A Simple Image Segmentation Approach to Overcome Microstructure Analysis and Homogenization Problems

• Main Author: Visconti, João Pedro Portella Guedes
• Format: Dissertation
• Language: English
• Published: ProQuest Dissertations & Theses 01-01-2021
• Subjects: Boundary conditions; Cooling; Homogenization; Industrial engineering; Manufacturing; Materials science; Mathematics; Mechanical properties; Mechanics; Microstructure; Software; Stainless steel; Steel alloys
• ISBN: 9798382141848

Multiscale modeling has proven to be a very useful tool in the production of new manufacturing methods, in the development of quality products, reducing production costs, and creating new materials. The need to improve some characteristics of this process drive the motivation for the current study that was developed at the Faculty of Engineering of the University of Porto (FEUP) together with INEGI - Institute of Science and Innovation in Mechanical and Industrial Engineering. The work developed in the thesis is complementary with a broader project (msCore) that aims at the development of a multiscale analysis model for advanced material manufacturing.Therefore, the main goal of this work is to build a multiscale approach to derive macro properties of Duplex Stainless Steel from experimental images of the microstructure. In this way, this study uses image processing tools to recognize the different phases in the microstructure and adapt them to a file format that allows the integration to a finite element software. In turn, having this file, micro properties are homogenized to obtain information regarding the macroscale properties of this material.In order to develop the referred software, a bibliographic study was made in order to observe similar problems and their resolutions. The advantages and disadvantages of different authors’ studies were taken into account. The limitation regarding microstructure analysis of the material was addressed with a script made in Matlab that identifies the features in the microstructure image and converts them into numerical data to be used in a Finite Element software. One of the main issues that the msCore project presented was the transcription of the visual experimental data into a model for numerical simulation. A Python script is used to extract the needed information from the results simulations obtained using ABAQUS commercial software. Thereat, another Matlab script was created to obtain the homogenization of the properties and predict the elasticity matrix for the material in the microstructure.In order to evaluate the developed software, a case study is presented. The steps of microstructure analysis, Finite Element simulation, and assembly of the elasticity matrix with real images of the microstructure of a Super Duplex Stainless Steel sample were applied. The results demonstrate the influence of the percentage of each phase present in the microstructure in the elastic material properties. The influence on inelastic material properties should be addressed in future work.