K-means aided broadband ultrasonic guided wavefield processing for damage detection in composite panels

This is the authors' abstract. We don't add key points for this paper.

• A novel method is proposed to combine various damage characteristics that are sensitive to different frequencies. • A K-means aided wavenumber automatic filtering technique is proposed for robust filter mask generation. • No reference data, material properties, or damage-sensitive frequencies are required. • The method accurately reflects the severity and shape of multiple damages in the GFRP plate. Guided wave full wavefield data have been widely used for detecting, localizing, and visualizing damage in composite structures. However, the sensitivity of detection maps to excitation wave frequency and the inefficiency caused by manual parameter selection limit their practicality. To address these challenges, this study presents a K-means aided automatic wavenumber filtering method, leveraging broadband excitation signals to identify local anomalies in composite plates. This approach eliminates the need for material parameters and specific excitation frequencies which enables the automatic generation of filter masks. The method’s effectiveness was verified through experiments and finite element simulations, using broadband and narrowband excitations on glass fiber-reinforced polymer plates with Teflon inserts of varying shapes and depths. Results demonstrated that broadband guided waves, filtered into multiple frequency bands, significantly enhance damage visualization in wavefield images compared to narrowband excitation, highlighting the robustness and efficiency of the proposed technique.