Back to search

Georgia Institute of Technology

GPR-based Process-Structure-Property Linkages in The MKS Framework

Abstract

dc:description.abstract

Establishment of low-cost high-fidelity process-structure-property (P-S-P) linkages are critical to dramatically accelerate the discovery and development of advanced materials. This task is made challenging due to the inherently high dimensional, problem-specific features which describe the material structure. The recently developed Material Knowledge System (MKS) framework offers a data-driven framework for developing low-dimensional, information rich representations of the microstructures which are subsequently used to extract the P-S-P linkages of interest. However, the extraction of the reduced-order P-S-P linkages in not a trivial task. This is mainly attributed to (1) lack of a priori knowledge on the model form, (2) often small/sparse material datasets, (3) need for rigorous quantification of prediction uncertainties and (4) need for guiding experiments/simulations. These efforts are made even more challenging when considering P-S linkages. In the development of new/improved materials, it is essential and advantageous to consider the entire evolutional path of the material structure. Thus, corresponding P-S linkage must capture the entire temporal evolutionary path of a microstructure; a problem which is inherently more challenging than the establishment of a S-P linkage. The complexity is further increased due to the need for exploration of the often multi-dimensional processing space. In consideration of all the above, the naïve implementation of many surrogate modelling strategies is impractical. This thesis aims to establish a versatile Gaussian Process-MKS framework for systematic formulation of high-fidelity P-S-P linkages across diverse materials phenomena while leveraging highly informative microstructure quantification. The gaussian process regression model (GPR) is particularly attractive as it does not require a specific model form enabling it to capture the underlying complex nonlinearities of a given material system/phenomena. Furthermore, GPs are well-suited to handle the small/sparse amounts of data often available for materials design. Finally, GPs provide a rigorous quantification of the uncertainty with respect to the model prediction which provides an avenue for intelligent exploration of the aforementioned vast design space of materials. However, the applicability of GP in extraction of P-S-P linkages have thus far been limited to relatively simpler microstructures (e.g., two-phase microstructures) or simplified descriptions of complex microstructures (e.g., grain size distribution). There has yet to be a systematic GP-based effort in studying a broad range of highly complex materials phenomena/systems encountered in advanced materials. In response, this work focuses on the utilization of GPs for suitable classes of materials problems: (1) When material dataset is rather small/sparse (and covers a vast parameter space) (2) When material dataset is rather large (often treated as ill-suited for GP usage) (3) When time evolution is to be accounted for (P-S evolution linkages). The applicability of the GP-MKS framework is shown through multiple case studies. More specifically, the charge transport properties of organic photovoltaic polymers (S-P linkage) and microstructure evolution in polycrystalline materials undergoing static recrystallization and large-scale plastic deformation (P-S evolution linkages) are studied. In each case study, suitable extensions to MKS were made to best suit the specifications of the material problem at hand. The remarkably accurate robust P-S-P linkages established for the diverse set of studied problems attest to the versatility and tremendous potential of the developed GP-based MKS framework.

Degree

thesis:*
Level thesis:degree_level
Doctoral
Department dc:contributor.department
Mechanical Engineering
Grantor dc:publisher
Georgia Institute of Technology
Year dc:date.issued
2022

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Hashemi Yazdi, Sepideh Sadat
Advisor dc:contributor.advisor
  • Kalidindi, Surya R.
Committee members dc:contributor.committeemember
  • McDowell, David
  • Neu, Richard
  • Barrales Mora, Luis
  • Garmestani, Hamid

Subjects

dc:subject × 6

Rights

Language dc:language.iso
en_US

Identifiers

dc:identifier.*
Handle dc:identifier.uri
https://hdl.handle.net/1853/72466

Chain of custody

source
Harvested from
Georgia Tech
Base URL
repository.gatech.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Hashemi Yazdi, Sepideh Sadat. GPR-based Process-Structure-Property Linkages in The MKS Framework. Doctoral thesis, Georgia Institute of Technology, 2022. https://hdl.handle.net/1853/72466