Rub-Impact Fault Diagnosis Using an Effective IMF Selection Technique in Ensemble Empirical Mode Decomposition and Hybrid Feature Models

. 2018 Jun 26;18(7):2040. doi: 10.3390/s18072040

Algorithm 1: EEMD Algorithm

Generate an artificial observation $x^{j} [n] = x [n] + ω^{j} [n]$ , where $ω^{j} [n] (j = 1, \dots, J)$ are various realizations of Gaussian noise.
Using traditional EMD, completely decompose the ensemble $x^{j} [n]$ with added white Gaussian noise into IMFs, ${IMF}_{k}^{j} [n]$ . Here, $k = 1, \dots, K$ indicates the modes.
Repeat steps 1 and 2 with different white noise realizations each time.
Obtain the result of decomposition as an ensemble means of the corresponding IMFs from all decomposition trials.
Reconstruct the original signal as $x [n] = \sum_{k = 1}^{K} {IMF}_{k} [n] + r_{K} [n]$