Data science and material informatics in physical metallurgy and material science: An overview of milestones and limitations

Abstract

Data science and material informatics are gaining traction in alloy design. This is due to increasing infrastructure, computational capabilities and established open-source composition-structure-property databases increasingly becoming available. Additionally, the popularization of data science techniques and the drive to reduce overall material life-cycle cost by ∼60% have necessitated increased use of the technique. Alloy design is a multi-optimization problem hence the Edisonian approach is no more viable from cost, labour, and time-to-market perspectives. Although, there have been successful application of data science and material informatics in alloy design, there are drawbacks. This review provides a critical assessment of limitations associated with data science and materials informatics to alloy discovery and property characterization. Among these are cost, false positives, over – and underestimation of properties, lack of experimental data to validate simulated results, lack of state-of-the-art facilities in most developing countries and uncertainty modelling. The implications and areas for future research directions are highlighted.

Keywords

• Alloy development
• Computational alloy design
• Machine learning
• Material informatics
• Rational alloy design