Computational Approach to Optimization of Heterogeneous Metallic Materials Using Microstructure-Based Constitutive Model
- Title
- Computational Approach to Optimization of Heterogeneous Metallic Materials Using Microstructure-Based Constitutive Model
- Authors
- 박형근
- Date Issued
- 2021
- Publisher
- 포항공과대학교
- Abstract
- Microstructural features in metallic materials control the bulk properties and the performances by interacting each other during deformation. New alloy-design always has great potential to reach the outstanding area on the conventional property space through totally different deformation mechanism, but at the same composition, microstructure is the most dominant factor in physical properties. In order to achieve the best performance within a confined property space, many studies has been active on improving physical properties by controlling microstructure through various process including severe plastic deformation (SPD), surface modification, and heat treatment. Recently, new attempts have been made to manufacture more complex microstructures in order to surpass the limitations of conventional microstructures. These complex microstructures with the superior performances by artificially generated ‘spatial heterogeneity’ are called heterostructures. Although heterostructures have higher potential for improvement than homogeneous structures, it is more difficult to determine the correlation between structure and properties as the structure becomes complicated.
Today, computer aided engineering (CAE) is the essential task in accelerating material design by narrowing down the area of interest in the infinite number of cases. Three-dimensional (3D) microstructure-based model offers an understanding of the mechanical behavior of materials from microscopic level, which allows more accurate prediction of final properties in bulk level. Nevertheless, there are some practical issues in implementing this model such as the complexity of modeling the unique deformation mechanism considering microstructural features and the high computational cost, especially for heterostructures. From this point of view, this thesis is dedicated to two goals: i) to develop a constitutive model for heterostructure based on its unique deformation behavior, and ii) to optimize microstructures with efficient investigation algorithms combined with finite element (FE) simulation.
To establish the deformation mechanism of heterostructure, deformation behavior of harmonic structured materials (HSMs), a type of 3D isotropic heterostructure, is investigated in microscopic level. The high strain partitioning and the steep strain gradient is observed along the grain size gradient, due to the plasticity incompatibility during deformation. This incompatibility is compensated by geometrically necessary dislocations (GNDs), which cause hetero-deformation induced hardening. The quantitative analysis is performed for this additional hardening effect in the heterostructure, and a constitutive model is developed based on the relationship between the spatial heterogeneity and GND evolution.
Microstructure-based simulation requires not only a reliable model but an appropriate representative volume element (RVE). 3D RVE is acquired by synthesizing 3D microstructure as statistically equivalent to real microstructure. To evaluate the effectiveness of 3D microstructure modeling, the thermo-mechanical FE simulations of Al-SiC composite are performed using statistically synthetic microstructures and other conventional models. It is confirmed that this scheme with statistical information is very efficient in terms of the accuracy, flexibility, and computational cost.
Finally, the structure-property (SP) linkage for a heterostructure is constructed by Bayesian inference using microstructure-based simulations. Gaussian process regression (GPR) is executed for probabilistic investigation based on expected improvement (EI) and uncertainty of predictions. Gaussian process is utilized as a surrogate model between the low-dimensional representation of microstructures using correlation statistics and the physical properties predicted by FE simulation. As a result, SP linkage for SS304L HSM is successfully built using a small number of FE simulations, exhibiting high accuracy over a wide range of microstructures. In addition, the investigation based on the EI proposes the optimal heterostructure with target properties using even smaller number of full-field simulations, which is identical to the microstructure revealed by FE simulation for all candidates. The proposed framework provides an efficient means of accelerating the design of general heterostructures.
- URI
- http://postech.dcollection.net/common/orgView/200000368036
https://oasis.postech.ac.kr/handle/2014.oak/111417
- Article Type
- Thesis
- Files in This Item:
- There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.