News Archive Item
Paper shines spotlight on materials informatics
A paper recently published in International Materials Reviews provides an insightful look at materials informatics and how it is being employed to overcome obstacles in computational materials engineering. “Data science and cyberinfrastructure: Critical enablers for accelerated development of hierarchical materials,” by Surya R. Kalidindi of the Georgia Institute of Technology, also identifies specific areas of research that offer the most promise based on the success of earlier work.
As the title suggests, the methods and techniques discussed in the paper apply to a special class of materials, the simplest of which have observable features at two distinct length scales. What’s more, these features are hierarchically related and, collectively, provide a window through which various material properties can be controlled by appropriate processing steps. Until recently, materials informatics has been limited to explorations at a single length or structural scale, but in the quest to accelerate materials discovery, this emerging new discipline is being used with increasing regularity to analyze data at multiple length scales, helping researchers like Kalidindi better understand the process-structure-property (PSP) relationships inherent in hierarchical materials.
Identifying trends on the evolution of structural features is just the beginning, however. At the next higher level – using data mining techniques to account for variances and uncertainties – researchers can acquire much more rigorous, reliable, and complete information, generating what Kalidindi says can be characterized as materials knowledge. And beyond that, facilitated by invertible PSP linkages that allow for bidirectional simulations, lies the realm of automated design. A diagram that illustrates the various stages of data transformation equates this highest level of abstraction to “wisdom.”
In addition to taming the vast amounts of data involved in materials design, materials informatics is also helping researchers decide what data to collect. Data-driven decision support tools, as discussed in the paper, are being used to guide effort investment in both measurements and computations at various stages in the materials development process. Data science is also proving useful for simplifying calculations and reducing compute times. An entire section of the paper in which the author develops the concept of an extensible framework for structure quantification describes several of these data-reduction techniques.
Another topic covered in detail is the importance of models in the context of design optimization. Models can be a make-or-break factor and if they don’t meet certain standards in terms of maturity, interoperability, and inversion, they can do more harm than good. Standards and protocols related to infrastructure are discussed as well. To properly support collaborative design, a data infrastructure should include automated protocols for capturing and tracking data provenance and identifying salient data attributes (i.e., metadata). Equally important is the development of ontologies and domain lexicons to facilitate meaningful exchange of ideas, data, tools, and knowledge.
In closing, the paper suggests several areas of research where additional attention could yield significant benefits. The list is quite diverse, addressing everything from data science and modeling methods to novel test and measurement techniques.