PSI - Issue 52

Yuhang Pan et al. / Procedia Structural Integrity 52 (2024) 699–708 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

707

9

As shown in Fig.9, some of the predicted location by using single-signal features have large differences from the real damage location, in which the mean value of MND predicted based on PCA-FRF is 2.05 cm and the minimum value is 1.18 cm, while the mean value of MND predicted by using the Eq.(2) in the guided-wave method is 1.07 cm and the minimum value is 0.87 cm. Results are summarized in Table 2.

Table 2. Results obtained by different methods.

Different method

Active Sensing result

Vibration-based

Hybrid method

Features

DI(1)

DI(2)

DI(3)

DI(2) +DI(3)

PCA-FRF

PCA-FRF+DI(2)+DI(3)

MND (cm) Min (MND)

1.86 1.55

1.07 0.87

1.04 0.80

0.91 0.66

2.05 1.18

0.79 0.38

Table 3 shows the average MND value and the minimum MND value obtained by different features and different methods. For the guided-wave based method, localization results obtained by the Damage index 2 and 3 (Eq. (2-3)) are better than that of damage index 1(Eq. (1)), and the average MND value and minimum MND value results obtained by combining them is 0.91 and 0.66, respectively, which is better than better each of them individually. In addition, the PCA-FRF methods has the largest error compared with the guided-wave method. Based on the above analysis, the guided wave method demonstrates optimal performance when combining damage index 2 and 3 (Eq. (2-3)), therefore, these two features in the guided wave are input into the neural network along with nine principal components obtained from FRF, and the predicted and true values of the damage location are summarized in Fig. 10(a). More importantly, the predicted average MND value is showed in Fig. 10(b) is 0.79 cm and the minimum MND value is 0.38cm. These results indicate that the hybrid method is capable of achieving a better performance compared with the individual methods, which verifies the effectiveness of the hybrid method proposed in this research.

Fig.10. (a) Damage localization results obtained by the combined method. (b) The average MND and minimum MND obtained by the different methods. 5. Conclusion A hybrid method combining vibration and guided-wave responses of the structure is proposed to detect and localize damage in this research for an aluminum plate. Firstly, some damage sensitive features are extracted from the signal. Specifically, three damage indices that are capable of describing the change of guided wave before and after damage are extracted, and FRF are used the basic features for the vibration signal. Then the damage detection and localization model are developed based on the neural pattern recognition and BPNN, respectively. To verify the proposed method, different damage locations are considered. Experimental results show that both of the individual vibration and guided wave method is capable of achieving 100% accuracy for damage detection. However, for the damage localization, the best performance with average MND value and minimum MND value obtained by guided wave is 0.91 and 0.66, and