Deep Learning Unlocks X‐ray Microtomography Segmentation of Multiclass Microdamage in Heterogeneous Materials
Year: 2021
Type: article
Abstract: Four-dimensional quantitative characterization of heterogeneous materials using in situ synchrotron radiation computed tomography can reveal 3D sub-micrometer features, particularly damage, evolving under load, leading to improved materials. However, dataset size and complexity increasingly require time-intensive and subjective semi-automatic segmentations. Here, the first deep learning (DL) convolutional neural network (CNN) segmentation of multiclass microscale damage in heterogeneous bulk materials is presented, teaching on advanced aerospace-grade composite damage using ≈65 000 (trained) human-segmented tomograms. The trained CNN machine segments complex and sparse (<<1% of volume) composite damage classes to ≈99.99% agreement, unlocking both objectivity and efficiency, with nearly 100% of the human time eliminated, which traditional rule-based algorithms do not approach. The trained machine is found to perform as well or better than the human due to "machine-discovered" human segmentation error, with machine improvements manifesting primarily as new damage discovery and segmentation augmentation/extension in artifact-rich tomograms. Interrogating a high-level network hyperparametric space on two material configurations, DL is found to be a disruptive approach to quantitative structure-property characterization, enabling high-throughput knowledge creation (accelerated by two orders of magnitude) via generalizable, ultrahigh-resolution feature segmentation.
Source: Advanced Materials
Institution: Massachusetts Institute of Technology
Language: English
Subfield: Materials Chemistry
Field: Materials Science
Domain: Physical Sciences
Open Access status: green
Funders: National Science Foundation, National Aeronautics and Space Administration, Strong, U.S. Army Combat Capabilities Development Command Soldier Center, University of Southampton, Saab, Embraer, Smithsonian Tropical Research Institute, Quest for Intelligence, Massachusetts Institute of Technology, Army Research Office, Army Research Laboratory, National Science Foundation Graduate Research Fellowship Program
Awards: G1335864876, THE INSTITUTE FOR ULTRA-STRONG COMPOSITES BY COMPUTATIONAL DESIGN (US-COMP) WILL BE FOCUSED ON THE MODELING-DRIVEN DESIGN OF A NEW CLASS OF ULTRA-HIGH-STRENGTH-LIGHTWEIGHT (UHSL) MATERIALS FOR FUTURE MANNED MARS MISSIONS. THESE MATERIALS WILL MEET THE REQUIRED MECHANICAL PERFORMANCE GOALS SET FORTH BY NASA AND EXCEED THOSE EXHIBITED BY CURRENT STATE-OF-THE-ART CARBON-FIBER COMPOSITES. US-COMP S VISION IS TO SERVE AS A FOCAL POINT FOR PARTNERSHIPS BETWEEN NASA OTHER AGENCIES INDUSTRY AND ACADEMIA TO: (1) ENABLE COMPUTATIONALLY-DRIVEN DEVELOPMENT OF CARBON NANOTUBE (CNT)-BASED UHSL STRUCTURAL MATERIALS AND (2) EXPAND THE RESOURCE OF HIGHLY SKILLED ENGINEERS SCIENTISTS AND TECHNOLOGISTS IN THIS EMERGING FIELD TO ENHANCE THE U.S. LEADERSHIP IN CRITICAL LIGHTWEIGHT STRUCTURAL MATERIALS. THIS VISION WILL BE ACHIEVED THROUGH THE FOUR PRINCIPLE OBJECTIVES: ESTABLISH A NEW COMPUTATIONALLY-DRIVEN MATERIAL DESIGN PARADIGM FOR RAPID MATERIAL DEVELOPMENT AND DEPLOYMENT DEVELOP A NOVEL UHSL STRUCTURAL MATERIAL FOR USE IN DEEP SPACE EXPLORATION. THE PANEL-LEVEL TESTS AND DEMONSTRATION OF THE NOVEL MATERIALS WILL BE CARRIED OUT TO MOVE THE DEVELOPED TECHNOLOGY TO A TECHNICAL READINESS LEVEL (TRL) OF 4. DEVELOP NOVEL MODELING PROCESSING AND TESTING TOOLS AND METHODS FOR CNT-BASED COMPOSITE MATERIALS ESTABLISH A POOL OF HIGHLY SKILLED ENGINEERS AND SCIENTISTS TO CONTRIBUTE TO THE MATERIALS DEVELOPMENT WORKFORCE. AN INTERDISCIPLINARY AND DIVERSE TEAM OF RESEARCHERS FROM ACADEMIA INDUSTRY AND NATIONAL LABS WILL PARTICIPATE IN THE PROJECT. THE COMPUTATIONAL DESIGN OF THE MATERIAL WILL BE DRIVEN BY A MODELING EFFORT TO INTEGRATE TOPOLOGICAL OPTIMIZATION ATOMISTIC MODELING MOLECULAR MODELING MESOSCALE MODELING AND CONTINUUM-BASED COMPUTATIONAL MECHANICS. INNOVATIONS IN MATERIALS SYNTHESIS AND MANUFACTURING TECHNIQUES WILL ENSURE THE PERFORMANCE AND SCALE-UP FABRICATION OF AEROSPACE-QUALITY TEST SAMPLES AND PANELS. MULTISCALE TESTING AND CHARACTERIZATION CAPABILITIES WILL BE ESTABLISHED AND INTEGRATED TO VALIDATE THE MODELING AND MANUFACTURING EFFORTS AND TO COMPLETE THE PROOF-OF-CONCEPT CYCLE. PARTICIPATION OF THE INDUSTRIAL PARTNERS WILL PROVIDE AND ENSURE THE SCALABILITY AND AEROSPACE-GRADE QUALITY OF THE DEVELOPED COMPOSITE MATERIAL. THE DEVELOPED MATERIALS AND MATERIALS DEVELOPMENT METHODS WILL HAVE A MAJOR IMPACT ON THE AEROSPACE COMMUNITY. FIRST UHSL MATERIALS WILL BE DEVELOPED WITH THE RIGOROUS STRENGTH MODULUS AND FRACTURE TOUGHNESS PROPERTIES NECESSARY FOR MANNED MARS MISSIONS. SECOND A NEW COMPUTATIONALLY-DRIVEN MATERIALS DESIGN PARADIGM WILL BE ESTABLISHED TO DEVELOP THE UHSL MATERIAL OF INTEREST AND FOR FUTURE RAPID MATERIALS DESIGN AND DEVELOPMENT EFFORTS. THIRD A FUNDAMENTAL UNDERSTANDING OF LOAD TRANSFER AND MULTISCALE FAILURE MECHANISMS OF CNT-BASED COMPOSITE MATERIALS WILL BE ESTABLISHED TO ACHIEVE THEIR THEORETICAL PERFORMANCE. FOURTH A REPRODUCIBLE ENGINEERING PERFORMANCE DATA FROM AEROSPACE-QUALITY AND SCALE-UP PANEL TEST RESULTS BUILDING UPON AEROSPACE-GRADE RESIN SYSTEMS AND HIGH-QUALITY COMMERCIALLY AVAILABLE CNT MATERIALS TO ENSURE SCALABILITY TO CONDUCT ASTM STANDARD TESTS. ANOTHER IMPORTANT EMPHASIS OF THE INSTITUTE WILL BE IN WORKFORCE DEVELOPMENT. STUDENTS WILL BE TRAINED FOR DEVELOPING AND UTILIZING ADVANCED COMPUTATIONAL AND EXPERIMENTAL APPROACHES FOR LIGHTWEIGHT MATERIALS. FUNDS WILL BE RESERVED FOR THE STUDENTS TO HAVE EXTENDED VISITS TO NASA FACILITIES DURING SUMMER MONTHS FOR DIRECT MENTORSHIP BY NASA RESEARCHERS. THESE ACTIVITIES WILL STRENGTHEN THE PARTNERSHIPS BETWEEN THE INSTITUTE MEMBERS AND NASA. WITH THE HELP OF THE HBCU PARTICIPANT (FLORIDA A&M UNIVERSITY) THE INSTITUTE WILL ESTABLISH A DIVERSE GROUP OF BOTH RESEARCHERS AND GRADUATE STUDENTS., G2660901070, G2999579082, G3429140574, G3491862500, Graduate Research Fellowship Program, G3732666562, G5259331294, G6274897657, G7452299184
OpenAlex ID: W3213327264