Classification of defects for guided waves inspected pipes by a neural network approach

In this paper the effectiveness of a procedure that allows the flaws characterization of pipes inspected by a long range guided waves is investigated. The method performs the extraction of correlation coefficients between the x, y, z components of the displacement of simulated guided waves reflected by defects on pipes. These features feed a neural network classifier which evaluates the dimensions of well defined geometry defects on the pipe under test. The results show lower error rates in the evaluation of both angular and axial extent of a defect. © 2007 IEEE