| dc.description.abstract | Microstructure sensitive design (MSD) has become an essential component of Integrated Computational Materials Engineering (ICME). There are effectively two components to MSD, multi-scale, microstructure sensitive materials models and design frameworks.
The models used in materials science show much variety, from atomistic Density Functional Theory models to Finite Element mechanical or Phase Field models and beyond. However, all these models attempt to construct Process-Structure-Property-Performance relationships that allow for predicting material microstructure, properties, and performance from either microstructure descriptors or material processing parameters.
Design frameworks can take many forms. However, for the current work, the aim is to use a Reification-Fusion framework. This Framework uses multiple information sources (i.e., models) and fuses them after estimating the model correlation using a process called Reification. This fused model is then used to optimize the material properties in a Bayesian Optimization framework.
One of the more recent developments in the materials science community has been the building of high-throughput experimental methods. These methods have typically relied on thin-film approaches. However, more recently, additive manufacturing methods have started to play a more prominent role. One of the significant challenges is how to use design methods to guide high-throughput experimentation by making batch predictions. A recent batch Bayesian optimization approach has shown promise in this regard.
This work details the construction, testing, and application of a novel design framework. The Batch Reification/Fusion Optimization (BAREFOOT) Framework combines the Reification/fusion and Batch Bayesian Optimization approaches. The Framework is developed in Python and can conduct optimizations using both the individual approaches and the combined approach. The optimization can be paired with computational models or experimental tests as the target source that is being optimized. Furthermore, the results show that all the approaches in the Framework are capable of reducing the time required to optimize the property of interest and provide significant benefit to the materials design community.
As such, the BAREFOOT Framework has been proven to be a flexible tool that can be used easily in materials design approaches to speed up material design and development. Despite this, much work can be done to bolster the options available in the Framework further. | en |
