The use of X-ray as an inspection technique to ensure the integrity of industrial products dates from the beginning of the 20th century. Therefore, it is a tool of non-destructive inspection widely known. Nowadays, however, there are several researches forward on the optimization of such inspection technique, mainly for the development of an automatic system of radiographic image analysis. That is, a system that can identify and classify the defects in the radiography. An important step in the construction of this system is the classification of defects, which is usually done by using some of their features. The purpose of this work is to study the relevance of some defect features in order to classify some of the main classes of defects. The employed technique is the linear correlation between the defect features and the classes of defects. A non-linear pattern classifier is used, implemented by a neural network, to evaluate the performance in the classification. The results showed the efficiency of the method used.
nondestructive tests; radiography; welding defects; neural networks
