Improved Sparse Representation Localization Method for Pipeline Weld Cracks With Bud-Shaped Electromagnetic Acoustic Transducer

Abstract

Shear horizontal (SH) guided waves with nondispersive low-order modes are well suited for small crack detection in pipelines, playing a crucial role in maintaining pipeline safety. However, the curved structural characteristics of pipeline welds make traditional SH guided wave electromagnetic acoustic transducer (EMAT) face significant limitations in the directivity of the excitation signal. This reduces the signal-to-noise ratio and increases the difficulty of extracting defect features from complex signals and achieving their localization. In this article, an improved dictionary construction method for pipeline defect localization is proposed based on a new bud-shaped coil structure design with a periodic permanent magnet (PPM) EMAT configuration. The method involves four primary steps: First, a circumferentially propagating SH guided wave is generated using the proposed bud-shaped EMAT, producing defect echo signals through its interaction with circumferential defects. Then, an over-complete dictionary incorporating wavenumber information and weakly damped attenuation characteristics of SH waves is constructed to mitigate wave packet overlapping and noise. Next, a convex optimization model is established, and the split augmented Lagrangian shrinkage algorithm (SALSA) is employed for fast signal reconstruction. Finally, the defect is localized based on the time–space characteristics of the reconstructed signals. Experimental results show that the proposed method localizes defects with high precision. The average error is about 3.1 mm when calculated using theoretical velocity, and under 1 mm with actual velocity, even with overlapping packets and simulated weld noise. The proposed method provides a novel approach for extracting SH waves traveling through a pipe and localizing defects.