Localized damages like flaws, due to its finite number, are essentially sparse in the spatial domain and result in a few echo signals that exist sparsely in a received pulseecho signal. Therefore, flaws may be effectively represented by sparse representation techniques. In this study, a specified sparse inversion method for ultrasonic flaw signals in Bayesian framework is proposed. Based on a sparse Bayesian representation method, noisy ultrasonic data can be represented effectively based on a specially designed over-complete dictionary. The appropriate coefficients for flaw echo signals are chosen by pruning operation and can be further enhanced by thresholding operation. The resulting coefficients are used to reconstruct flaw echo signal. The capability of the proposed method is evaluated by a noisy data acquired from experiment. Results show that the proposed method has very good performance for flaw detection and flaw signal reconstruction against noise.
doi: 10.12783/SHM2015/327