Application Of Sparse Decomposition Theory To Ultrasonic Nondestructive Testing Signal Processing

Ultrasonic testing technology, as currently the most widely used nondestructive testing technique, is an important tool to ensure the safe operation of equipments.Therefore, ultrasonic signal processing has been occupying an important position in reducing diagnostic accuracy of the internal defects in materials. The appearing of each new signal processing theory will have a profound impact on the field of ultrasonic signal processing, and cause new discoveries and new methods in the field of ultrasonic signal processing. With the development of the signal processing field,ultrasonic signal processing also develops corresponding algorithms, such as Fourier transform, adaptive algorithm, neural network algorithm, wavelet transform algorithm,etc.Since the ultrasonic testing signal has the non-stationary characteristic, the ultrasonic echo signal denoising and feature wave inspection and identification need further research. The appearing of matching pursuit algorithm introduces a new starting point for ultrasonic echo signal processing. With the development of signal processing, matching pursuit algorithm also innovates constantly. Orthogonal matching pursuit algorithm and compressed sensing matching pursuit algorithm have been widely used in the field of image and signal processing. Therefore, it is particularly important for further study on matching pursuit algorithm.Over-complete dictionary is an important part of the matching pursuit algorithm,during the characteristic wave detection of ultrasonic echo signals, the paper cited the over-complete dictionary theory. This method uses the over-complete redundancy function system instead of the traditional orthogonal basis functions, which provides great flexibility for signal adaptive sparse extension. Sparse decomposition can achieve high efficiency of data compression, more importantly, it can use the redundancy feature of the dictionary to capture the essential feature of the signal.Ultrasonic echo signal is the non-stationary signal, it is difficult to fully extract the features of ultrasonic echo signal, so this paper uses the over-complete dictionary with the same essential feature as the ultrasonic echo signal. The over-complete dictionary trained by the K-SVD algorithm has good separability for the essential feature of the ultrasonic echo. Separability is critical for the feature detection of ultrasonic echo signal, it can clearly extract all the useful ultrasonic echo signals and make accurate identification for the internal defects.How to correctly extract the essential feature of the ultrasonic echo signal and remove all kinds of interfering noise is the main topic of this research. Sparsedecomposition theory provides a good idea in dealing with these problems. The matching tracking algorithm based on the improved Gabor dictionary was used to conduct dictionary atom selection for the above mentioned over-complete dictionary.Then these outstanding atoms were combined linearly to obtain the essential feature of the ultrasonic echo signal. Finally, the original signal was obtained. In this paper,the proposed method and the traditional wavelet thresholding denoising method were compared by simulation. Experimental results show that the method has better ultrasonic echo signal denoising effect than the wavelet thresholding denoising method, and the larger the noise, the more obvious the contrast. This method can not only more effectively filter the white Gaussian noise in signal and improve signal to noise ratio, but also retain the useful information in the original signal as far as possible.