Building a Robust Framework to Crowdsource Material Property Data for Structural Alloys Processed Through Advanced Manufacturing Routes
Open Access
- Author:
- Wietecha-Reiman, Ian
- Graduate Program:
- Materials Science and Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- February 19, 2025
- Committee Members:
- John Mauro, Program Head/Chair
Jay Keist, Major Field Member
Todd Palmer, Chair & Dissertation Advisor
Darren Pagan, Outside Unit Member
Albert Segall, Outside Field Member - Keywords:
- materials science
metallurgy
statistics
machine learning
fatigue
tensile
austenitic stainless steel
316L
titanium
alloy
alloys
Ti-6Al-4V
Ti64
design allowables
additive manufacturing
AM
laser-powder bed fusion
powder bed fusion
L-PBF
directed energy deposition
DED
meta-analysis
meta-data
data quality
data stewardship
database
microstructure
inclusions
secondary phases
porosity
defects
reliability
composition
powder feedstock
supply chain
data science
oxides
nitrides - Abstract:
- Design allowables for mature manufacturing processes, such as forging and casting, are typically derived from material property databases constructed using a set of highly controlled procedures. Comparable data repositories for advanced manufacturing processes, such as additive manufacturing (AM), are far less developed and would be costly and time consuming to the point of infeasibility. One alternative approach for obtaining comprehensive estimates of materials properties produced through these advanced routes is to aggregate historically available data. Given the complexity of these processing routes, post processing, material, and testing parameters need to be compiled as additional meta-data. Since these meta-data tend to go unreported, a robust framework for processing data has been developed by profiling meta-data structure and imputing unreported values. Once compiled, various statistical and machine learning models can then be applied to datasets for AM 316L austenitic stainless steel and Ti-6Al-4V. Marginal contributions to fatigue life from microstructural- and porosity-driven failure mechanisms were extracted and sources of fatigue life scatter could be identified. The framework can be extended to assimilate quantitative microscopy, fractography, and non-destructive evaluation results into the meta-data structure to improve model precision and direct more efficient future investigations. By extending it to the development of crowdsourced materials property databases, the larger community can use these data sets to develop baseline material properties for important alloy systems and to train advanced machine learning models, helping accelerate the eventual development of the high pedigree and accuracy data sets required to create design allowables.