Files in this item



application/pdfBICHARAVIEIRA-DISSERTATION-2021.pdf (13MB)
(no description provided)PDF


Title:Strain inhomogeneities during plasticity and creep of metals: representative volume elements and microscale strain predictions
Author(s):Bichara Vieira, Renato
Director of Research:Lambros, John
Doctoral Committee Chair(s):Lambros, John
Doctoral Committee Member(s):Chasiotis, Ioannis; Chew, Huck Beng; Sehitoglu, Huseyin
Department / Program:Aerospace Engineering
Discipline:Aerospace Engineering
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):Digital Image Correlation
Representative Volume Element
Neural Network
Microscale Strain Filed
Abstract:This work investigated microscale strain inhomogeneities during plasticity and creep of an austenitic stainless steel alloy, namely stainless steel 709. Digital image correlation (DIC) was employed to experimentally measure strain fields at the microstructural level (defined as a level comparable to the grain scale) and combined with electron backscatter diffraction (EBSD) data of the microstructure. The overarching goal of this work was to investigate the dependence of such microscale strain inhomogeneities on loading type, temperature, and microstructural parameters. The first part of this work explored the microscale response of austenitic stainless steel 709 to plasticity and creep loading. Macroscale experiments were performed in which far-field load-displacement data and in situ DIC measurements were made. These macroscale experiments were used to determine specific loading conditions to be investigated in more detail with high-resolution microscale measurements. Subsequently, a high-resolution digital image correlation technique (HiDIC) was successfully applied to measure strain fields at the grain level of samples subjected to plasticity and creep loading conditions over a range of maximum stresses, temperatures (room temperature, 300°C, 500°C, 650°C), and hold times (from 15 s to 30 min). The measured microscale strain fields were compared, and showed that both creep and plasticity produce similar highly inhomogeneous strain fields. Furthermore, localization of strains primarily near grain boundaries was observed for all cases investigated, with no visible difference in the patterns of strain accumulation for creep and plasticity. In the second part of this work, the strain-based representative volume element (RVE) resulting from the plastic and creep loading of alloy 709 samples was explored. The RVE is a key concept behind homogenization techniques used to correlate microscale behavior with macroscale (i.e., far-field) response of the material. Therefore knowledge of the RVE size is an essential piece in many multiscale modeling frameworks. A robust experimental method to measure the strain-based RVE size from HiDIC microscale strain fields was developed through the exploration of the statistical nature of the RVE. The proposed stereological approach takes randomly selected boxes from within a measured strain field and determines the statistical distribution of the size of the strain-based RVE from the average strains for different box sizes. The proposed method was then applied to measure strain-based RVE sizes on samples subjected to creep and plasticity loading conditions, over a range of temperatures (room temperature, 300°C, 500°C, and 800°C), maximum stresses (25% below and 25% above yield strength) and hold times (5 min to 1 h), with the relationship between the loading conditions and the strain-based RVE size being explored. Through a relevant coordinate transformation, local strain accumulations near grain boundaries (mantle regions) were resolved and their correlation to the strain-based RVE size was also explored. Samples with higher local normal to shear strain ratios were seen to have larger strain-based RVE sizes. The reason that both the RVE size and the local normal to shear strain ratios near grain boundaries behave in a similar fashion is because they are both controlled by the same underlying deformation mechanisms. In the final part of this work, a methodology for applying neural network algorithms to predict microscale strain fields was employed. The approach used each correlation point from HiDIC strain fields as an input data point for the training of a neural network, allowing for a large quantity of training data (up to 70,000 data points in each case) to be obtained from a relatively small number of experiments. The grain boundary inclination angle to the loading direction was shown to be a good predictor for the average residual strain accumulated inside the mantle regions (i.e., near grain boundaries) for all cases investigated, even when used as the sole input parameter.
Issue Date:2021-03-12
Rights Information:Copyright 2021 Renato Bichara Vieira
Date Available in IDEALS:2021-09-17
Date Deposited:2021-05

This item appears in the following Collection(s)

Item Statistics