Andrew Marshall

Advisor: Prof. Surya Kalidindi

will defend a doctoral thesis entitled,

Gaussian Process-based Workflows for Efficient Extraction of Process-structure-property Linkages

On

Wednesday, May 14 at 8:00 a.m.

via 

Microsoft Teams Video Conferencing 

https://teams.microsoft.com/meet/2497895949991?p=cIaB8dc8HJyiseOQ4H

Committee
             Prof. Surya Kalidindi – School of Materials Science and Engineering (advisor)
            Prof. Matthew McDowell – School of Materials Science and Engineering
             Prof. Hamid Garmestani – School of Materials Science and Engineering      

            Prof. Roshan Joseph – H. Milton Stewart School of Industrial and Systems Engineering

            Dr. Balasubramaniam Radhakrishnan, Oak Ridge National Laboratory

            Dr. Jim Belak, Lawrence Livermore National Laboratory, Retired

Abstract:

Materials processing consists of a vast range of operating conditions, which can have a profound impact on the resulting material structure. In addition, the range of possible microstructural features and morphologies is also extremely large. Therefore, an approach is required to explore these spaces and produce materials knowledge in an efficient manner. This work seeks to address this problem with a new process-structure-property (PSP) framework based on Gaussian Processes (GPs). Gaussian processes are able to produce predictions with uncertainty quantification, which may be utilized for the adaptive selection of new data. The goal of this work is to demonstrate the ability of GPs to extract PSP linkages efficiently for a variety of materials and physical processes.

The first project demonstrates a novel strategy for the autonomous development of a machine-learning model for predicting the equivalent stress–equivalent plastic strain response of a two-phase composite calibrated to micromechanical finite element models. The model takes a user-defined three-dimensional, two-phase microstructure along with user-defined hardening laws for each constituent phase, and outputs the equivalent stress–plastic strain response of the microstructure modeled using J2-based isotropic plasticity theory for each constituent phase. It is demonstrated that the use of Gaussian process regression (GPR) together with a Bayesian sequential design of experiments can lead to autonomous protocols for optimal generation of the training dataset and the development of the model. It is shown that this strategy dramatically reduces the time and effort expended in generating the training set.

The second project demonstrates a novel Gaussian process-based approach for the prediction of final microstructures for thermoelectric Mg2SixSn1-x alloys in a spinodal decomposition process. A chained Gaussian process framework is introduced, with a classification model predicting microstructure homogeneity vs. heterogeneity, and a regression model to predict the final structure for heterogeneous cases. Additionally, multiple active learning strategies are investigated for updating the training datasets for both models. It is demonstrated that the combination of the chained Gaussian process framework with active learning is able to extract process-structure linkages with high accuracy, even while using far less training data compared to previous approaches.

The third project seeks presents a Gaussian process-based framework to the extraction process-structure linkages during the time evolution of multiphase microstructures. Gaussian Process Autoregression (GPAR) is used to predict solid-state phase transformations of Ni-based superalloys in the context of an additive manufacturing process. These alloys contain multiple inclusion phases, with each inclusion phase having multiple crystallographic variants. It is shown that, even for a very large number of local states, reduced-order microstructure representations show clear differences based on their corresponding processing parameter values. A sparse-variational GPAR model is constructed and demonstrates highly accurate predictions of the microstructure evolution paths, capturing the competitive growth between inclusion phases based on the Nb concentration.