Efficient automated extraction of local defect resonance parameters in fiber reinforced polymers using data compression and iterative amplitude thresholding. (22nd December 2019)

• Record Type: Journal Article
• Authors: Segers, Joost; Hedayatrasa, Saeid; Verboven, Erik; Poelman, Gaétan; Van Paepegem, Wim; Kersemans, Mathias
• Is Part Of: Journal of sound and vibration. Volume 463(2019)
• Publication Date: 2019-12-22
• Subjects: Local defect resonance (LDR) -- Non-destructive testing (NDT) -- Fiber reinforced polymers -- Automated defect detection; Sound -- Periodicals; Vibration -- Periodicals; Son -- Périodiques; Vibration -- Périodiques; Sound; Vibration; Periodicals; Electronic journals; 620.205
• DOI: 10.1016/j.jsv.2019.114958
• Languages: English
• ISSNs: 0022-460X

Abstract

Abstract: Local defect resonance (LDR) employs a specific high frequency, the LDR frequency, to get a localized strong resonant activation of the defect. However, one of the major difficulties for applying LDR as a non-destructive testing technique, is the proper identification of the required LDR frequency, and the subsequent LDR localization. In this study, post-processing methods in both time and frequency domain are applied to low-power broadband vibration data in view of automated extraction of LDR parameters, i.e. LDR frequency and LDR location. In order to reduce the computational effort for large datasets (>1 GB), various data compression methods have been considered: power spectral density (PSD), principal component analysis (PCA) and operational modal analysis (OMA). The actual LDR parameter extraction from the (compressed) data is based on an iterative procedure to threshold the vibrational amplitudes. The LDR parameter extraction procedure is demonstrated on different carbon fiber reinforced polymers with various defect types: flat bottom holes, inserts and low velocity impact damage. It is further demonstrated that the procedure can equally handle multiple defects. A comparison of the performance of the various data compression methods is provided.