Influence of stopping criterion for sifting process of Empirical Mode Decomposition technique (EMD) on roller bearing fault diagnosis

Empirical mode decomposition (EMD) is a self-adaptive data driven technique for analyzing nonlinear and non-stationary signals and decompose them into some elementary Intrinsic Mode Functions (IMFs). Although EMD method has been applied in various applications successfully, this method has some drawbacks, i.e. lack of a mathematical base, no robust stopping criterion for sifting process, mode mixing and border effect problem. Under the practical point of view, the most relevant is possibly the sifting stop criterion. Although sifting as many times as possible is needed to decompose the signal, too many sifting steps will reduce the physical meaning of IMFs. To preserve the natural amplitude variations of the oscillations, sifting must be limited to as few steps as possible. The proposed criteria so far are: Cauchy-type convergence, three-threshold, energy difference tracking, resolution factor, bandwidths, and orthogonality criterion. There is not a thorough study yet regarding the fault diagnosis application, to determine the effects of stopping criteria on the fault detection performance. In this paper the influence of different criteria to this purpose is investigated.