Abstract：T-type welds serve as critical load-bearing components in large-scale engineering structures such as bridges, ships, and offshore platforms. The structural integrity of these welds directly influences the safety and service life of the entire facility, necessitating the development of efficient and reliable nondestructive testing (NDT) methods. Conventional techniques, including bulk wave ultrasound, eddy current, and magnetic flux leakage, rely on point-by-point scanning and prove inefficient for inspecting weld lines extending tens of meters. This limitation underscores the need for rapid, long-distance inspection approaches. In this work, we propose a rapid inspection method for T-type welds based on feature-guided waves. Through semi-analytical finite element analysis, we discovered a torsional-like guided wave mode that exhibits two distinctive characteristics: its energy remains highly confined within the weld region, and it demonstrates nearly non-dispersive behavior above 200 kHz. These properties make it particularly suitable for long-range defect detection. Based on the displacement distribution of this mode, we designed an excitation scheme using thickness-shear piezoelectric patch arrays positioned symmetrically on both weld sides. Three-dimensional finite element simulations confirmed that the torsional-like wave can be efficiently generated and propagates stably along the weld with well-confined energy. Subsequent simulations revealed that this wave mode is sensitive to surface transverse cracks, with the reflected wave amplitude increasing monotonically with crack depth—a feature that enables defect quantification. Experimental validation on a steel T-type weld specimen successfully reproduced the torsional-like wave, and the measured wave velocity showed excellent agreement with theoretical predictions. Raw experimental signals, however, contained defect echoes masked by background noise and multi-mode interference. To address this challenge, we implemented a matched filtering algorithm based on cross-correlation analysis, which successfully extracted weak defect reflections and accurately located the crack position. This study systematically demonstrates the feasibility of employing torsional-like guided waves for rapid defect detection in T-type welds. By enabling a single excitation to scan a long weld section, the proposed method transforms inspection efficiency from point-wise to linear scanning, offering a promising solution for high-throughput NDT of welded structures in practical engineering applications.
2026, Vol. 47 
