Investigation of the approximate decomposition of alternating current field measurement signals from crack colonies

Abstract

Cracks, such as those caused by stress corrosion, are often present as colonies. The pick-up signals from the cracks in colonies obtained through the alternating current field measurement technique are composed of the signals from the total cracks. Using a superposed signal directly causes large errors in depth prediction. Therefore, this study investigates the approximate decomposition of alternating current field measurement signals from crack colonies. The effect of the number and the relative position of the cracks on the measured signals is studied. A method based on the signal circle is proposed to decompose the superposed signals of crack colonies. Further, a depth quantification algorithm, which is based on a generalized regression neural network, is developed for crack colonies. The results of the simulation and experiment indicate that the proposed algorithm can preliminarily predict crack depth in a colony, especially for deep cracks, which are dangerous. The predicted results can also be used as initial values for accelerating phenomenological inversion methods.