A Fitted empirical demodulation for low frequency electrical signal evaluation

Abstract

Empirical Demodulation (ED) is a technique used to build a discrete signal, called an empirical envelope, from a modulated time sequence. For example, in three-phase induction motors, this envelope can carry fault frequency data that allows the machine health status to be evaluated during a spectral analysis. However, due to mathematical reasons, the method is very sensitive to the amplitude oscillations within the signal. When these oscillations are unwanted, as in the presence of measurement noise, the results can be strongly affected. This work proposes an iterative and adjustable version of the ED that considerably reduces its sensitivity to the presence of high frequency noise, thus eliminating the need for signal pre-filtering. To prove the effectiveness of the Fitted Empirical Demodulation, the authors applied the new tool in motor current signals for analysis of the rotor bars conditions.