MULTIDIMENSIONAL DECOMPOSITION OF TIME-VARYING VIBRATION RESPONSE SIGNALS IN ROTATING MACHINERY

A new method for decomposition of speed-dependent vibration signals me asured on rotating machinery is presented. The method uses a number of measured signals from an array of sensors to decompose the vibration response into: (1) the direction of the progression of motion; (2) its frequency content; (3) the various wavelengths; and (4) speed of rota tion (or time). For the simple case of shaft vibration, two sensors sp aced at 90 degrees are required to separate the vibration pattern into forward- and backward-travelling components. For the case of a rotati ng disc, more sensors are required to decompose the response further i nto different spatial wavelengths and their directions of progression. This decomposition of the response allows monitoring and identificati on of potentially dangerous vibration patterns which exhibit large bac kward- or forward-progressing components. The new method is shown to p rovide better resolution in the time-frequency (speed-frequency) repre sentation where usually overlapping patterns are separated. Several an alytical and experimental results that demonstrate the usefulness of t he proposed method are shown. (C) 1997 Academic Press Limited.