Lamb Wave Based Structural Damage Detection Using Stationarity Tests

Lamb waves have been widely used for structural damage detection. However, practical applications of this technique are still limited. One of the main reasons is due to the complexity of Lamb wave propagation modes. Therefore, instead of directly analysing and interpreting Lamb wave propagation modes for information about health conditions of the structure, this study has proposed another approach that is based on statistical analyses of the stationarity of Lamb waves. The method is validated by using Lamb wave data from intact and damaged aluminium plates exposed to temperature variations. Four popular unit root testing methods, including Augmented Dickey–Fuller (ADF) test, Kwiatkowski–Phillips–Schmidt–Shin (KPSS) test, Phillips–Perron (PP) test, and Leybourne–McCabe (LM) test, have been investigated and compared in order to understand and make statistical inference about the stationarity of Lamb wave data before and after hole damages are introduced to the aluminium plate. The separation between t-statistic features, obtained from the unit root tests on Lamb wave data, is used for damage detection. The results show that both ADF test and KPSS test can detect damage, while both PP and LM tests were not significant for identifying damage. Moreover, the ADF test was more stable with respect to temperature changes than the KPSS test. However, the KPSS test can detect damage better than the ADF test. Moreover, both KPSS and ADF tests can consistently detect damages in conditions where temperatures vary below 60 °C. However, their t-statistics fluctuate more (or less homogeneous) for temperatures higher than 65 °C. This suggests that both ADF and KPSS tests should be used together for Lamb wave based structural damage detection. The proposed stationarity-based approach is motivated by its simplicity and efficiency. Since the method is based on the concept of stationarity of a time series, it can find applications not only in Lamb wave based SHM but also in condition monitoring and fault diagnosis of industrial systems.