Low-frequency vibration inspection of sandwich composite

Abstract

The inspection of thick-section sandwich structures, such as those with honeycomb, balsa, or foam cores, often relies on low-frequency vibration techniques to detect defects by identifying changes in amplitude or phase response. However, existing methods are often tailored to detect specific defect types, potentially overlooking others, and lack the capacity to provide detailed information on the defect characteristics due to limited frequency spectrum analysis.

This study addresses these limitations by investigating both homogeneous (aluminium) and sandwich structures to enhance defect detection and characterisation. Using a pitch-catch setup, defects in aluminium plates were excited with a 5-50 kHz broadband chirp signal to monitor the local defect resonances (LDR). Finite element analysis (FEA) was used to create a numerical model which was iteratively refined to reduce the error between the model and the experimental results. Adjusting for the probe's stiffness and damping effects significantly improved the accuracy of predicted resonance frequencies, reducing errors from approximately 3 kHz to below 1 kHz. This improvement enabled the potential for accurate inference of the defect characteristics by comparison with the modelled results.
This demonstrated the effectiveness of a full-spectrum analysis approach for defect detection, establishing a robust foundation for further testing on sandwich structures.

In addition, the characterisation and classification of the defects in sandwich composite panels were investigated by statistical analysis of the frequency-domain responses. Distributions of amplitudes and phases from non-defective and defective regions were compared, utilising receiver operating characteristic (ROC) curves for optimal frequency selection. At a 90\% probability of detection (PoD), this methodology effectively distinguished defective responses from non-defective responses, enabling the detection of defects as small as 10 mm and achieving reliable imaging through pixel-by-pixel classification with low levels of artefacts. Accurate defect sizing was achieved for defects larger than twice the probe pin-spacing, with results indicating a dependency on both the defect size and depth in aluminium and composite specimens.

Date of Award	13 May 2025
Original language	English
Awarding Institution	University of Bristol
Supervisor	Robert A Smith (Supervisor) & Bruce W Drinkwater (Supervisor)