A dictionary-reconstruction approach for separating helical-guided waves in cylindrical pipes

Abstract

Pipes, which have a hollow cylindrical structure, are very susceptible to corrosion in industrial environments. Pipe wall loss can be evaluated by measuring helical-guided waves; however, various paths of these waves are easily superposed in the received signals due to the cyclic nature of the pipe’s geometry. The overlap phenomenon can be more serious when the guided wave contains multimodes. A signal-processing approach based on dictionary reconstruction for separating helical-guided waves in cylindrical pipes is established to get rid of this dilemma. The full-wave packet with a single mode and single path can be extracted successfully utilizing this method. In this work, the pipe wall is expanded into an infinite plane and a linear wave number model is adopted with the assumption of a narrowband excitation at the initial stage. After that, an over-completed dictionary containing all modes and propagation paths is designed based on this model to decompose the received signal. The specific process of decomposition is implemented through a single-layer neural network to solve the mode and path weight factors. Simulated and experimental cases are carried out to verify the effectiveness of the proposed approach. The results show that this method can completely separate the wave signals of all modes and paths even when the process is full of noise. This signal-processing approach can provide an important technical reference for high-precision imaging of helical-guided waves.