ICME 2023: Mat Data & Platform: II
Program Organizers: Charles Ward, AFRL/RXM; Heather Murdoch, U.S. Army Research Laboratory
Monday 3:20 PM
May 22, 2023
Room: Caribbean IV
Location: Caribe Royale
Session Chair: Raymundo Arroyave, Texas A&M University
3:20 PM
Database Design Strategies for Coordinated Simulation and Testing in Additive Manufacturing: Andrew Kitahara1; George Weber2; Samuel Hocker2; Brodan Richter2; Joshua Pribe1; Edward Glaessgen2; 1National Institute of Aerospace; 2National Aeronautics and Space Administration, Langley Research Center
The qualification and certification (Q&C) process presents a significant challenge for widespread adoption of additive manufacturing (AM) materials and processes for aerospace applications. A relational database framework will be presented as a tool for data curation of coordinated experimental and computational materials modeling research activities. A comparison of relational and hierarchical data structures in this domain will be emphasized through the evolution of a database design strategy. This framework’s mission is to support the advancement of computational materials-informed Q&C by providing the necessary data infrastructure to trace reliability and reproducibility measures through unified AM materials simulation and experimental testing. FAIR (findable, accessible, interoperable, and reusable) data will be highlighted as a necessary precursor for automation of specific actions, which ultimately reduces the time and expense burden for Q&C. The discussion will be mostly limited to back-end design elements, though a few front-end user experience examples will also be shared.
3:40 PM
Applications of CALPHAD Based Tools to Additive Manufacturing: Amer Malik1; Minh Do Quang1; Johan Jeppsson1; Andreas Markstrom1; Paul Mason2; 1Thermo-Calc Software; 2Thermo-Calc Software Inc
Viscosity, surface tension of liquid, thermal/electrical conductivity, and their composition and temperature dependence, are all important properties for modeling the AM process together with heat and heat capacity. While the treatment of heat and fluid flow is considered as state of the art in current FEM models, material properties are treated in a highly simplified manner. During the last few years new models to predict thermophysical material properties using the CALPHAD method have been developed and by adopting a unified treatment of both process parameters and alloy dependent thermophysical properties, there is the potential for more accurate solidification simulations for AM. This presentation will describe how such an approach has been adopted to implement into Thermo-Calc a module for Additive Manufacturing. This combined approach can be thought of as an example of ICME (integrated computational materials engineering). Examples will be shown together with other ongoing/completed developments related to AM.
4:00 PM
Databasing Through the AM Pipeline: From Powder to Part: Srujana Rao Yarasi1; Elizabeth Holm1; Amir Barati Farimani1; Anthony Rollett1; 1Carnegie Mellon University
There is a need for consolidation of multi-modal datasets obtained from sensors within AM machines, build planning, and post-build imaging and analysis, going through the entire AM pipeline. We propose a database repository, built on the FAIR principles, to assist the application of machine learning tools that can harness the immense amount of data that has already been generated and set up protocols to deal with future data generation. This ranges from powder characteristics to build data to part analysis. The challenges of effectively using machine learning tools often come from a lack of sufficient and structured data. Enabling this structure throughout the AM pipeline will set a standard for future ML analysis and by knitting together disparate data sources, we can facilitate data fusion and analysis across the entire AM pipeline. We address the challenges of storage capacities, file naming conventions, and data retrieval for analysis.
4:20 PM Cancelled
CRADLE a Data Infrastructure for Printable Corrosion-resistant Alloys: Xiaoli Yan1; Pikee Priya1; Phalgun Nelaturu2; Dan Thoma2; Santanu Chaudhuri3; 1University of Illinois at Chicago; 2University of Wisconsin–Madison; 3Argonne National Laboratory
Corrosion-Resistant Alloy Design and Lifetime Evaluations (CRADLE) platform is designed to combine first-principles data, machine learning, and data visualization for improving corrosion resistance. The platform consists of a database server, a data processing server, a machine learning (ML) server, a multi-objective optimization code, and a web-based interactive GUI. The MongoDB database can use in-house first-principles calculation results and open public access databases. The closed-loop corrosion prediction and alloy design using surface energy, work functions, and texture in a machine-learning model can be improved by users using their own data. The CRADLE framework can fuse data sets, use ML to predict higher corrosion-resistant phases, and launch first-principles calculations as needed. We will demonstrate the case of printable high entropy alloys in search of more corrosion-resistant and single-phase high-entropy alloys in the Fe-rich and low-Ni phases. The CRADLE is open access with web-based with local Docker installation.