Use this URL to cite or link to this record in EThOS:
Title: Image processing for the analysis of TI-6AL-4V microstructures
Author: Campbell, Andrew
Awarding Body: University of Strathclyde
Current Institution: University of Strathclyde
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Access from Institution:
The primary aim of the work presented in this thesis is to develop new and improved methods for analysing the microstructure of titanium alloys, specifically Ti-6Al-4V. This is achieved through the introduction of a software tool which incorporates novel image processing techniques to automate the measurement of a wide range of microstructural features in microscopic images of Ti-6Al-4V. It is shown that these measurements are performed in a faster, more repeatable way, with minimal input from expert materials scientists, when compared with existing methods. The microstructure of a material consists of individual grains of different shapes and sizes. Precise analysis of these provides information about the properties of the material and thus is valuable for quality assurance and development of new materials, models or manufacturing processes. However, performing this analysis usually depends manual analysis techniques, requiring extensive input from expert materials scientists. Attempts at more efficient automated methods have so far had limited success, largely due to the wide variations that can occur in microstructural images. In this thesis, a robust set of image processing techniques are proposed to automatically identify and measure key features of microstructural images. Due to the unique challenges posed by different microstructure types, two separate techniques are proposed; one aimed at measuring globular grains within microstructures and another aimed at measuring elongated grains, known as platelets. Measurements of globular grains are obtained using a novel segmentation algorithm that partitions the image such that each grain is individually labelled. Once identified in this way, the size and shape of each grain can then be measured, allowing both individual and aggregated measurements to be reported. The algorithm includes a variety of pre- and post- processing steps that dramatically reduce measurement errors, common in other segmentation methods. However, it was found that this method could not be reliably applied to segment and measure platelets. Measurements of these are instead determined by shape fitting techniques. Similar approaches are already used for analysing this type of elongated object in similar datasets. However, several limitations exist that negatively affect both the accuracy and usefulness of these methods when applied to the dataset in this study. This thesis proposes novel adaptations to such techniques to improve reliability and extend the range of properties that can be measured. The resulting technique can be applied to measure platelet width, orientation, and morphology. A separate algorithm is also proposed that uses this data to identify and measure colonies of platelets. A software tool is designed to allow these tools to be deployed by materials scientists. This tool provides simple, intuitive feedback to the user to allow the proposed algorithms to be properly parameterised without image processing experience, or a good understanding of their operation. This work is validated through a range of trials conducted on real world datasets of Ti-6Al-4V microstructural images. The dataset uses images from different microscopy technologies as well as different morphological types of microstructure. Results are validated through comparison with measurements performed by expert materials scientists using the most reliable procedures currently available, and demonstrate accurate results can be achieved automatically, significantly faster than before. The techniques proposed in this thesis also have general value in the wider domain of image processing, due to their robustness in challenging datasets. This is demonstrated by a detailed comparison with existing image processing tools, never before testing on microstructural data.
Supervisor: Marshall, Stephen ; Murray, Paul ; Yakushina, Evgenia ; Ion, William J. Sponsor: Not available
Qualification Name: Thesis (Eng.D.) Qualification Level: Doctoral