table 1) are compared with those of the traditional Fast ICA method (Table 1) and the comparison shows preliminarily that, for the three performance Indies, TFA-BD performance index values are about twice as high as Fast ICA ones. Subsection 3.2 gives the results of measurement experiments; for fault diagnosis, the power spectra of two types of signals separated by TFA-BD method, given in Figs.3 (a) and (b), are more helpful than the power spectrum of unprocessed mixed signals, given in Fig.4.