Back to search

Virginia Tech

ML-accelerated pipeline for understanding atomistic hardening of MPEAs and predicting hardness in additive manufacturing

Abstract

dc:description.abstract

MPEAs, including HEAs and MEAs, have demonstrated exceptional mechanical performance, such as high hardness and high strength-to-weight ratios under extreme conditions. However, designing MPEAs is challenging due to the high dimensionality of the compositional and processing search spaces. Active learning methods have recently been applied to address the challenges in designing MPEAs, leading to discoveries such as the TaCoSi alloy, which exhibits a remarkably high Vickers hardness of 1269. On the other hand, at a larger scale, accurately predicting mechanical properties remains challenging in additive manufacturing due to the heterogeneous, high-dimensional nature of the manufacturing data. In this thesis, we present two complementary computational frameworks that together advance the understanding and design of high-performance alloys. In the first work, we develop a machine-learned force field (MLFF) for the TaCoSi system and conduct molecular dynamics simulations to explain the underlying atomistic hardening mechanisms. Training data is generated through a combination of single-point DFT and AIMD calculations, and used to fit a machine-learning force field. The resulting force field is implemented to investigate short-range ordering, bulk defects and crystal formation that contribute to hardness and is evaluated with a universal force field. In the second work, we address the challenge of predicting the mechanical properties of additively manufactured materials produced via laser powder bed fusion (LPBF). We propose a multimodal machine learning framework that integrates microstructural images with processing parameters to predict the hardness of the manufactured alloys. Convolutional autoencoders are used to encode spatial microstructural features, which are then combined with processing conditions in a supervised predictive model. Results show that incorporating microstructural features alongside process parameters improves prediction accuracy relative to a unimodal model. And highlights the influence of the microstructure on property prediction. Together, these works demonstrate a machine learning and atomistic calculation pipeline that accelerates the understanding of MPEAs and a multi-modal framework for additively manufactured alloys to understand process-structure-property relationships.

Degree

thesis:*
Name thesis:degree_name
Master of Science
Level thesis:degree_level
masters
Discipline thesis:degree_discipline
Aerospace Engineering
Department dc:contributor.department
Aerospace and Ocean Engineering
Grantor dc:publisher
Virginia Tech
Year dc:date.issued
2026

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Imtiaz, Nafiz
Chairs dc:contributor.committeechair
  • Sharma Priyadarshini, Maitreyee
  • Fu, Yao
Committee member dc:contributor.committeemember
  • Bai, Xianming

Subjects

dc:subject × 7

Rights

dc:rights
Statement dc:rights
  • In Copyright
Language dc:language.iso
en

Identifiers

dc:identifier.*
Dc Identifier Other
vt_gsexam:46816

Chain of custody

source
Harvested from
Virginia Tech
Base URL
vtechworks.lib.vt.edu/oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Imtiaz, Nafiz. ML-accelerated pipeline for understanding atomistic hardening of MPEAs and predicting hardness in additive manufacturing. masters thesis, Virginia Tech, 2026. https://hdl.handle.net/10919/143258