Study of the IMF Selection Methods using Kurtosis Parameter for Bearing Fault Diagnosis

Fault diagnosis in induction machines is an essential phase to maintain good system health. Selecting effective and sensitive Intrinsic Mode Functions (IMFs) is one of the important steps for rolling bearing fault diagnosis. Kurtosis is the most commonly used parameter for the selection of these IMFs. Generally, two selection methods for IMFs based on the kurtosis criterion are used in the fault detection field. The first one is to select the IMFs with high kurtosis, and the second selection method is based on the best IMF with the highest kurtosis value (Max Kurtosis value). This paper pointed out the detailed study of the usual IMF selection methods based on kurtosis value to detect rolling element bearing faults. This study treated the experimental data (vibration signal) using Ensemble Empirical Mode Decomposition (EEMD). In addition, this work used two IMF selection methods carefully to find the best decision about bearing fault. The envelope spectrum of the selected IMFs is well analyzed to define the characteristic frequencies of the bearing fault. The results clearly showed what we could face using the IMF selection methods based on kurtosis criterion.