Data-Driven Physics-Based Modeling Tools to Determine Effective Mechanical Properties of As-Built Composite Structures

This opportunity targets the development of a software toolkit to automate the generation of nonlinear, anisotropic mechanical properties for as-built composite structures, including the effects of defects, to accelerate finite element (FE) analysis for fleet repairs and aircraft production non-conformal dispositions.  To meet this objective, Navy wants to implement a data-driven, physics-based, modeling approach including material structure and full-field deformation by analyzing the parts in the as-built condition with their own unique configuration, including manufacturing defects and service damage. Accordingly, the advances in the state of the art of the in-situ Computed Tomography (CT), Scanning Electron Microscopy (SEM), and Digital Image Correlation (DIC) / Direct Strain Imaging (DSI) data driven multiscale micro-meso-macro mechanics modeling are proposed to meet the material input data requirements for predicting residual strength and fatigue life of the as-built composite structures.  The data/measurement resolution requirements include the appropriate length scale(s) associated with material system components (e.g., ply thickness/orientation, fiber path/bundle/volume, fiber/resin, and adhesive interfaces) and manufacturing defects (e.g., porosities/voids, wrinkles, delamination, and fiber waviness). The most critical defects, affecting structural integrity, include combinations of wrinkles, porosity/voids, and resin-rich or adhesive-rich zones, which will be captured by the model with an effective relationship to the FE mesh and intended analysis.  The proposed toolkit would also account for material degradation due to repeated loadings and Hot/Wet (H/W) operating environments.  Due to the size of the data for a full-scale component, speed and accuracy issues relating to data acquisition, image processing, and data storage and retrieval will also be addressed, including the use of machine learning (ML) techniques. The Phase I effort will demonstrate technical feasibility of the proposed concept to develop a computationally efficient, multiscale, physics-based, modeling toolkit coupled with CT-scanned data, machine learning, and computer vision techniques to generate in-situ, quasi-static, and dynamic effective mechanical properties (stiffness, strength, and strain energy release rate) for as-built, thick laminate composite structures, including effects of defects, repeated loadings, and expected H/W operating environments. Demonstrate the proposed workflow to auto-populate the input data for different 2-D and 3-D FE meshes, including various element sizes and types to support progressive damage analysis of thick laminate composite structures.  Develop a verification and validation (V & V) test plan for the proposed concept, including, at a minimum, the use of DIC.