Higher order statistical frequency domain decomposition for operational modal analysis

Experimental methods based on modal analysis under ambient vibrational excitation are often employed to detect structural damages of mechanical systems. Many of such frequency domain methods, such as Basic Frequency Domain (BFD), Frequency Domain Decomposition (FFD), or Enhanced Frequency Domain Dec...

Full description

Saved in:
Bibliographic Details
Published in:Mechanical systems and signal processing 2017-02, Vol.84, p.100-112
Main Authors: Nita, G.M., Mahgoub, M.A., Sharyatpanahi, S.G., Cretu, N.C., El-Fouly, T.M.
Format: Article
Language:English
Subjects:
Modal analysis
Sample to model ratio estimator
Spectral kurtosis estimator
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Experimental methods based on modal analysis under ambient vibrational excitation are often employed to detect structural damages of mechanical systems. Many of such frequency domain methods, such as Basic Frequency Domain (BFD), Frequency Domain Decomposition (FFD), or Enhanced Frequency Domain Decomposition (EFFD), use as first step a Fast Fourier Transform (FFT) estimate of the power spectral density (PSD) associated with the response of the system. In this study it is shown that higher order statistical estimators such as Spectral Kurtosis (SK) and Sample to Model Ratio (SMR) may be successfully employed not only to more reliably discriminate the response of the system against the ambient noise fluctuations, but also to better identify and separate contributions from closely spaced individual modes. It is shown that a SMR-based Maximum Likelihood curve fitting algorithm may improve the accuracy of the spectral shape and location of the individual modes and, when combined with the SK analysis, it provides efficient means to categorize such individual spectral components according to their temporal dynamics as coherent or incoherent system responses to unknown ambient excitations. •Statistical properties of the PSD estimates are discussed in the context of OMA.•Spectral kurtosis is used to discriminate harmonic and stochastic oscillations.•The sample to model ratio (SMR) estimator is adapted for use in OMA applications.•A SMR-based self-adaptive maximum likelihood modal fitting algorithm is developed.•SMR-estimated modal parameters are compared with corresponding state-space estimates.
ISSN:0888-3270
1096-1216
DOI:10.1016/j.ymssp.2016.07.004