Electromagnetic vibrations estimation of an induction motor by nonlinear optimal filtering

Stator frame radial vibrations of an induction motor are composed of the sum of three different components: aerodynamic, mechanical and electromagnetic vibrations. The separation of these components could be useful in order to quantify their respective vibratory influence. Moreover, each of these components carrying different physical informations, such a processing could be interesting to further analyze each component independently, and finally diagnose induction machine faults more easily. This paper deals with a new processing algorithm able to extract electro-magnetic vibrations of an induction motor from measured signals. To this end, a nonlinear quadratic optimal filter is used to estimate these vibratory components from stator currents and radial vibrations measured at one location on the stator frame. This algorithm is based on the physical quadratic link between stator currents and electromagnetic vibrations, which is first detailed. The algorithm used to estimate the optimal nonlinear quadratic filter is then determined and analyzed. Finally, the proposed algorithm is applied to real signals, and is shown to be very efficient, whatever frequency band.