There has been significant growth in applying composite materials for various engineering applications in the past several decades. Composites allow engineers to tailor material microstructures and constituents for desired structural properties that usually outperform traditional engineering materials such as alloys. However, due to the heterogeneous nature of composites' microstructure, composite structures exhibit complex behavior and variations in performance. Moreover, the current composite structures design process is still largely an experimental one, which is costly and requires a considerable amount of experience. This becomes an invisible barrier between composites’ extraordinary properties and short design concept to production time. Efficient and accurate numerical tools can facilitate the experimental process by narrowing down the scope of the experimental inputs by eliminating unacceptable designs leveraging the numerical predictions. In this dissertation, a framework aiming to construct a computational composite materials design platform is presented. A specific engineering composite material, Carbon Fiber Reinforced Polymer (CFRP), is considered, and various modeling strategies are proposed to predict properties CFRP microstructures. Starting with basic CFRP microstructures (e.g., unidirectional and woven composites), microstructure modeling techniques, and efficient Reduced Order Modeling method for predictions CFRP nonlinear responses are introduced. Then, a multiscale modeling framework is designed to leverage the microstructure information (e.g., elasto-plastic response directly computed by microstructure instead of constitutive laws) for the prediction of composite structural performance. The multiscale modeling framework provides the virtual testing capability of CFRP structures, demonstrated by two practical problems. Also, possible extension of the framework is discussed, such as predicting thermo-elastic responses of composites and incorporating microstructure uncertainty for robust composite design. It is believed that the methodologies presented in this dissertation can accelerate the traditional composite design process through the virtual testing capability for composite structures.

Creator

• Gao, Jiaying

DOI

• https://doi.org/10.21985/n2-t5y9-kv70

Subject

• Engineering

Language

• en

Alternate Identifier

• http://dissertations.umi.com/northwestern:15053
• etdadmin_upload_735144

Keyword

• UD composite
• woven composite
• reduced order modeling
• structural performance prediction

Date created

• 2020-01-01

Resource type

• Dissertation

Rights statement

• In Copyright