A fractal-based flaw feature extraction method for ultrasonic phased array nondestructive testing

Abstract

Ultrasonic phased array technology is a promising industrial nondestructive testing (NDT) method for its excellent abilities of electronic steering, deflection and focusing. Nowadays, the ultrasonic phased array systems can locate the detected flaw, but flaw classification is still determined by the inspection operator. Therefore, human errors are unavoidable. This article proposes a fractal-based flaw feature extraction method, used in automatic flaw classification to eliminate the artificial errors. Lifted wavelet transform (LWT) is applied to decompose the received ultrasonic echo. Soft-threshold denoising processing is applied to remove the background noises. Then, the echo is reconstructed based on inverse transform of LWT. Box-counting algorithm is selected to compute fractal dimension of the reconstruction signal. Experimental results show fractal dimensions of different flaws are distributed in different scale invariance intervals. Therefore, fractal feature can effectively characterize the essential nature of flaw type. Although this work is motivated by the need for characterizing defects in pipeline girth weld, the system and corresponding inspection scheme can also find applications in other fields of similar defect detection.