A method to improve and automate flat defect detection during ultrasonic inspection

In the nondestructive testing of materials, ultrasonic imagery can detect and characterize defects that are present in a structure. Data are displayed in the form of images, and processing algorithms can be applied for automatic detection and characterization. However, when using diffracted waves, the amplitude is often too low, and the signals are difficult to distinguish from the noise. Other times, the volume of data requires significant computation time. In this paper, we propose a method that can avoid image formation by replacing it with a sparse matrix and significantly reducing the amount of data to process; this allows for the enhancement and the automation of the detection of thin and flat defects such as cracks. The elements of the sparse matrix form a curve, which is sufficient to characterize defects in many cases. These elements are selected from diffracted signals using the split-spectrum processing method. In this way, the signal-to-noise ratio is improved, and the position of the echo signal is accurately determined. When a crack is present in a material, the points of the sparse matrix form a parabola and classical tools of pattern recognition such as the Hough transform can detect it, thus providing significant help in decision-making processes. Copyright (c) 2011 John Wiley & Sons, Ltd.