Artificial intelligence for help in decision making during non Destructive Testing of Materials

In non Destructive Testing of Materials, diffracted ultrasonic waves are used in semi automated techniques for crack detection. Signals are displayed as images and artificial intelligence allows automated interpretation in order to give a help in the decision making, especially when large structures are inspected (pipelines, reactor vessels...). This paper describes a new approach for data storage, namely sparse matrix structure that avoids image formation. Only the coordinates of pertinent samples regarding a detected defect are stored in a 2D array. In addition to reducing significantly the amount of data to store and to process, sparse representation make possible the exploitation of pattern recognition methods such as Least Mean Square (LMS) algorithm to automatic interpretation. Indeed, when a crack is presented in a controlled structure, the graph formed by the sparse matrix elements has a parabolic form and its summit location deals with the crack summit position.