ENHANCED BEARING FAULT DIAGNOSIS USING INTEGRATED ICEEMDAN-TKEO-ENVELOPE ANALYSIS UNDER VARIABLE LOAD CONDITIONS

Abstract

This study proposes a hybrid methodology for early bearing fault diagnosis in rotating machinery, particularly asynchronous motors. The approach combines three advanced techniques: ICEEMDAN (Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise) to extract intrinsic mode functions from vibration signals, TKEO (Teager-Kaiser Energy Operator) to amplify fault-related impact characteristics, and envelope analysis to identify characteristic frequencies. Experimental validation was conducted using the Case Western Reserve University database with SKF 6205-2RS bearings exhibiting 7-mil single-point defects on inner race, outer race, and ball elements. Tests were performed under no-load (0 HP) and loaded (2 HP) conditions. The methodology involves decomposing the original signal using ICEEMDAN, selecting the most correlated IMF component, applying TKEO processing, and analyzing the envelope spectrum. This methodology provides a robust solution for clearly distinguishing healthy from faulty states and enables precise diagnosis of different bearing fault types under various load conditions.