A multiple improved envelope spectra via feature optimization gram (MIESFO-gram) for diagnosis of compound fault signatures

Detecting compound faults in rotating machinery is challenging for fault diagnosis due to simultaneous occurrences of multiple faults, hindering the isolation of specific fault signatures. This is particularly relevant in the expanding field of bearing diagnostics, which focuses on complicated rotating machinery with diverse components operating under variable conditions (e.g. speed and load). Meanwhile, some components with weak signatures may remain hidden while others with intensive defects are detected. Therefore, the ability to detect combined faults in machinery, having different cyclic frequencies is critical. Envelope Analysis is a popular method for bearing diagnostics, however, as several damaged bearings may excite not only different but also several frequency bands simultaneously, band-pass filtering around only one frequency band may not be sufficient to detect all bearing faults in a machine, especially if it operates under varying conditions. Recently, IESFOgram has been proposed, utilizing Targeted and Blind features and being based on either the Cyclic Spectral Correlation or the Cyclic Spectral Coherence, in order to select the optimal frequency band and extract the corresponding Improved Envelope Spectrum. When there are more than one bearing faults exciting different natural frequencies, selecting only the single most dominant carrier may prove insufficient to detect other damages present in the signals. In this paper, Multiple Improved Envelope Spectra via Feature Optimization gram (MIESFO-gram) is introduced with the aim of finding all possible unique frequency bands occupied by cyclic frequencies and identifying different types of faults. The method is applied and evaluated on simulated and experimental data with different types of faults under steady and varying speed conditions in a complicated system. Finally, the results are compared with the conventional Targeted and Blind IESFOgram, demonstrating the superiority of the approach.