PSI - Issue 77

Claudia Barile et al. / Procedia Structural Integrity 77 (2026) 3–10 Barile and Kannan/ Structural Integrity Procedia 00 (2026) 000 – 000

9 7

Accordingly, the AE data acquired from the flexural tests are classified using k-means++ data clustering algorithms. Representative results of the classified data from damaged (DS-1 and DS-2) and healed (DS-1-Healed and DS-2 Healed) groups are shown in Fig. 5. In the damaged specimens, shown in Fig. 5(a), cluster 1 representing matrix-dependent failure modes dominates the failure progression. Conversely, in the healed specimens, shown in Fig. 5(b), cluster 2 representing fibre-dependent failure modes dominates the failure progression. These results agree with the microscopic analysis. In the damaged specimens, failure is dominated by kink formation and longitudinal cracking that contribute to the interlaminar failure between the plies (see Fig. 4(a) and 4(b)). Due to the resin bleed in the healed specimens, the failure is dominated by local shear failure, longitudinal and through thickness crack growth. Therefore, the classified AE data can be associated with the failure progression modes in the self-healing composites. The difference between the AE data cluster distribution confirms the difference in the failure progression path between the damaged and the healed specimens. In order to verify the classified AE data, few representative signals from each cluster are analysed in the time frequency domain using CWT. The results are shown in Fig. 6.

Fig. 6. (a) and (b) Representative CWT Spectrograms of Cluster 1 Signals; (c) and (d) Representative CWT Spectrograms of Cluster 2 Signals.

The CWT results agree with the findings in the microscopic analysis and the clustered AE data shown in Fig. 5. The spectrograms of AE signals in cluster 1, as shown in Fig. 6(a) and 6(b), have low frequency and the energy is distributed in the time domain. These are typical characteristics of AE signals from matrix-dependent failure modes (De Groot et al., 1995). The spectrograms of cluster 2, as shown in Fig. 6(c) and 6(d), have high frequency and the energy in a much narrower time window. The fibre-dependent failure modes often have narrower time domain distribution and higher frequency due to the large and rapid strain energy release. This is also associated with the high