Data-driven Multi-scale Analyses of Materials and Structures

Data-driven research has the potential for groundbreaking achievements in the design of materials and large-scale engineering structures. A new data-driven computational framework for the design of heterogeneous materials is created. The framework recursively integrates design of experiments, efficient computational analyses of each design, and data mining. This enables the discovery of the influence of the microstructure and its building blocks on the macroscopic material behavior by creating a property-structure-performance feedback loop. The framework is applied to advanced composites by developing a high-fidelity multi-scale model for these materials and then using a general method that reduces the computational cost of the high-fidelity analyses called self-consistent clustering analysis. The overarching goal of this dissertation is to enable the future design of new materials with new capabilities.