Dissertation/Thesis Abstract

Statistical microstructure generation and 3D microstructure geometry extraction
by Sintay, Stephen D., Ph.D., Carnegie Mellon University, 2010, 175; 3421737
Abstract (Summary)

3D microstructure modeling is a powerful way to study mesoscale mechanisms and phenamena and to explore the effect that the microstructure may (or may not) have on material performance. This work focuses on processes for generating statistically representative implicit microstructure models of polycrystalline materials, and extracting explicit geometries from implicit microstructure data. The generation methods are based on quantifying grain size and shape, grain orientation distribution, and grain misorientation distribution, which are obtained from orthogonal Electron BackScatter Diffraction (EBSD) scans of polycrystalline materials. This is followed by generation of a representative volume of synthetic material whose distributions match those of the observed microstructure. An example of statistical microstructure generation for aluminum alloy AA7075—T651 is given, where the distribution of the synthetic microstructure features are a close match to that of the EBSD observations. The synthetic aluminum alloy can then be used for physics—based modeling of microstructurally small fatigue cracks.

Synthetic materials generation, as described above, defines the geometry of the polycrystalline microstructure implicitly and obtaining an explicit geometry is expedient for generating a volumetric mesh for future finite element analysis. A novel method is presented that uses the centers of mass of linear portions of the dual grid polygon to define the geometry of the triple line network. The location of the triple line network is constrained to be within the acceptable error bounds as defined by the implicit data. The triple line network is then used as a framework for triangulating the interfaces between each region. Using the dual grid method to define the triple line network essentially reduces the multi—region data into patches of binary data. The interfaces between two regions are modeled with triangulated meshes. Trimming, stitching, and deformation with a moving finite element method are steps used to create the surface meshes. The partial entity structure boundary representation is used as a framework for defining the interface geometry of the non—manifold, multiple—region microstructure data.

The dual grid center of mass method provides a well defined set of rules such that the uncertainty of the inclination angle of a 2D geometric feature obtained from this method is explicitly defined. Further, the entire 3D multi—region geometric modeling strategy is tested for accuracy and fitness by using 3D Phantom geometries. Implicit data sets are generated from the explicit phantoms by sampling the phantoms through a range of resolutions, and these implicit data set are then reconstructed. The reconstructed models are tested for error against the phantoms to characterize the accuracy of the reconstruction techniques as a function of resolution. The error of the reconstructed geometries is reduced with increasing resolution. However, the mean width of the reconstructed regions are consistently lower than the phantoms.

The geometry extraction methods are used on the digital microstrucres for AA7075-T651 and for data obtained from molecular dynamic simulations.

Indexing (document details)
Advisor: Rollett, Anthony
Commitee:
School: Carnegie Mellon University
School Location: United States -- Pennsylvania
Source: DAI-B 71/10, Dissertation Abstracts International
Source Type: DISSERTATION
Subjects: Materials science, Remote sensing
Keywords: Fatigues, Grain orientation, Microstructure geometry extraction, Microstructures, Polycrystalline materials
Publication Number: 3421737
ISBN: 978-1-124-21636-2
Copyright © 2019 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest