A Computer Aided Multiscale Material Design Optimization Framework for Composite Materials Tailoring

The need and the opportunity for significant savings in both time and cost for the engineered development of advanced nanomaterials coupled with the tremendous growth in the past couple of decades in computational materials science has not yet materialized into significant material design tool developments. Of particular importance in the engineering design of composite materials for various applications is the ability to tailor the constituent materials and the internal architectures. The inverse problem of determining an optimal microstructure for a desired application is a challenging task. This procedure has been traditionally accomplished by trial-and-error and depends considerably on the designer's intuition and experience. For this reason, obtaining new materials has been a time consuming and an expensive process. Accordingly, a systematic method capable of synthesizing the optimal microstructure that will satisfy the design requirements, while reducing cost and time, is desired. The intensive computational cost of numerical tools for material behavior analysis makes the use of iterative design and optimization procedures based on such simulations prohibitively expensive to perform. One therefore, requires a design approach that can incorporate multiple simulations of varying fidelity in design iterations, in an iterative manner, while simultaneously reducing the design cycle time. The present investigation focuses on the development of a simulation-based design optimization methodology to predict the most suitable microstructures of Silicon Carbide - Silicon Nitride (SiC-Si3N4) nanocomposites for desired high temperature properties. This work presents a systematic optimization methodology to predict optimal material microstructures, while considering uncertainties in the microstructural representations with simultaneous reduction in the design cycle time. Also, a trust region managed variable fidelity optimization framework is proposed in this investigation to address the computational challenges and model management issues that are inherent to multiscale material design. Although the material of interest in this investigation is Silicon Carbide - Silicon Nitride (SiC-Si3N4), the presented methods are not restrictive and could be an invaluable design tool to support the development of any type of materials. Overall, the result of the present investigation is a systematic method capable of predicting optimal microstructure that will satisfy the design requirements of targeted properties, while reducing cost and time.