ATA Engineering, Inc., (ATA), in collaboration with the National Institute for Aviation Research (NIAR) at Wichita State University, proposes to develop Volume Image-based Virtualization for Allowables (VIVA), a software toolkit for the automated physics-based prediction of multiscale in-situ effective mechanical properties of as-built laminated composites reconstructed numerically from X-ray computed tomography (CT) data. The toolkit takes a multistage approach to property generation, starting with a state-of-the-art rule-based and data-driven machine learning-based image segmentation procedure to generate multiscale sub-surface geometric models from input CT data that include laminate structure features, such as ply waviness, orientation, interfaces, and drops, and critical defects, such as ply wrinkles, voids, and resin-rich regions. Multi-fidelity finite element (FE) models will then be created, resulting in the generation of a set of multiscale as-built representative volume element (RVE) models. Virtual characterization of the RVEs will generate a relationship between the material scales that is finally propagated through the component scale as a set of local mesoscale and global macroscale in-situ, nonlinear, anisotropic effective mechanical properties encompassing various hot/wet conditions and collectively supporting static, dynamic, and progressive damage analyses. In the Phase I Base, a prototype toolkit will be developed and demonstrated. The Option period will then prepare for verification and validation tests and software enhancements, including automated property assignment for different commercial FE types, that will take place in Phase II. Once complete, VIVA will provide an efficient, extensible, scalable, and accurate physics-based simulation capability that will enable more definitive and cost-effective decision-making for repair limits and disposition of non-conforming composite components.
Benefit: The large-scale laminated composite parts used in rotary- and fixed-wing aircraft structures are expensive, microstructurally complex components with long manufacturing lead times. Recent advances in nondestructive evaluation technologies, specifically X-ray CT scanning, have significantly improved the ability to detect non-conforming composite parts through imaging of sub-surface features, defects, and damage. However, despite several research-based investigations, there remains no clear link between measurable defect properties and the likelihood of fatigue failure of these costly parts, which is exacerbated by the significant uncertainty among the defects generally known to be acceptable versus unacceptable. This results in an unclear and inefficient decision-making process for the disposition of non-conforming parts, which impacts newly manufactured, field-damaged, and repaired components. The envisioned capabilities of the proposed VIVA toolkit will enable damage and durability analysis of as-built parts using commercially available FE analysis software at an acceptable computational expense. Results of these simulations can be used across the Naval Aviation enterprise, as well as in aerospace and automotive markets, to provide quantitative support of decision-making for the disposition of non-conforming composite components as well as significant program cost and schedule savings through reduced scrap rates.
Keywords: Machine Learning, Machine Learning, Manufacturing Defects, Composites, computed tomography, Progressive Damage Analysis, Multiscale Material Characterization, Finite Element Analysis, Fatigue
