Structural health monitoring exploiting MIMO ultrasonic sensing and group sparse Bayesian learning

In this paper, we propose the exploitation of sparse Bayesian learning in multiple-input multiple-output (MIMO) systems to account for the multi-dimensional group sparse nature of extended defects in guided ultrasonic wave based structural health monitoring. The multi-dimensional group sparsity in the underlying reconstruction problems arises due to the clustered spatial occupancy of extended defects and the multiple-aspect MIMO observations. Sparse Bayesian learning techniques have been shown to provide robustness for high-resolution signal reconstruction due to their insensitivity to dictionary coherence and have the flexibility of effective exploitation of the signal structure. The superiority of the proposed technique over the state-of-the-art sparse signal reconstruction techniques is demonstrated through simulations and preliminary experiments.