Multivariate statistical analysis for damage and delamination in composite structures

Garcia, David and Trendafilova, Irina (2013) Multivariate statistical analysis for damage and delamination in composite structures. In: 11th International Conference on Vibration Problems, 2013-09-09 - 2013-09-12, Instituto Superior Tecnico.

[thumbnail of Garcia D & Trendafilova I - Pure - Multivariate statistical analysis for damage and delamination detection in composite structures Sep 2013] PDF. Filename: Garcia_D_Trendafilova_I_Pure_Multivariate_statistical_analysis_for_damage_and_delamination_detection_in_composite_structures_Sep_2013.pdf

Download (726kB)


The article is devoted to the analysis of the vibration response of composite laminates .Our aim is to develop a method for analysis of the vibration response of structures made of composites which will also be used to develop a vibration-based health monitoring procedure for such structures. Composite materials and composite laminates in particular, exhibit complex dynamic behaviour which on most occasions cannot be modelled linearly. Delamination introduces additional nonlinearities in the vibration behaviour of the structure as a result of the interrupted contact between the layers or the opening and closing of the delamination. Thus conventional linear structural dynamics methods like modal analysis cannot be applied. In this study, the vibration response signals are recorded from damaged and non-damaged (healthy) laminated composite beams. The frequency domain signals are subjected to a special type of Principal Component Analysis, known as Multichannel Singular Spectrum Analysis (MSSA). This type of analysis is known to uncover oscillation patterns and was suggested in the investigation in place of modal analysis. The idea is to establish a new feature based state-space for the vibration response signal. The response of the healthy structure is used as a baseline to which all the responses are compared. MSSA decomposes the signal into new components which are lineal combinations of the original frequency series components. The first several components are responsible for most of the variance of the original signal. The new space is with a much smaller dimension as compared to the original data and creates new variables which can be used as damage features. The results demonstrate strong potential for using MSSA for the purpose of structural health monitoring.


Garcia, David ORCID logoORCID: and Trendafilova, Irina ORCID logoORCID:;