A multivariate data analysis approach towards vibration analysis and vibration-based damage assessment : application for delamination detection in a composite beam

Garćia, David and Trendafilova, Irina (2014) A multivariate data analysis approach towards vibration analysis and vibration-based damage assessment : application for delamination detection in a composite beam. Journal of Sound and Vibration, 333 (25). pp. 7036-7050. ISSN 0022-460X

[thumbnail of Garcia-Trendavilova-JSV-2014-A-multivariate-data-analysis-approach-towards-vibration-analysis]
Preview
Text (Garcia-Trendavilova-JSV-2014-A-multivariate-data-analysis-approach-towards-vibration-analysis)
Garcia_Trendavilova_JSV_2014_A_multivariate_data_analysis_approach_towards_vibration_analysis.pdf
Accepted Author Manuscript
License: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 logo

Download (7MB)| Preview

    Abstract

    This paper introduces a novel methodology for structural vibration analysis and vibration-based monitoring which utilises a special type of Principal Components Analysis (PCA), known as Singular Spectrum Analysis (SSA). In this study the methodology is introduced and demonstrated for the purposes of damage assessment in structures using their free decay response. The method's damage assessment properties are first demonstrated on a numerical example for a two degree-of-freedom (2DOF) spring-mass and damper system with non-linear stiffness. The method is then applied to an experimental case study of a composite laminate beam. The method is based on the decomposition of the frequency domain structural variation response using new variables, the Principal Components (PCs). Only a certain number of the new variables are used to approximate the original vibration signal with very good accuracy. The presented results demonstrate the potential of the method for vibration based signal reconstruction and damage diagnosis. The healthy and the different damaged scenarios are clearly distinguishable in the new space of only two reconstructed components where a strong clustering efect is observed.

    ORCID iDs

    Garćia, David ORCID logoORCID: https://orcid.org/0000-0002-3841-6824 and Trendafilova, Irina ORCID logoORCID: https://orcid.org/0000-0003-1121-7718;