Matching pursuit with novel dispersive dictionary for mode separation in guided wave signals obtained from pipes

Pipes that carry gas into residential buildings are one of the most important infrastructures in cities. These pipes are prone to corrosion after a period of time because of different environmental factors. In order to avoid any pipe rupture that may cause explosion, the integrity of such pipes much be checked on a regular basis. Ultrasonic guided wave is an effective nondestructive testing (NDT) method for structural health monitoring (SHM) of plate like structures and pipes. However, the signals that are obtained from guided waves can be very difficult to interpret due to mode conversion, overlapped modes and low signal to noise ratio. In order to extract meaningful information of guided wave signals, using an advanced signal processing technique is inevitable. Most of the signal processing methods that have been developed so far, neglect the dispersion characteristic of guided waves. Dispersion means that the velocity of the propagating wave is a function of its frequency. As in most guided wave applications for inspecting pipes, we use narrowband tone-burst signal, dispersion causes guided wave signal to spread in time axis as it travels and changes its shape. Matching pursuit is one the greedy algorithm that can be used to approximate signals. This method iteratively approximates a signal by its predefined dictionary. However, the dictionary that usually contains redundant number of fundamental waveforms usually Gaussian signals may not be suitable for guided wave inspection of pipes. In this paper we designed a dictionary based on real guided wave signals by taking finite element method (FEM) approach.