Fatigue in Materials: Fundamentals, Multiscale Characterizations and Computational Modeling: Applications of Machine Learning and Data Science to Fatigue Studies
Sponsored by: TMS Materials Processing and Manufacturing Division, TMS Structural Materials Division, TMS: Mechanical Behavior of Materials Committee, TMS: Computational Materials Science and Engineering Committee, TMS: Integrated Computational Materials Engineering Committee, TMS: Advanced Characterization, Testing, and Simulation Committee, TMS: Additive Manufacturing Committee
Program Organizers: J.C. Stinville, University of Illinois Urbana-Champaign; Garrett Pataky, Clemson University; Ashley Spear, University of Utah; Antonios Kontsos, Drexel University; Brian Wisner, Ohio University; Orion Kafka, National Institute Of Standards And Technology

Monday 2:00 PM
March 20, 2023
Room: Sapphire H
Location: Hilton

Session Chair: Ashley Spear, University of Utah; Orion L. Kafka, National Institute of Standards and Technology


2:00 PM  Invited
Capturing Spatial Fields of Deformation ahead of Fatigue Cracks in Alloys Using Dictionary-based Data Reduction Strategies on In Situ High-energy X-ray Diffraction Data: Kelly Nygren1; Daniel Banco2; Akihide Nagao3; Shuai Wang4; Matthew Miller1; Eric Miller2; Darren Pagan5; 1Cornell University; 2Tufts University; 3Air Liquide; 4Southern University of Science and Technology; 5Pennsylvania State University
    To capture the plastic deformation field ahead of a fatigue crack tip in 316L stainless steel, high-energy X-ray diffraction is combined with a new, dictionary-based data reduction algorithm. Strategies for reducing diffraction data on an area detector typically require either fully isolated diffraction peaks or fully overlapping diffraction rings, conditions often met by modifying a materials microstructure. The dictionary-based data reduction algorithm was designed to address the growing need to study “real” materials whose microstructures are complex and produce partially overlapping diffraction peaks. Utility of this method is demonstrated with a study of hydrogen pre-charged 316L austenitic stainless steel to elucidate the role of hydrogen on the development of plasticity ahead of a fatigue crack tip. Hydrogen is found to dramatically localize and enhance the defect generation ahead of the crack tip. The application of this technique for in situ fatigue crack growth studies will be discussed.

2:30 PM  
Using Computer Vision to Identify Crack Initiation and Link to Fatigue Life: Katelyn Jones1; Paul Shade2; Reji John2; William Musinski2; Elizabeth Holm1; Anthony Rollett1; 1Carnegie Mellon University; 2Air Force Research Laboratory
    This work seeks to collect SEM images of Ti-6AL-4V fatigue fracture surfaces and apply Convolutional Neural Networks (CNNs) to make a connection between fracture surfaces and fatigue life. SEM images of the crack initiation site, short crack region, and steady crack regions from varying fatigue life samples were taken at multiple magnifications to determine which length scale allows the machine learning algorithms to infer physically meaningful information. The images are compared using first unsupervised and then supervised machine learning methods to additionally determine which part of the fracture surface provides the information that links the fracture surface to the fatigue lifetime. The images taken, the algorithms used, identified fatigue properties, and fracture characteristics will be presented.

2:50 PM  
A Machine Learning Model to Predict Fatigue Progression Using 3D Topology Data of Materials Obtained from X-ray Microscope: Gunjick Lee1; Leslie Tiong2; Donghun Kim2; Seok Su Sohn1; 1Korea University; 2Korea Institute of Science and Technology
     It is very difficult to quantitatively determine how much fatigue failure has progressed. In particular, from the macroscopic point of view, it is more difficult to notice the progress of fatigue failure until almost immediately before failure occurs. However, from a microscopic point of view, void and micro-cracks change as the fatigue progresses. Therefore, it is necessary to correlate with microscopic information to determine how much fatigue failure has occurred, and to predict how much life is left.X-ray microscope can measure the microstructure of a material in 3D form without damaging the material. The measured 3D topology data includes information of voids and micro-cracks inside the material, and it has a very complex shape for human analysis. In this study, we develop a machine learning model that predicts the fatigue progress and remaining life by using this complex 3D topology data.

3:10 PM  
Intelligent Data-guided Process Design for Fatigue-resistant Steel Components with Bainitic Microstructure (iBain): Ingo Steinbach1; 1Ruhr-University Bochum
     The fatigue life of martensitic and bainitic steels is heavily influenced by the microstructure, which is in turn highly dependent on the manufacturing process. These microstructures are hierarchically multi-scale and exhibit structural properties down to the nanoscale. As a result, the experimental characterization is already exceedingly complex and heavily dependent on the subjective experience of the experimenter itself. Consequently, these microstructures are ideal case studies for materials integration, including experiments, data-driven techniques, and direct microstructure modelling. Here, we present: • A phase-field approach for modelling martensite and bainite microstructure formation. • A systematic approach to the characterization of microstructures. • Correlation process microstructure material properties – service lifetime. • Prediction of fatigue lifetime via a microstructure informed Kocks-Mecking type fatigue model.The project, iBain, is supported within the German research initiative Material Digital [https://www.materialdigital.de/].

3:30 PM  
Machine Learning Segmentation Methods for Fatigue Fracture Surface Defect Analyses: Austin Ngo1; David Scannapieco1; Oluwatumininu Adeeko1; Shuheng Zhang1; Shuyue Bian1; Collin Sharpe1; John Lewandowski1; 1Case Western Reserve University
    Machine learning algorithms for feature segmentation were utilized on fracture surfaces of AM Ti-6Al-4V fatigue samples. Process-induced defects were identified and quantified by training an image classification system using SEM images of fatigue fracture surfaces. Process defects were found to range in size, shape, and population due to variations in AM build process parameters. Different types of process defects (i.e. lack of fusion, keyhole) were found to be more prevalent based on particular process parameter sets. All process-induced defects across each fracture surface were quantified, with ‘killer’ fatigue crack initiating defects being identified. The fracture surface defect characteristics are compared to the corresponding S-N fatigue data for defect-based fatigue life modeling in a defect-based Kitagawa-Murakami-type approach. In addition, this computer vision method is compared to manual quantification of fracture surface defects. The advantages of implementing ML algorithms to streamline fracture surface defect quantification will be discussed.