A Computationally Driven Approach to Nano-engineered Composite Structures

A Computationally Driven Approach to Nano-engineered Composite Structures A Computationally Driven Approach to Nano-engineered Composite Structures We propose an atomistically driven multiscale computational framework to investigate the capabilities of nanocomposites with novel architecture to enable concurrent (bottom-up and topdown) traceability in analyses. The emphasis is to characterize material behavior at the nanoscale and the effective transfer of this information to model structural scale response, damage and durability. The framework integrates models and algorithms that integrate and transfer information from the atomistic to the mesoscale, uncertainty quantification, computationally efficient reduced order techniques and bench mark laboratory experiments. Atomistic scale analysis of each constituent including the fiber, matrix, carbon nanotubes (CNT) and their related architecture, coating/substrate will be addressed and modeled to capture possible interatomic/intermolecular interactions in various CNT fiber composites with novel architecture. Both linear and nonlinear dynamic behavior will be captured. A recently developed hybrid MD simulation methodology, which combines classical force fields for elastic response and bond order based force fields for inelastic response analysis, will be further extended to simulate the complex mechanochemistry phenomena. The MD simulations will be used to generate a physical cohesive law that describes damage initiation and propagation at interfaces and in the bulk material. Furthermore, a continuum framework, spanning micro to mesoscale that investigates the static and dynamic damage initiation and propagation will be modeled based on results from the nanoscale; the stochasticity in the relevant parameters will also be considered at this length scale. A novel machine learning based Bayesian approach will be used to formulate adaptive reduced order models (ROMs) of the high fidelity atomistic and microscale analysis. The ROMs will be used to bridge the gap and optimal transfer of information across the length scales (between the nano-, micro-, and mesoscale); this will ensure computational efficiency. The ROMs will also be used to conduct sensitivity analysis with respect to problem parameters which will help evaluate the parameters that affect the output distributions; this will reduce the dimensionality of the ROMs and further improve the computational efficiency. Following are the innovative aspects of the proposed research: i. Atomistic model of various CNT fiber composite architectures to fully describe the interaction between all the phases including carbon fiber, CNTs, coating or substrate layer (if any), and polymer matrix ii. iii. iv. v. High fidelity explicit models of the unique CNT architectures using the method of embedded finite elements Physics-based damage evolution laws, merging the nano and sub-microscale damage effects, to investigate damage and failure at the continuum scales under both quasi-static and dynamic loading Novel computationally efficient bridging using adaptive ROMs to accurately predict the output distributions of relevant stochastic parameters at each length scale Scale-specific experiments for closed-loop validation. The proposed framework will lead to a comprehensive understanding of the effects of various atomistic and microscale parameters that leads to the design of optimized nanocomposite materials with improved mechanical properties and multifunctionality. It will help establish a computational material design framework with optimal selection of design parameters, including weight fractions, the height, density, type of CNT, polymer coating for carbon fiber protection, etc., to meet the design objectives in mission specific applications. Methodological developments in this research will be steered by a closed-loop design and validation plan that incorporates both computational modeling and experimental approaches to understand the underlying physics. The project is organized into the following research tasks: Task 1: Atomistic Modeling of CNT Fiber Architecture; Task 2: Continuum Scale Modeling and Structural Integration; Task 3: Adaptive Reduced Order Models for Bridging of Length Scales; Task 4: Scale-dependent Testing and Model Validation.