الفهرس 
Chapter one introductionOnly 14 pages are availabe for public view
 |  | from | 100 |  |  |
| | | from | 100 | | |
Abstract
The microstructure of a material is one of the main influences on its mechanical properties, and therefore it participates in determining its uses and possible applications in the field of manufacturing. Ultrahigh-carbon steels (UHCSs) have remarkable structural properties when processed to achieve fine ferrite grains with fine spheroidized carbides. The spheroidization annealing, which results in a microstructure of fine spherical cementite particles in a soft ferritic matrix, is of significant interest for industrial applications. Understanding the effects of heat treatment variables on final carbide morphology can increase the efficiency of the produced steel of optimal cold formability. So, it`s an urgent need to correctly identify and classify these spheroidite morphologies. The automatic recognition which used for microstructure classification is a major challenge. Nowadays, Deep learning is the most exciting method that used to classify the microstructure of matter automatically. In this work, six outstanding Fully Convolutional Neural Networks (FCNNs) architectures were used to study their capabilities in classifying Spheroidite microstructural images into classes (6classes and 3classes). Three tasks were accomplished to test the classification of different classes of Spheroidite micrographs that were divided into based on different annealing conditions. The constructed Datasets were comprised of the images that are taken over a range of magnifications. The six networks were compared to assess their performance during the supposed tests. The comparison includes the all possible combinations of training datasets size, the learning rate, the cropping method of images, number of classes and the magnification scale.