Abstract
In recent years, image processing has been applied to metallurgy but its applications were mainly confined to iron. This research instead focuses on recognizing phases present in steel. Phase recognition is very important because it aids in specifying the properties of steel specimens. This research is concentrated on recognizing Pearlite, Ferrite, Martensite and Cementite present in steel. A neural network was used to recognize these four phases. For a neural network to recognize any object it requires input values called features. This research introduces a combination of features which were not used before: texture based features of entropy, energy, contrast, and homogeneity along with a count of significant peaks on the histogram and the percent of black pixels present if the image is converted to a binary format. A neural network was successfully trained to recognize Pearlite, Ferrite, Martensite and Cementite. In addition information was provided regarding problems faced with Martensite and Cementite.
Date of publication
Fall 11-12-2013
Document Type
Thesis
Language
english
Persistent identifier
http://hdl.handle.net/10950/178
Recommended Citation
Kesireddy, Adarsh, "Artificial Intelligent Metallurgical Grain Detection" (2013). Mechanical Engineering Theses. Paper 2.
http://hdl.handle.net/10950/178