The Research On Ultrasonic Signal Processing Method Based On Time-Frequency Analysis

Modification layers of materials surface are applied extensively in aerospace, petrifaction, micro-electronics and automobile manufacture engineering fields etc, ultrasonic nondestructive characterization for its properties, such as porosity, microcracks, geometry scale, elastic modulus, bonding strength and residual stress etc, was very significant in quality evaluation of modification layers. Because the thickness of modification layers was small and included heterogeneous thin layers, ultrasonic signals from heterogeneity interfaces overlapped and interfered. The signal is so non-stationary that it is hard to acquire ultrasonic characteristic parameters of modification layers. Although the traditional Spectral Analysis has many advantages and has been used in materials ultrasonic nondestructive characterization and evaluation extensively, it can not provide information about time domain. In this paper, the method of Time-Frequency Analysis was used to analyse coating materials in ultrasonic nondestructive characterization and evaluation, from two aspects in time and frequency domain.Through learning about Time-Frequency analysis, it is found that this method can provide both time and frequency information which is fit to analyse non-stationary signals. After comparing the property of Fourier transform, Short-time Fourier transform(STFT) and Wavelet transform, it is founded that Fourier transform is fit to analyse stationary signal, Short-time Fourier transform is real-time reaction, so it can be used in certain fields after good length of window and type were chosen, Wavelet transform is fit to analyse Break Signal. Procedure and visualization are actualized in Time-Frequency Analysis by matlab.Technique of db4 wavelet was used in the researching of noises removing from ultrasonic signals of DZ4 alloy with recrystallization. Because of the noise exists in high-frequency, db4 wavelet was used to decompose the signals. The signals were reconstructed after the high-frequency signals were filtered. Time signals are more smoothly after noise removing, not only remain its character but also improve analysis accuracy. The result indicates that Wavelet transform has strong ability in noise removing and is fit to remove noise from faint signals.This paper analyses the ultrasonic time domain waveform of ZrO₂ coatings using STFT method, the depth of the coatings are 270μm and 295μm separately. After comparing the time-frequency Spectrogram of STFT results which were transformed using rectangular,...