Abstract: The outputs of existing blind signal separation(BSS) are signals which can present spectral characters of the separated signals but fail to provide valid amplitudes and phases. Consequently, their application in mechanical area is basically limited to detecting vibrating sources in fault diagnosis. Here an analytical BSS method used for getting full-info signals is introduced, aiming at that the prevailing BSS is unable to satisfy the needs for frequency, amplitude and phase information necessary for dealing with dynamic problems in mechanical engineering. The analytical method gets rid of the roundabout separation strategy based on objective functions and optimization algorithms taken by traditional BSS. It primarily estimates the mixing matrix on the basis of sparseness in signal frequency domain and by using available amplitude and phase information in the frequency domain, then carries out analytical operation of linear equations by using the solved amplitude ratio and relative time shift parameter, and lastly restores the signals expected. Experiment results show that the separated signals contain the effective information of amplitude, phase and frequency, which are the so called full-info signals suitable for dynamic analyses in mechanical engineering, such as the calculation of vibration intensity and mechanical impedance.

Key words: Analytical blind signal separation(BSS), Dynamic analysis, Full-info signal, Signal separation