Pedestal looseness fault diagnosis of rotating machine using CEEMDAN

Abstract

A novel fault diagnosis approach for rotor systems experiencing pedestal looseness is introduced, utilizing complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) to address this mechanical issue. In cases of pedestal looseness, rotor system vibration signals predominantly consist of three constituent elements: periodic components reflecting rotational frequency and harmonic content, transient impact components carrying fault characteristics, and stochastic noise interference. The CEEMDAN technique is employed to isolate and enhance the diagnostically critical impact elements within these complex vibration signals, followed by application of envelope demodulation analysis to process the extracted transient features. The diagnostic identification of pedestal looseness is ultimately achieved through characteristic periodicity detection in the resulting envelope spectrum. Validation through both simulated signal analysis and practical experimental investigations confirms the technical efficacy of this methodology for detecting pedestal looseness faults in rotating machinery rotor systems, maintaining fidelity to original operational data throughout the analytical process.