The lack of time resolution restricts the quantitative detection of shallow subsurface defects with ultrasonic time-of-flight diffraction (TOFD) technique due to the superposition between lateral wave and diffracted waves from upper and lower tips. In this article, the frequency-domain sparsity-decomposability inversion (FDSDI) method was proposed to enhance the time resolution in TOFD based on the sparsity and decomposability of the ultrasonic reflection sequence. An optimization problem was formulated in the frequency domain by combining l1 - and l2 -norm constraints. The simulation was performed with a carbon steel model containing a series of shallow subsurface cracks at the depths of 2.0, 2.5, 3.0, 3.5, and 4.0 mm. The relative measurement errors of defect depths and heights were no more than 6.57%, and the depth of the dead zone was reduced by 70%. Subsequently, the feasibility of the FDSDI method was experimentally verified on a carbon steel specimen with an artificial defect. The defect depth and height were calculated with relative errors within 6.0%. Finally, the detection capacity of the FDSDI method was discussed, and the effects of frequency bandwidth, regularization parameter, and noise on inversion results were analyzed by experiments. It is concluded that the FDSDI method decouples the multiple overlapped signals and significantly improves the time resolution to quantify the small defects in the dead zone.

中文翻译：

---

使用频域稀疏可分解性反演 (FDSDI) 方法提高超声波飞行时间衍射技术的时间分辨率。

由于横向波和来自上下尖端的衍射波之间的叠加，时间分辨率的缺乏限制了超声飞行时间衍射（TOFD）技术对浅层次表面缺陷的定量检测。在本文中，基于超声反射序列的稀疏性和可分解性，提出了频域稀疏-可分解性反演（FDSDI）方法来提高TOFD中的时间分辨率。通过组合 l1 - 和 l2 - 范数约束，在频域中制定了一个优化问题。模拟是使用碳钢模型进行的，该模型包含一系列深度为 2.0、2.5、3.0、3.5 和 4.0 毫米的浅表下裂纹。缺陷深度和高度的相对测量误差不超过6.57%，并且死区的深度减少了70%。随后，在具有人工缺陷的碳钢试样上实验验证了 FDSDI 方法的可行性。计算缺陷深度和高度的相对误差在 6.0% 以内。最后讨论了FDSDI方法的检测能力，并通过实验分析了频宽、正则化参数和噪声对反演结果的影响。得出的结论是，FDSDI 方法解耦了多个重叠信号，显着提高了时间分辨率，以量化死区中的小缺陷。讨论了FDSDI方法的检测能力，并通过实验分析了频宽、正则化参数和噪声对反演结果的影响。得出的结论是，FDSDI 方法解耦了多个重叠信号，显着提高了时间分辨率，以量化死区中的小缺陷。讨论了FDSDI方法的检测能力，并通过实验分析了频宽、正则化参数和噪声对反演结果的影响。得出的结论是，FDSDI 方法解耦了多个重叠信号，显着提高了时间分辨率，以量化死区中的小缺陷。