Date of Award
Open Access Thesis
Computer Science and Engineering
Mechanical and structural properties of ultra-high carbon steel are determined by their microstructures composed of constituents such as pearlite and spheroidites. Locating micro constituents and quantitatively measuring its presence is key for material researchers to study the physical properties of the carbon steel materials. This micrograph analysis is currently done manually and subjectively by material scientists, which is tedious and time-consuming. Here we propose to apply the image segmentation algorithm called U-Net to achieve automated labeling of steel microstructures on a subset of ultra- high carbon steel image dataset containing pearlite and spheroidite as the primary micro constituents. Our work provides an automated way to micrograph segmentation using the deep learning algorithm. Our prediction model only needs annotating a few micrographic images manually, which are used to train the segmentation algorithm. The trained model will help the researchers to automatically annotate new micrograph images. In this work, 20 micrographs containing pearlite and spheroidite micro constituents are first manually annotated. Then this dataset is used to train the conventional U-Net segmentation model. The trained U-Net model successfully performed segmentation on new micrograph images containing pearlite and spheroidite with an accuracy of 87.39%. We also contribute the 20 annotated image dataset to public access. Our approach can be further extended to the rest of the UHCS dataset and it will help the material researchers to automate the process of locating and analyzing complex microstructure which otherwise needs a lot of manual labor.
Suresh, S. K.(2019). Machine Learning Based Ultra High Carbon Steel Image Segmentation. (Master's thesis). Retrieved from https://scholarcommons.sc.edu/etd/5519