Degree

Doctor of Philosophy (PhD)

Department

Engineering Science

Document Type

Dissertation

Abstract

Refractory high entropy alloys (RHEAs) are gaining prominence because of their balanced mechanical properties and are regarded as suitable alloys for a variety of applications such as aerospace, structural material, nuclear, and hydrogen storage. The first part of the dissertation explored the mechanical and thermal properties of RHEAs, specifically HfNbTaTiZr and Hf0.5Nb0.5Ta0.5Ti1.5Zr, which have potential uses in the health industry. The first-principles method using 100 atom supercell was carried out to compute the lattice constant, density, elastic modulus, Poisson ratio, and Vickers hardness. These calculated results were consistent with the available experimental reports, which reveal the accuracy of the supercell used in investigating the properties of RHEAs.

The elastic constants of RCCAs are crucial for understanding their mechanical and thermal properties. However, conventional experimental methods to obtain these constants are not efficient, as they require a lot of complex work and are expensive. To overcome these challenges, the elastic constant of different RCCAs was calculated by integrating machine learning (ML) algorithm with the first-principles calculation. Pearson correlation analysis was applied to detect the highly correlated features and improve the quality of the dataset. The final dataset was trained and tested with three regressor models, namely random forest, gradient boosting regressor (GBR), and XGBoost, to determine how well they predict the elastic constants. In comparison to the other two regressor models, the GBR model showed the best performance in predicting the elastic constants of RCCAs, based on performance metrics such as R-squared, root mean square error, and mean average error. Additionally, the reliability of the GBR model was verified through its successful predictions of elastic constants on four new RHEAs MoNbTiVZr, MoNbTiZr, NbTiVZr, and NbTaTiV, that had not been used in the original train and test data. The predicted elastic constants for the four given RHEAs closely match the experiment and simulation data, demonstrating the robustness of the GBR model. The current results illustrate that the GBR model could be applied to estimate the elastic constants of novel RHEAs.

The implication of ML model is extended to experimental data. The alloys manufactured with additive manufacturing (AM) are highly applicable in high temperature applications. However, the experimental measurements of thermal conductivity (TC) are laborious and costly. The composition of alloys can be tailored to huge numbers, and it is difficult to select the right compositions for a targeted set of thermal properties by conducting experiments for each alloy. To overcome the challenges of experiments and speed up the method for novel alloy design with specific thermal property values, ML method is investigated to estimate the TC of AM alloys. To accomplish this, an extensive TC dataset for AM alloys was generated for several AM alloy families (Nickel, Copper, Iron, Cobalt-based) over various temperatures (300–1273 K). Among the five different regression ML models trained with the dataset, extreme gradient boosting (XGB) performed the best with an R2 score of 0.99. Furthermore, the accuracy of this model was tested using Inconel 718 and GRCop-42 fabricated with laser powder bed fusion (L-PBF) based AM, which has never been observed by the XGB model, and a good match between the experimental results and ML prediction was observed. This proves that the thermal conductivity of novel AM alloys could be predicted quickly based on the dataset and the ML model.

Additive friction stir deposition (AFS-D) is considered as effective method of additive manufacturing (AM) due to its capability of producing dense parts at a high deposition rate when compared to other AM technologies. Al6061 alloy is extensively used in the nuclear and aerospace sectors and is exposed to high-energy particles during its application causing damage to mechanical properties. However, the behavior study of Al 6061 parts produced from AFS-D when exposed to radiation remains unexplored. In this work, the AFS-D as-deposited Al6061 alloy and the Al6061 feedstock rod were implanted with 10 dpa of He+ ion at ambient temperature. To explore the microstructural and mechanical changes brought by irradiation of He+, SEM, EDS, TEM, and nano indentation studies have been performed. This study reveals that at 10 dpa of irradiation, AFS-D as-deposited Al6061 produced less density of He bubbles than the feedstock Al6061. The nano-indentation tests confirmed that the hardening effect of irradiation on feedstock Al 6061 is more pronounced than the AFS-D as-deposited Al 6061 sample.

Date

10-19-2023

Committee Chair

Shengmin Guo

Download

Available for download on Sunday, October 18, 2026

Share

COinS