Classification of ductile cast iron specimens based on image analysis and support vector machine

Abstract

Abstract: The ductile irons discovery in 1948 gave a new lease on life to the cast iron family. In fact, these cast irons are characterized both by a high castability and by high toughness values, combining cast irons and steel good properties. The high mechanical properties (especially ductility) are mainly due to the peculiar graphite elements shape: thanks to the addition of some elements like Mg, Ca, Ce, graphite elements shape can be near to spheres (nodules) instead to lamellae as in “normal” grey cast irons. In this work, the problem of classification of ductile cast irons specimens is addressed; first the nodules present in each specimen are identified determining their morphological shapes. These characteristics are suitable used to extract global features of the specimen. Then it is outlined a procedure to train a classifier based of these properties.