Abstract
In this paper, Empirical Mode Decomposition (EMD) is employed to eliminate noises existing in
experimentally acquired ultrasonic guided Lamb wave signals. The detailed procedures are provided and
several examples involving both narrowband and wideband signals are given to show the validity of the
proposed de-noising method. Comparisons are made with the results obtained by the well-known Wavelet
method. It is found that both EMD-based and Wavelet-based de-noising methods can yield acceptable
results. It is also found that EMD-based de-noising method can separate noises from noise-polluted signals
more efficiently due to its adaptive nature in the de-noising process, and that less empirical information is
required as compared with the Wavelet-based de-noising method. Based on the results reported herein, one
may conclude that EMD-based de-nosing method can be effectively used to get a clear and significant
response signal, an important step in the area of structural health monitoring. Further research should be
made to establish a reliable criterion for separating a useful IMF from noise components in general cases.
