Joint Sessions of AIM, ICME, & 3DMS: Industrial Case Studies
Program Organizers: TMS Administration

Wednesday 9:10 AM
June 18, 2025
Room: Platinum Ballroom 7&8
Location: Anaheim Marriott

Session Chair: Evan Adcock, Carnegie Mellon University

9:10 AM  
Efficient, Coupled Process-Structure-Property Simulations of Additive Manufacturing Using the “Materialize” Framework: Brodan Richter1; Joshua Pribe2; George Weber1; Edward Glaessgen1; Evan Adcock1; 1NASA Langley Research Center; 2Analytical Mechanics Associates
    Process-structure-property (PSP) simulations have the potential to guide and supplement experiments through physics-based insight into the additive manufacturing (AM) process. However, implementing PSP simulations often requires difficult, bespoke coupling of various software packages that use a range of programming languages. This presentation introduces “Materialize”, a Python-based framework recently developed by NASA Langley Research Center, to implement coupled physics-based PSP models and support integrated computational materials engineering workflows. Materialize was conceived to efficiently link computational materials models across length scales and PSP space, with a particular emphasis given towards supporting exploratory studies in AM applications. The use of Materialize to perform GPU-accelerated process-structure simulations of transient temperature fields and microstructure evolution during AM is presented. The linking of process-structure simulations to structure-property simulations is then discussed, and the full PSP pipeline will be highlighted. The results demonstrate capabilities of Materialize that are intended to streamline physics-based PSP simulations of AM.

9:30 AM  
Pinax: A Machine Learning Platform for Data-Driven Materials Development: Satoshi Minamoto1; Takuya Kadohira1; Masahiko Demura1; 1National Institute for Materials Science
    The National Institute for Materials Science (NIMS) is developing a robust platform to accelerate data-driven materials research. To date, NIMS has built an ecosystem that includes “RDE”, an IoT-based data collection infrastructure; "MatNavi", a comprehensive materials database; and "MInt", a computational platform for solving complex scientific problems. Recently, we have developed "pinax", which functions as a centralized hub for data aggregation and machine learning playground. This platform provides advanced analysis provenance management, which improves the efficiency, reproducibility, and accuracy of model development. We aim to strengthen the interaction between these platforms and maximize the comprehensive capabilities of the research ecosystem.

9:50 AM  
Materials Microstructure Design Integrated With Image-Based Simulation: Nelly Nunheim1; Oliver Rimmel1; Mike Marsh1; 1Math2Market
    Imaging workflows that create digital twins of physical samples and then computationally measure the material properties of the samples through simulation are well-demonstrated and valuable characterization techniques. It is desirable to then digitally adjust the parameters of the sample (e.g. fiber volume content, fiber orientation, void content, etc) to derive new materials with improved properties. Here we show how to make digital twins of empirical samples, statistical twins with identical parameters, siblings with single-parameter changes, and cousins with multi-parameter changes. This approach informs the engineering design team of the material properties of a wide variety of samples. In this example, we systematically vary fiber volume content and void content in a fiber-reinforced composites application and solve the mechanical properties for each digital experimental variant, reaffirming the “Digital Transformation” pattern that provides clear economic and scalability advantages. The framework shown extends to conduction, flow, acoustic, and other properties characterization.

10:10 AM  
Architecture for Developing an Image Recognition Model Workflow for Workplace Safety Application: Kyle Toth1; Monika Singhal1; Chenn Zhou1; Chason Ault2; Matt Liddick3; 1Purdue University Northwest; 2Steel Dynamics, Inc.; 3Charter Steel
    This research presents a computer vision-based safety system using multiple models, utilizing the YOLOv8 architecture, to enhance safety by detecting workers and ensuring compliance with personal protective equipment (PPE) requirements. While the second model, trained on industry-specific and open-source data, detects PPE such as safety jackets and helmets with accuracy, the first model identifies workers. Additionally, the system monitors marked static hazard zones, issuing real-time alerts when workers enter these dangerous areas. This multi-model approach offers a practical solution for improving safety protocols and preventing accidents in steel manufacturing.

10:30 AM Break

10:50 AM  
Smart Sustainable Packaging for Local Fruits—TRACE Your Food, KNOW Your Food, TAKE CARE of Trash: Mudra Kapoor1; Parijat Deshpande1; Henil Panchal1; Hitanshu Sachania1; Shrikant Kapse1; Shankar Kausley1; Beena Rai1; Riddhi Jain1; 1TCS
    This study proposes an innovative approach to sustainable fruit packaging to ensure optimal freshness, traceability, authenticity and responsible consumption while reducing environmental impact. We integrate smart sensors on fruit packaging that provides consumers with transparent, authentic information about the product, including harvest date, farm name, nutritional properties, and disposal methods. Using Computational Materials Engineering approach and augmented AI tools, we design packaging material with high barrier functionality, sustainability, and compostability, addressing performance and environmental impact. We achieve this by screening the material requirement from our in house library of sustainable materials. The smart sustainable packaging promotes high end consumer experience to trace their food, gain nutritional information, real time assessment of quality and responsible disposal of food waste and packaging waste. This is achieved by digital tools that scan the food information in real time as input, processes it in backend and gives requisite output using mobile application.

11:10 AM  
Uncertainty Quantification, Error Propagation, and Sensitivity Analysis for Synchrotron X-Ray Residual Stress Measurements: Diwakar Naragani1; Chris Budrow2; Kelly Nygren1; Paul Shade3; 1Cornell University; 2Budrow Consulting; 3Air Force Research Lab
    High-energy synchrotron X-rays can measure lattice strains in structural materials under an energy-dispersive or an angle-dispersive modality. During the experimental data collection and analysis workflows several choices are made about the instrumentation, scanning procedure, profile fitting, and material constants that can significantly affect the strain calculated from the diffraction signal. We can explore the impact of these choices within a structured Bayesian framework to deliver accurate and reliable measurements. First, Bayesian uncertainty quantification is used to augment the typical process of calibrating the experimental setup. Second, the determined uncertainty is propagated through the reconstruction software to ascertain error bars on the reported strains. Third, uncertainty due to the material, specifically due to the reference lattice and elastic constants, is also propagated to the predicted residual stress. Finally, a global sensitivity analysis is used to understand the relative importance of these choices for the reported uncertainty on residual stress.