Dissertation/Thesis Abstract

Gray-cast iron classification based on graphite flakes using image morphology and neural networks
by Yeliyur Siddegowda, Darshan, M.S., California State University, Long Beach, 2016, 55; 10017846
Abstract (Summary)

Gray-cast iron is an iron carbon alloy which is regularly used in manufacturing processes. Carbon is distributed in the iron material in the form of graphite. The distribution of the graphite flakes in the alloy contributes greatly towards the chemical and physical properties of the metal alloy. Thus it is important to identify and classify the Gray-cast iron based on the morphological parameters of the graphite flakes. Gray-Cast iron is classified into five types in ISO-945 represented with the letters A through E. These five classes possess different structures or distributions of the graphite flakes. The current project presents an automated classification method using image processing and machine learning algorithms. The method presented here obtains the required parameters from the microstructure through image morphological operations. The image information is subsequently fed through a supervised machine learning algorithm which is trained using parameters such as area of the flakes, perimeter, minimum inter-particle distance and chord length from over twenty samples. The algorithm calculates the percentage of the type of the flakes present in the given image. The simulation is done in MATLAB and was tested for six images in each class. Class C and D were classified with 100 percent accuracy, Class A and B were classified with accuracy of 82 percent and Class E was identified with accuracy of 68 percent.

Indexing (document details)
Advisor: Chassiakos, Anastasios
Commitee: Ary, James, Tran, Boi
School: California State University, Long Beach
Department: Electrical Engineering
School Location: United States -- California
Source: MAI 55/03M(E), Masters Abstracts International
Source Type: DISSERTATION
Subjects: Electrical engineering
Keywords: Image processing, Machine learning
Publication Number: 10017846
ISBN: 9781339505411
Copyright © 2019 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest