Damage detection using empirical mode decomposition method and a comparison with wavelet analysis

This study focuses on data analysis methods for damage detection based on structural response data. Specifically, the application of the newly-emerged empirical mode decomposition (EMD) method to damage detection is investigated and compared with the wavelet analysis (WA) approach. Numerical studies are performed for synthetic structural response data generated from a three-story shear building with a sinusoidal external excitation and a sudden stiffness loss in the first story columns during the monitoring period. Results indicate that both the EMD and WA approaches can be used to sharply identify the time at which structural damage occurs. One known strength of the EMD method is its adaptive nature. It decomposes the original signal into a number of intrinsic mode functions (IMFs) with each IMF often containing distinct physical significance. The WA approach is purely mathematical and non-adaptive in nature. However, the application of discrete WA is very effective for extracting from acceleration signals the singularities caused by sudden damage.