| dc.description.abstract | Technological advancements such as efficient jet engines and nuclear reactors rests on our ability to design and discover new materials. Historically, discovery of new materials has relied on Edisonian approach, but it is doubtful that this approach will efficiently meet our future needs. Thus, the aim of this dissertation is to enable goal-oriented material design and discovery. A goal-oriented closed-loop material design framework rests on our ability to correlate the design space (input) and objective (output). Most design frameworks tend to utilize a single source of data or information to exploit the input-output correlation even though often there are multiple sources of information. Herein, a closed-loop multi-fidelity Bayesian optimization framework is used to efficiently exploit a wide variety of information sources to design a dual-phase material with a targeted property. While fundamentally materials science involves the study of processing/chemistry - microstructure - property correlations, in practice, material design involves finding optimum processing/chemistry that yields desired properties, and the microstructure information is used to rationalize the observations. This raises a fundamental question, can the intermediate microstructural information aid in a material design campaign. To answer this, a novel microstructure aware design approach is proposed and compared against the traditional microstructure agnostic approach. The results show that the knowledge of material microstructure does not only rationalize an observation but can also accelerate the design process. Furthermore, some of the most critical material performance metrics depend on the detailed description of the length-scales associated with the material microstructure e.g., crack growth resistance. Intuitively, crack growth resistance of a material can be enhanced by microstructural design. However, microstructural design calls for a computationally efficient method to assess material’s crack growth resistance within a closed-loop iterative design framework. To this end, a novel computationally efficient method utilizing evolving graphs and microstructural unit events is developed and validated against the results of microstructure-based finite element calculations of ductile fracture. The fully validated method is then used to design material microstructures with enhanced intergranular crack growth resistance. | |
