Broadband Sparse Decomposition of Lamb Wave Signals

Lamb wave is a promising tool for structural health monitoring and nondestructive testing. During propagation, dispersion and multi-mode characteristics generally result in overlapping of wave packets. Sparse decomposition is developed to address the overlap problem. Traditional dictionary design in sparse decomposition is to search optimal matching between dictionary atoms and propagating wave packets. However, design accuracy is limited in broadband Lamb wave because propagating wave packet is difficult to predict with severe amplitude modulation and velocity dispersion. To overcome this limitation, broadband Lamb wave sparse decomposition is proposed. Firstly, tone burst models are used as dictionary atoms to match waveform segment rather than global wave packet. Subsequently, solved sparse vector by basis pursuit algorithm is reshaped into a time-frequency matrix. Finally, non-zero values in the matrix are extracted for reconstruction of separated wave packets. Experiments are given to demonstrate the effectiveness of proposed method.