Wavelet Transform On The Signal Of Pile Integrity Testing

The pile integrity testing is a new-style and meaningful basic testing technology, but there are many aspects still need to be perfected such as its principle, experiment technique,postprocessing and analyzing means. The postprocessing methods are mainly researched in this dissertation. Firstly, the propagations of one-dimensional compression waves in pile and the influences of the soil around the pile are discussed, as well as the effects of impulses to the real signal and some digital signal analyzing methods. Based on the theory of wavelet, Fourier transform, and Gabor transform (short time Fourier transform), time-frequency localization properties of wavelet transform and multi-resolution analysis are discussed. Meanwhile, orthogonal, support internal, symmetric, smoothness and regularity which affect the characters of wavelet are explained. Secondly, according to the characters of wavelet, the bior5.5 as a wavelet base of biorthogonal series is chosen. Grounded on the existent methods of wavelet filtering, the suitable wavelet filter is selected and the MATLAB program for filtering noises of signal and testing singular points of signal is compiled. The results show that the wavelet transform is effective for the testing of piles. A program with VC which implements wavelet decomposition and wavelet-pack decomposition is compiled; furthermore, this program is brought into the RS-pile testing system to extend its application, which can be used for removing baseline drift, wavelet filtering and picking up the characters of shallow defect of reflection wave signal. Finally, a unique neural network for classification of reflection wave signal is compiled with MATLAB. Some effective characters of signal picked-up by wavelet decomposition and other characters are regarded as input data of the network. This neural network can improve the intelligence and availability of postprocessing means.