Singular spectrum analysis for trend extraction in ultrasonic backscattered echoes

In this investigation, singular spectrum analysis (SSA) is explored to decompose and analyze ultrasonic signals in nondestructive evaluation (NDE) applications. Unlike transform-based algorithms, SSA is a time-series analysis algorithm which is completely driven by the signal itself. As a result, ultrasonic signal is decomposed through a four-step processing, that is embedding, singular value decomposition (SVD), grouping and diagonal averaging. The decomposition result can be used to characterize ultrasonic signals for the NDE of materials. Simulation results show that SSA reveals signal trend related to defects and grain scatters. The performance of the algorithm is also compared with other algorithms of ultrasonic signal decomposition for signal analysis and feature extraction. Numerical and analytical results demonstrate that SSA is effective in ultrasonic signal analysis. Especially the data-driven nature makes SSA unique for signal analysis. The algorithm can be utilized for flaw detection, signal classification, and pattern recognition in NDE applications.