Extreme environments are ubiquitous in military systems, which during service may experience chemical attack, mechanical stress, irradiation, extremes in temperature and pressure, or combinations of these conditions. Materials determine the performance of platforms like hypersonic aircraft, naval vessels, and nuclear reactors, defining maximum and minimum allowable usage temperatures, limiting service lifetime, and dictating maintenance requirements. For example, the permissible speed of a hypersonic aircraft is directly limited by the thermal properties of the leading edge and thermal protection system in atmosphere. (Sziroczak and Smith 2016) Meanwhile, many platforms will experience cracks and mechanical failures after repeated or long-term exposure to...

In this section, we focus on only a few examples of data science or statistical learning methods that have been applied in materials research and appear promising for future work on materials in extreme environments. The purpose of this section is to illustrate ways in which these methods have been successful, rather than to fully review the topic. For more complete coverage of previous materials data science work, we refer the reader to a group of recently published comprehensive reviews (Kalidindi and Graef 2015; Rajan 2015; Agrawal and Choudhary 2016; Ramprasad et al. 2017).

Classification is a common goal of...

Materials data science tools fall into several broad classes: databases, data repositories, and analysis tools. This chapter will briefly review available resources in each area, citing strengths, weaknesses, and application as appropriate. The challenges of hosting, maintaining, and curating such databases are well-described in multiple reviews (Tenopir et al. 2011; Kalidindi and Graef 2015; DMMI 2014). The resources mentioned here are listed in Appendix A, but the highlighted resource registries should provide more comprehensive lists.

Databases for material science may contain computational data, experimental data, or both, and may be commercial, open access, or limited access. This discussion will not...

In this section, we describe the factors that contribute to a materials data science problem, and highlight a few target problems in the area of materials in extreme environments that appear well-suited to a data science approach.

When structuring a study using data science tools, the following list of questions should help guide choices related to data gathering, analysis methods, and variable selection. These questions need not be asked in the order listed, as analysts may not have control over data collection.

An explanatory study seeks to understand which knobs to turn to control a particular outcome, and to explain...

In this chapter, we apply a few classification methods to experimental data on cast high entropy alloys. The data come from a table available in previous work that sought to classify alloys by their ability to form solid solutions based solely on composition (Yang and Zhang 2012). That study found that the average atomic size difference δ and the ratio of entropy and enthalpy of mixing Ω could be used to discriminate between solid-solution and mixed-phase compositions. Here, we focus on discriminating between particular phases—face-centered cubic (FCC), body-centered-cubic (BCC), and intermetallic phases labeled by α and σ.

Phase selection...

This report covered a broad spectrum of topics in order to build a picture of the current status, limitations, and opportunities of data science and statistical learning methods for understanding, modeling, and designing materials for extreme environments. We conclude with key observations on areas where these approaches could be of the most immediate value, with some comments on barriers to adoption and the necessary shifts in institutional policy and workforce training needed to facilitate this kind of work.

A single, universal database that applies to all materials of interest and incorporates all relevant experimental and computational results in a validated...