Aplicación de la herramienta svm en matlab para la cuantificación de ferrita y perlita en el acero AISI/SAE 1045

Abstract

The following project opens a more detailed door to using the SVM support vector machine tool with Kernel gaussiano in matlab for data that is not linearly separable, to classify two metallographic properties, particularly ferrite and pearlite, in addition to working on matlab with image processing through the Image Processing Toolbox on a metallographic image of AISI / SAE 1045 steel in its commercial state through the identification of colors to quantify the amount of ferrite and pearlite present in the steel. The metallographic image to work in matlab is taken from the institutional repository of mechanical engineering degree projects of the Francisco José De Caldas District University of the project Microstructural comparison of 1020, 1045 and 8620 steels tempered from intercritical and tempered temperatures. In general, the analysis of metallographic image processing is done through other software such as statistical R or Excel, likewise the application of the SVM (support vector machine) is also done analytically or through software such as Excel, but In the case of matlab, the use of this tool is not very common, due to this, deepening the study of the SVM tool and the processing of metallographic images in matlab will optimize processes in the study of steels. Usually, matlab can work with the SVM tool for data that are linearly separable, and for data that are not linearly separable, it can make use of different types of Kernel, in this project due to the Cartesian arrangement of the ferrite and pelite points. The Gaussian kernel was worked. It is important to establish that the method to work for the processing of the metallographic image to quantify the amount of ferrite and pearlite is based on the application of the image Processing Toolbox. With this matlab tool, a logical matrix is ​​first established by binarizing the metallographic image In order to determine some characteristics of the colors, the number of pixels determines the quantification of the ferrite and pearlite of the metallographic image for AISI / SAE 1045 steel in its commercial state.