De-noising and compression of SHM signals with wavelet transform

Structural Health Monitoring (SHM) signals usually are recorded with noise. Moreover, in some applications, saving the extracted signals is of great importance. Thus de-noising and compression of SHM signals is of great importance. Wavelet transform is one of the most popular tools for de-noising and compression of SHM signals. In this work, de-noising and compression of SHM signals, based on wavelet transform, are studied. A steel beam is used as the experiment structure. A saw-cut slot is created in the middle of the beam as damage. Several signals are captured by means of Piezoelectric Wafer Active Sensors (PWAS) as the experimental signals. Furthermore, simulation signals are computed using Finite Element Method (FEM) simulations. The signals are de-noised with Discrete Wavelet Transform (DWT) by means of different wavelets. Furthermore, simulation and de-noised experimental signals are compressed using Wavelet Packet Transform (WPT) by means of different orthogonal and bi-orthogonal wavelets. In both, de-noising and compression subjects, bi-orthogonal wavelets show better efficiency than the orthogonal ones. The results indicate that in low noisy signals, SNR of the de-noised signal by means of bi-orthogonal wavelets is almost 10-15dB better than SNR of the de-noised signal by means of orthogonal wavelets. Also, with the same retained energy parameter in compressed signals, numbers of coefficients that converted to zero, in compression by means of bi-orthogonal wavelets are almost 0.5-1% more than similar numbers in compression by means of orthogonal wavelets

Keywords:

Structural health monitoring (SHM) , wavelet transform (WT) , signal de-noising , signal compression