点击次数：

论文类型：期刊论文

发表时间：2022-04-16

发表刊物：JOURNAL OF NONDESTRUCTIVE EVALUATION

卷号：41

期号：2

页面范围：37

摘要：The improvement of time resolution is commonly concerned in ultrasonic testing, while the reference signal significantly influences the decoupling and interpretation of multiple overlapped signals. In this paper, the frequency-domain sparse blind deconvolution (FSBD) method is proposed to enhance the time resolution of ultrasonic signals without using any reference signal. The matching pursuit (MP) algorithm is introduced to remove noise for signal reconstruction. On this basis, homomorphic transformation is applied to the de-noised signal, and the l1 and l2 norm constraints based on sparsity are combined to construct a frequency-domain objective function for improving time resolution. The multilayer specimens and small defect were quantitatively detected using the FSBD method by experiments. The results demonstrate that the to-be-measured object with a size of half-wavelength can be identified by decoupling the multiple overlapped signals. Finally, the FSBD method is compared with minimum entropy deconvolution and homomorphic deconvolution. The relationship between the number of scattered/reflected echoes K and the regularization parameter μ is discussed to solve reasonably the regularization problem.