Damage assessment for wind turbine blades based on a multivariate statistical approach

11th International Conference on Damage Assessment of Structures

物理学;材料科学

García, David^1 ; Tcherniak, Dmitri^2 ; Trendafilova, Irina^1

Mechanical and Aerospace Engineering, University of Strathclyde, 75 Montrose street, Glasgow

G11XJ, United Kingdom^1

Bruel and Kjæir Sound and Vibration Measurement A/S, Skodsborgvej 307, Næirum, Denmark^2

关键词: Delamination detections;    Detection and localization;    Free vibration response;    Multivariate statistical approaches;    Number of components;    Real time monitoring;    Singular spectrum analysis;    Vibration-based structural health monitoring;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/628/1/012086/pdf
DOI  :  10.1088/1742-6596/628/1/012086

学科分类：材料科学（综合）

来源: IOP

【 摘 要 】

This paper presents a vibration based structural health monitoring methodology for damage assessment on wind turbine blades made of composite laminates. Normally, wind turbine blades are manufactured by two half shells made by composite laminates which are glued together. This connection must be carefully controlled due to its high probability to disbond which might result in collapse of the whole structure. The delamination between both parts must be monitored not only for detection but also for localisation and severity determination. This investigation consists in a real time monitoring methodology which is based on singular spectrum analysis (SSA) for damage and delamination detection. SSA is able to decompose the vibratory response in a certain number of components based on their covariance distribution. These components, known as Principal Components (PCs), contain information about of the oscillatory patterns of the vibratory response. The PCs are used to create a new space where the data can be projected for better visualization and interpretation. The method suggested is applied herein for a wind turbine blade where the free-vibration responses were recorded and processed by the methodology. Damage for different scenarios viz different sizes and locations was introduced on the blade. The results demonstrate a clear damage detection and localization for all damage scenarios and for the different sizes.